LTE User Equipment, network efficiency and value Mobile broadband radio access economics - pre requisites for

profitability Published September 1st 2010
Executive summary – the LTE network cost and performance challenge The future of the mobile broadband industry is very much in the balance. In the next five years traffic volumes are forecast to grow by a factor of 30, from 3 to 90 exabytes. 1 On present tariff trends 2 revenue is projected to grow by a factor of 3. The projected efficiency savings from new technology such as LTE are insufficient to bridge this gap. New solutions to this dilemma are therefore required in order to enable the sustainable growth of the mobile broadband industry. This study identifies crucial links between user equipment performance and network economics and argues the case for year-on-year improvements to be made to UE performance over and above the minimum conformance requirements. A business model is presented which shows that a small increase in UE cost has a positive return when measured across the mobile broadband ecosystem. Key findings by industry sector For operators and carriers - the study and economic model validate that a move to LTE, if coupled with investment in band flexibility and improved year on year user equipment efficiency yields sufficient radio network efficiency gain and incremental user value to provide a positive and improved return on present and future spectral and radio network investment, or put another way, allows operators and carriers to make money out of data volume growth.
3 We argue that the combination of user equipment efficiency and extended multi band support will have a more positive impact on mobile broadband radio access economics than commonly assumed or presently modelled. Implementing additional bands should however only be considered if performance in existing bands is maintained 4

Over and above this, a year-on-year performance improvement of 1dB over and above the single band baseline but applied to multi band and extended multi band user equipment would yield year-on-year user experience gain (user value) and a reduction of direct and indirect radio network cost, delivering a positive investment cycle even if additional user equipment BOM costs are factored in to the equation. A 3 dB improvement over three years would double the average data rate per user or double the user data duty cycle, a combination of link budget and scheduler gain. This translates directly into additional realisable user experience value and reduced network cost per subscriber and device supported. For infrastructure hardware and software vendors We show how improving user equipment performance transforms scheduler efficiency and by implication the ROI model for infrastructure hardware and software investment. For user equipment hardware and software vendors We argue the case for a shift from user equipment cost optimisation to a combination of cost and performance optimisation and show how this yields a positive EBITDA return. For component hardware and software vendors and the test equipment hardware and software community.
1

An exabyte is a million terabytes

By extended multi band we mean more than the standard 5 bands presently supported and the eight bands presently on most vendor road maps. WiFi at 2.4 and 5 GHz, Bluetooth at 2.4 GHz, GPS receive at 1.5 GHz, FM receive at 100 MHz and (for larger form factor devices) ATSC and DVBT receive at 700 MHZ may/will also need to be supported. 4 Conformance specifications are already being degraded to accommodate multi band designs. The goal should be to exceed existing requirements rather than maintain par with requirements that are getting laxer over time.

3

Geoff Varrall

Page 1 LTE UE efficiency and network value Final

15/09/2010

We identify presently under exploited opportunities to add value to the whole industry supply chain. For industry, financial and investment analysts We demonstrate that provided certain preconditions are met, the demand side growth potential of mobile broadband is capable of delivering greater fiscal returns than presently reflected in mobile broadband industry equity valuation. LTE delivers spectral efficiency benefits over and above HSPA Release 6 by roughly an order of three 5 which is impressive but short of the increase needed to redress this traffic cost/income disparity. 6 An alternative is to reduce the cost of network hardware. Some network hardware costs scale with Moore’s law but some do not, RF hardware in the radio access network being one example. Although hardware cost reduction yields a reduction in capital cost investment, there are associated increases in through life software cost, an operational expense. Operators and carriers can scale up their present spectral assets to accommodate the additional data volume. However this amount of spectrum is unlikely to be made available, any spectrum made available will be expensive and, probably most important, present and planned user equipment lacks the band flexibility to access the spectrum effectively or efficiently. Unless an operator is bidding for spectrum for speculative gain, which is rightly discouraged by regulators, there is no point in buying bandwidth if it cannot be economically used. Another approach is to increase network density. This has the effect of improving the link budget which increases capacity, improves data rates and reduces user power budgets up to the point at which the radio network becomes interference limited.
7 The problem with this from an operator EBITDA perspective is that capital and operational costs increase, a composite of site acquisition, site rental, site energy cost, hardware and software investment and backhaul costs. 8 Using subscriber ADSL 9 lines as backhaul via femtocells partly provides a solution to the back haul cost issue and can increase local area network density in a cost-effective way at the subscriber’s expense. However femtocells address local area access economics not wide area or high mobility user access economics 10 . Femtocells do not deliver wide area or high mobility user value.

Similarly MIMO 11 achieves high peak data rates in small cells but if poorly implemented in user equipment compromises SISO and SIMO performance in larger diameter micro and macro cells where average data throughput is more important, having a more direct impact on user experience value. 12 User experience value is not realised just from connectivity but the relevance of the connectivity. Search algorithms have become worth more than handover algorithms. This is why operator data value is increasing at a slower rate than data volume. However mobile connectivity, including wide area and high mobility connectivity 13 , has the potential to realise additional value if specific user experience metrics can be addressed. This study quantifies the relationship between user equipment 14 (UE) performance, quality of service, quality of experience and operator and industry EBITDA return and demonstrates that a relatively small investment in user equipment performance optimisation yields a positive non linear gain in radio network efficiency and value. We argue that this is a prerequisite for industry profitability and a determining factor in mobile broadband radio access economics.
5

Analysed in Appendix 4 Page 63
The LTE gain over HSPA Rel-8 for example may be closer to 20%

6
7 8

Earnings Before Interest Tax Depreciation and Amortisation Backhaul costs are a particular issue in mobile broadband networks as the offered traffic will generally be more asynchronous and asymmetric than voice which means that in general bandwidth needs to be over provisioned to protect quality of service for latency sensitive traffic. 9 Asymmetric Digital Subscriber Lines 10 There is some off loading from macro and micro cells to femtocells but the noise shedding benefits of this are open to debate and probably more efficiently achieved with WiFi. 11 Multiple Input Multiple Output antennas. Single Input Single Output antennas 12 Diversity gain, also known as SIMO, Single Input Multiple Output can be shown to deliver downlink gains of the order of 50% in present HSPA networks, see antenna section 22 for more detail. 13 High mobility includes users travelling between 15 and 350 kilometres per hour. 14 User equipment (UE) is a generic term used to describe any form factor of mobile and portable device – the terms ‘phone’ or ‘terminal’ are also used more colloquially.

Geoff Varrall

Page 2 LTE UE efficiency and network value Final

15/09/2010

The present opportunities and challenges for the industry can be summarised as: A demand side opportunity A 30 fold increase in data volume demand over the next five years. A supply side opportunity An additional 500 MHz of spectrum on offer in some markets over and above the 1 GHz of spectrum already allocated and auctioned on a global basis. The problem with LTE user equipment Accessing these two opportunities cost effectively is however critically dependent on user equipment cost and performance. Over the past ten years, spectral auctions have taken over $200 billion dollars out of the industry which might have been alternatively spent on user equipment R and D and engineering and manufacturing investment. This study highlights three areas where LTE user equipment performance constraints, which are partly due to engineering under investment, compromise LTE network efficiency and increase network cost, potentially compounding the cost/income problem. A lack of band flexibility in LTE user equipment means that existing and new spectrum cannot be accessed effectively or efficiently. A difference between best and worse performance of at least 7 dB – a radio network cannot be accurately or efficiently dimensioned to meet cost and defined user experience expectations with this degree of user equipment performance uncertainty. Compromised conformance standards – the need to support new radio bands is forcing conformance standards to be compromised – each new radio band results in a further loss of RF performance – this reduces peak and average throughput and shortens the user’s data duty cycle, decreasing user experience value. A year on year loss of RF performance in user equipment increases the demand for spectrum and radio network density with associated cost but no associated income, increasing network cost, decreasing network value. These constraints together make it harder for operators and carriers to achieve a return on present and future spectral and radio network investment. An increase in radio network efficiency is therefore needed in order to ensure that present and future mobile broadband data volumes can be viably supported. Although the relationship between user equipment RF performance and network cost and efficiency is understood, there have been fragmented claims as to the magnitude of the effect and its fiscal impact. In particular some claims of potential performance gains have failed to take full account of system implementation issues. The study brings together inputs from the RF component industry and associated industry supply chain including user equipment, infrastructure and test equipment hardware and software vendors and presents a collaborative and candid consensus of what can be achieved, how much can be achieved and when. We show how a well implemented increase in RF and baseband efficiency in user equipment and a small increase in cost, translates into a relatively large fiscal gain for the whole industry. Users benefit too. The study is relevant to chief technical officers, chief executive officers and chief financial officers and their designates in the operator community and the whole industry supply chain including infrastructure hardware and software vendors, user equipment hardware and software vendors, component hardware and software vendors and the test equipment hardware and software community. The EBITDA modelling is also directed towards industry, financial and investment analysts with an interest in mobile broadband investment valuation and opportunity.

Geoff Varrall

Page 3 LTE UE efficiency and network value Final

15/09/2010

Implications of future mobile broadband data demand The measured growth rates (see below) suggest that the present rate of increase of data traffic on mobile broadband networks is increasing over time, a function of the growth in subscriber numbers, number of attached devices including connected machines and the volume of data generated per device, particularly from smart phones, tablets/slates and lap tops with dongles or embedded mobile broadband connectivity. 15 Opinions vary as to the scale of the increase. According to Ericsson 16 , data volumes as at the end of the second quarter of 2010 are running at 225,000 terabytes per month which is equivalent to 2.7 exabytes per year 17 . Ericsson point out that mobile broadband subscribers only account for 10% of the global subscription base. Global mobile data traffic has nearly tripled in the last year and is growing ten times faster than voice. Ericsson also forecast a connected device population 18 of 50 billion by 2020, a ten fold increase from the present five billion subscribers. Nokia Siemens Networks 19 suggest that by 2015, mobile voice will have increased by 50% from present levels, laptop data will have increased by 1,000% and “smart device” data will have increased by 10,000%. As an example, a You Tube clip is the equivalent of sending 500,000 text messages which highlights the growing disparity between data volume and data connectivity value. According to NSN, these subscribers together with the device population will generate 23 exabytes of data in 2015 compared to a present level of two or three exabytes. We forecast 87 exabytes of offered traffic data volume in 2015. This is in line with other higher end industry estimates. On the basis of present growth trends, and the trend of the trend (increasing over time), these projections would prove to be conservative 20 On the basis of present tariff trends 21 over the same period data revenues will have increased by at most a factor of 3, an income/cost disparity which must be regarded as a major challenge for the industry. Delivery cost is a function of the type of data traffic. Conversational, interactive and streamed traffic is more expensive to deliver in terms of bandwidth and power consumed than best effort traffic. This mix is also largely determined by the hardware and software form factors of user devices but it seems very possible that average application latency both in terms of first order latency (end to end delay) and second order latency (worst to best difference) will need to be more closely controlled. This effectively means that radio bandwidth and power will need to be over provisioned in order to meet user experience expectations. These are all factors that together add to the cost of delivery. While the demand projections and data revenue forecasts could well be realistic, there are two reasons why the growth in demand and revenues might be slower than anticipated and costs may be greater than expected (the downside equation). 1)The relatively slow rate of revenue growth relative to traffic growth will mean that operators will find it hard to raise the capital needed for the network density and or spectrum required to meet user experience expectations. This will result in poor voice quality, poor coverage, high blocked call rates, high dropped call rates, unacceptable application and task latency, high session failure rates and unacceptably short mobile broadband session duty cycles (time between battery recharge). This will result in high product returns and high churn rates introducing additional delivery and support cost and making it hard/impossible to realise any relative income gain on a per subscriber per device basis. 2) Battery capacity and heat density constraints will make mobile broadband duty cycles even shorter, choking offered traffic volume and value. While some users will be connected to a mains supply for 22 example in trains, these applications are by definition portable rather than mobile broadband.

15

Note that by definition mobile broadband connectivity implies not being connected to a mains supply. Portable broadband users might be more tolerant of a recurrent need to recharge a device or have easier access to a 12 volt or mains supply. 16 Press release Hakan Eriksson, Ericsson,12th August 2010 17 Vodafone (annual report page 15) reported a European data traffic load of 90 petabytes per year for the same period which scales to a global total of 2831 petabytes or 2.8 exabytes per year. This includes machine to machine (M2M) connections NSN CEO Rajeev Suri briefing at the Mobile World Congress February 2010 20 The range of forecast numbers is in itself a symptom of commercial uncertainty which when coupled with technical uncertainty is a disincentive to investment. Technical progress and innovation reduces commercial uncertainty and creates ROI opportunity.
19 18

21 22

Based on present tariff trends and operator income analysis undertaken by The Mobile World. Users are also more likely to have access to WiFi in these ‘tethered’ environments.

Geoff Varrall

Page 4 LTE UE efficiency and network value Final

15/09/2010

This study directly addresses these interrelated challenges of delivery cost, delivery value and mobile user experience expectations. We show that the solution (the upside equation), not surprisingly, is a function of mobile broadband radio network efficiency improvement. User equipment performance is a more important part of this efficiency equation than commonly assumed. Our data volume forecasts are based on a year on year improvement in the user experience which in turn delivers a 1% uplift in ARPU which in turn yields the additional EBIDTA needed to justify the investment required to sustain the year on year user experience gain - a positive investment cycle for operators and the whole of the industry supply chain. To achieve this, however, implies an associated need for materials and process innovation particularly in the radio front end. Antenna design for example is crucial to getting optimum performance out of SISO, SIMO and MIMO techniques and is a defining component in wide area and local area access economics. Antenna design and the associated matching requirement, is apparently simple but actually complex and dependent on the application of a mix of material and algorithmic innovation, knowledge and expertise. Similarly Silicon on Sapphire based components and systems provide an example of materials and algorithmic innovation applied to realise efficiency gain in switch path and adaptive tuning applications but with wider application potential. The study discusses and quantifies other similar innovations including RF power amplifiers and related enabling components. Generally it is also important to address the growing disparity between bench top measurements and real life user equipment performance. This is resulting in best to worst differences of more than 7 dB in user equipment performance 23 . A network cannot be accurately or efficiently dimensioned to meet defined cost and user experience expectations with this degree of performance uncertainty. RF and baseband innovation help to reduce this difference. Changes to conformance and performance measurement methodologies are also required to reduce the gap between what is measured on the bench and real world performance. There is clearly a demand side opportunity if delivery cost, investment, innovation and test issues can be addressed. Acknowledgements The fiscal impact economic modeling is based on a wide range of inputs from the network vendor and radio planning community. The study draws on work undertaken originally for the GSMA spectrum management group on RF cost economics and a technology study for a US operator bidding for 700 MHz spectrum. Thanks are also due to Vodafone for allowing us to reuse research undertaken for them last year, to Peregrine Semiconductor and Ethertronics for providing initial funding and sponsorship for this study and to the many companies and organisations that have provided inputs, comments and advice.

Of the 30 vendors contacted we would like to thank the following for providing particularly comprehensive information. Atheros (WiFI performance), Agilent (testing including DigRF), Anadigics (RFPA), Antenova (antennas), Avago (FBAR filters), Cavendish Kinetics,(RF MEMS), Cognovo (SDR architectures), Ethertronics (isolated magnetic dipole antennas), Icera (advanced receivers and baseband algorithmic innovation), Nokia Siemens Networks (physical layer scheduler efficiency), Paratek (BST based adaptive matching and tuning), Peregrine Semiconductor (CMOS on sapphire based switching and tuneable capacitors), Quantance, (RF PA linearization), Quintel (smart antennas and LTE link optimisation), RFMD, (RF PA), Sand9 (temperature compensated MEMS oscillators), Skyworks (RF PA), Skycross (antennas), ST Ericsson (baseband, RFIC and reference platforms), Triquint (RFPA) and Wispry (digital capacitors for adaptive matching and tuning).
23

Analysed in Section 22 page 32 and Section 29 page 43.

Geoff Varrall

Page 5 LTE UE efficiency and network value Final

15/09/2010

‘LTE user equipment, network efficiency and value’ summarises RF and baseband innovation opportunities but also quantifies the industry supply chain justification for improving RF and baseband performance in LTE user devices over and above present conformance standards. In parallel, a more detailed technical study on tuneable components and architectures is being undertaken by a working group within the IWPC. The objective of this group is to identify where tuning will be generically useful in the RF front end. Cases being investigated for a report due in September include 1) Reducing antenna size but still supporting lower bands by using active components for tuning 2) Improving MIMO and SIMO performance for the second antenna 3) Improving efficiency in the RF chain via impedance matching –on/near the antenna or antenna switch module 4) Improving efficiency of the PA by adjusting the load line impedance, PA bias and/or low-loss switches to use different PA stages and sections 5) The technology necessary to make tunable filters technically feasible and economically affordable 10+band phone possibilities are proposed to be addressed in a follow on report

 
All research undertaken for this study has been shared with the IWPC working group but if you would like to provide inputs to the IWPC or receive pre advice of the outputs from this group please e mail larry.morrell@cavendish-kinetics.com Other research partners for this study include the National Microelectronics Institute(NMI). The market and business analysis and economic modelling in this study have been undertaken by The Mobile World. The model is on the final pages of this document but is referred to throughout the study. Lead author of the study - Geoff Varrall Subject specialists Roger Belcher - RF Design Earl McCune – Linearization Queries, comments and questions to geoff@rttonline.com 00 44 208 744 3163 www.rttonline.com Additional resources relevant to this study are available from the linked web site www.makingtelecomswork.com Index Section

1 2 3 4 5 6 7 8 9

Title Executive summary- the LTE network cost and performance challenge Key findings by sector Implications of future mobile broadband data demand Acknowledgements Risk and value distribution in the industry supply chain – the problem and the solution The value dynamics of the industry User equipment RF Power budgets, link budgets and the relationship with user experience value – the energy cost of mobile broadband connectivity Requirements of the band plan China and global market requirements and fast track options FDD versus TDD Particular requirements of the US 700 MHz band plan Extended Multi Band User equipment options The business proposition- cost saving and revenue potential of extended

Page 1

5 7 8 10 12 14 16 17 19 20

Geoff Varrall

Page 6 LTE UE efficiency and network value Final

15/09/2010

10 11 12 13 14 15 16 17 18 19 20 21 22

23 24 25 26 27 28 29 30 31 32 33 Appendix 1 Appendix 2 Appendix 3 Appendix 4

Appendix 5 Appendix 6 Appendix 7

multi band user equipment The market proposition The technology proposition The engineering proposition What bands should be supported in future LTE handsets? RF MEMS and other innovations MEMS resonators Filtering using Surface Acoustic Wave and Bulk Acoustic Wave Devices NRE cost of RF MEMS Tuneable capacitors and adaptive TX and RX matching techniques Power amplifier material options Broadband power amplifiers Path switching and multiplexing Antenna innovations - impact on LTE in sub one GHz bands - general comments on smaller form factor devices - SISO - MIMO testing - SIMO gains Adaptive matching power savings RF PA and front end switch paths RF Baseband interfaces Conformance test costs Conformance and performance test standards opportunities Software Defined Radio and baseband algorithmic innovation Cost and performance comparisons The role of band flexibility in reducing delivery cost – the performance gain from multi band macro scheduling Band plan options 2011 to 2014 Band plan summary Pre requisites for industry profitability The impact of user device RF performance on the user experience and user and network value LTE user equipment power drain comparisons with Wi FI connectivity The impact of user equipment performance on spectral requirements and future band plans The impact of user equipment RF performance on network cost - the relationship with scheduler efficiency LTE QOE and compression MOS and QOS The cost of store and forward (SMS QOS) and buffering MIMO and spectral efficiency Operator cost/benefit analysis including energy cost Network infrastructure cost benefit analysis User equipment vendor cost benefit analysis Mobile broadband radio access economics - THE EBITDA MODEL – Study sponsor profiles and research partner profiles RTT contact points and supporting resources Risk and value distribution in the industry supply chain – the problem and the solution

20 21 22 23 23 25 25 26 26 27 28 29 32

37 37 39 40 40 40 41 46 47 48 49 51 53 55 59

66 68 68 69 72 73

1

The distribution of risk and value in the industry supply changes with time and is influenced by regulatory and competition policy, the price paid for spectrum and the cost of accessing that spectrum. It could be argued that the 3G auctions and subsequent spectral sales have taken $200 billion dollars out of the industry that might have been better spent on R and D and engineering and manufacturing investment but we are where we are. Opinions vary in the industry but the consensus seems to be that data traffic volume could increase by several orders of magnitude over the next five to ten years, and various demand side equations can be used as collateral for this assumption.

Geoff Varrall

Page 7 LTE UE efficiency and network value Final

15/09/2010

However these assumptions are only valid if the industry is technically and commercially able to meet this demand cost effectively. This means that delivered cost per bit including energy cost per bit has to reduce at a rate that is faster than the industry can presently manage. Additionally any new spectrum introduced has to be capable of being accessed cost efficiently with performance competitive user equipment. Timely availability of user equipment for new spectrum and new technologies has always been a problem and continues to be a problem for the industry today. How this influences risk distribution in the supply chain is debatable but there is a compelling case to be made that RF components are an important part of the problem and an important part of the solution to the problem. This part of the supply chain is fragmented and competitive pressures have meant that gross margins are presently insufficient to deliver the level and pace of innovation needed by the rest of the industry. In this study we argue that adding a dollar per year per band to the user equipment RF BOM helps to resolve this issue but some other mechanism for encouraging sector investment is also needed. Compared to web based businesses and green energy this is not a fashionable sector in which to invest. Of course this probably implies an investment opportunity. 2 The value dynamics of the industry

Irrespective of the precise demand projections, most operators are or should be concerned that mobile broadband offered traffic volumes are increasing at a faster rate than revenues. The effect is that average margin per user (AMPU) and average margin per device (AMPD) 24 reduces over time. This can be offset by improving network efficiency, reducing delivery cost, and or increasing income on a subscriber per bit delivered basis or by managing product and service mix by price. Income and margin is ultimately determined by quality of service and the quality of the user experience. The functionality and effectiveness of user equipment has a direct impact on the user experience and the value and cost of the traffic offered to a network. A lap top with a large screen and keyboard and LTE dongle or embedded connectivity, a touch screen tablet or slate, a smart phone and a standard phone are different both in terms of the amount and type of traffic they generate, realizable user experience value and delivery cost. Smart phones for example generate additional signalling load which adds to the cost of delivery.
25 User experience value is increasingly determined by application and task latency . Application and task latency is a function of end to end network latency but is also determined by the performance of the application processor, memory and buffer functions and the efficiency of the RF and baseband functions in the user’s device. This relationship is well understood but presently not well quantified.

The LTE business model is dependent on delivering a user experience which creates value from higher data rates delivered efficiently both in terms of cost and power consumption. LTE improves core network efficiency by a combination of multiplexing gain and routing flexibility. LTE improves radio access efficiency by a combination of traffic and channel and band 26 multiplexing. User equipment performance is a part of this efficiency equation. The connectivity efficiency of user equipment is defined by the effectiveness of the baseband and RF functions in the device. Improving RF functionality is generally dependent on materials and process innovation. This translates into additional component cost and R and D and manufacturing risk that needs to be amortized over significant market volumes. Improving baseband performance is generally dependent on algorithmic innovation and therefore translates less directly into additional cost other than accommodating extra clock cycles of processor overhead and high speed memory functions. In practice RF and baseband processes are closely related and
The objective of Average Margin Per Device is to capture all of the cost and income metrics that are specific to a device - this includes user equipment and connected machines. This can either be inferred from device specific network performance statistics or pre calculated before devices are accepted on to a network. (Range planning and acceptance).It provides an objective basis for subsidy and pricing decisions. 25 Applications consist of tasks and sub tasks some of which run in parallel. Individual task latency determines application latency. 26 Handover for example is a form of frequency domain multiplexing.
24

Geoff Varrall

Page 8 LTE UE efficiency and network value Final

15/09/2010

interdependent and solve different parts of the same problem. An efficiency gain in the RF front end will generally yield a reduction in baseband processor overheads which translates into improved downlink and uplink throughput and a longer user data duty cycle (the battery lasts longer). Generally it can be said that user equipment is cost engineered to ensure that conformance standards are met but not greatly exceeded. An alternative approach is to performance engineer user devices to yield year on year performance improvements over and above the conformance specification. It can be shown that these performance improvements reduce network cost and increase user value but introduce a small uplift on user equipment costs. The value of the performance gain and network cost savings achieved can be demonstrated to be greater than the additional costs. This in turn increases operator EBITDA but also delivers a net value gain to the whole industry supply chain. Users benefit as well. The user equipment performance gains have to be realized in both single and multi band devices. Multi band devices are needed to achieve global scale economy and to realize an improved (more valuable) user experience by providing the best connect multi band broadband connectivity needed to realize application value and meet future user experience expectations. Additional bands have to be added in as new spectrum is introduced to the market. However each additional band introduced into the front end of a user’s device introduces additional insertion loss and hence reduced sensitivity and reduced band to band isolation. The challenge particularly from an RF perspective is therefore to deliver performance gain and extended multi band support. This is dependent on a combination of RF component innovations and baseband algorithmic innovations that for both technical and commercial reasons are being adopted at a slower than optimum rate. A more aggressive adoption of RF component and RF architectural innovation in user devices combined with baseband algorithmic innovation is a necessary precondition for the long term commercial viability of LTE spectral and network investment. LTE networks are being deployed with cells that have a coverage radius that can be anything from a few meters to tens of kilometers and can be in any band from 700 MHz to 2.6 GHz with the longer term option to deploy at 3.5GHz. Some countries, most recently the USA, have announced plans to auction substantial 27 additional spectrum over and above present international allocations. However in practice almost one GHz of radio spectrum has already been allocated or auctioned, not all in the same place admittedly but in most countries it is true to say that at any moment in time a significant amount of mobile broadband spectrum is under used or inefficiently used. One reason for this is that user equipment presently lacks the flexibility needed to access this composite bandwidth on demand. More comprehensive inter band handover would yield substantial gains both in terms of coverage, capacity and data throughput but the amount of flexibility needed in the user’s device both to access this bandwidth and monitor the bandwidth is hard to realize in terms of present RF component availability. Efficient extended bandwidth monitoring for example implies a cost and energy efficient and frequency agile dual receiver. Such a device does not presently exist. Irrespective of the band or bands into which it is deployed, the spectral and power efficiency of a mobile broadband radio network can be shown to be closely coupled to scheduling efficiency. Improving the receive(RX) and transmit (TX) RF performance of single and multi band LTE user devices coupled with baseband algorithmic innovation significantly improves scheduling efficiency and throughput in single and multi band LTE networks. If this can be combined with extended multi band capability then additional scheduling gain can be achieved. Delivering RF performance gain particularly in multi band and extended multi band user equipment is challenging and requires a combination of new component technologies combined with optimized versions of existing technologies.

27

US Presidential memorandum June 28 2010

Geoff Varrall

Page 9 LTE UE efficiency and network value Final

15/09/2010

However relatively small improvements in terms of receive path insertion loss (sensitivity) and isolation (selectivity) in single band and multi band and extended multi band user equipment can yield a non linear gain in physical layer scheduling efficiency. This translates directly into higher data rates, more capacity, lower cost and a more consistent user experience (user value) per MHz of allocated and auctioned spectrum. User RF power drain and associated baseband and application processor overheads can also be significantly reduced. Network efficiency is traditionally coupled to spectral efficiency measured in bits per hertz. However it can also be argued that it is becoming increasingly important to benchmark radio network power efficiency and user equipment power efficiency in joules per bit and or watt hours per megabyte or, more optimistically, megabytes per watt hour, the megabyte per watt hour metric.
28 Improving RF efficiency in the e node B reduces network energy cost and carbon footprint. Improving RF efficiency in user equipment on the RX path (sensitivity and selectivity) and modulation accuracy on the TX path (error vector magnitude) helps reduce e node B energy cost and network carbon footprint but also reduces user energy cost and increases the user data duty cycle (the battery lasts longer).

This translates directly into additional offered traffic volume and value. Consider what a user might have done if his/her battery had not gone flat half an hour before. Consider the negative value of not being able to make or receive a vital call or send or receive a vital email or text. Any function that improves user experience should be capable of being translated into network value either in terms of direct value, for example tariff premium, or indirect value/reduced cost, for example lower product return rate and churn reduction. Telecoms remains a highly regulated market. The nightmare for a mobile broadband network operator is to have over bid for spectrum, usually to meet short term financial market expectations, to over or under invest in network hardware and software, both options being equally expensive, and to have pricing and quality of experience metrics imposed that fail to reflect actual delivery cost. 3 User equipment RF Power budgets, link budgets and the relationship with user experience value – the energy cost of mobile broadband connectivity

The RF power budget in user equipment has been considered in recent years as less important than the power drain from baseband and application processor functions and touch screen displays. The display on its own can be responsible for 40% of the energy consumed in the user’s device. In dense networks built for capacity, phones transmitting voice or text are often transmitting at a fraction of their potential peak power output. However this is less true with data duty cycles and particularly not the case in LTE 700 or 800 MHz deployments in larger cells in rural areas. Additionally in LTE, if voice is supported, the energy budget is an issue that also needs to be addressed. 29 For many data exchanges, packets received on the downlink will be acknowledged on the uplink creating an additional load on the transmit chain. Always-on applications like push email, location-based services and social networking create signalling load which absorbs power, particularly in mobile smart phone devices. IP networks were designed to be resilient rather than power efficient. As a wireless IP network, LTE has to deliver a user experience that is equal to or preferably better than that available from existing user equipment on existing networks that typically use a mix of IP and traditional circuit switching. This applies to all forms of data including voice. Fourth generation radio systems deliver high physical layer peak data rates by using higher order modulation (typically OFDM, nPSK or nQAM) and wider channel bandwidths. These improve bandwidth efficiency but require more linearity so are not inherently energy efficient.
28

We refer to the e node B as an illustration of performance gain but the comments can be generally applied across all base station options including micro and macro cell topologies. 29 LTE voice avoids most of the packet overheads of pure IP voice by using Packet header compression techniques . However compression and decompression uses fast memory and absorbs processor bandwidth and introduces processing delay. Additionally the process of extracting a narrow band information stream, for example a 12 kbps voice codec, from a broadband channel, for example 20 MHz, is computationally complex and results in an energy drain that is higher than present GSM handsets. This is, or at least should be, in the ‘things to be addressed’ in box of most LTE design teams.

Geoff Varrall

Page 10 LTE UE efficiency and network value Final

15/09/2010

For these reasons, RF transmission efficiency will have a greater impact on user duty cycles (voice minutes and data megabytes per watt hour of battery capacity) and peak and average data throughput. This in turn affects the user experience 30 and by implication operator profitability – the energy cost of connectivity. Similarly, receiver front end sensitivity, selectivity and stability determine downlink data rates. Poor sensitivity will increase baseband coding and error correction overheads, trigger high retry rates and compromise scheduling efficiency. This will further reduce the user duty cycle and absorb useful network bandwidth. 31 User devices have to discriminate an OFDMA waveform across multiple frequency sub carriers modulated with higher order modulation. As with the transmit path, these mechanisms achieve high peak data rates but are noise and distortion sensitive and not inherently power efficient. Alternatives such as higher capacity batteries offer only a partial answer. Battery energy 32 and capacity by volume and weight are generally increasing at an annual rate of 10% to 15%. Mobile system power requirements are increasing at an annual rate of 35 to 40% 33 . The user data duty cycle therefore decreases over time, a disincentive to product replacement. It would be absurd to have engineered a new generation of mobile broadband device that needed to be tethered to a mains or 12 volt socket to meet user expectations. Even if batteries suddenly improved, RF device and system efficiency gains would be needed to avoid heat dissipation issues. 34 Heat dissipation will of course also be a problem for portable devices connected to an external power supply. LTE user equipment will have to be capable of working at 700 and 800 MHz, 35 2600 MHz, and any one or all of several existing bands including 850/900 MHz, 1800 and 1900 MHz and 3GPP Band 1(1900/2100 MHz). There will also be additional ATC 36 allocations in L band at 1.4, 1.5 and 1.6 GHz and S band at 2 GHz. Legacy connectivity standards will also need to be supported in these bands and future iterations of LTE including LTE Advanced (100 MHz channel bandwidths) may also be allocated spectrum in C band between 3.6 and 4.2 GHz. User equipment capable of supporting seven bands or more could be shipped to a broader mix of markets to achieve scale economy and inventory management benefits. The ability to provide a best connect capability over multiple bands (national roaming) would also increase throughput rates and reduce power drain, substantially improving the quality of the user experience. As we shall also show, this also improves operator EBITDA. However each additional band supported in a traditional RF front end introduces potential additional insertion loss and results in decreased isolation between multiple signal paths within the user’s device. Some bands, for example the US 700 MHz band, place particular demands on RF component

31 In recent years sensitivity has been considered as less important as connections are often made at medium power in interference limited conditions. However applications such a rural broadband at 700 and 800 MHz will increase the likelihood of a noise limited channel. 32 One of the constraints in the design of portable systems is the run time between recharging. Batteries are a significant fraction of the size and weight of most portable devices. The improvement in battery technology over the last 30 years has been modest, with the capacity (Watt-hours kg) for the most advanced cells increasing only by a factor of 5-6, to the present value of 110 Wh/kg for lithium-ion. The energy density is beginning to approach that of a high explosive, at 1000 Wh/kg, so that safety considerations will likely limit any increases beyond another factor of 2 or 3. Since battery technology will not provide the needed improvement in run time, energy minimization of the electronics and devices (and their connectivity cost) is critical. Advances in fabrication technology, low-voltage operation, and circuit design can be coupled with low-energy system and architectural design to reduce the energy requirements of the electronics. Reference: Broderson, R. W., “The network computer and its future,” 43rd IEEE International Solid-State Circuits Conference (ISSCC), 6-8 Feb 1997. Digest of Technical Papers. pages 32 – 36. Additional resources are available on this subject- if you would like these forwarded please e mail geoff@rttonline.com

Source- Semiconductor Association – see also Appendix 5 for battery density comparisons. Short duty cycles will also mean the already short overall life span of the battery will be reduced. It might be argued that this encourages people to buy new devices but this is not necessarily beneficial to the industry value chain. For instance churn rates are likely to increase particularly if the replacement decision is triggered by user dissatisfaction or premature product failure. 35 At June 2010 there were a total of 80 LTE national network commitments in 33 countries with Verizon stated as being one of the first most likely to deploy commercially with a nation wide footprint – source Global Mobile Suppliers Association 36 Auxiliary Terrestrial Component – hybrid LTE/satellite networks see later references.
34

33

Geoff Varrall

Page 11 LTE UE efficiency and network value Final

15/09/2010

performance. Supporting additional bands also increases the number of RF components with associated real estate and cost implications. There is therefore a need to cost efficiently achieve significant improvements in user device RF component and system efficiency. This can be achieved by improving the performance of existing active and passive components and or through the introduction of new components based on materials innovation which in turn enable architectural and system innovation. The optimum implementation of new and improved components is also of paramount importance. Examples of new materials or mechanical structures being introduced into or considered for RF front ends include gallium nitride for RF power amplifiers, silicon on sapphire for switch paths, digital capacitors and tuneable filter banks, barium strontium titanate for adaptive matching, ceramic substrates for antennas and RF MEMS for adaptive matching, filtering and tuning or as oscillators to replace the TCXO, the frequency reference at the heart of all RF devices. These new materials when combined with new packaging techniques provide the basis for new families of RF components that potentially introduce band flexibility without cost or performance compromise. These components can be coupled to innovative antenna solutions that minimise additional space requirements. Any or all of these innovations are capable of delivering a performance gain. They can be made to be technically efficient. However to be successful these devices also have to make commercial sense for RF component and baseband component and algorithmic vendors, user equipment manufacturers, infrastructure vendors, network operators and users. In other words, the devices have to be commercially efficient. Inherent technical and commercial inefficiencies have been introduced as a consequence of well intentioned but misguided regulatory policies that have resulted in a multiplicity of band plans that are hard 37 to support in the front end of a user’s device. These inefficiencies need to be resolved to ensure the long term economic success of the mobile broadband transition. The band allocations cannot easily be changed or harmonized so technology solutions are needed. 4 Requirements of the band plan

Table 1 lists eleven of the ‘legacy bands’ that presently constitute the majority of global spectral allocations. The band combination most commonly supported in a 3 G phone is the original 3G band at 1900/2100 MHz described as Band 1 in the 3GPP numbering scheme and the four GSM bands, 850, 900, 1800 and 1900 MHz (US PCS).

Table 1 3GPP Band allocations by region
Band 3GPP I II III IV V VI VII VIII IX X
37

Allocation Uplink

Duplex spacing

Downlink 2110-2170 1930-1990 1805-1880 2110-2155 869-894 875-885 2620-2690 925-960 1845/1880 2110-2170

Region Present 3G US PCS GSM Europe, Asia, Brazil US AWS US and Asia Japan New Europe and Asia Japan As Band 1 with lower mobile TX, but also

2100 1900 1800 850 800 2600 900

2x60 MHz 1920-1980 190 MHz 2x60 MHz 1850-1910 80 MHz 2x75 MHz 1710-1785 95 MHz 2x25 MHz 824-849 2X10 MHz 2X35 MHz 830-840 45 MHz 45 MHz

1700/2100 2x45 MHz 1710-1755 400 MHz

2x70 MHz 2500-2570 120 MHz 880-915 45 MHz

1750/1845 2x35 MHz 1750-1785 95 MHz 1700/2100 2X60 1710-1770 400 MHz

The ‘front end’ is an industry term to describe the signal paths in and out of a user’s device including the antenna, filters, oscillators, mixers and amplifiers, a complex mix of active and passive components that are not always easily integrated. The physicality of these devices often means they do not scale as efficiently as base band devices. It should be remembered that any radio wave is analogue as are many of the front end functions.

Geoff Varrall

Page 12 LTE UE efficiency and network value Final

15/09/2010

MHz XI 1500 2X32 MHz 1427.91452.9 48 MHz 1475.91500.9

effectively an extension of Band IV Japan (was PDC1500)

However this is an incomplete list. In total there are 25 bands identified as E UTRA 38 frequency bands These are itemized in Table 2

Table 2 Global band plans 3GPP Release 8 June 2010 E-UTRA frequency bands 39

Release 9 of the 3GPP specifications define four further bands, three being extensions or variations of the 850 MHz band and the other being around 1500 MHz. Bands 18 and 19 extend the existing 850 band adding 10 MHz at the lower end of the present Band V allocation to provide a 35 MHz rather than 25 MHz pass band. Band performance relative to Band V would be degraded by about 1 dB. 40 In Release 10 of the specifications, several more bands are being considered. For example 3GPP is finalizing the 790 to 862 MHz allocation in Europe, implemented as a standard duplex and defined as Band 20. On the 28th June 2010, the US Obama administration issued a memorandum 41 stating an intention to auction a further 500 MHz of spectrum. Some but not all of this will line up with rest of the world band plans and may or may not include spectrum in L band at 1.4,1.5 and 1.6 GHz and S band allocations presently ring fenced for hybrid satellite/ terrestrial networks. 42 Additionally in some form factors, for example tablets, slates and lap tops, there may be a future need to integrate ATSC TV receive and or DVB T or T2 receiver functionality in the 700 MHz band. For certain it can be said that an already complex set of band plans will only become more complex over time. Within existing 3GPP band plan boundaries there are already significant differences in handset performance which are dependent on metrics such as the operational bandwidth as a percentage of centre frequency, the duplex gap and duplex spacing in FDD deployments and individual channel characteristics determined, for example, by band edge proximity.

38
39

Extended Universal Terrestrial Radio Access

Bands 15 and 16 are only defined by ETSI for use in Region 1 (Europe) and are not recognised by the global 3GPP organization but are listed here for completeness. 40 This directly illustrates the relationship between operational bandwidth, the pass band expressed as a ratio of the centre frequency of the band, and RF efficiency. 41 http://www.whitehouse.gov/the-press-office/presidential-memorandum-unleashing-wireless-broadbandrevolution For example part of this newly available spectrum may overlap the European 900 MHz band. 42 Details on these ATC allocations can be downloaded here. See also Light Squared and Inmarsat press announcements August 2010

Geoff Varrall

Page 13 LTE UE efficiency and network value Final

15/09/2010

It is not therefore just the number of band plans that need to be taken into account but also the required operational bandwidths, duplex spacing and duplex gap of each option expressed as a percentage of the centre frequency combined with the amount of unwanted spectral energy likely to be introduced from proximate spectrum, a particular issue for the 700 and 800 MHz band plans. The performance of a phone from one end of a band to the other can also vary by several dBs. 43 Adaptive matching 44 techniques can help reduce but not necessarily eliminate this difference. At present, a user device parked on a mismatched channel on the receive path will be or at least should be rescued by the handover process but the associated cost will be a loss of network capacity and absorption of network power (a cost to others). Wider bandwidth channels for example 10 or 20 MHz, reduce the opportunities for an operator to do in band frequency domain handovers particularly in lower band allocations with only 10 or 20 MHz of paired spectrum. This strengthens the case for inter band inter operator handovers (see later sections on macro scheduling gain). 5 China and global market requirements and fast track options

China has pursued its own standards with TD SCDMA and band allocation policy with Band 34 at 2010 to 2025 MHz, Band 39 at 1880 to 1920 MHz and Band 40 at 2300 to 2400 MHz. 45 There is a perceived market need to support TD LTE in Band 40, Band 38 at 2570 to 2620 MHz and Band 39 and LTE FDD in Band 1 1920-1980/2110-2170 MHz, Band 7 2500-2572/2620-2690 MHz and Band 13 at 777-787/746-756 MHz (US 700 MHz Verizon spectrum). Backwards compatibility for GSM/GPRS/EDGE is assumed for Band 2 1850-1910/1930-1990 MHz, Band 3 1710-1785/1805-1880 MHz, Band 5 824849/869-894 MHz and Band 8 880-915/925-960 MHz. There is recognition 46 that it may be challenging to persuade global handset vendors and their supply chain to support TD SCDMA at least in the near term but confidence exists that the present market volume and future volume and value potential of the China market will be sufficient to progress a nine or eleven band TD LTE, LTE FDD and EDGE global handset to meet all other band requirements both for the China market and most other global markets far faster than the industry has ever moved before. To deliver on this vision, the vendor supply chain would have to provide the following deliverables • • • Baseband offerings that support TD-LTE, FDD LTE and EDGE in a single chip. RF IC solutions for an 11 band phone integrating all functions except the PA and antenna, supporting one RF path at a time with switching between bands. A single antenna covering the 700 MHz to 2.7 GHz bands with antenna efficiency at 700 MHz of 3 70%. The device would have a target volume of 2 cm to give a low band resonance below 1 GHz, a high band resonance close to 2 GHz and a third resonance at 2.6 GHz. A MIMO antenna would of course double this volume. The additional bands would need to be achieved by antenna tuning through a multi port switch or MEMS tunable capacitor. Note that antenna efficiency would need to be maintained across all bands. An 11 band SAW filter bank module using LiTaO3 material combined with FBAR filters for Wi Fi and Band 40 co existence.

Source Paratek March 2009. The ability to dynamically optimise the efficiency of signal power transfer between one section of the phone and the next. 45 China potentially has the market scale to force the pace of change (the great leap forward) both in terms of hardware and software innovation, India can potentially do the same in ICT innovation. However non standard standards and non standard spectral allocations have hampered the pace of past progress. The decision by Japanese regulators in the late 80’s/ early 1990’s to introduce PHS (an alternative to DECT and the UK’s ill fated CT2 cordless standard) and PDC, a non standard implementation of GSM into non standard spectrum at 800 and 1500 MHz was designed to create a protected internal market which could be used by local vendors to amortise R and D and provide the basis for innovation incubation. In practice it proved hard to translate this innovation into global markets and the R and D opportunity cost made Japanese handset vendors and their supply chain less rather than more competitive internationally. Korean vendors have faced similar challenges from nationally specific mobile broadband and broadcast standardisation. This has introduced unnecessary opportunity cost without a proportionate market gain. 46 Our thanks for China Mobile for sharing their technology vision with us on these topics (discussions end July 2010)
44

43

Geoff Varrall

Page 14 LTE UE efficiency and network value Final

15/09/2010

• •

Multiple power amplifiers integrated into a single module. An external multi port switch for example an SP10T switch combined with an SP4T switch to support UE mode configuration and inter working measurement.

Any cost increase is considered as being negligible and likely to be more likely due to the complexity introduced by MIMO implementation. The recommendations may go forward as an NGMN industry supply chain expectation. The timeframe for executing on the above plan has been stated to be as short as six to twelve months. This is substantially more aggressive than we would have anticipated on the basis of our own research. However if a cross section of RF component vendors have committed to these targets and time scales, and apparently they have, then we must assume there is a high level of confidence that the technical and commercial challenges can be met. We recommend that close attention be paid to the IWPC study outputs to validate progress over the next two to three months to provide reassurance that the six month to twelve month implementation time scale remains achievable. For the purpose of this study we have taken a more cautious approach and modeled the financial benefits 47 that would accrue from adding one additional band per year rather than five bands in the next twelve months while in parallel delivering a 1 dB sensitivity and selectivity gain per year over and above an agreed LTE conformance standard. 48 To emphasize this point, we think that it is necessary to deliver extended multi band AND year on year performance improvements in order to meet future LTE user experience expectations. A failure to achieve this would invalidate many or most of the present LTE spectral and network investment models. This implies that an extended multi band RF front end only makes sense if combined with improved RF performance. Being positive about this, we argue that adding one band per year and improving RF and base band 49 performance by one dB per year produces a net gain for all involved parties and creates a positive rather than negative LTE investment cycle. The model also factors in a dollar of additional cost for each additional band which is substantially higher than the IWPC and NGMN work is assuming. If it proves possible to add five bands in the next twelve months and deliver one dB per year of performance gain 50 both in this twelve month period and for subsequent years and have virtually no increase in cost then the fiscal gain could of course be even greater. Getting there in one leap (the great leap forward) probably also implies a higher level of integration than a step (band) at a time approach which is more likely to be based on an evolving mix of discrete solutions. Whether the fast track approach is achievable depends entirely on how realistic the RF component vendor community is being about the practical implementation challenges and whether there is a willingness from shareholders and investors to underwrite the risk. This in turn is determined by the perceived volume and value of the opportunity. For various understandable reasons, the supply chain does not have an unblemished record of meeting to market time scale commitments An alternative would be to accept that some user equipment will not meet conformance specification at market launch at least in some channels in some bands. The argument here is that poor channel specific and or band specific performance will be accommodated by handover processes and other radio air interface adaptive processes. Poor overall performance in user devices will also be less noticeable in new network roll outs in new spectrum that are by definition initially lightly loaded with a low noise floor.

The industry’s historic rate of progress with new band implementation has been rather slower. GSM phones were single band in 1991, dual band by 1993, tri band a few years later, quadruple band a few years later and quintuple band a few years later. A five band phone remains a common denominator product. Self evidently if a band per year had been added we would have twenty band user devices by now, the average has been a new band every three years or so.. This implies it might take five years or more for eight band user equipment to become a common denominator market dominant product. This is a problem for operators investing in new spectrum as it limits user product choice and adds cost due to a lack of global scale economy. 48 Note that maximum power is also affected by multi band and also needs to be addressed particular for larger cell higher power deployments which tend to be uplink power limited. 49 This includes base band functions such as enhanced equalisers, diversity gain and interference cancellation. 50 Across all channels, across all bands.

47

Geoff Varrall

Page 15 LTE UE efficiency and network value Final

15/09/2010

However this is just hiding the fact that poorly performing user equipment has a real cost both in terms of network bandwidth (handover signaling, error coding overheads, packet retries and retransmissions) and a degraded user experience. Handovers increase user equipment energy drain, increase the risk of session discontinuity and compromise application and task latency which as we have stated will increasingly determine user experience value. Over time, networks will have the problem of legacy devices that take up a disproportionate amount of network bandwidth and absorb a disproportionate amount of network power. This inflicts an opportunity cost that ends up being imposed on other users. The next few sections look at these user equipment and network efficiency challenges in detail and reflect the technical (engineering) and commercial (market and business) inputs that we have received. These inputs highlight that there is a safe absorption rate for spectrum and a safe absorption rate for technology. Both are closely coupled. The more closely this rate can be calculated the more industrially efficient we become. Before diving in to the technical detail, a final thought on R and D resource allocation. From a return on investment perspective it is often better to concentrate on making what you already have work better rather than producing something new. Say you have 100 design engineers and you set them to work on an existing five band design to do performance and cost optimization. The team produces a device that costs less and works better and gives you a competitive advantage in a dominant market of known volume and value. The alternative is to take the same team and deploy them on a project, for example an extended multi band design, that could be perceived as higher risk technically (new materials with minimal experience of production yield) and commercially (addressing a market of unknown volume and value). This gravitational effect makes it harder to motivate innovation in the industry supply chain and can only be overcome by providing investment incentive. More generally it can be stated that if a supplier base narrows and margins decline then the focus will be on extracting as much as possible from existing technologies rather than investing in innovation. Within the operator community, leaving range planning and product choice to marketing teams and failing to factor in LTE user equipment RF and base band performance variation is likely to be extremely costly. More generally it can be said that the largest industry disconnects tend to happen when marketing teams decide engineering and product policy and economists decide regulatory, spectral and standards policy. 6 FDD versus TDD

Table 2 shows the present mix of FDD and TDD band plans. In TDD bands the uplink and downlink are on the same frequency. This makes MIMO and adaptive matching easier to implement and more effective. Time division duplexing also means that the front end duplex filters can be replaced by a switch, potentially reducing cost and insertion loss. Present 3G handsets in 3GPP Band 1 are however all FDD. Providing frequency domain separation between transmit and receive improves sensitivity and mitigates the risk of user-to-user or cell-to-cell inter symbol interference (ISI) in larger cell deployments at the expense of additional filtering. TDD systems can be deployed in large cells but a longer time domain guard band is needed. This reduces spectral efficiency. For these reasons FDD probably remains the best choice for networks at 700, 800, 850 or 900 MHz and for higher bands where larger cells are deployed. Most FDD and TDD decisions are anyway influenced by local regulatory policy. Some radio air interfaces, GSM being the most ubiquitous example are FDD and TDD, sometimes described as half duplex. Filtering can be in the frequency and or time domain. Many low cost GSM phones dispense with duplex filters and use a switch to perform the duplex function. This could be feasible in TDD LTE provided TX/RX time slots do not overlap though the switch path would need to be highly linear and combine high isolation with low insertion loss. In both FDD and TDD the two resources being shared amongst multiple users are bandwidth and power allocated in the time domain, code domain or frequency domain (radio channels and/or in LTE, frequency sub carriers). In GSM handsets, the maximum peak and average power is either two watts at 900 MHz or 51 one watt at 1800 MHz based on a one in eight slot duty cycle. Multi slot configurations support GPRS and EDGE system implementation.
51

It is harder to achieve gain from active devices at higher frequencies hence the lower output power for GSM 1800.

Geoff Varrall

Page 16 LTE UE efficiency and network value Final

15/09/2010

3G phones have lower average power, typically 250 milliwatts for a 900 MHz device but can be transmitting continuously. However the PA design must be based on the signal peak power. For newer 3G signals the peak power is about 6 dB (four times) higher than the average power, which is one watt. Worse, to maintain acceptable linearity this peak power must still be below the maximum saturated power the PA can provide, usually by 1 to 3 dB (1.2x to 2x). So even though the average 3G power is 250 milliwatts, the PA must be designed for 1.5 to 2 watts – the same saturated power as GSM. The overall maximum available transmit energy is therefore similar though the linearity requirements are more onerous but it can be seen that the TX path in an LTE user’s device is not inherently power efficient. Energy efficiency thus has to be achieved in other ways. By definition a TDD channel will have half the time domain bi directional transmission bandwidth and time domain power of an FDD device though to be fair the FDD device is occupying a channel pair not a single channel. There is a subtler relationship. As implied above, LTE TDD and FDD power amplifiers should theoretically 52 be backed off by between 10 and 12 dB though are typically clipped to 7 dB . This is the ‘cost’ of achieving high peak data rates in close to cell centre conditions. It does however make it more likely that user equipment becomes uplink limited in edge of cell conditions in larger cells both in terms of available power and the accuracy of the applied modulation. This in turn will impact average throughput rates. The back off can be reduced by compromising EVM but this decreases the modulation accuracy. In general however it can be stated that FDD should at least theoretically be able to deliver a more robust up link in larger cells. 7 Particular requirements of the US 700 MHz band plan

On the basis of present market announcements, Verizon 53 will be the first operator to deploy an LTE network into Digital Switch Over spectrum at 700 MHz with a commercial launch in November this year. This is new technology deployed into a new band. This means that the timely availability of cost and performance and market competitive user equipment is critical to the financial success of the investment. The lower and upper 700 MHz auctions in the US resulted in Verizon owning nationwide upper C block with a two by ten MHz (plus 1 MHz of guard band) paired allocation of 746 to 757 and 776 to 787 MHz. The band is known as Band 13. They also have some lower band paired A Block, 698 to 704 MHz, 728 to 734 MHz and lower band B block, 704 to 710, 734 to 740 MHz. Lower band A B and C blocks are collectively known as Band 12. AT and T have mainly lower band B 704 to 710, 734 to 740 MHz and lower band C spectrum, 710 - 716, 740 to 746 MHz. Verizon’s lower A block spectrum will only be deployed when the adjacent Channel 51 (one megawatt terrestrial broadcasting) is not present and they will rely on their national Upper C Block to provide nation wide coverage. There is an open access block D in the upper band at 758 to 763, 788 to 793 MHz the future of which is yet to be determined. This is half of Band 14. (But see Motorola note on the LTE 700 MHz public safety network below). Band 12 operates as standard (forward) duplex with the mobile transmitting on the low side of the duplex as it was felt to be easier to filter out high power TV signals at 698 MHz and below at the base station. Band 13 operates as a reverse duplex (mobile transmits on the high side of the duplex). The Verizon network will be based on a two by ten MHz duplex with 1 MHz of guard band. Media FLO broadcasting in channel 55 at up to 50 kW is immediately adjacent to lower band C which is problematic for the UE receiver. In the longer term there may be other users in the band particularly if the White Space lobby gains momentum. The graphic below shows the US 700 MHz band plan - Source Quintel based on FCC original 54

The SC FDMA used on the LTE uplink helps a bit here. Verizon is not the first LTE network to be commercially deployed, Telia Sonera having deployed already in Sweden and Finland but Verizon is definitely the first to deploy at 700 MHz. 54 Quintel. have a range of base station antenna solutions that allow the deployment of multiple access technologies in proximate bands onto a single antenna array with independent beam tilt control, for example LTE700+GSM 850 for the US Market.
53

52

Geoff Varrall

Page 17 LTE UE efficiency and network value Final

15/09/2010

In 2008 Verizon paid $10 billion dollars for their upper band and lower band holdings. 55 LTE network investment will at least double this. This implies that a substantial return on investment needs to be achieved. 56 The service proposition is being pre marketed as ‘4G LTE’ with promised peak downlink speeds of 40 to 50 megabits per second, peak uplink speeds of 5 to 12 Mbps, and average data rates of 5 to 12 Mbps on the downlink and 2 to 5 Mbps on the uplink. ‘Most users should expect to see 8 Mbps on the downlink and 3 Mbps on the uplink’ (Verizon web site May 2010). 57 The RF performance of user equipment in terms of sensitivity and selectivity will be crucial to delivering against these market, user and investor expectations. The requirement to filter out large in-band or close-to-band signal energies from terrestrial or mobile TV and avoid interference to the public safety channels adds complexity and cost and increases DC power drain. Considerable work has been done within 3GPP to define solutions for how to make Band 13 co exist with public safety. This includes power reductions on the channel edge resource blocks and moving the control channels from the edge to nearer the centre 58 . Solutions to the band 12 issues and in particular Block A are still being investigated. The band plan introduces a number of conflicting RF cost and performance objectives. The best performance in terms of selectivity and sensitivity for a Verizon only user device would be achieved by just supporting a 10 MHz channel pair in Band 13 (upper C band). However Verizon owns spectrum in the lower band and anyway a C band only solution would be unlikely to have sufficient economies of scale to deliver long term scale economy. It may also be commercially, operationally and politically beneficial for AT and T and Verizon to have the capability to support each other’s customers on each other’s networks (best connect national roaming) if the cost and user experience benefits of such an option exceed the perceived collaborative risk.
55

AT and T paid just under $5 billion but may need to invest additional sums on infrastructure to manage some of the filtering challenges inherent in the lower band allocations. To alleviate some of this and to simplify product development AT and T led the creation of Band 17 which is the same as band 12 but without the problematic A block adjacent to Channel 51 TV and Block E Media FLO. This fragments the market further and leaves significant commercial and technical difficulties for owners of Block A. 56 http://www.cellularmaps.com/700_auction.shtml provides a detailed analysis of who owns what in the US 700 MHz band. 57 This will be dependent on network loading. 3 to 8 Mbps will be close to the capacity of each entire sector, enough for two continuously connected users! 58 Band 13 has coverage challenges since it will require a significant power back off at the channel edge and a narrowing of the reverse duplex physical uplink shared channel (the PUSCH) caused by having to move the physical uplink control channel (the PUCCH) further into the traffic channel to avoid spurs falling into the public safety band base station receive band. (Implemented as a standard duplex with mobile TX on the low side).

Geoff Varrall

Page 18 LTE UE efficiency and network value Final

15/09/2010

However having an RF front end that can cover 698 MHz to 787 MHz implies an operational bandwidth that in ratio terms (12%) is significantly wider than the existing 850, 900, 1800 or 1900 MHz bands. This will result in a decrease in sensitivity and selectivity which will translate into lower and more inconsistent data throughput rates offsetting the theoretical benefits of using 700 MHz (lower free space loss and better in building penetration). 59 Similarly, economies of scale considerations and roaming considerations would suggest that there would be merit in having an RF front end that could also cover the European 800 MHz DSO band plan extending up to 862 MHz and integrated with US 850 band functionality. However this implies an even more ambitious operational bandwidth as shown below. (Source Quintel)

As stated earlier, every band added to an existing five or six band phone or mobile broadband device if implemented with traditional components will introduce insertion loss and will make isolation harder to achieve through the switch matrix. There will also be additional cost. Increasing the number of channels supported within a band will have a similar impact. For example, Motorola have also announced (August 2010) pilot funding for an LTE public safety network in D block at 700 MHz. The market logic would be for this equipment to be compatible with the Verizon and AT and T 700 MHz networks both to achieve scale economy and to provide access to all of the composite bandwidth available in the band but this would involve a significant loss of sensitivity and selectivity for all devices. Fairly obviously it would be useful to find a way to include all of these additional bands without cost or performance loss. A starting point is to decide which bands and how many bands need to be supported and to attempt to quantify what the user experience benefits might be from extended multi band user equipment including the fiscal benefits that this improved user experience might realize for operators and other parts of the industry supply chain. 8 Extended Multi Band User Equipment options

The table below compares five band user equipment with a presently theoretical ten band implementation. Note here we are using ten band as a proxy for any extended band phone by which we mean at present a nine, ten or 11 band device.

59 Opinions on this vary – some propagation research points to building penetration being broadly frequency independent from 700 MHz through to 2 GHz.

Geoff Varrall

Page 19 LTE UE efficiency and network value Final

15/09/2010

Option 1) LTE 5 Band Band Band Band Band V VIII III II 850 900 1800 1900 Existing GSM Edge Quad band transceivers 50 70 150 120 MHz MHz MHz MHz Band I 2100 UMTS 120 MHz

Total Bandwidth

Bandwidth by band

510 MHz

2) LTE 10 Band

Bandwidth

Band 13 US DSO 746 756 777 787 20 MHz Band 12 698 716 728 746

Europe DSO 790 862

Band 40 2.3 GHz China LTE TDD 100 MHz Band VI Japan 830 840 875 885 20 MHz Band XI Japan 1428 1453 1476 1501 50 MHz Band IV US AWS 1710 1755 2110 2155 90 MHz

Band VII LTE 2.6 GHz band

Band 38 2.6 GHz LTE TDD

60 MHz

140 MHz

50 MHz

916 MHz

3) LTE 10/13 plus

Bandwidth

36 MHz

1112 MHz

As can be seen, option 2 almost doubles the amount of total bandwidth available 60 . Option 3 would provide an additional increase though includes bands that are either tricky to implement (Band 12 adjacency to high power TV implies a need for increased dynamic range which implies additional DC power drain and or additional filter insertion loss) or address nationally specific markets (the AWS band, Band IV, in the US and Band VI and XI in Japan). Option 3 would need to support 12 or 13 bands and would imply additional component and performance cost with a relatively small increase in addressable market size. (100 million Japanese subscribers and 32 million T Mobile subscribers in the US split across GSM 1900 and AWS). Option 2 has a number of positive aspects that make it a potentially attractive business, market, technology and engineering proposition with short term cost but substantial longer term gain for all involved parties. These positive aspects can be summarized as follows 9 The business proposition- cost saving and revenue potential of extended multi band user equipment

Although roaming revenue premiums for voice and data are eroding over time, international business and leisure travellers both in developed and emerging markets represent an important and expanding sector which generates definable revenue streams which can indirectly help off set R and D and engineering development budgets.

60

The bandwidth is not all available simultaneously but does represent accessible bandwidth when considered on a global basis.

Geoff Varrall

Page 20 LTE UE efficiency and network value Final

15/09/2010

Intuitively this target market is less price sensitive than some nationally specific markets. In turn this offers more flexibility on the RF bill of materials. Thus although market volumes may not be comparable with emerging markets and core national markets, for example China and India, the combined volume achievable from a large and growing sector of roaming subscribers combined with the potentially higher realized prices for user equipment suggest that the sector provides a potentially profitable market for nine, ten or eleven band mobile broadband devices. 10 The market proposition

This however only holds true if network operators can meet the expectations of these target user communities. Mobile users expect that the service they receive when roaming is at least as good as the service they receive from their home network - the best connect broadband proposition. Operators therefore need to ensure that handsets can access higher data rate services in all markets. Relying on a fall back to band limited GSM or EDGE services or other legacy access options will become less acceptable over time and will result in a disappointing and inconsistent user experience. This is particularly true for mobile broadband users who expect a service that is equal or close to ADSL with coverage that is equivalent to existing voice provision. This implies high peak data rates but also an ability to sustain acceptable average data rates into buildings and in edge of cell conditions.
61 To an extent these expectations can be met by increasing network density but operators also need to have user equipment that delivers better RF performance (sensitivity and selectivity) than previous and present generation devices to allow them to work in low signal and/or high interference conditions. The devices also need to deliver consistent performance over a wide range of operational and channel conditions and to have an operational duty cycle that is at least equivalent and preferably better than existing devices - needing to recharge user equipment more than once a day limits user acceptance and compromises operator revenue potential.

Ten band user equipment 62 with over 900 MHz of accessible bandwidth should be capable of delivering a very different (better/more valuable) user experience when compared with a five band phone with 500 MHz of accessible bandwidth. The differentiation is not just a function of available bandwidth, although Band VII and Band 38 on their own add 190 MHz but also bandwidth quality with the 700 MHz and 800 MHz band offering benefits in terms of free space link budget and in building penetration. 63 Ten band user equipment with multi mode capability (see below) allows operators to provide a ‘best connect’ voice and broadband data proposition in all international markets. However producing ten band user equipment at an acceptable cost with a power drain and RF performance that is equivalent to an optimized single band or two, three, four or five band phone is a non trivial design task. The efficient exploitation of extended multi band handover opportunities requires dual receiver functionality. Such devices do not presently exist. The combination of SIMO and extended multiband handover could transform the RF cost and performance economics of the LTE uplink and downlink. 11 The technology proposition

A ten band phone (UE) provides a test bed for a number of new technologies and techniques which will help to minimize the additional cost of adding frequency bands to cellular handsets while delivering improved or at least equivalent rather than degraded performance.
61

Interference cancellation at baseband can also be beneficial in edge of cell conditions or in building where internal and external base stations may be simultaneously visible. 62 We refer at various times in this study to nine, ten or eleven band devices – these are cellular or mobile broadband bands and do not include other transmit/receive or receive paths for example Bluetooth, GPS or FM radio. 63 This assumes antenna and matching efficiency are equivalent to the performance achieved at higher frequencies and other operational bandwidths. Some propagation studies also suggest in building penetration may be frequency independent between 700 MHz and 2 GHz which would mean that 700 and 800 MHz networks have less inherent propagation gain than commonly assumed.

Geoff Varrall

Page 21 LTE UE efficiency and network value Final

15/09/2010

These techniques will be needed in order to address future global market needs, which will include inevitably more band options. This is not an optional requirement but a function of the way in which spectrum has been allocated and auctioned over the past twenty five years and a function of the Release 8, 9 and 10 standards process. A ten band LTE handset is arguably a necessary stepping step towards broader band devices that can be economically produced and shipped to all markets including markets with non standard band plans of which the US is one. The associated inventory savings would help off set any additional component costs. That said there remains a significant gap between the global market needs of the cellular operator community and present user equipment and RF component industry implementation desire and capability. The enabling technologies and techniques that will help bridge this gap can be summarized as follows; On the transmit side The development of efficient linear broadband power amplifiers. The challenges of producing efficient broadband linear amplifiers are well understood –one option is run the power transistors at a higher voltage into a higher impedance, improve efficiency through a combination of feed back and feed forward linearization and adaptive power matching and ensure that this efficiency and linearity and isolation is maintained through the filter chain and multiple switch paths and out to a broadband and/or tunable antenna. Higher power transistors may imply new materials technology (see engineering section below) and or more broadly applied adaptive matching techniques.
64 The benefits of a broadband amplifier would be greater if the devices could be coupled with adaptive duplex filters or switchable filter band modules but substantial R and D is still needed before such devices can be made commercially available.

On the receive side The realization of a broadband receiver can be significantly more challenging than a broadband transmitter. Adaptive noise matching for example is dependent on accurately measuring low energy signals that may be corrupted by locally generated noise. Adaptive matching can deliver substantial theoretical benefits when correcting for poor VSWR caused by using channels at the edge of a band or from hand capacitive effects or adapting to low battery conditions but are not necessarily applicable across multiple bands. Fundamental filtering both on the transmit and receive paths is presently reliant on surface acoustic or bulk acoustic wave filters. The performance in terms of Q and insertion loss continues to improve and new packaging techniques (two filters in a single package) save space but these are not tunable devices. As on the transmit path (see above) adaptive or self tuning filters that can be reconfigured for each band and or for specific channels within each band have yet to achieve mass market viability and may require 65 material science innovation to which the industry has little or no present visibility. These are solvable problems but the present rate of progress is unlikely to be sufficient to meet short to near term, mid term or longer term multi band cost, performance and market availability requirements. This is an area where substantially more short term R and D resource and engineering and manufacturing investment is clearly needed. On the transmit and receive side An efficient ten band antenna is challenging both in terms of basic physics (size versus frequency) and the constraints of present broadband matching networks and complex switch paths. New materials have helped improve performance and reduce coupling effects. See antenna section 20 for more detail. 12
64 65

The engineering proposition

A broadband amplifier is often in practice several amplifiers packaged together on a common die There are a range of industry opinions on this –see earlier section on China which draws on vendor assurances that these devices could be available within the next twelve months.

Geoff Varrall

Page 22 LTE UE efficiency and network value Final

15/09/2010

Some of the above are based on fundamental developments in material technologies. Examples include gallium nitride for high power transistors, CMOS on sapphire for high performance switches and doped barium strontium titanate thin film ceramic for voltage tunable devices. Voltage tunable devices have to deliver sufficient capacitance density, linearity, harmonic performance, low current leakage, dynamic range and switching speed. These devices can potentially find their way into tunable filters, tunable antennas, power amplifier matching circuits, adaptive impedance matching modules and phase shifters. There is the additional non-trivial requirement of selecting and measuring a parameter that will give the ‘quantity’ of tuning required to effect an optimum change. This function will absorb further baseband power. The development of these materials has taken fifteen, twenty or thirty years or more and required (and continue to require) substantial investment in specialist manufacturing. This investment has to be recovered. Adding NRE and manufacturing cost is easier to justify initially in a higher value design. The benefits of these new technologies and techniques can then be translated across into lower cost mass market products. 13 What bands should be supported in future LTE handsets?

The decision as to which RF bands should be supported is a crucial part of the negotiation process between an operator and their handset vendors and has a direct knock on effect on the RF component and semiconductor supply community. The process has been made more complex as a result of additional band allocations combined with the need to provide UMTS (3G) and LTE functionality together with GSM backward compatibility. Individual operator requirements are similar but not identical and it is important to ensure that handset vendors are willing and able to deliver a full range of products that specifically meet immediate and longer term cost and performance expectations. This includes handsets and mobile broadband devices that when roaming can deliver ‘best connect’ service in all markets. The diagram below illustrates the complexity introduced even when a relatively small sub set of 3GPP bands is supported.

Radio Multiband Architecture Example
MIMO FEM Diversity/MIMO front-end Main FEM LTE, HSPAevo, EDGE Rx MIMO signal processing

LTE, HSPAevo, EDGE Rx signal processing

Baseband i/f

‘Core’ WCDMA combination with international roaming

Main PA

- Band 1, 2, 4, 5 (6), 8

LTE, HSPAevo, EDGE Tx signal processing

Additional band support: e.g. band 3, 7, 11, 12-14, 20 or 40

Add on FEM

Add on FEM

Geoff Varrall

Page 23 LTE UE efficiency and network value Final

15/09/2010

Core Band Combination Band 1 plus other bands for international roaming support Band 1 3G WCDMA/LTE 1900/2100 MHz Band 2 US PCS 1900 1800/1900 MHz Band 4 US AWS 1700 1700/2100 MHz Band 5 and 6 US 850 band (plus Asia 824/840 869/894 and Japan) Band 8 Europe and Asia 900 880/915 925/960 MHz Additional bands Band 3 1800 MHz Europe and 1710/1785 1805/1880 Asia Band 7 2.6 GHz band Band 11 1.5 GHz for Japan Bands 12 and 14 US 700 MHz Band 20 or European 800 MHz Band 40 2.3 GHz for China The table shows the core bands supported and the additional bands and band (either or) combination options. Note the separate MIMO front end module (FEM) to support the diversity receive path. With thanks to ST Ericsson 66 To support progress beyond this point faster progress would be desirable or further optimization is needed in the following areas; 14 RF MEMS and other innovations

The application of Radio Frequency Micro Electro Mechanical System micro scale devices is being expanded to include MEMS switches, resonators, oscillators and tuneable capacitors and (in the longer term) inductors that potentially provide the basis for tuneable filters. RF MEMS are not a technical panacea and as with all physical devices have performance boundaries particularly when used as filters. If any device is designed to tune over a large frequency range it will generally deliver a lower Q (and larger losses) than a device optimised for a narrow frequency band. 67 An alternative approach is to exploit the inherent smallness of RF MEMS devices to package together multiple components, for example packaging multiple digital capacitors on one die. The purpose of these devices to date has been to produce adaptive matching to changing capacitance conditions within a band rather than to tune over extended multi band frequency ranges. 68 In an ideal world additional switch paths would be avoided. They create loss and distortion and dissipate power. More bands and additional modes therefore add direct costs in terms of component costs and indirect costs in terms of a loss of sensitivity on the receive path and a loss of transmitted power on the transmit path. MEMS (micro electrical mechanical system) based switches help to mitigate these effects. The idea of building micro electrical mechanical switches has been around for twenty years or so but is now becoming increasingly practical and has the benefit of sharing available semiconductor fabrication techniques. MEMS components are manufactured using micro machining processes to etch away parts of a silicon wafer or to construct new structural layers that can perform mechanical and electromechanical functions. A MEMS based switch properly implemented can exhibit low insertion loss, good isolation and linearity and can be small and power efficient. In addition it is essentially a broadband device. It is electro statically activated so it needs a high voltage which is inconvenient but at low current (so is more practical).

Used with permission for this study but please do not replicate for use in other material. The theory and practice of RF MEMS is treated in a new book from Cambridge University Press http://www.cup.es/uk/catalogue/catalogue.asp?isbn=9780521897716 68 Two suppliers, Cavendish and WiSpry, now have variable capacitor devices that can be used in a band-switching application to optimize the antenna gain by changing a capacitive load on the antenna to change the resonant frequency. This can also be implemented with GaAs switches and fixed passive capacitors in a simple network. Although the switches consume power and add parasitics, manufacturers such as Nokia have been using this technique to optimize band performance for at least two years. With the greater capacitive tuning range of RF MEMSbased variable capacitors, efficient 10-band designs have been demonstrated.
67

66

Geoff Varrall

Page 24 LTE UE efficiency and network value Final

15/09/2010

MEMS devices are sensitive to moisture and atmospheric contaminants so have to be hermetically sealed, rather like a quartz crystal. This packaging problem disappears if the device is sealed at the wafer level during manufacture with additional over moulding to provide long term protection. 69 Integrated MEMS devices are therefore a plausible candidate for band switching and mode switching. TX/RX switching (for GSM or other time division multiplexed systems) is more ambitious due to the duty cycle requirements but still possible using optimised production techniques. There is also a potential power handling and temperature cycling issue. The high peak voltages implicit in the GSM TX path can lead to the dielectric breakdown of small structures, a problem that occurred with early generations of SAW filters. Because MEMS devices are mechanical, they will be inherently sensitive to temperature changes. This suggests a potential conflict between present ambitions to integrate the RF PA on to an RFIC and to integrate MEMS devices to reduce front end component count and deliver a 70 spectrally flexible phone. The balance between these two options will be an important design consideration. The optimal trade off is very likely to be frequency specific. For example, if the design brief is to produce an ultra low cost handset, then there are arguments in favour of integrating the RFPA on to the RFIC. However this will make it difficult to integrate MEMS components on to the same device. 71 15 MEMS resonators

MEMS are also being suggested as potential replacements for present quartz crystal based sub systems. The potential to use micro electrical mechanical resonators has been the subject of academic discussion for almost 40 years and the subject of practical research for almost as long. The problem with realising a practical resonator in a MEMS device is the large frequency coefficient of silicon, ageing, material fatigue and contamination. A single atomic layer of contaminant will shift the resonant frequency of the device. As with MEMS switches and filters, the challenge is to achieve hermetically robust packaging that is at least as effective as the metal or ceramic enclosures used for quartz crystals but without the size or weight constraint. There are products now available that use standard CMOS foundry processes and plastic moulded packaging. While some 2G handsets make do with a lower cost uncompensated crystal oscillator, 3G and LTE devices need the stability over temperature and time, accuracy and low phase noise that can only presently be 72 achieved using a voltage controlled temperature compensated device. Temperature compensated MEMS oscillators (TCMO), particularly solutions using analog rather than digital compensation, are now becoming a credible alternative and potentially offer significant space and performance benefits. A MEMS resonator is a few tenths of a millimetre across. A quartz crystal is a few millimetres across, one hundred times the surface area. MEMS resonator performance is a function of device geometry. As CMOS geometries reduce, the electrode gap reduces and the sense signal and signal to noise ratio will improve, giving the oscillators a better phase noise and jitter specification. As MEMS resonators get smaller they get less expensive. As quartz crystals get smaller they get more expensive. MEMS resonators therefore become increasingly attractive over time.
69

Cavendish has introduced a MEMS process that does not require hermetically sealed packages since the MEMS devices are manufactured in situ during the metallization steps in a standard semiconductor flow. The MEMS are sealed within the semiconductor device, not wafer-sealed afterwards. Other MEMS suppliers are working on similar technology which will eliminate a large cost burden now born by RF MEMS devices.

The combination of RF power amplification and an RFIC is also problematic with GSM as the excess heat generated by the PA can cause excessive drift in the small signal stages of the receiver, including baseband gain stages and filtering. At the lower levels of UMTS and LTE this becomes less of a problem but still requires careful implementation. 71 Combining different technologies in a small board area dictates the use of a module type packaging approach. The more elements that are integrated, the larger and more costly the assembly. The manufacturing process of RF MEMS mentioned above could be used in the same process used to manufacture the RFIC. The question would be whether integration at the chip level is a less expensive solution than a multi-chip module.
72 In GSM some base bands compensate digitally avoiding the need for voltage control. VCTCXO devices are very low cost components so replacements must offer significant performance gain or flexibility benefit.

70

Geoff Varrall

Page 25 LTE UE efficiency and network value Final

15/09/2010

A technical study from Sand 9 provides information on the present status of these devices. 16 Filtering using Surface Acoustic Wave and Bulk Acoustic Wave Devices

SAW filters use semiconductor processes to produce combed electrodes that are a metallic deposit on a piezoelectric substrate. SAW devices are used as filters, resonators and oscillators and appear both in the RF and IF (intermediate frequency) stages of present cellular handset designs. SAW devices are now being joined by a newer generation of devices known as BAW (bulk acoustic wave) devices. 73 In a SAW device, the surface acoustic wave propagates, as the name suggests, over the surface of the device. In a BAW device, a thin film of piezoelectric material is sandwiched between two metal electrodes. When an electric field is created between these electrodes, an acoustic wave is launched into the structure. The vibrating part is either suspended over a substrate and manufactured on top of a sacrificial layer or supported around its perimeter as a stretched membrane, with the substrate etched away. The devices are often referred to as Film Bulk Acoustic Resonators (FBAR). The piezoelectric film is made of aluminium nitride deposited to a thickness of a few tens of microns. The thinner the film, the higher the resonant frequency. BAW devices are useful in that they can be used to replace SAW or microwave ceramic filters and duplexers in a single component. BAW filters are smaller than microwave ceramic filters and have a lower height profile. They have better power handling capability than SAW filters and achieve steeper roll off characteristics. This results in better ‘edge of band’ performance. The benefit apart from the roll off characteristic and height profile is that BAR devices are inherently more temperature resilient than SAW devices and are therefore more tolerant of modules with densely populated heat sources (transceivers and power amplifiers). However this does not mean they are temperature insensitive. BAR filters and SAW filters all drift with temperature and depending on operational requirements may require the application of temperature compensation techniques. However present compensation techniques compromise Q.
2 Twelve months ago a typical BAW duplexer took up a footprint of about 5 by 5mm with an insertion height of 1.35mm. In the US PCS band or 1900/2100 band these devices had an insertion loss of about 3.6 dB on the receive path and 2.7 dB on the transmit path and delivered RX/TX isolation of 57 dB in the TX band and 44 dB in the RX band.

A typical latest generation FBAR duplexer takes up a footprint of about 2 by 2.5mm2 and has an insertion height of 0.95mm max. In the US PCS band or UMTS-1900 band these devices have a typical insertion loss of about 1.4 dB on the receive path and 1.4 dB on the transmit path and deliver RX/TX isolation of 61 dB in the TX band and 66 dB in the RX band. More miniaturised versions (smaller than 2 by 1.6mm) are under development. SAW filters also continue to improve and new packaging techniques (two filters in one package) offer space and cost savings. FBAR filters and SAW filters could now be regarded as mature technologies but as such are both still capable of substantial further development. Developments include new packaging techniques. For example FBAR devices are encapsulated using wafer to wafer bonding, a technique presently being extending to 74 miniaturised RF point filters and RF amplifiers within a hermetically sealed 0402 package. SAW filters are however also continuously improving and many handsets and present designs assume SAW filters rather than FBAR in all bands up to 2.1 GHz though not presently for the 2.6 GHz extension band. Some RF PA vendors are integrating filters with the PA. This allows for optimised (lower than 50 ohm) power matching though does imply an RF PA and filter module for each band. There are similar arguments for integrating the LNA and filter module as a way of improving noise matching and reducing component count and cost. A present example is the integration of a GPS LNA with FBAR GPS filters into a GPS frontend module. 75
Epcos and Murata have ongoing development programmes for BAW devices to complement their existing highly optimised SAW filter product offerings. Avago are already an established supplier of FBAR filters. 74 Information from Avago July 7 2010 75 Information from Avago July 7 2010– these devices are presently shipping in volumes of more than 100 million units per year
73

Geoff Varrall

Page 26 LTE UE efficiency and network value Final

15/09/2010

RF PA vendors often ship more than just the power amplifier. Triquint for example in addition to GaAs based power amplifiers, switches and SAW filter/duplexers also supply BAW filters/duplexers using in house technology. Most of their BAW products ship inside their own integrated front end modules. 17 NRE cost of RF MEMS

In comparison to SAW, BAW and FBAR filters, RF MEMS realistically still have to be characterised as a new technology in terms of real life applications in cellular phones and mobile broadband user equipment. RF MEMS hold out the promise of higher Q filters with good temperature stability and low insertion loss and could potentially meet aggressive cost expectations and quality and reliability requirements. However the NRE costs and risks associated with RF MEMS are substantial and need to be amortised over significant market volumes. Recent commercial announcements 76 suggest that there is a growing recognition in the RF component industry that RF MEMS are both a necessary next step and a potential profit opportunity. The issue is the rate at which it is technically and commercially viable to introduce these innovations. 18 Tuneable capacitors and adaptive TX and RX matching techniques

In common with a number of other vendors including RF Micro Devices, Wispry, Peregrine and Paratek see below), Epcos are targeting tuneable capacitance applications in cellular phones, initially for adaptive matching within individual bands. This represents an extension from their original business (SAW filters). The rationale for adaptive matching is that it can deliver TX efficiency improvements by cancelling out the mismatch caused by physical changes in the way the phone is being used. A hand around a phone or the proximity of a head or the way a user holds a tablet or slate will change the antenna matching, particularly with internal and or electrically undersized antennas. A gloved hand will have a different but similar effect. Mismatches also happen as phones move from band to band and from one end to the other end of a band. TX matching is done by measuring the TX signal at the antenna. In practice the antenna impedance match (VSWR) can be anything between 3.1 and 5.1 (and can approach 10 to 1). This reduces to between 1.1 77 and 1.5 to 1 when optimally matched. Adaptive matching of the TX path is claimed to realise a 25% reduction in DC power drain 78 in conditions where a severe mismatch has occurred and can be corrected. However in a duplex spaced band matching the antenna on the TX path to 50 ohms could produce a significant mismatch. 79 Ideally therefore the RX path needs to be separately matched (noise matched rather than power matched). This is dependent on having a low noise oscillator in the handset such that it can be assumed that any noise measured is from the front end of the phone rather than the oscillator. Matching can then be adjusted dynamically. 80 Note that low phase noise is also needed to avoid reciprocal mixing to avoid desensitisation. Peregrine, Wispry, Paratek and Cavendish Kinetics all have tuneable capacitor product offerings. Peregrine utilizes a simple monolithic FET based switch approach, other technologies such as BST (Paratek), MEMS, (Wispry and Cavendish Kinetics) rely on analog bias control and DC to DC conversion which in themselves consume energy and create noise and thus represent a non trivial design and implementation task. 81

In April 2008 Epcos, the former Siemens Matsushita SAW filter business, acquired the RF MEMS assets of NXP Semiconductors. The stated objective is to change their business from essentially a one product company (SAW filters) to become an integrated vendor of adaptive front end modules. RF Micro Devices have stated similar ambitions. 77 Peregrine 22nd May 2009 78 Epcos press release as above 79 Information from Peregrine 22nd May 2009 80 A low noise oscillator is however expensive and current hungry. 81 The counter argument is that MEMS require very little energy to hold state since electrostatic force is used to hold the mechanical devices. Only leakage current is consumed which is quite small (in the nano amp range). In the Cavendish implementation there are no non-linear elements in the device since the capacitive plates are never used in the “free standing” state, rather are always held to the floor or ceiling of the MEMS cavity. These devices are therefore inherently immune to non-linearity from the RF signal. The advancement is that the capacitor plate can be pulled either to the floor – for a high C state, or to the ceiling – for a low C state. A large array , >1000, of these very small devices is used to produce a VariCap function by having some “down” and some “up” – but all locked to a fixed state.

76

Geoff Varrall

Page 27 LTE UE efficiency and network value Final

15/09/2010

RF Micro Devices have adaptive TX matching on their RF 6285 multi band power amplifier module. The device is intended to be capable of supporting Band 1(present UMTS 1900/2100 MHz), Band II US PCS at 1900 MHz, Band III (GSM/UMTS 1800), Band IV (the AWS 1700/2100 band in the US), Band VI (the 800 MHz band in Japan), Band VIII (the GSM/UMTS900 band) and Band IX (the 1700 MHz band in Japan). Analog Devices is developing a range of MEMS based devices for antenna tuning and matching. Sand9 Inc, a US based start up has received government funding to apply MEMS technology in to similar applications including potentially in the longer term tuneable RF filter applications (adaptive filter banks)and MEMS oscillators (see link above). Adaptive matching is by present indications one of the more promising new techniques for use in cellular phones. The technique is essentially agnostic to the duplexing method used. It does however offer the promise of being able to support wider operational bandwidths though this in turn depends on the tuning range of the matching circuitry, the loss involved in achieving the tuning range and the bandwidth over which the tuning works effectively. Note that most of these devices and system solutions depend on a mix of digital and analog functions. A strict definition of tunability would imply an analog function whereas many present device solutions move in discrete steps – eighty switched capacitors on a single die being one example. Of course if the switch steps are of sufficiently fine resolution the end result can be very much the same. There are some informed industry observers that consider a tuneable front end defined strictly as a well behaved device that can access any part of any band anywhere between 700 MHz and 4.2 GHz could be twenty years away. This begs the question as to what the difference would be between say a twenty five band device in twenty years time (assuming one band is added each year) and a device with a genuinely flexible 700 MHz to 4.2 GHz front end. From a user experience perspective the answer is probably not a lot but the inventory management savings of the infinitely flexible device would be substantial suggesting that this is an end point still worth pursuing. 19 Power amplifier material options

RF efficiency is determined by the difficulty of matching the power amplifier to the antenna or multiple antennas and to the switch path or multiple switch paths across broad operational bandwidths. However transmission efficiency is also influenced by the material and process used for the amplifier itself. Commercialisation of processes such as gallium arsenide in the early 1990’s provided gains in power efficiency to off set some of the efficiency loss implicit in working at higher frequencies, 1800 MHz and above, and needing to preserve wanted AM components in the modulated waveform. In some markets, GSM for example, CMOS based amplifiers helped to drive down costs and provided a good trade off between cost and performance but for most other applications GaAs provided a more optimum cost/performance compromise. This is still true today. WCDMA or LTE is a significantly more onerous design challenge than GSM but CMOS would be an attractive option in terms of power consumption, cost and integration capability if noise and power consumption could be reduced sufficiently. Noise can be reduced by increasing operating voltage. This also helps increase the passive output matching bandwidth of the device. However good high frequency performance requires current to flow rapidly through the construction of the base area of the transistor. In CMOS devices this is achieved by reducing the thickness of the base to microns or sub microns but this reduces the breakdown voltage of the device. The two design objectives, higher voltage for lower noise and a thinner base for higher electron mobility are therefore directly opposed. At least two vendors are actively pursuing the use of CMOS for LTE user devices but whether a cost performance cross over point has been reached is still open to debate. Low voltage PA operation also is a direct contributor to PA bandwidth restriction. Because it is output power that is specified, low supply voltages force the use of higher currents in the PA transistors. This only happens by designing at low impedances, significantly below the nominal 50 ohms of radio interfaces.

Geoff Varrall

Page 28 LTE UE efficiency and network value Final

15/09/2010

Physics requires that impedance value shifts happen with bandwidth restrictions: larger impedance shifts involve increasingly restricted bandwidths. At a fixed power, broadband operation and low voltage operation are in direct opposition. Gallium arsenide in comparison allows electrons to move faster and can therefore generally deliver a better compromise between efficiency, noise and stability, a function of input impedance. However the material has a lower thermal conductivity than CMOS and is more brittle. This means that it can only be manufactured in smaller wafers which increases cost. The material is less common than silicon, the world’s most common chemical element, and demands careful and potentially costly environmental management. But as at today, gallium arsenide remains dominant in 3G user devices for the immediately foreseeable future and is still the preferred choice for tier 1 PA vendors. Gallium nitride, already used in base station amplifiers is another alternative with electron mobility 82 equivalent to GaAs but with higher power density allowing smaller devices to handle more power. However gallium nitride devices exhibit memory effects which have to be understood and managed. Agilent have a useful application note on how to use X parameters to model these effects. Silicon germanium is another option, inherently low noise but with significantly lower leakage current than GaAs and silicon. However an additional thin base deposition of germanium (a higher electron mobility material) is required which adds cost. Breakdown voltage is also generally low in these devices. 20 Broadband power amplifiers

Irrespective of the process used it is plausible that a greater degree of PA re use may be possible. For LTE 800 to use the same power amplifier as LTE900 would imply covering an operational bandwidth of over 100 MHz (810 to 915 MHz) at a centre frequency of 862 MHz. This is 11.6% of the centre frequency. 83 As a comparison the 1800 MHz band is 4.3%. It would likely be unacceptable to accept any efficiency loss for a UMTS/LTE 900 handset incurred as a consequence of needing to also support LTE 800. This implies incremental R and D spending in order to deliver an acceptable technical solution. The general consensus is that 15% is OK, 30% is stretching things too far which would mean a power amplifier covering say 698 to 915 MHz would be unlikely to meet efficiency and or EVM and or spectral mask requirements. At present RF Micro Devices public statements would suggest that they will aggressively address these areas with announced plans to build a dedicated 200mm wafer fabrication plant for tuneable power amplifier modules with integrated TX/RX switches and mode switching. 84 The PA will need to be characterised differently for GSM (higher power) and for TDD LTE. The power in full duplex has to be delivered through a highly specified (linear) RX/TX duplexer with a TDD switch with low harmonics. The design issues of the PA and switch paths are well understood but if both FDD and TDD paths need to be supported then the design and implementation becomes problematic particularly when all the other existing switch paths are taken into consideration. Companies such as Nujira 85 and Parker Vision and Quantance are promoting the possibility of wide band amplifiers where one PA (with suitable linearization and adaptive matching) can replace two or three existing power amplifiers. The approach promises overall reductions in DC power drain and heat dissipation and cost and board real estate but will need to be optimized for any or all of the process choices. However, even if this broadband PA problem can be solved, within the foreseeable future each full-duplex band requires an individual duplexer. Connecting one broadband PA to multiple duplexers is a lossy proposition with complex circuitry. To match a good broadband PA, a switched (or tuneable) duplexer is required.
Gallium nitride is currently used in a number of military applications where efficient high power RF is needed – this is a Triquint example 83 PA operational bandwidths of up to 20% of centre frequency are now possible though with some performance compromise. This is an area where substantial recent progress has been made (conversation with Avago 13 June 2008) for example Skyworks and RFMD have multiband power amplifiers that cover 1710 to 1980 MHz, a bandwidth of 13%. 84 RF Micro Devices Introduces MEMS technology for functional integration of RF 5th December 2008 Product news 85 http://www.nujira.com/news.asp?pageID=42
82

Geoff Varrall

Page 29 LTE UE efficiency and network value Final

15/09/2010

21

Path switching and multiplexing

Unlike power amplifiers which drive current into a low impedance, switches must tolerate high power in a 50 ohm system. If connected to an antenna with impedance other than 50 ohms the voltage will be even higher, commonly above 20Vpk. The switch is in the area of highest voltage and has the most severe power handling and linearity requirements with a need to have an IP3 of >70dBm. The table below from Peregrine Semiconductor gives typical power handling requirements.

System Standard Two Tone GSM EDGE CDMA2000 WCDMA LTE 802.11g VHF 2-way

Average Power1 30 33 27 23 24 19.4 15 40

Peak Power1 33 33 30.2 28.1 27 27 22 40

PAR 3 0 3.2 5.1 3 7.6 6.6 0
3

VPK 14.1 14.1 10.2 8.0 7.1 7.1 4.0 31.6

Operating VSWRs n/a 4:1 4:1 4:1 4:1 4:1 3:1 8:1

VPK @ VSWR2 n/a 22.6 16.4 12.9 11.3 11.3 6.0 56.2

Note 1 -- Maximum average power at antenna in a 50-ohm system per the system specification Note 2 -- At angle which produces the highest voltage Note 3 -- Assumes crest factor reduction

Low voltage FETS can be ‘stacked’ in series for high voltage handling in the OFF condition but this requires a low loss insulating substrate, either GaAs or silicon on sapphire. Apart from power handling, switches must be highly linear and have low insertion loss. Switch losses have more than halved over the past five years which has allowed for higher throw counts and cascaded switches but substantial performance improvements are still needed due to the steadily increasing throw count of the antenna side switch, the addition of a PA side switch and the need to support bands at higher frequencies.

The diagram from Peregrine shows this additional complexity. The CMOS Sapphire technology used provides good isolation 86 and the use of CMOS means that the off chip logic can be integrated which is generally harder to achieve in GaAS switches.
86

However there is trade off between insertion loss and isolation and both are needed.

Geoff Varrall

Page 30 LTE UE efficiency and network value Final

15/09/2010

The nine throw switch integrates TX low pass filters. The diagram below shows the comparison with a GaAs device. .

GaAs SP9T ASM
Integrates two TX LPFs

UltraCMOS™ SP9T ASM
Integrates TX LPFs

3.0 mm

CMOS Decode/Bias

GaAs SP9T

LPF IPD

2.5mm

2.5 mm 3.8 mm

>45% reduction in area
CMOS scaling combined with finer metal pitch technology and improved transistor performance should however realise future benefits both in terms of size and insertion loss, a 0.18 um process would halve the insertion loss and core area and double the bandwidth. The insertion loss and size benefits are shown below.

The overall progress of silicon on sapphire is shown below. Note that on the basis of these measurements the insertion loss of silicon on sapphire has now achieved parity with GaAs.

Geoff Varrall

Page 31 LTE UE efficiency and network value Final

15/09/2010

These devices are used in >30% of WCDMA handsets presently shipping. Peregrine has also launched its first two digital tunable capacitor DTC products with high volume production slated for later this year. The DTC products will be produced on the same process and at the same fabrication facilities as its high volume switch products. Longer term applications extend to digitally controlled monolithic tunable circuits. More detailed information on these devices can be found on the Peregrine Semiconductor web site. A technical case study on RF front end tunability for LTE handset applications can be downloaded here. Paratek are suggesting similar significant flexibility options and cost and performance advantage can be achieved using their proprietary Parascan process. This is a thin film ceramic material, a doped version of Barium Strontium Titanate (BST). The dielectric constant of the material varies with the application of a DC voltage. The device is claimed to have a Q of 100 at 1 GHz and better than 80 at 2 GHz, and is claimed to be highly linear and provide good harmonic performance with low power consumption, high capacitance density, an IP3 of greater than 70 dBm, fast switching speed and low current leakage. Last year, Paratek announced a joint venture with ST Micro Electronics to commercialize these products for cellular handsets. 87 The process is potentially usable in tunable filters and antennas, power amplifier matching circuits, adaptive impedance matching modules (AIMM), and phase shifters. Successful implementation would imply that ST Ericsson could develop new wide band RF solutions that could be integrated with their LTE modem products (previously Ericsson Mobile Platforms). Note that all parts of the TX/RX path have to evolve in parallel. For example one RF MEMS vendor commented that they had reached an advanced stage of development with a digital duplexer for a handset manufacturer but the project was moth balled due to lack of progress with tuneable and/or broadband linear power amplifier modules. 88 22 Antenna innovations - impact on LTE in sub one GHz bands

The design of the antenna system for LTE mobile devices involves several challenges; in particular if the US 700MHz (Bands 12, 13, 14, 17) bands have to be supported alongside the legacy cellular bands frequencies in handset-size user equipment. In general, the upper frequency bands (>=1710MHz) don’t create particular difficulties to the antenna design. The two main challenges are: - Bandwidth: The antenna must have enough bandwidth to cover the additional part of the spectrum, and this is difficult to achieve in the 700MHz bands as size of the device is small compared to the wavelength - MIMO (Multiple-input and Multiple-output): Multiple antennas, at least 2, must be integrated in the device to support the MIMO functionality. Below 1GHz isolation becomes an issue and the correlation between the radiation pattern is high, limiting the MIMO performance. Bandwidth The maximum theoretical bandwidth available for a small antenna is proportional to the volume, measured in wavelengths, of the sphere enclosing the antenna. A typical antenna for mobile devices (PIFA, monopole, etc) excites RF currents on its counterpoise, typically the main PCB of the device or a metal frame; therefore the PCB/frame of the device becomes part of the “antenna”, and the maximum linear dimension of the device should be used as the diameter of the sphere. Therefore, in the low frequency bands the bandwidth of the antenna is proportional to the third power of the size of the device using the antenna, and as the frequency is reduced the bandwidth reduces quickly. More bandwidth can be traded off at the expense of antenna efficiency. For a typical handset with length around 100mm, it can be already difficult to cover the legacy 3GPP bands 5+ 8 (824-894MHz and 880- 960MHz), so adding support for an additional 100MHz (698-798MHz) at an even lower frequency is a major challenge for small devices; for larger devices, for example tablets, the problem is reduced.

87

January 13 2009 Press Release Paratek and ST Micro Electronics announced a strategic relationship to supply RF tuneable products (dynamic impedance matching) from the ST facility in Tours. 88 Conversation with Wispry 20th May 2009

Geoff Varrall

Page 32 LTE UE efficiency and network value Final

15/09/2010

To overcome the above challenge it is widely believed a tuneable/switchable antenna is required; in fact, most antenna manufacturers, including Pulse, Antenova, Ethertronics and Skycross have announced tuneable LTE antennas. Although a tuneable/switchable antenna can greatly help in meeting the LTE requirements, it comes with its own set of challenges. In particular, the antenna has to satisfy stringent linearity requirements and introduces added cost. Different choices of active elements are possible. Tuneable digital capacitor arrays have been announced by Peregrine Semi Conductor (DuNE™ Antenna Tuning Technology, based on Ultra CMOS FET switches), Wispry (using MEMS) and EPCOS (using MEMS) and others; the resulting “variable capacitors” have ranges between fractions of pF to tens of pF. Other companies, for example Antenova and NEC, propose the use of conventional high linearity RF switches (SPDT, SP3T) to create switchable antenna functionality. The active elements can be used in the matching circuit, in which case the function partly overlaps with that of the adaptive matching circuits for the PA discussed below or can be part of the antenna, in which case the active element can be used to change the effective electrical length of the antenna. In any case, the physical dimension of the antenna cannot be reduced below a certain point even using a tuneable antenna: - the instantaneous bandwidth of the antenna must be well above that of the channel (20MHz for bands above 1 GHz) - if the bandwidth is narrow, a sophisticated feedback loop must be implemented to make sure that the antenna remains tuned correctly in all possible conditions where an external object (head, hand…) could shift the resonant frequency. The feedback loop is typically required if capacitor arrays are used, as the fabrication tolerance can be quite high. - The above feedback loop is easier to implement in the uplink (UE TX) than the downlink (UE RX) however the antenna must have enough instantaneous bandwidth so that if it is optimally tuned in the uplink it is automatically also tuned in the paired downlink frequency. Some companies have shown that it is possible to cover the US 700MHz plus all the legacy 3GPP bands plus the LTE band 7 (2500-2700MHz) band using a small antenna with only 3 states. As the bandwidth of the antenna is mostly dictated by the size of the device through some fundamental physical relation, it is unlikely that innovations in materials can bring groundbreaking advancements in the miniaturization of the antennas. MIMO To meet the assumed market need for high peak data rates, LTE uses MIMO. The baseline UE capability is two receive antennas and one transmit antenna. In the downlink, the UE must support receive diversity and dual stream MIMO. Initial network deployments will be 2x2 or 4x2 in the downlink. In order to work effectively, the two receive antennas must have similar gain and high isolation, preferably better that 10dB, and their far field radiation pattern must be as much as possible orthogonal (envelope correlation coefficient 89 <<1). If these requirements are not satisfied, the signal received from antenna 2 will be very similar to that from antenna 1, and so the receiver will not be able see two independent channels created by the multi path environment, collapsing into a single stream MISO (Multiple-input and Single-output) transmit diversity system. For a typical handset-size device, designing a MIMO antenna system is not particularly challenging in the upper frequency bands. However, in the lower frequency bands (<1GHz), the requirements are very difficult to achieve due to the small size of the device compared to the wavelength. The root of the problem is the same as the one causing the limited bandwidth. In general, a separation of about a quarter wavelength between two antennas is required to achieve a good isolation and low pattern correlation. Although apparently this requirements can be met even at 700MHz (quarter-wavelength = 107mm) in a 100mm long device, for instance placing the antennas at opposite ends of the device, this result is just apparent. In fact, as described above, both antennas will excite the same or a very similar radiation mode on the main PCB of the device, being the counterpoise for both antennas, 90 resulting in very similar (dipole-like) radiation patterns and so high correlation (close to 1). In general, the
89 Of course, all the other usual requirements (good input matching, high efficiency, broad bandwidth) must be met at the same time. 90 At higher frequency it is possible to excite very different radiation modes, resulting in low correlation however although it is possible to excite higher order modes even at frequencies <1GHz, with fields confined near the antenna

Geoff Varrall

Page 33 LTE UE efficiency and network value Final

15/09/2010

best that can be achieved is to excite two radiation modes aligned with each diagonal of the main PCB: the two resulting dipole-patterns will be skewed by an angle about 40-50 degrees, resulting in an envelope correlation coefficient about 0.3, which might be sufficient. The isolation of the antenna will depend on the details of the antenna designs, and methods are available to increase the isolation, for instance artificially introducing a coupling line between the two antennas that partially cancel in phase the natural coupling of the two antennas. As at least two antennas have to be embedded, the MIMO requirement has a significant impact on the dimensions and mechanical design of the user terminals. We have seen above that one cannot expect a reduction in the size of the antennas even if the they are made tuneable or switchable: so, the mechanical designers have to allocate enough space for two antennas of reasonable size (for instance 40mm (W) x 8mm (L) x 6mm (H) if full ground plane is placed under the antennas), while taking into account all the constraints required to make the MIMO antenna system work properly. In particular, the placement must be such that the isolation between antennas is sufficient and pattern correlation is low; moreover, the antennas must not be covered or otherwise affected by the user hand and head in the typical use conditions, and comply with specific absorption rate (SAR), and hearing aid compatibility (HAC) and other regulations. It is worth noting that performance in presence of the hand(s) but not the head will be of greatest importance in high speed data mode (e.g. streaming videos); moreover, it is not sufficient if just one antenna is not affected by the hand(s), as otherwise the 2x2 MIMO system will effectively collapse into a lower rate 2x1 MISO channel. All this means that it will be even more important than before that the antenna designers are involved from the concept stage of any new terminal design. To summarize, the decision to deploy LTE in the lower frequency (<1GHz) bands has created enormous pressure on the antenna design for the user terminals, and set off a race between designers to produce effective solutions, although most of these solutions have yet to be proved in real devices and real networks. Hopefully, the high data rate will demand large displays, and so the typical device size is bound 91 to grow, somewhat easing the problems and challenges. General comments on antennas in smaller form factor devices including smart phones A traditional PIFA antenna uses the ground plane as a significant part of the antenna. PIFA antennas tend to have a natural broadband response due to their effective large size due to the interaction with the ground plane but this increases interaction between other antennas and increases susceptibility to hand capacitance effects. Unwanted coupling becomes a particular issue at lower frequencies. Antennas in space-constrained designs are a fraction of the resonance length, and radiation efficiency and bandwidth are a function of volume. Space needs to be left around the antenna to avoid power being shorted out of the device. The antenna is shielded by the user’s hand or head and the antenna needs enough bandwidth (a function of volume) to resist detuning. Antenna designs also need to comply with regulatory requirements for SAR and HAC. A technical paper on these constraints is available from Quintel. A number of companies have introduced material and packaging innovations that have helped reduce real estate occupancy in smaller form factor phones. 92 Ethertronics for example now have a have a penta band antenna 800, 900, 1800, 1900, 2100 MHz based on their Isolated Magnetic Dipole technology. IMD relies on the magnetic field as main source of radiation (most antennas use the electric field between the antenna element and ground plane as the main source of radiation). This localizes current to the antenna rather than the ground plane

and with a radiation pattern different from that of the fundamental mode, these modes have a extremely narrow band and are not suitable for this application. 91 The counter argument to this is that 700 and 800 MHz networks will probably need to support smaller form factor devices including dongles, smart phones and standard phones in order to amortise spectral and network investment over a broad user product mix. Not to do so would place these networks at a commercial disadvantage. 92 The launch of the iPhone4 in summer 2010 highlighted some of the issues of small form factor devices where the antenna design has been dictated by aesthetic rather than functional considerations. Although the pattern for form factors is not set, the current trend is smart phone and tablets with no “tweeners”. The Dell Streak (5” screen) is the current exception, but early market data indicates this form factor has not taken off and most other vendors have gone to a 3.5” or 4.3” display for a smart phone and a much larger display for a tablet. In both cases, the amount of available volume for an antenna is getting challenged as all the dimensions of a phone (iPhone z-axis is ~10mm) are getting smaller to stay competitive with consumer requirements. In USB dongles, growing the length of the dongle larger than in the past is problematic. Lastly, in laptop, the bezels around the display are shrinking to 10mm for aesthetic reasons and again, making the antenna challenge greater.

Geoff Varrall

Page 34 LTE UE efficiency and network value Final

15/09/2010

Antenova have an 8 band antenna based on their High Dielectric Antenna (HDA) technology based on the principle of inducing radiating currents in dielectric materials. The devices deliver good resistance to detuning, high radiation efficiency and good isolation from other antennas. 93 Skycross base their antennas on their Isolated Mode antenna technology iMAT. This makes a single antenna structure behave like a multiple structure – based on maintaining isolation between multiple feeds to the same structure – useful for MIMO 94 and higher gain diversity systems Provided sufficient isolation can be achieved between feed points then the single antenna option avoids the effect of unwanted coupling between antennas and can improve the total radiated power by several dB. Skycross also promote their meander line and spiral shaped antennas as an option for creating ultra compact narrow band antennas or ultra wide band antennas (of the order of several octaves). In parallel, more conventional antenna structures from other vendors such as Pulse Engineering (previously LK Products) continue to improve. SISO Testing 95 Antennas in user equipment have morphed from a folding or whip antenna in first generation devices to smaller helical designs to the first integral patch antenna in 1994, the first of many patch antenna implementations. From an aesthetic perspective patch antennas were a great leap forward, from a performance perspective a potential significant step back. Antenna performance concerns prompted a set of single input single output test standards which were ratified by CTIA in 2001. These define methods for testing total radiated power (TRP) and total isotropic sensitivity (TIS). No similar antenna test standards presently exist for MIMO. In terms of functional design the volume required for one good antenna provides sufficient space for two poor antennas and the performance has to be verified over multiple cellular and non cellular frequency bands. Most user equipment testing is performed by connecting a cable to a ‘temporary antenna connector’ which bypasses the antenna. The assumption used to be that the device under test antenna could be fairly represented by an isotropic antenna with 0dBi gain. This was relevant when an antenna was a dipole tuned to a single band of interest but is no longer appropriate with most present antenna implementations. This means that conducted tests for reference sensitivity and maximum output power do not reflect real life radiated performance. This can be seen by comparing the requirements for conducted tests with those for radiated tests. The 2001 CTIA tests were followed in 2008 by the 3GPP test specification TS34.114 which define requirements for TRP and total reference sensitivity (TRS), which is the same as TIS in the CTIA standard. There is a significant amount of measurement uncertainty in these tests of the order of plus or minus 1.9 dB for TRP and plus or minus 2.3 dB for TRS, the consequence of an error model with over twenty terms. A large measurement error would allow some poor devices to pass the test and some good devices to fail. The solution to this for 3GPP was to relax the minimum requirements set out in the test specification by about half of the acknowledged test uncertainty, and in parallel introduce a more strict non mandatory recommended performance target. The tests are quite thorough covering azimuth and elevation angles, polarization, device mechanical and application modes (open, shut, vertical, and horizontal, voice and data) and include the low, middle and high channels of each frequency band. Phantom heads and phantom hands are used to replicate typical usage conditions. Phantom hands are currently right hand but could be extended to the left hand as the interaction between the device and hand can be highly asymmetric. Some device form factors will have contact from two hands at once and possibly

93 94

Short antennas are capacitive and need inductive loading to tune them. The result is low Q and efficiency. Good MIMO performance is dependent on a low antenna pattern correlation coefficient 95 Our thanks to Moray Rumney of Agilent Technologies for these insights into SISO and MIMO performance verification. More details are available in the August 2010 issue of the Microwave Journal.

Geoff Varrall

Page 35 LTE UE efficiency and network value Final

15/09/2010

more in future 96 Characterizing one device over one band requires thousands of measurements and can take up to two weeks in an expensive anechoic facility 97 . Tests undertaken by Orange 98 showed that free space measurements of radiated performance across a range of devices varied by 7 dB for TRS and 4.5 dB for TRP and head and hand loading will increase this spread further. All the devices tested had passed narrower conducted tests suggesting that the larger spread was due to the physical antenna implementation. The positive point to make is that operators now have access to objective and standardized over the air (OTA) radiated domain performance measurements that provide comparative measurements that are closer to real world performance than conducted domain measurements (made via a temporary antenna connector). The introduction of minimum performance requirements has however been difficult. Operators wanted to set high targets but user equipment vendors wanted to protect existing designs and small form factor devices that could be termed non compliant. The result was a compromise which allowed the bulk of legacy devices to remain compliant but with a non mandatory recommended average performance typically 3 dB better. The recommended TRP performance is still about 6 dB below the nominal power for the conducted test which suggests substantial room for improvement were it not for shrinking device sizes. If average data throughput, for example in micro and macro cells is as important or more important than peak data rate for example in pico or femtocells in terms of impact on application performance and user experience value then it would seem that optimising SISO antenna performance should be a priority over and above MIMO optimisation. Presently this would not appear to be the case. It is particularly important to ensure that an over aggressive pursuit of peak data rates in small cells does not compromise average data throughput in larger cells. A MIMO antenna design ideally needs to deliver a net gain in a wide range of operational conditions. This is often not the case. MIMO testing MIMO OTA standardization started in the CTIA and later the 3GPP standards process. The rather complex relationship between MIMO antenna performance, spatial channel gain and scheduling efficiency makes it difficult to agree on an objective figure of merit that captures real life conditions and real life throughput performance. One approach is to undertake SISO measurements and then emulate arbitrarily complex spatial channel conditions and frequency selective scheduling gain. MIMO gain relative to SISO is dependent on low antenna correlation and high signal to noise ratios. Low antenna correlation is more likely to be achieved at higher frequencies. Such gains are only likely to be significant in cells where the doppler speed and channel conditions change slowly enough to allow adaptive modulation and coding to be used. In a loaded network, the median SNR will be in the order of 5dB. This suggests real life MIMO versus SISO throughput gains will be nearer to 20% than the 100% often claimed. Ideally a test regime has to be developed that emulates practical rather than theoretical conditions. If SISO OTA radio measurements can be controlled within plus or minus 2 dB then MIMO measurements are unlikely to be more accurate. A large uncertainty could be argued to invalidate the value of the test. Relaxing the MIMO OTA minimum requirement by some proportion of the test uncertainty would really not help very much. What all of the above highlights is that a measurement approach needs to be agreed for MIMO devices that results in a methodology that can distinguish good and bad performance. Such an approach has now been agreed after considerable effort in SISO measurements but has yet to be agreed for MIMO measurements.

96

Future multi player gaming devices being one example. See also CTIA/PTRCB test procedures for the US market http://www.cetecom.com/us/en/certification/northamerica/ctia/over-the-air-performance.html. The test procedures state a requirement of a minimum of 50% antenna efficiency. Optimum antenna designs can approach 70% efficiency (measured in an anechoic chamber). Some handsets shipped in to some markets have been measured at 10% efficiency which would appear to validate the need for a minimum acceptable performance threshold to be applied in all markets. 98 Orange, “OTA TRP and TRS requirements for GSM 900 and 1800,” 3GPP R4-091762
97

Geoff Varrall

Page 36 LTE UE efficiency and network value Final

15/09/2010

As of today it is possible to do comparative measurements of SISO user equipment that are meaningful and provide a benchmark against which subsequent year on year performance improvements can be measured. This is not presently possible in MIMO user equipment. The industry might pause to consider whether the expectations for spatial multiplexing gains in ideal conditions might possibly be disproportionate and unduly influenced by a marketing fascination with headline peak data rates in small diameter cells. Improvement in robustness from RX and TX diversity may well be of more significance. Some aspects of MIMO, for example RX and TX diversity and beam steering, may deliver enough gain over a sufficiently broad spread of operational conditions to be worthwhile but generally spatial multiplexing gains in the lower bands and in larger diameter cells will be marginal and it would seem that effort is being expended in areas that will have less impact on the user experience in wide area access conditions than commonly assumed. Appendix 2 also visits this topic in the context of comparing LTE and Wi Fi connectivity throughput speed and efficiency suggesting that LTE MIMO in local area applications may also be of marginal value particularly if it compromises wide area performance. WiFi might be a better (more cost and energy efficient) option for load shedding. Whatever the timing and relative R and D focus, at some point a rational method for the comparative testing of MIMO devices needs to be agreed. Those measurements can then go forward as a basis for establishing real rather than theoretical year on year MIMO user equipment performance improvements. SIMO gains SIMO refers to the user/device path from the base station from where a modulated waveform is transmitted using specific resource blocks delineated in the time domain and frequency domain (frequency sub carriers within a wide band channel). This ‘single input’ to the radio channel is then transmitted and is received by multiple antennas at the receiver each with its own version of the signal based on the multipath channel. The receiver then combines these signals to create a better signal than is possible from one receiver. Generally, in almost all small and large cell propagation conditions, there will be strong multi path components (with a longer time delay in larger cells). In WCDMA (and GSM) multi-path signals are equalised out at baseband. In HSPA (see Advanced Receivers in Appendix 4) and LTE the additional option of diversity gain is supported (multiple outputs from the channel). If implemented with a separate receive path then additional gain can be achieved in most channel conditions provided antenna efficiencies have not been compromised by the SIMO implementation. Diversity also mitigates fast fading though there are other processes, for example channel sensitive scheduling, that have the same or similar effect at least in small cell environments. Quite often one technique is credited with delivering more gain than it really achieves in practice, it is the combination of adaptive averaging mechanisms that is important. Note that none of these techniques are a substitute for good RF performance but should be seen as techniques for getting higher throughput though a complex rapidly changing multi path channel where fast and slow fading and Doppler effects need to be managed. As stated above, good RF performance in larger cells may be as important as high peak data rates in small cells and an over emphasis on small cell performance may compromise large cell performance metrics. Although the generalized benefits from spatial multiplexing might be questionable there is strong 99 evidence from operator tests that receiver diversity, which can be described as SIMO (single input multiple output) can result in a useful gain in downlink capacity, of the order of 50% to 100% assuming all terminals are diversity enabled. This would suggest that theoretically at least a second antenna and second receiver are well worthwhile. A full bidirectional MIMO implantation of course also requires a second transmit path in the user’s equipment which is altogether more challenging. The second receiver in a SIMO enabled UE could also be used to do inter band CQI measurements to support extended multi band handover which could deliver substantial macro scheduling gain. 100 23 Adaptive matching power savings

We have stated earlier that a cross section of RF component vendors are presently working on the implementation of adaptive techniques, in particular adaptive matching on the TX and RX paths and

99

100

Inputs from a Tier 1 operator 20th August 2010 There is no present standards work ongoing on this but this would seem to be a useful future work item.

Geoff Varrall

Page 37 LTE UE efficiency and network value Final

15/09/2010

adaptive linearization on the TX path. This includes SAW filter vendors such as Epcos, RF MEMS vendors and RF PA vendors, usually in partnership with one or more of the RF MEMS suppliers. Adaptive matching can either be used to improve RX and TX performance within a specific band or be used in multi band RF architectures to minimise performance loss from wider band TX/RX devices. Epcos 101 provide an example of present work on adaptive matching. Modules presently under development use an electro statically variable capacitive RF MEMS switch (acquired from NXP’s RF MEMS activity two years ago) with a Q of 250 at 1 GHz. Five switches are used to provide 32 capacitance values to give a 10.1 tuning ratio . Power consumption of the adaptive tuner is presently 4.4 milliwatts and will reduce to 1 milliwatt in the future, insertion loss is 0.5 dB, harmonics H2 H3 are <91 dBc, and intermods IM3 are <117 dBm. The device will handle 3 billion cold switching cycles for GSM. Every 1 dB of mismatch loss equates to 1 dB less output power and 1 dB loss of sensitivity. For a WCDMA phone at 1900/2100 MHz, a 1dB loss will increase current drain by 70 mA. In low band GSM a 1 dB loss will increase current drain by 250mA due to its higher average power. (Two watts rather than one watt) 102 24 RF PA and front end switch paths

There are a range of views from RF PA component vendors as to how a multi band phone supporting 700, 800 and 2600 MHz will be implemented. Options 103 include Option 1- a single RF PA and band optimised matching network for each band Option 2 - one PA module for Band V (US LTE 850 or E850), Band VIII (LTE 900), Band III (LTE1800), Band II (US LTE1900) and Band I (UMTS/LTE1.9/2.1). A second module would cover Band 40 (China LTE TDD at 2.3 GHz), Band VII (UMTS/LTE 2.6 GHZ extension band) and Band 38 (LTE TDD in the Band VII duplex gap). A third module would cover the US and European DDR bands at 700 and 800 MHz. Option 3 would be as Option 2 but extended down to cover the 700 and 800 MHz DDR bands. Option 4 would be as option 3 but extended up to cover the 2.6 GHz band
104 would be to have a PA module for bands below 1 GHz, a second module for bands up to and Option 5 including Band I and a third module for frequencies above 2.1 GHz. However this option would be significantly different from existing solutions available in the market.

Option 6 would be one wideband power amplifier to cover all bands. Table 5 PA options Option 1 Option 2 Option 3 Option 4 Option 5 Option 6 Separate power amplifiers and individually band optimised TX/RX paths Three separate modules, one for core bands from Band V 850 MHz up to 2.1 GHz, a second for 2.3 and 2.6 GHz, and a third for US and European DDR at 700/800 MHz As option 2 but with the core band module extended down to 700 MHz As option 3 but with the core band module extended up to 2.6 GHz Three separate modules, one for sub one GHz, one for existing bands between 1GHz and 2.1 GHz, one for 2.3/2.6 GHz One broadband PA for all frequencies

101 102

Epcos acquired the RF MEMS assets of NXP Semiconductors in 2008 Source Peregrine July 2010 103 Options based on discussions with RFMD, Triquint, Anadigics and Skyworks Feb 2009 104 Based on conversations with Nujira 18 Feb 09 though see earlier comments on LTE energy drain

Geoff Varrall

Page 38 LTE UE efficiency and network value Final

15/09/2010

Option 6 would presently be ambitious both from a PA design perspective and would be dependent on implementing wide band phase lock loop architectures with acceptable noise performance, availability of sufficiently efficient wide band adaptive matching and availability of sufficiently efficient wide band linearization techniques. Note that PA bandwidth is limited not by the active device but by the bandwidth of the output matching circuit. An output matching circuit is necessary because of the inherent impedance difference between the antenna and the PA device. Five volts (or less) causes the RF output device to have an output impedance of well below 50 ohms. Devices following the PA for example directional couplers, filters, antennas and isolators have 50 ohm impedance and so an impedance matching circuit (passive) is required. The realization of this network will limit the bandwidth. Using a supply above five volts will enable the PA output transistor to be operated nearer to a 50 ohm output impedance thus relaxing but not entirely removing the requirements on the bandwidth limiting impedance transformation network. By this means, PA bandwidths of an octave become possible. In order to linearise PA operation, some vendors (Nujira, Quantance, Parker Vision) use envelope tracking. This is achieved by generating a ‘tracking control voltage’ to modify the supply voltage to the PA transistor. A DC-to-DC buck/boost converter is used both as a control function of the supply and to raise the voltage to greater than 5 Volts. 105 Additionally the PA drive is produced at very low output power levels in order to assist the tracking process through the area where the PA transfer function is extremely non-linear. Using this technique ideally results in a relatively constant PAE over a widely varying PAPR. The control voltage function is reconstructed from digital baseband information and due to the PA device variation will require calibration at handset assembly stage. The assumption is that the improvement in efficiency will out weigh the inconvenience. Note that as stated earlier higher data rates require better control of transmit and receive linearity in order to avoid unacceptable EVM on the transmitted signal and/or demodulation errors in the receiver. Linearization techniques have to work harder as operational bandwidth increases (see matching comments above), as channel bandwidth increases (from 3, 5, 10 to 20 MHz), as data rates increase and as distance from the cell site increases. After taking these factors into account the consensus is that Options 1 through 5 should all be realisable for market introduction by 2012/2014. Options 2 or 3 are the favoured option from two/three of the major RFPA vendors as this leverages existing R and D investment.
106 is that component count reduction, one of the objectives of options 2 The generally accepted principle through 6, is only commercially worthwhile if performance is at least as good as individually band optimised solutions.(Option 1). A present benchmark for this could be considered to be Anadigics 107 , advocates of optimised power amplifiers for each band. Optimisation includes adaptive matching (Quantance are a present partner for some of this work).

Present devices are known as ‘Help’ (high efficiency linear power) devices. The company suggests that power amplifiers are similar in price to switches and if efficiency can be optimised then heat gain can be minimised and package size can be minimized. Their present generation (Help3) devices have two operating modes one for high output and one for medium low output. A built in voltage regulator avoids the need for external voltage regulation and load switching. The devices come in a 3mm by 5mm by 1mm (low profile) surface mount package incorporating matching networks for optimum output power, efficiency and linearity into 50 ohms. The devices have a peak HSPA output power of 28.5 dBm, power added efficiency of 38.5% at + 29.5 dBm and 22% at + 16dBm in the PCS 1900 band. They suggest that 35 to 40% efficiency should be an achievable LTE bench

105 When closed loop (adaptive) feedback linearising techniques are used, a precision receiver is needed within the transmitter to accurately measure the actual PA output. This has complexity and power consumption implications. 106 Conversation with Triquint 18th February 09 107 Conversation with Anadigics 16th February 2009

Geoff Varrall

Page 39 LTE UE efficiency and network value Final

15/09/2010

mark at max power reducing to 25% at lower power. Quiescent current is presently 8 milliamps but with a target to reduce this to 4 milliamps when next generation (Help4) devices are available. Skyworks 108 provides a similar benchmark and has single-band or multi-band LTE devices supporting Options 1-3 and 5. The SKY77441 FDD/TDD LTE PA module is a fully matched 16 pin surface mount multi-band module developed for LTE FDD band VII (2.6 GHz extension band) and TDD bands 38 (duplex gap in the extension band) and 40 (China LTE TDD at 2.3 GHz) giving 27.5 dBm of linear power output based on a full LTE resource block allocation using QPSK or 16 QAM modulations. It integrates the input and output matching networks, the PA stages and power detection in a 4 x 4 x 0.85 mm package and is an advanced indium gallium phosphide (InGap) bipolar FET (BiFET) device capable of operating down to 3 volts. Similar size and power level modules exist for LTE Bands 13/14 (Verizon and Public Safety bands in the United States) – the SKY77449, and LTE Bands 12/17 (AT&T band in the United States) – the SKY77453, as well as the line-up of smaller size 3 x 3 mm single band devices, such as the SKY77706 Band VII module, which powers one of the first commercial LTE devices – the Samsung GT-B3710 USB modem. The equivalent output power for WCDMA for these devices is 28.5 dBm. This implies the PA is backed off for SCFDMA which is consistent with simulations that show 1 to 2 dB of additional linearity is needed for the LTE TX in user equipment over and above WCDMA. This is still less back off than would be required for a full OFDMA uplink which is why SCFDMA was chosen for LTE. Skyworks has also developed a line of LTE front-end modules, such as the SKY77455 (Band I), the SKY77456 (Band IV), the SKY77445 (Band VII) and the SKY77458 (Band VIII) supporting M7xx reference designs from ST-Ericsson. These are variations of Option 1 architecture, but are the more integrated devices and in addition to the above power amplifier module (PAM) content also include inter-stage SAW filters at the input and duplexers at the output of the PA in the 4 x 7 mm package footprint. These devices deliver 24.5 dBm of LTE or 25.5 dBm of WCDMA power output. To be financially and commercially viable from an operator perspective, options 2 through 6 would have to have a similar or better RF performance than option 1 and would have to deliver additional differentiation on the basis of cost and or component count reduction, real estate occupancy and power drain. Note that apparent efficiency gain in RF power amplifiers can be achieved by trading efficiency against modulation accuracy (error vector magnitude). This is not a real gain as the link budget will be the same or possibly worse. From both an individual and group perspective it is better to talk quietly and clearly rather than shout loudly and indistinctly. 25 RF Baseband interfaces

Mobile RF devices have an RFIC front end with a digital bus on the back end that connects to a baseband IC. There has been substantial discussion as to how, whether or when to standardise these control and communication interfaces between the RF and transceiver and baseband functions in the handset. One present initiative is the Dig RF work group which is supported within the MIPI alliance. Version 4 of this standard increases bus speed from 300 Mbps to 1.5 Gbps so that it can be compliant in terms of throughput, timing and latency control with LTE protocols. The functioning of these protocols requires robust testing, an area where Agilent presently have a dominant presence. Testing involves the use of active probes with ultra low capacitive loading (less than 0.15 pF) and high sensitivity (needed to work at 1.5Gbps and 200mVp_p). Although the required test investment is of the order of $50,000 dollars or more, the net benefits for handset manufacturers in terms of sub system interchangeability seem to be substantial. Handset vendor support of Dig RF is consistent with initiatives in other areas (for example the standardization of baseband RF and smart antenna interfaces in base stations) as a mechanism for achieving a broader base of multi source interchangeable components. Many of the new adaptive techniques (adaptive matching and adaptive linearization) require broader visibility to the physical layer than is presently available. This could include for example knowing whether peripheral devices such as Bluetooth are active. The lack of a standardized approach to what information is needed, what metrics should be measured, how they should be measured and how the measurements should be used is a significant barrier to the general
108

Information supplied as at August 2010

Geoff Varrall

Page 40 LTE UE efficiency and network value Final

15/09/2010

deployment of adaptive techniques. 109 A parallel and related issue is the multiplication of the control lines needed to support these adaptive modules both on the transmit and receive paths. To address this there is a new working group within MIPI called RF-FE, chaired by Nokia, working to replace the multitude of control lines in modern front-ends with a single serial bus. The first version of this specification is recently approved and released. 26 Conformance test costs

There are a number of features included in the LTE standards that are targeted at reducing conformance test costs both in terms of hours and dollar costs, automatic reset between test runs being one example. Despite these efforts, the hour and dollar cost conformance cost indications which we have received 110 for additional band costs are substantial even before development test and manufacturing test costs are included. An indicated 100 hours per band for RF conformance and protocol test implies a six month test programme. This does not include interoperability and network testing or MIMO testing (see earlier section 20). Aside from the direct costs, a six month cycle introduces substantial time to market delay and reduced range availability, an indirect cost. This is an area which definitely demands collaborative attention from handset vendors and the operator and test community. 27 Conformance and performance test standards opportunities

Within the HSPA standards process, 3GPP and interested vendors have identified and promoted opportunities for realizing throughput gains from enhanced levels of baseband processing. This has been captured in the 3GPP standards for advanced receivers. For LTE there is arguably a more compelling case to be made for enhanced RF performance to be captured as a design goal and for the benefits of performance gain over conformance compliance to be more explicitly stated, in effect a standard for an advanced front end. Note there is no reason why this should not be complementary to present and future baseband algorithmic innovation including for example interference cancellation. Conformance standards are set to be achievable with acceptable manufacturing yield at the time when user products first become available. In practice it should be possible to realize year on year user equipment performance gains from year one of a technology being introduced for at least five years though these gains may and often are traded against cost saving opportunities. Operators and standard bodies working together have the opportunity to influence these trade off decisions. The design and measurement challenges of LTE are addressed in detail in LTE and the Evolution to 4G Wireless edited by Moray Rumney and published by Agilent Technologies. 28 Software Defined Radio and baseband algorithmic innovation

The baseband processing of Release 99 3G technologies meant that the standard DSP architectures that had begun to dominate GSM and EDGE could no longer be used. A Release 99 handset needed about 1.5 million gates of dedicated hardware logic and associated memory to perform rake receiver, combining and channel decoding. The introduction of HSPA and new equalizer designs added another million gates. LTE introduces further complexity. The layer 1 multiple high throughput FFT/iFFT 111 operations, and FIR 112 and MIMO detection needed to decorrelate the OFDMA sub channels requires a hardware implementation of the LTE physical layer of the order of 3 to 4 million gates coupled to a megabyte of fast memory of a 50 Mbps class device. This has to be coupled with L2/L3 CPU requirements of the order of 150 to 200 MHz to support 100 Mbps operation. To be viable in terms of cost and power consumption, LTE and LTE-A devices will therefore need to use 32nm silicon geometries. A 65nm implementation costs over 40 million dollars, may not work first time and
109 110

Phone conversation with Karen Jackson Business development Director of Paratek Thursday 26th February 09 Discussions with AT4 Wireless March 2009 and prior cost estimates from RFI Global 111 Fast Fourier and inverse fast Fourier transforms. 112 Finite impulse response filtering

Geoff Varrall

Page 41 LTE UE efficiency and network value Final

15/09/2010

will be obsolete within less than eighteen months. A 32 nm device will cost more than twice as much, be harder to implement and be even more likely to not work first time round introducing additional cost and time to market delay. The use of dedicated ASICS therefore may be technically efficient but increasingly economically risky and dependent on achieving a stable standards environment, unlikely unless LTE advanced standards become finalized rather faster than present progress would suggest. As a result there are opportunities to optimize DSP architectures to support the specific requirements of LTE when implemented as a multi mode device with backwards compatibility to present 3 G and 2G layer one handsets. 113 The multi mode ‘problem’ is therefore essentially solved or at least almost solved. This highlights the need for parallel progress with a solution for the multi band ‘problem’ or more positively, multi band market opportunity. 29 Cost and performance comparisons

Three years ago (May 2007) the ‘common denominator’ dual mode GSM/UMTS handset was a quad band GSM (850/900/1800/1900) with UMTS Band I at 1900 and 2100 MHz Today’s equivalent common denominator is quad band GSM and anything between single and quintuple band UMTS/LTE, adding in US1900 Band II, UMTS 1800 Band III, Band V US850 and Band VIII UMTS900. In a study published in May 2007 114 we reported an industry consensus that the major RF cost in a handset was the amortized development cost. The RF BOM had remained remarkably stable over time at approximately 10% of the overall BOM cost of the handset. This might seem counter intuitive for higher end phones with expensive displays and application processors but in practice the additional RF functionality in these devices meant that RF costs tracked other costs in the phone remarkably closely. This implied a budget for RF components from anything between 3 and 7 dollars per phone. This had reduced from just over 12 dollars per phone for a tri band handset in 2003. Adding non standard bands had a direct cost consequence of the order of between one and two dollars but this was inconsequential when compared to the R and D amortization cost. The study also highlighted large differences in sensitivity across a number of commercial UMTS handsets from different vendors, of the order of 4 to 5 dB. The worst performing handsets only just met the conformance specification, the best were 4 to 5 dB better. Such differences are undesirable for voice service but very undesirable for mobile broadband where a 4 to 5 dB margin improvement over conformance specification should be regarded as a design requirement rather than a market ambition. Similarly the industry consensus was that talk times and stand by times with early UMTS handsets were unacceptable and were probably getting worse over time. Volume related performance gain and maturity effects have always been a feature in our industry. The table below shows the impact of rising market volume and maturity effects on the RF performance of GSM phones. Note that these performance gains were being delivered in the context of cost decreases of typically 15% per year.

Our thanks to Cognovo for providing these data points www.cognovo.com and to Icera Scheduler theory and practice is covered accessibly in ’3G Evolution HSPA and LTE for Mobile Broadband’ Dahlman, Parkvall, Skold, Beming, Associated Press. The mathematics behind the theoretical gains achievable with MIMO are covered very thoroughly in LTE for 4G Mobile Broadband by Farooq Khan Cambridge University Press 114 RTT study RF Cost economics for handsets for the GSM Association May 2007 downloadable from the resources section of the RTT web site www.rttonline.com

113

Geoff Varrall

Page 42 LTE UE efficiency and network value Final

15/09/2010

The table shows measured year on year gains in sensitivity in GSM handsets through the 1990’s. At network launch, GSM handsets barely met the conformance specification, five years later they were on average at least 5 dB better than specification. As volumes increased, tolerances on RF components were tightened without reducing yield. Note that aggressive cost engineering including a transition to direct conversion combined with smaller form factor phones and early iteration GPRS functionality then had the effect of reducing sensitivity towards the end of the decade. Given by that time most networks were being built to deliver additional capacity rather than for coverage and given that most applications were voice then this loss of sensitivity largely went unnoticed by users except at the edge of a cell. Edge of cell problems would show up as poor voice quality and high dropped call rates but improvements in handover algorithms largely hid this problem at the cost of some loss of capacity.
Static Sensivity dBm -110 -109 -108 -107 -106 -105 -104 -103 -102 -101 -100 X Volume Based Performance Gains* typically 1 dB per year Introduction of smaller form factor phones, dual band, tri-band, direct conversion. Sources: Ad Hoc measurements by RTT. Industry sources 1992 to 2000. Introduction of GPRS X X X X X X X X X X X X 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

The graphic below identifies another factor, in this case the difference between worst and best handsets comparing the sensitivity of four Release 9 WCDMA handsets.

Conducted tests undertaken by Spirent The difference between worst and best case performance should reduce over time as a technology matures and market volumes increase. If this is not the case, it suggests that insufficient performance engineering is being applied. In addition to an underlying performance difference, additional areas of performance uncertainty include band to band variation, channel to channel variation, capacitive effects (head, hand and hands) and polarization (depending on whether the product is used vertically by the head or horizontally on a flat surface).

Geoff Varrall

Page 43 LTE UE efficiency and network value Final

15/09/2010

It is of course natural for handset procurement teams to concentrate on cost optimization unless other factors are identified as having equal or at least relative importance. A shift of emphasis could be achieved if operators and handset vendors decided together that performance optimization was fiscally worthwhile. The performance of user equipment in real life is often worse than tested performance. This was highlighted in Section 22 but it is the scale of the possible difference that is unsettling. The Spirent diagram above shows a 6 dB difference between worst and best handsets but this is for measurements taken at the ‘temporary antenna connector’ (conducted tests) and thus assume an isotropic gain of 0 dBi. The measurements in section 22 of TRP (page 35) which included the effect of the antenna showed a spread of 7 dB. The conducted performance is more likely to lie within a 3 dB range suggesting the antenna variation is at least 4 dB. These results are based on GSM SISO dual band devices. If we jump forwards and consider the additional challenges of multi-band MIMO or SIMO devices, particularly in the lower bands like 700 MHz, it should be expected that the variation in antenna performance could be much wider. Just how wide remains to be determined as MIMO and SIMO radiated test methods are still under development. You cannot plan or efficiently dimension a network to meet target user experience quality metrics with this degree of performance uncertainty. Any relaxation of the conformance specification, which is presently being discussed in the context of bringing extended multi band products to market expeditiously would widen the difference with legacy performance still further. It could be argued that the network has visibility to user equipment performance. For example in a legacy network, CQI and throughout metrics including dropped call rates and blocked call rates are captured at the HLR. In an LTE network similar performance metrics can be captured on a device by device basis including packet loss, packet delay and packet retransmission attempts. However this only provides a picture of relative performance once devices are already out there on the network. The alternative is to do RF domain testing (anechoic chamber tests) and accommodate the measurement uncertainties in ‘’performance recommendations’, essentially as suggested in Section 20 though presently this seems rather complex and potentially costly and slow (introducing time to market delay). It would seem that conducted tests are becoming so remote from real life that their usefulness is decreasing over time, anechoic chamber tests are closer to real life but with substantial measurement certainty. Another possible approach is to have a set of benchmark applications for each category of device (lap top/slate/smart phone/standard phone) which are run in a series of benchmark channel conditions which would need to be simulated with comparisons made on the basis of task latency, speed, energy consumption and occupied bandwidth (a measure of delivery cost). More work would seem to be needed in this area and a combination of both approaches would seem to be the most robust approach. For operators not to have visibility to one of the most important parts of the end to end channel (the user equipment) would seem to be a very bad idea. One of the interesting suggestions made to us as part of the study input process was for a measured performance rating to be added to user equipment packaging giving “expected data rate performance”, 115 similar to a gas mileage rating or a rating from Consumer Reports . This puts the consumer in the driver’s seat to select devices with higher ratings that may cost more – a differential advantage important to those consumers who care about data rates, for instance smart phone users. However measured or marketed, the closely coupled relationship of RF performance and the user experience cannot be ignored. The recent iPhone4 launch highlighted that its not just poor performance but inconsistent performance that impacts perceived and actual user experience value. 116 This has not stopped people buying the iPhone4
115

A similar but different scheme has just been introduced to rate mobile phones in terms of their environmental footprint. http://www.forumforthefuture.org/node/13373 116 The variation in received signal level in the iPhone4 has been generally attributed to users’ hands grounding two parts of the multi part antenna embedded in the outer case of the device. Users will generally rate consistent but average performance higher than inconsistent performance with large throughput variations particularly if these are shown as a received signal strength indication.

Geoff Varrall

Page 44 LTE UE efficiency and network value Final

15/09/2010

and most of the time the user will not notice any particular inconsistency but the power drain will be higher, the application latency will be more variable and occasionally very poor, the traffic load and signalling load on the network will be higher and if some users churn or return products due to poor and or inconsistent connectivity then customer support costs will increase. These cost factors are hard to quantify and may be incurred many months after product shipment. This of course makes them all the more dangerous. The present fashion for ultra slim form factors often results in non optimum mechanical lay out and a restricted choice of components which further compounds the problem. In summary it can be said that there is very little incentive to produce user devices that do any better than meet conformance standards apart from operator comparative testing and work is still needed to make comparative testing faster, more accurate, less complex, and lower cost. However if achievable, a year on year improvement in LTE device performance over and above an agreed conformance and performance benchmark would have a significantly positive impact on the user 117 Or put negatively, a failure to deliver year on year performance improvements would result experience. in disappointing and inconsistent data rates at the edge of cell. High session drop rates in particular would be disruptive both to the user and to the network and definitely not good for IP voice quality which will need to be at least as good as existing voice service offerings. The particular concern for mobile broadband is that poor sensitivity and selectivity and stability (over temperature over time) translate directly into unacceptable retry rates (the stack effect). Send again signaling on the TX path will compound the RF and baseband related DC power drain of the device. The issue of signalling load is discussed in a number of Nokia Siemens Network White Papers The impact of mobile user device form factor on network loading is reasonably well understood. For example a dongle in a lap top typically presently generates about 2 GB a month of data traffic, a smart phone about 115 MB per month and an iPhone about 165 MB. 118 Presently this creates between two or three exabytes 119 of data per year 120 . Forecasts for 2015 range from 23 exabytes per year (NSN) to 87 exabytes per year (TMW) to 90 exabytes per year. 121 Interestingly and rather obviously when you think about it, data connection attempts are increasing at a faster rate than bandwidth volumes. In old fashioned network planning terminology the impact of this would be higher blocked call rates. In more modern terminology it is described as session set failure and session set up delay. Additionally smart phones absorb more signaling bandwidth than lap tops partly because they are more mobile and partly because the more mobile applications that they use demand to be updated more often. On present trends signaling volume will have increased by 40% between 2008 and 2012. Frustratingly smart phones absorb signaling bandwidth even when the applications are not being used Nokia Siemens mitigate this effect by using the 3GPP paging channel by putting user devices into power saving mode while maintaining a data connection. This is claimed to reduce the signalling load by a factor 122 of 3 to 5 times. However the case study serves to illustrate the close link between user device form factor, user data consumption, network loading and user experience metrics including failed session set up attempts and variable session throughput. Now it might be an old fashioned thing to say but irrespective of the mix of data and signaling traffic, the user experience is ultimately determined by the link budget. The link budget can be increased by increasing the number of base stations (network density) up to the point at which interference becomes the dominant constraint. This will increase coverage and capacity expressed as the number of users for a given data
One vendor pointed out that if 1 dB of performance gain was achievable on a year by year basis there would be a strong financial case for replacing all handsets every year though the logistics of this might be rather forbidding. 118 Operators are suggesting this may be nearer 300 to 500 Mb per month – source NSN August 2010 119 An ExaByte is equivalent to 1,000 PetaBytes or a million TeraBytes It is revealing to note that two vendor estimates for data volumes this calendar year range between two and thee exabytes – quite a large difference! This reflects the speed at which the voice to data transition is taking place. 121 See also the executive summary at the start of this study and the EBITDA model at the end – by definition we will not know actual numbers until after the event, it is the direction and magnitude of the trend that is important. 122 Information on Nokia Siemens Network solutions can be found by following the link http://www.nokiasiemensnetworks.com/news-events/in-the-news.
120 117

Geoff Varrall

Page 45 LTE UE efficiency and network value Final

15/09/2010

throughput. The point at which interference becomes the dominant constraint can be managed by various mechanisms including cell sectoring and baseband interference cancellation in the base station and user device. Scheduler efficiency will also be an influencing factor (addressed in more detail in Appendix 4). It is critical to remember that increasing the number of base stations is great for increasing coverage and network capacity but each base station must be connected back into the network. Backhaul, connecting base stations to the network can be a significant and expensive issue. 123 Site costs including acquisition, rental and energy cost also increase. Link budgets assume user devices meet but do not exceed the sensitivity and selectivity specifications stated in the conformance standard. The question to answer is whether an improvement in sensitivity and selectivity in user equipment over and above the conformance specification is technically possible or commercially worthwhile. Generally it is assumed that it is commercially and technically easier to improve a link budget by improving the performance of the base station. Commercially this is because there are fewer base stations, and the base stations last longer than user equipment. It is technically easier because a base station is less power and space constrained than a mobile or portable device. Adding a dollar of RF BOM cost to the billion user devices presently shipped per year costs a billion dollars which equates to a lot of base station hardware and software/algorithmic investment. Take for example the usual problem that in a rural cell, coverage will often be uplink limited. Mobile devices have a maximum allowed transmit power so run out of TX headroom. The only option is to increase the gain on the receive path of the base station. But we would argue that this is the exception that proves the rule. In most other conditions, improving selectivity and sensitivity in a user device at a range of cost premiums will deliver a net fiscal gain at network level. Even in the example above, improving transmit efficiency and transmission accuracy (error vector magnitude) in the user device will be an essential pre condition for delivering an acceptable user data duty cycle. Also it has been stated that all operators are faced with the problem that data rates and associated signaling load per user and the associated bandwidth cost including energy cost are not being matched by an equivalent increase in user revenue. Femtocells are offered as a partial solution to this problem but improving the RF performance of user equipment is arguably a technically and commercially efficient complementary option particularly for wider area coverage. (See earlier comments about the backhaul capital and operating costs associated with increasing wide area network density at micro cell and macro cell level). Let us test this thesis with some practical examples. We have stated that in the first five years following the introduction of GSM, sensitivity improved by about 1 dB per year over and above the conformance specification Users experienced a year on year improvement in coverage and capacity which was due to a combination of increasing network density, improved base station sensitivity and selectivity and a year on year improvement in user equipment RF performance. In 3G to date, five to seven years after initial roll out, network density had increased, base stations have improved but user equipment has not improved on a year on year basis. There is also significant variability between user devices (as shown by the Spirent measurements).This is partly due to the fact that there are few incentives for handset vendors to performance optimize user equipment to deliver devices that provide better than conformance performance. For a start the variety and often short life of user devices makes it uneconomic for operators to undertake performance testing. Operators are therefore reliant on vendor self certification and generally consider price to be more important than RF performance. This means that user device vendors achieve competitive advantage through cost rather than performance engineering. At component level there are similar but different disincentives. Replacing six RF power amplifiers with one broadband amplifier or six duplex modules with one tunable or switchable duplex filter bank is not attractive
123

Backhaul can account for up to 30% of capital and operational costs.

Geoff Varrall

Page 46 LTE UE efficiency and network value Final

15/09/2010

if the realized per unit price stays the same. As a result the industry has user devices that do not have enough band flexibility and do not improve in terms of RF performance on a year by year basis. As a consequence coverage in many countries does not match the footprint of older networks and the connectivity experience is at best inconsistent. Note that users will generally rate a consistent though relatively low data rate service as being better than an inconsistent service that occasionally delivers high uplink and downlink data rates. So at operator level we have a dual problem of delivery cost and delivery quality that needs to be resolved but what we need to prove is that improving user device RF performance defined as band flexibility, improved sensitivity and improved selectivity combined with baseband algorithmic innovation provides a technically and commercially efficient solution to the problem. There is also an issue that the industry needs to agree measurement methodologies particularly on LTE user equipment with MIMO antennas (see earlier section 22) There is at least a consensus on how SISO equipment comparisons can be made which provides a plausible basis for comparative testing to validate year on year performance improvements. At time of writing a similar approach to MIMO comparative measurement has yet to be agreed. 30 The role of band flexibility in reducing delivery cost – the performance gain from multi band macro scheduling

The cost problem at the radio physical layer is intriguing because it is an opportunity cost problem. It can be simply stated by taking the extreme but not statistically unusual example of a heavy data user close to the edge of a cell. It is up to the scheduler to decide how to service this user and the scheduler will know the quality of the allocated channel pair from the RX channel CQI values (including happy bits and sad bits 124 ) received from the user’s device and the TX CQI values measured at the base station. If the scheduler decides to meet the real time throughput demand the user device will be at maximum TX power and occupying the maximum allowable channel bandwidth (resource blocks as a composite of time slots and frequency sub channels). The single user is at this point absorbing a significant percentage of the available bandwidth and TX power in the cell and creating the maximum amount of interference into adjacent cells. All other users suffer as a consequence. This is a maximum pain, minimum gain equation. The scheduler has the option of reducing the quality of service to the edge of cell user but this just adds to the general feeling that the connectivity experience is at best inconsistent. However from an operator perspective, the opportunity cost of serving that single user greatly exceeds the revenue received. Even if the bandwidth is made available the user equipment battery will rapidly die of exhaustion. See appendix 4 for a more detailed analysis of these scheduler performance trade offs. It is of course quite possible that the user will be closer in RF terms to another base station owned by another operator working on a different band. If this is the case then it will be a significantly lower cost option to support that user from that other cell, in effect this is national roaming, already used in machine to machine to machine connectivity and in some rural coverage areas, 3G in rural Sweden being one example. Applying this principle to present 700 MHz roll outs it is arguable that consistent energy and cost efficient connectivity across an acceptable coverage area is more likely to be delivered if users can be handed over to other bands, the more band options and by implication operator options available the better in terms of connectivity efficiency and effectiveness. For the sake of simplicity let us refer to this as macro scheduling gain. This applies equally in all markets. In the UK in common with Europe the DSO band will be at 800 MHz rather than 700 MHz but there may be a potential to use the 700 MHz spectrum at certain places at certain times using unused interleaved TV channel ‘white space’ bandwidth. Providing LTE coverage at the 2012 London Olympics to US visitors with LTE 700 MHz lap tops would be one example opportunity.

124 ‘Happy bits’ are an indication that a UE has sufficient buffer capacity and connectivity bandwidth to support an ongoing application. ‘Sad bits’ indicate this is not the case.

Geoff Varrall

Page 47 LTE UE efficiency and network value Final

15/09/2010

However this requires LTE devices to have extended multi band functionality and for these devices to have an RF performance that is at least as good as single or existing multi band devices. This will only be achieved if faster progress is made with RF innovation – in an environment which will clearly adequately reward the innovator for taking the financial and technical risk to develop the innovation. Note that theoretically the biggest single improvement that could be made to improve the spectral and energy efficiency and investment return from existing and future spectral allocations would be to repurpose all present spectrum into four ‘super bands ‘ which would be individually mapped to cell topology as shown below. < 1 GHz Macro Rural and suburban and urban outdoor to indoor penetration, also for high mobility (up to 350kph) Low band <2.1 GHz Micro Suburban 2.1-2.6 GHz Pico and femto Suburban/ outdoor hotspots and indoor 3.6-4.2 GHz Tico cells Suburban/ outdoor hotspots and indoor

Mid band

High band

Super high band

Whether this could be practically achieved in terms of present competitive market structures must be open to doubt. 31 Band plan options 2011 to 2014

More likely band plan options in a 2011 to 2014 time frame can be stated as follows: It is reasonable to assume that the common denominator dual mode GSM/UMTS/LTE platform will be what we have today (quad band GSM and quintuple band UMTS/LTE covering Band I 1.9/2.1 GHz, US1900 Band II, UMTS 1800 Band III, Band V US850 (possibly extended by 10 MHz) and Band VIII UMTS900 but with the addition of Band VII at 2.6 GHz. There will be a need for some operators to cover the AWS band in the US (1.7/2.1 GHz), and a need to cover the US and European and Asian DDR bands at 700 and 800 MHz. This suggests that Option 2 in Table 6 (core module plus DDR module plus extension band module covering 2.3 and 2.6 GHz) may be the best basis for a common denominator which could leverage global market volumes to achieve an optimum cost point and still provide sufficient flexibility to meet operator specific needs. However Option 2 should have an RF performance (sensitivity, selectivity, stability and TX efficiency) across all bands that is at least equivalent to Option 1(separate RX/TX chains for each band), a cost base and real estate overhead that is substantially lower than Option 1 with an ambition to be the 125 lowest cost option. Note that it might be assumed that Option 6, the single transceiver approach for all bands would be the lowest cost partly due to a lower component cost and component count and partly due to the scale efficiency benefits accruing from a single transceiver approach. 32 Band plan summary

Operators require mobile broadband user devices (lap tops or notebook with embedded LTE modem or dongle or smart phones 126 ) that can provide ‘best connect’ access to existing frequency bands plus 700, 800 MHz and 2600 MHz. Other devices including standard phones with IP voice will need to be supported. Important metrics include cost, risk, power consumption, size and performance, receiver sensitivity, TX efficiency and consistency (band to band, channel to channel within band, centre cell to edge of cell performance).

If anything is to be traded off, perhaps relaxation in performance in the legacy GSM bands would be the best choice as UMTS becomes more widely available and offers cost saving in PA implementation. Although it is sometimes assumed that mobile broadband is mainly for larger form factor devices this in no way excludes the necessity to support smart phones and more basic standard phones. As stated earlier smart phones impose additional signalling load on the network which absorbs power and network bandwidth. Standard phones if using IP voice have additional processing overheads which subtract from the link budget and add to the power consumption of user equipment.
126

125

Geoff Varrall

Page 48 LTE UE efficiency and network value Final

15/09/2010

There are a range of views in the industry as to how future cellular phone and mobile broadband RF architectures will evolve. To an extent this is determined by how many power amplifiers on the TX path and low noise amplifiers on the RX path will be needed. At one extreme (Option 1) the approach would be to use separate RF power amplifiers and LNA’s for each band, using adaptive techniques, for example adaptive matching on the RX and TX chain and adaptive linearization techniques on the TX path to deliver close to theoretical best RF performance. At the other extreme (Option 6) there would be a single transceiver implemented to cover all bands from 700 MHz to 2.6 GHz. This could provide the basis for a cost optimized and space optimized solution which could leverage global market volumes to realize substantial scale efficiencies. Option 2, with a core module addressing bands between 850 MHz and 2.1GHz, a second module covering the 700/800 MHz digital dividend bands and a third module covering the 2.3 and 2.6 GHz bands would appear to be a favored option at least for initial implementation. Incorporating the DSO 127 bands into a single module would be the next step. With any luck this will coincide with a final resolution of the European and Asian DSO band plan. Integration of the 2.3 and 2.6 GHz bands into a single 850 to 2.6 GHz module would be a further step presently further out than the 2015 time frame covered by this study. This allows RF component vendors to service operator specific needs at 2.3 or 2.6 GHz. Such a module could for example be extended to cover the AWS band in the US. A 2.3 and 2.6 GHz specific module would also allow for an optimized implementation of TDD LTE. This would deliver user experience benefits into a major global market (China). China would also provide sufficient market volume to amortize the R and D costs of TDD which in turn would improve the economics of ROW 2.6 GHZ TDD LTE deployment. The SKY77441 Skyworks module referenced earlier is an early example of this approach. For all bands it can be stated that cost or space saving at the expense of RF performance needs to be avoided particularly because RF performance will be very directly coupled to the DC power drain of the device. However there will be a continuing need to deliver year on year RF component cost reductions to support future market growth. In practice this may mean the RF BOM remains fairly constant as any year on year cost reductions tend to become absorbed by the need to provide additional RF functionality either in terms of additional band support and/or due to the need to support more aggressive peak and average data rate requirements. This performance has to be delivered on the TX path without compromising adjacent channel performance. On the receive path, there may be significant unwanted signal energy that needs to be managed both in terms of front end dynamic range and management/cancellation of time correlated interference. The use of higher order modulation (16 QAM on the TX path and 64 QAM on the RX path) places further demands on TX/RX transceiver performance. It is over simplistic to assume that baseband processes can compensate for poor RF performance. Improving RF performance has a beneficial multiplier effect both in terms of baseband coding and processing gain and in terms of stack gain (referenced earlier). This provides a convincing rationale for the belief that RF performance of LTE handsets should be better than, rather than equivalent to present HSPA devices. This needs to be delivered with a measure of band flexibility that suits operator specific requirements. To achieve this without adding significant component manufacturing cost or complexity may imply a need to implement new technologies and techniques faster than present R and D resource constraints allow. In the RF domain the techniques that allow us to implement adaptive processes, specifically adaptive matching and adaptive linearization are already beginning to be implemented. Presently these techniques offer band specific performance gain and have not resulted to date in any step function improvements in terms of cost or real estate or performance in multi band applications. The next step is to scale these
127

Digital Switch Over – the vacated TV bands at 700 or 800 MHz

Geoff Varrall

Page 49 LTE UE efficiency and network value Final

15/09/2010

processes so that band flexibility can be realized without loss of performance and without a cost or real estate premium. We suggest there is an emerging industry consensus as to how this should be achieved over the next three to five years. The solution is not dependent on any particular innovation either in terms of materials or architecture but is rather a combination of existing devices which over time continue to improve in terms of performance and cost (SAW and FBAR filters being examples) and new devices and materials that offer flexibility combined with cost and performance advantage. In this context it is encouraging to see that RF MEMS are finally becoming sufficiently reliable, robust and low cost to become an accepted part of the RF BOM, initially in adaptive matching (variable capacitance) applications and in switch path optimization. Similarly other process innovations such as silicon on sapphire are already delivering performance gain in 3G transceiver front ends. Within three to five years it is reasonable to assume that these devices will gain a wider application footprint and any measures taken to speed this adoption would be likely to yield a positive return to all parts of the industry value chain. Oscillator applications and channel bandwidth tunable filters are an obvious though non trivial next step. These will be needed in a range of devices other than cellular phones, for example in next generation portable and pocketable DVB T receivers. These parallel but interrelated requirements will broaden market demand for these devices. In parallel, baseband interference cancellation techniques promise step function improvements in receive selectivity that could potentially be translated across into additional user value. RF performance can no longer be viewed in isolation and must be seen effectively as a closely integrated baseband ‘channel partner’. Linearization and active matching techniques are two examples of the developing interdependency between RF and baseband functionality. However these technologies and techniques will be unacceptably slow to market unless sufficient commercial incentives can be introduced to justify significantly higher levels of R and D investment. 33 Pre requisites for industry profitability

User equipment RF and baseband efficiency and band flexibility together have the potential to transform the economics of the mobile broadband industry supply chain. Users are motivated to buy user devices that deliver a better experience in terms of average throughput rates and data duty cycle which makes applications run faster while using less network bandwidth and power. Voice quality and coverage are also improved. At operator level, satisfaction rates increase which means return rates and churn rates reduce and tariff premiums are more easily achieved. Using less network bandwidth and network power means the megabyte per user bandwidth cost goes down. This improves operator EBITDA performance. Band flexibility allows inter band inter operator macro scheduling to be deployed thereby transforming mobile broadband delivery cost and user value economics. This increases as additional bandwidth becomes available provided this can be accommodated in user equipment front ends without a loss of performance. The ability to ship standard user products to all markets drives down inventory costs. Improving the RF efficiency in user devices therefore benefits the whole industry supply chain. RF efficiency has of course to be closely coupled with baseband efficiency. There is substantial opportunity to improve user device performance through algorithmic innovation. Interference cancellation on the receive path and phase and error vector correction on the TX path are two examples. The LTE transition has the potential to deliver a step function gain in network efficiency and user value provided user devices can deliver steady year on year performance improvement – RF components and RF innovation closely coupled with baseband algorithmic innovation have an important part to play in the success of that transition. Between 1992 and 1997 the sensitivity and selectivity of GSM user devices improved by at least one dB per year over and above the conformance standard. This, combined with increasing network density and improvements in base station performance delivered year on year improvements in the user experience in terms of voice quality, coverage, consistency (lower blocked call/dropped call rates) and duty cycle (minutes of use between recharge). This created a virtuous

Geoff Varrall

Page 50 LTE UE efficiency and network value Final

15/09/2010

cycle with delivery value per subscriber increasing over time and delivery cost per subscriber decreasing over time. As the industry has matured cost pressures have meant that user devices have been cost engineered to meet rather than exceed conformance standards. Products are conformance tested but generally not performance tested at vendor or operator level. The economic modelling in this study suggests that performance engineering LTE user devices to be band flexible and deliver a 1 dB gain per year over and above the RX and TX conformance specification would provide a substantial net gain to all parts of the mobile broadband industry supply chain. Specifically it would improve network efficiency more significantly than the link budget gain on its own would suggest. This is a function of the more aggressive scheduling used at the LTE physical layer combined with the higher multiplexing gain in the upper layers of the protocol stack (the LTE stack effect). The performance gain in the user device can be achieved by a combination of RF component and architectural innovation closely coupled with baseband algorithmic innovation. In parallel this implies a shift from user device conformance compliance testing to user device comparative performance testing at vendor or operator level. Operators range plan on the basis of aesthetic and functional value. Functional value is increasingly determined by the applications supported by or downloaded into the device. Application performance in terms of speed and consistency is substantially influenced by the RF performance of the device. This would suggest that RF performance should be a more heavily weighted factor in product and vendor choice decisions and that application and task responsiveness and an extended duty cycle will be an essential pre condition for meeting future mobile broadband user expectations. Some recent product introductions have highlighted that compromising RF performance in order to improve aesthetic appeal can be problematic for devices that aspire to deliver mobile broadband connectivity. The positive coupling between RF performance and baseband processing gain is also generally under appreciated. RF performance is at least in part a consequence of market volume. As volume increases, RF component tolerances can be tightened. The vendor with the largest market volume should therefore be in the strongest position to cost optimise and performance optimise user devices. It is a legitimate operator expectation that this should be reflected in terms of practical product performance. In some markets averaged per subscriber mobile broadband delivery cost is increasing faster than averaged per subscriber delivery value. If this is not the case in some markets it may be because operators are not investing in new spectrum and or infrastructure at a rate sufficient to ensure that future user experience expectations are fully met. This would mean that in these markets product return and churn rates are likely to increase Based on the model and assumptions described at the end of this study, a one dB gain per year in LTE user device TX and RX RF and baseband performance over and above the conformance standard combined with a year on year increase in RF band flexibility means that averaged per subscriber delivery value increases faster than averaged per subscriber delivery cost providing a sustainable and even reasonably compelling basis for future LTE spectral and network investment. Ends Appendix 1 The impact of user device RF performance on the user experience and user and network value

Sometimes we need to think about where we need to be rather than where we are at the moment. At present we are at the point where LTE networks are going live both in existing and new spectrum. The Verizon launch in November in the 700 MHz band in the US is an example of new spectrum deployment. However in order to achieve economies of scale and reduce inventory management costs it will also be necessary to support LTE in the 800 MHz band and 2.6 GHz band and within a three to five year time scale it is reasonable to assume that there will be a need to support LTE channels in the 850, 900, 1800, 1900 and 1900/2100 MHz bands.

Geoff Varrall

Page 51 LTE UE efficiency and network value Final

15/09/2010

This is an 8 band phone but by then there will be a need to support other bands as well. Operators need to justify their investment in LTE networks and the spectrum associated with those networks. Partly this can be done on the basis of improved spectral efficiency at the physical layer expressed in bits per hertz and improved multiplexing efficiency though the network. This should improve the ROI on network and spectral investment. This in theory at least can then be consolidated by a reduction in operating costs, including of course network energy cost. However experience tells us that any successful technology transition has to combine cost reduction with an improved user experience. As stated earlier, in GSM through the 1990’s, engineering investment and market volume drove down user device costs but also yielded a 1dB per gain in sensitivity per year over and above the conformance specification and steady improvements in TX efficiency. This combined with efficient digital speech encoding and more efficient signalling meant that phones could work for days rather than minutes or hours. A parallel reduction in weight and size meant that GSM phones quickly over took analogue phones in terms of user acceptance and user value. In an ideal world the same virtuous circle would have applied from 2000 to 2010 as 3G phones were introduced. However the transition from analogue to digital through the previous decade was in some ways a single step function and a similar performance leap was always going to be harder to achieve. In practice much has been achieved over the past ten years particularly in terms of peak data throughput and this has translated into the present mobile broadband market opportunity. The problem with mobile broadband is that as data rates have increased, revenue per bit has decreased. This would have been fine if dollar costs per bit had done the same but in practice delivery costs have reduced more slowly. There are two answers to this. One is to decrease delivery cost per bit and much effort is being expended to achieve this. The other answer is to increase revenue per bit or at least slow the decrease in revenue per bit to the point where it is slower than the decrease in cost per bit. However this can only be achieved if an improvement can be made to the user experience which is sufficiently substantial to justify a tariff premium. The two areas where the user experience could be tangibly improved are the user data duty cycle and the time taken for mobile applications to update (application and task latency). If the data duty cycle is improved by increasing average data throughput then mobile application speed will improve in parallel. Consider four different user device form factors, a basic GSM voice and text phone, an LTE smart phone, an LTE tablet or slate and an LTE lap top. The user expectation is that all four devices should have a similar duty cycle. The smart phone, the tablet, slate and lap top devices all support broadband connectivity 128 The voice quality and coverage expectation will be similar for all which of course may include IP voice. four devices. Let us say for sake of comparison the power envelope for the basic phone is about two watts, about 5 watts for the smart phone, ten watts for the larger than smart phone device and twenty watts for the lap top. The lap top will theoretically then have a battery capacity that is sufficient in terms of watt hours to support an acceptable user duty cycle which as we all know is already far too short. Wireless connectivity makes this shorter. There are some potential innovations in display technology and processor technology which could reduce the overall power consumption but this would mean that the RF power budget became a bigger rather than smaller part of the problem. As stated earlier the RF power ‘connectivity’ budget is already far more important than it was. The reasons for this from an application perspective are always on connectivity and applications that demand constant updating but other factors such as the need to support large cells at 700/800 and 900 MHz are also important. At the physical layer receive sensitivity determines the downlink data rate. As RX sensitivity increases coding overheads decrease and retry rates reduce. At IP level this reduces the IP acknowledgement/send again overhead which reduces the TX load on the device. Similarly improvements in TX efficiency achieved through lower insertion loss will translate directly into a lower RF power budget.
128

Note that IP voice in LTE will have packet header compression though this incurs processing and memory overhead and processing delay. Also as stated earlier extracting a 12 kbps voice stream from a wideband 20 MHz channel is computationally expensive and presently less energy efficient than GSM voice. This means that LTE voice services will absorb more rather than less network and power for the same volume of voice minutes, an additional reason to improve LTE user equipment sensitivity.

Geoff Varrall

Page 52 LTE UE efficiency and network value Final

15/09/2010

We stated that over the past ten years peak and average data rates have increased. However this has been achieved by improving multiplexing efficiency over wider channel bandwidths, initially 5 MHz. This has improved bandwidth utilization but has not improved the power efficiency of the user device. The power efficiency of user devices would have improved if a year on year improvement in RX sensitivity and TX efficiency had occurred. This helped drive GSM adoption a decade before. This has not happened for a combination of reasons. Most of the effort to reduce power consumption went in to processor optimisation, adaptive voltage scaling and low leakage memory being two examples. Better than conformance sensitivity was not seen as a prime design goal as network density had increasingly become a function of capacity gain which meant that more than adequate downlink signal level was available at least in urban networks. Also the assumption has been and often is today that the link budget is interference limited which means that additional sensitivity will not deliver performance gain. If the LNA is close to compression it might make things worse. However additional selectivity will help reduce unwanted out of channel energy and baseband interference cancellation can potentially mitigate the effect of unwanted on channel signal energy. The RF front end and baseband are not competing with each other; they are complementary to each other. It is sometimes assumed that each generation of cellular radio air interface has allowed a relaxation of front end performance. The assumption is partly based on moving from the 25 or 30 KHz channel bandwidths used in analogue cellular to the 200 KHz bandwidths used in GSM to the 5 MHz bandwidths used in 3G today and 20 MHz bandwidths now planned. In parallel however the need to deliver spectral efficiency gains has resulted in the adoption of higher order modulation schemes that are inherently more noise sensitive. These require greater accuracy (linearity) on the TX and RX path of the user device. Although it is true to say that substantial signal pre and post conditioning is achieved at baseband it is also true to say that it is the combination of improved front end selectivity and sensitivity and baseband processing and algorithmic innovation that delivers performance gain in a wide range of operational 129 conditions. In addition as stated the year on year progress that has been made with RF components in terms of lower insertion loss and or improved selectivity have been more than off set by the need to support additional bands, initially the 1900/2100 GHz to quad band phones but now with the addition of 700, 800 and 2600 MHz band support. In practice the demands on front end RF performance have thus increased rather than decreased over time. Appendix 2 LTE user equipment power drain comparisons with Wi FI connectivity

The table 130 shows how power consumption is distributed in a lap top between the computer functions and Wi Fi connectivity functions, in this case, 802.11g.
Lap Top
WLAN Card WLAN Host Total 70% 9% 21% 100%

Note this includes the RF power budget and directly related IP uplink and downlink packet processing overheads. The following table shows the RF power budget
TX RX Listen Sleep
129

2 watts .9 watts .8 watts 40 milliwatts

It could be argued that 700 and 800 MHz networks are more likely to be noise limited and therefore will benefit particularly from improved sensitivity. 130 With thanks to Atheros

Geoff Varrall

Page 53 LTE UE efficiency and network value Final

15/09/2010

The RF power consumed including TX and RX efficiency loss is a function of the duty cycle Here the assumption is as follows TCP Uplink power consumption Duty cycle Watts Total

.6 x TX + 2x listen+ .2x receive x.2 x sleep .6 x2 +.2x.8 + .2x .9 x.2 x.04 1.54 watts

It might be considered that 1.54 watts is inconsequential but in practice there is a further 2.85 watts of associated CPU power totalling 4.389 watts. 4 watts represents a 20 to 30% reduction in battery duty cycle, four hours is reduced to three hours. Note that the average power envelope is 20 watts for a lap top, 10 to 15 watts for a larger than smart phone device, 5 to 10 watts for a smart phone and sub 5 watts for a basic phone. The RF power budget (including related processor overheads) of a smart phone is likely to be higher than 30% due to the additional signaling load incurred by mobile applications. The final table shows typical Wi FI throughput against received signal level. The table also shows an Atheros implementation where higher throughputs have been achieved at these levels, the difference being the improved sensitivity in the Atheros front end. Received signal level dBm Industry norm Data rate achieved Atheros -85 -88 -89 -92 -93 -101 -105

11 Mbps

5.5 Mbps

2 Mbps

1 Mbps

6 Mbps

1 Mbps

250 kbps

There are several points to make here. Theoretically, in a noise limited channel a 3 dB increase in power or the equivalent increase in sensitivity translates into the error rate dropping a thousand fold. However as can be seen in the industry standard example, between -88 and -89 dBM the packet error rate triggers a change of modulation scheme which reduces the throughput from 5.5.Mbps to 2 Mbps. As with LTE, an increase in sensitivity in the user device will delay the point at which additional error coding and or a lower order modulation is needed. But one of the areas of particular interest for cellular operators is how LTE compares with WiFi both in terms of peak data rate and the throughput power budget for local connectivity (a few meters from the base station). On a like for like basis assuming similar TX power levels in a similar channel (20 MHz at 2.4 GHz or 20 MHz at 2.6 GHz) and a similar distance from the base station then the peak data rate and average data throughputs should be similar. LTE replaces the OFDM used on the WiFi link with an OFDM related modulation and multiplexing scheme described as Single Carrier Frequency Division Multiple Access (SC-FDMA). This reduces the envelope variation on the modulated waveform which at least in theory should allow the TX path to be marginally more efficient. However it can be said that this “improvement” is only from an onerous 12 dB or 16x to an only slightly less onerous 10 dB or 10x. The relative performance will also depend on whether LTE is used in TDD or FDD mode and what band is used for local connectivity but the assumption would be 2.6 GHz versus 2.4 GHz or 5 GHz for the Wi Fi connection. An alternative is to use LTE TDD devices that have been developed for the China Market in Band 40 However if operators are aspiring to use LTE as an alternative to Wi Fi for example in femtocells then any noticeable performance advantage will probably need to be realised by improving sensitivity and selectivity in the user device. From a user experience perspective the throughput at 2.4 or 2.6GHz will be very closely determined by distance as shown below. It can be seen that in these small cell environments user equipment sensitivity very directly determines the user experience in terms of distance from in this case the WiFi access point.

Geoff Varrall

Page 54 LTE UE efficiency and network value Final

15/09/2010

Path loss dB

10 m 86

20m 100

30m 107

40m 113

50 117

60 121

70 124

Source Atheros os Note that in many technical papers written both about WiFi and wide area mobile broadband, an assumption is made that high peak data rates equate to improved energy efficiency. This may be the case sometimes but not always and is dependent on how well the power done/power up algorithms are implemented in the user equipment (a function of software latency), hardware memory effects, for example in RF power amplifiers, the efficiency of the fast memory needed to cache short data bursts and the bandwidth and characteristics of the channel over which the data is transferred. The bandwidth characteristics of the channel will also determine the amount of gain realisable from MIMO operation. Wider channel spacing, for example the 20 MHz used in WiFi and small cell LTE deployments deliver high peak rates but the noise floor proportionately increases with bandwidth. In a large noise limited cell it is quite possible that the average data throughput with narrower channel spacing, for example 3 or 5 MHz could be significantly higher than a 20 MHz channel used in the same propagation conditions and the megabyte per watt hour metric may be lower. For certain the gain achievable from MIMO in larger cells will be marginal at best and may be negative in many channel conditions due to long multi path delay and or low signal to noise ratio. It is laudable and understandable that operators should expect and want LTE to be more effective and efficient than all other forms of connectivity in all operational conditions but whether this is an achievable or fiscally sensible short term ambition must be open to question. To get LTE to work as well or preferably better than Wi Fi for local connectivity implies a relatively aggressive adoption of MIMO techniques in LTE user equipment. The question then to answer is how often users will be close enough to an LTE base station to realize a real benefit from the composite MIMO LTE channel and how many of those users would actually be better served from a WiFi channel. The answer is probably not a lot or at least not enough to warrant the additional cost and complexity and weight and size and power drain added to some or most LTE devices. For certain it would seem that MIMO will deliver very marginal gains in larger diameter cells, for example anything larger than a pico cell or femto cell and as stated above may have a negative impact on throughput in many wide area access channel conditions. In terms of band allocations this would suggest MIMO might be useful at 2.6 GHz and above but less useful in lower bands and more or less pointless at or below 1 GHz. This is just as well given the problems of getting enough volume and distance to achieve effective spatial antenna separation in low band device designs. Additionally there are proposals in LTE Advanced to bond channels. This could either be adjacent channel bonding or bonding two channels from different bands which would imply a need for yet another RX/TX processing path. The rather pointless marketing pursuit of headline data rates is therefore resulting in the industry investing R and D dollars which from an end user experience perspective would be better being spent at least initially on optimising SISO performance within the existing band plans. In other words present R and D priorities are probably disproportionate to the proportionate user experience gain that can be achieved. Inefficiently focused R and D spending has an associated opportunity cost (indirect cost). The direct and indirect costs of MIMO investment (and or work on channel bonding) have to be recovered from all LTE devices. The Section 22 SISO and MIMO sections and related Agilent references provide a more thorough analysis of this topic. In the longer term, Wi Fi semiconductor vendors including Atheros are working on tri band solutions that combine 2.4 GHz, 5 GHz and 60 GHz connectivity suggesting that this may be the best option for delivering high and ultra high data rates in local area network environments. Note that some parts of the industry still regard WiFi as being competitive rather than complementary to cellular connectivity. The development of recent applications where mobile broadband is used to provide connectivity to multiple WiFi devices and hybrid WiFi and mobile broadband integration in high end automotive applications suggest this misconception will lessen over time.

Geoff Varrall

Page 55 LTE UE efficiency and network value Final

15/09/2010

In car automotive applications are a form of machine to machine communication, albeit highly mobile. On a historic basis, automotive two way area connectivity has been relatively narrow band and often predominantly down link biased, GPS based navigation being one example. Adding Wi Fi to the in car environment could change this. If mobile broadband devices are integrated at the manufacturing stage then they can be reasonably easily performance optimised with efficient external antennas. They are also connected to a generously dimensioned 12 volt supply. Devices introduced at a later stage may be much less efficient. A present example would be people using hand held smart phones to do in car navigation. These devices can also be connected to a 12 volt supply typically via the cigarette lighter (who uses these any more for the purpose for which they were intended?). This solves the duty cycle problem but the loading on the network can be substantial, a composite of signalling load to handle high mobility and the penetration loss into the car, particularly severe with tinted windows (of the order of 6 to 10 dB). Applications where maps are progressively downloaded into the device will be particularly bandwidth hungry. This is an example where a particular LTE device used in a particular LTE application, which may not be the originally intended application, has the potential to inflict serious collateral damage on the network. The damage (opportunity cost to other users) will be disproportionate to the realisable subscriber and application value. If the device additionally has poor sensitivity and selectivity the collateral damage will be greater. Appendix 3 The impact of user equipment performance on spectral requirements and future band plans

There is an ongoing debate as to how much spectrum the industry will need over the next ten to twenty years and whether future spectrum could be more efficient if allocated as contiguous bandwidth To take a simple example, if as per the Obama administration consultation, an additional 500 MHz of bandwidth was to be made available for allocation and auction, would it be better if that spectrum was made available as one contiguous allocation, for example as a 500 MHz ‘chunk’ in C band between 3.6 and 4.2 MHz, or ‘tacked on’ to the edge of existing bands or popped in at random anywhere and everywhere. Or more fundamentally could it be argued that operators are having difficulty absorbing the bandwidth already allocated to them. If this is the case what are the constraints that need to be addressed before it becomes fiscally sensible to bid for additional bandwidth. The traditional assumption used to be that additional bandwidth would provide capacity gain which would improve investment return for each base station and site but this assumes both the base stations and user equipment can be easily upgraded to support the additional band plans. It is intrinsically harder to add new frequency bands into user equipment because the devices are cost constrained, size constrained and power constrained. These constraints have been closely examined in earlier sections of this study and are substantially determined by the component cost and performance cost implications of adding bands or extending bands in compactly dimensioned mobile broadband user equipment. The data traffic forecasts suggest that more bandwidth is needed but if this bandwidth cannot be accessed cost effectively then something needs to change. The technology answer is to have a flexible RF front end that can access any band anywhere as efficiently as a single band phone. Industry estimates as to when this is achievable range from two to twenty years to never and the truth as always is somewhere in between. The position of this study is that present operator expectations that LTE user equipment supporting ten or eleven bands (which we describe as extended multi band user devices) can be delivered within twelve months at an acceptable cost delivering an acceptable performance are likely to be unrealistic for a range of technical and commercial reasons that we hope we have adequately explained. However if we were to be proved wrong, the benefits to the industry would be substantial. The risk is that an over aggressive implementation of such devices, particularly if associated with deliberately relaxed conformance standards, would be likely to result in a compromised and inconsistent user experience. While this might not be obvious initially to a user in a lightly loaded network and or in a dense network with small cells it would quickly become a constraint as network loading increased.

Geoff Varrall

Page 56 LTE UE efficiency and network value Final

15/09/2010

The cost of supporting these devices in terms of radio access overheads would be substantial and certain functionalities, for example IP voice, would be particularly hard to support.(See earlier footnotes about IP voice processing overheads). The rock and hard place scenario is that every year a user will expect better performance than the year before but in the meantime the noise floor of the network will have increased to the point where the deafness of the user’s equipment becomes painfully apparent to the user. Note that if these are lap tops or tablets or slates with embedded connectivity the user will probably be expecting to own that device for at least three years by which time the data throughput and or data duty cycle will at best be halved. You might argue that this is an upgrade motivation but users trading up due to a decreasing QOE will most likely blame the network and not the device and will churn to another provider. From a network perspective these legacy devices will be absorbing unnecessary network bandwidth and power and at some point this begins to compromise the overall QOS and QOE metrics of the network. This translates into additional opportunity cost. Our alternative model is that operators should plan on the basis of setting an agreed performance target for SISO and MIMO LTE equipment as at November 2010 (Verizon market launch) and then plan to add an additional band per year and deliver an extra dB per year over and above the agreed conformance specification. For the sake of simulation we have factored in a dollar increase in the user equipment BOM every year to help recover the R and D needed to achieve this. By year three this will mean that user equipment will be 3 dB better which will have at least doubled the 131 This gain is arguably understated as average data rate per user and or doubled the data duty cycle. each dB of user equipment link budget gain will have translated into a non linear increase in network scheduling efficiency. The flaw in this argument is that users will still need to buy a new device in order to enjoy better performance. This is true but at least the newer devices are absorbing less network bandwidth and power (see point above) which will mean the networks have a better chance of delivering at least an acceptable QOE to the earlier generation devices and other users. This points out the need for a rather more fundamental change. We have said earlier that within three years it will be possible to blow over the air (OTA) software upgrades to user equipment so that operators can offer customers the chance to upgrade to faster access options in the same way that ADSL modems get remotely up graded . It would be nice but not possible to do the same for additional bands. This is because the RF front end will still be intrinsically inflexible. User equipment by then may have switchable eight band duplex filter band modules but this is not the same as having tuneable filter modules that can access any band any where between 700 and 4200 MHz. However if tuneable filter solutions could be developed within say a five year time span then this would mean that operators could access newly allocated and auctioned bandwidth flexibly and cost efficiently. Additionally if RF architectures could be realised that allow dual receiver functionality then user equipment could measure channel quality across all available bands providing the basis for best connect algorithms which would transform the user experience in terms of application speed, latency and duty cycle and dramatically reduce per bit delivery cost for the operator community. In particular it will have minimised the opportunity cost of supporting high use edge of cell subscribers assuming that they can be moved to a closer base station possibly owned by another operator and probably on another band. This of course will mean operators will have needed to put national roaming in place and any competitive barriers to this will have had to be removed. By this stage (2015) LTE user equipment will be 5 dB better than the conformance specification which will mean it will become increasingly harder to achieve year on year performance gains (approaching the fundamental noise thresholds of the device). But fortuitously macro scheduling gain then provides the basis for year on year improvements in user average data throughput and data duty cycles. Let us describe this in link budget terms as maintaining the
131

Read on to see micro scheduling and macro scheduling gain explanation for this.

Geoff Varrall

Page 57 LTE UE efficiency and network value Final

15/09/2010

year on year 1 dB performance improvement benchmark. Even better, every time an operator adds new spectrum the macro scheduling gain increases and so does the operators’ EBITDA. At this point it may well be that mobile broadband is divided into low band <1 GHz for macro cells, mid band <2.1 GHz for micro cells, high band 2.3/2.4/ 2.7 GHz for pico cells and femto cells and super high C band 3.6 to 4.2 GHz for LTE Advanced tico cells. The cost of course is that 5 dollars has been added to the use equipment BOM but at least this will have provided the necessary incentive for the RF component and baseband industry to have delivered the basic preconditions of LTE economic success, year on year improvements in user equipment performance, a new band each year from now to 2015 and band flexibility from 2015 132 , providing the basis for sustainable spectral and network investment across the assumed fifteen to twenty year life cycle of LTE and LTE A (advanced) mobile broadband networks. In effect we have described a user equipment technology model that over ten years delivers ten dB of link budget gain and an associated non linear improvement in scheduling efficiency with those improvements being delivered as micro scheduling gain over the first five years and macro scheduling gain over the second five years. The macro scheduling gain would continue to increase every year thereafter with each additional band. The same result could be achieved by increasing network density at least up until an interference threshold was reached but it would be unlikely that such a gain could be realised just from improvements in base station performance. This would mean that at least some of the improvement would need to come from increased network density which would incur additional capital and operational cost. Precisely described, any loss of user equipment connectivity efficiency (sensitivity and selectivity on the RX path, accuracy and selectivity on the TX path) creates demand for additional spectral and network investment with associated cost but no associated income. The fiscal case for improving user equipment RF and baseband performance both over the short term (three to five to ten years) and long term (ten to twenty years) is therefore compelling. A note about the difference between quality of service and quality of experience LTE network hardware and software vendors and their operator customers are rightly proud of how end to end latency improves (reduces) in LTE networks. This includes first order latency (the average latency) and second order latency delay (the variance between minimum and maximum delay) the combination of which determine user quality of service. The reduced latency is generally attributed to the collapse of the seven layer OSI model to a three layer IP stack. But of course this only applies to those parts of the end to end link to which the network has visibility which does not include the journey between the user equipment antenna or antennae and the back end of the application processor. This part of the journey can introduce significant first order and second order latency particularly if burst error rates trigger buffer overflow on the TX or RX path. As you would expect we would like to point out than improving RF front end performance will greatly improve the latency performance of the user’s device and therefore improve the end to end latency performance of the network as defined from a user experience perspective. For these reasons round trip packet delay cannot be correlated directly to application and task latency as seen by the subscriber. From the point of view of defining quality of service and quality of experience we would therefore argue that quality of service refers to the parts of the end to end channel that the network can see and control. Quality of experience takes you to the back of the applications processor and is a more meaningful metric to which the operator community needs to have visibility. Summary Operators need to calculate how much spectrum will be needed over the next ten years and the type of spectrum needed, contiguous or non contiguous, low band or high band, regulated or deregulated. Most industry demand side models show an exponential growth in data traffic which can be assumed to translate into spectral demand.

Note some informed industry observers consider this time scale to be substantially over ambitious and suggest twenty years as a more feasible horizon taking into account present investment incentives. Our point is that if sufficient innovation incentives could be created, the time scale would reduce.

132

Geoff Varrall

Page 58 LTE UE efficiency and network value Final

15/09/2010

The amount and type of spectrum needed is however influenced by the choice and mix of technologies used to access the spectrum. This in turn influences the network density needed to meet or exceed user service and experience expectations. In a legacy circuit and packet switched radio network user expectations are determined by voice quality and coverage, blocked call and dropped call rates and to some extent text latency. Extended talk time (several hours) and stand by time (several days) is now assumed. In a mobile broadband network these same expectations apply but there is an additional expectation that large files can be exchanged with minimal latency and that a large number of simultaneous applications can be supported. These applications inflict additional signalling load on the network and drain energy from the user’s device, shortening session time and stand by time to a point where mobile applications (where users by definition cannot be connected to a main power socket) become unacceptably short. Wide area access technologies are traditionally benchmarked in terms of spectral efficiency measured in bits per Hz. Increasing spectral efficiency means more subscribers can be supported per Hz of allocated spectrum for a given network density. There is however a strong argument to be made that other metrics; particularly metrics that have an impact on quality of service and the quality of the user experience, need to be factored in to this economic efficiency equation. This includes the energy efficiency of the radio network but more directly the energy efficiency of the user equipment used to access the network measured in megabytes per watt hour or joules per bit. Improving energy costs at the e node B reduces operational costs (though site rental is still the dominant site specific cost) but reducing energy cost for the user will arguably become increasingly important over time. The standardisation work for LTE and LTE Advanced has been focused on delivering high peak data rates. This has focused considerable attention on techniques such as MIMO that provide spatial channel and multiplexing gain at least in small diameter cells in the higher bands. However high peak data rates do not necessarily translate into spectral or energy efficiency at subscriber level and it can be shown that average throughputs per user are rather more important particularly in larger cells in lower bands and technologies that maximise this metric are more likely to have an impact of future spectral needs. More recently considerable attention has been given to developing and marketing femto cell technologies. The operator business case for femto cell deployment is clear in that backhaul costs are reduced (loaded on to the subscriber’s ADSL line)) and heavy users, some of whom will be edge of cell are taken off outdoor sites, reducing offered traffic noise in the network. The user picks up the energy bill as well. However from a user perspective a femto cell will have to deliver access connectivity that is at least equivalent and preferably better than existing low cost WiFi solutions both in terms of data throughput and energy drain. WiFI connectivity is also of course often though not always free. This in turn adds urgency to adding MIMO to user equipment but this adds cost and complexity to all user devices including those used in wide area applications where femto cells are largely irrelevant. More frustratingly, MIMO introduces additional insertion loss and isolation loss which in many channel conditions will mean that average data rates will be lower. Dropping back to SISO in these conditions only partially helps. The objective in most present LTE user equipment designs is to pack two MIMO antennas into the space of one SISO antenna thus replacing one moderately efficient device with two dramatically inefficient devices. Good intentions sometimes result in an outcome that is opposite to the one intended.

Appendix 4

The impact of user device RF performance on network cost – the relationship with scheduler efficiency

This next section discusses and quantifies the impact of user equipment performance on scheduler efficiency. Note that scheduling gain is just one of many adaptive techniques used in any and all contemporary mobile broadband networks. These adaptive techniques can of course be used to compensate for poor channel conditions which can be viewed as a proxy for insufficient network investment

Geoff Varrall

Page 59 LTE UE efficiency and network value Final

15/09/2010

and or poor user equipment sensitivity and selectivity, a proxy for a failure to invest in user equipment performance optimisation. This is the glass half empty view of the world. Alternatively adaptive techniques can be seen as a mechanism for achieving performance extension based on network density that has been dimensioned to deliver specific user experience quality metrics across a wide range of loading conditions and user equipment where a compromise between cost and performance optimisation has been successfully achieved. This is the glass half full view of the world. We have stated that user device sensitivity and selectivity improves scheduler efficiency which in turn increases the user data duty cycle. The EBITDA model reflects this to show the fiscal gain that this can deliver to an operator on a per dB per additional band basis. Improving scheduler efficiency also improves e node B throughput which translates directly into lower energy cost per subscriber session supported and faster more responsive applications (improved application and task latency). This increases user value. In parallel, scheduler efficiency gain reduces network cost including energy cost. However both the increase in user value and the decrease in network cost per subscriber served needs to be described and dimensioned. A case study from Icera compares three receiver implementations • • •
Enhanced Type 1 UEs with receive diversity Enhanced Type 2 with chip level equalization Enhanced Type 3 with adaptive receive diversity and chip level equalization.

The spectral efficiency evolution for user equipment with two antennas is illustrated below showing the relative difference between HSPA Release 6, HSPA+ and LTE.

Source Nokia Siemens Networks An example of throughput rates measured in drive round conditions can be found here These examples show the gains that can be achieved from increasing the granularity of the scheduling. The scheduling algorithms in an e node B will be different to the scheduling algorithms used in a micro cell or macro cell. In larger cells round trip signaling delay means that the shorter scheduling options (for example half millisecond) cannot be used. Also in the lower bands it is unlikely that the 20 MHz channel bandwidths used to get high peak data rates in WiFi or higher band LTE networks could be deployed. This is not a problem but a function of fundamental channel properties at different frequencies. Note that channel sensitive scheduling doe not deliver gain in larger cells or for users moving at faster than walking pace. Fast moving users become part of the multi channel averaging process as they move through the continuously changing multi path channel.

Geoff Varrall

Page 60 LTE UE efficiency and network value Final

15/09/2010

The point about LTE is that it has been designed to be flexible enough at the physical layer to scale across a wide range of channel bandwidths and respond to a wide range of channel conditions. The ability to support 1.4 MHz, 3 MHz, and 5MHz, 10, 15 and 20 MHz channel spacing is one example. This should deliver substantial throughput gains. However we have said that LTE must also deliver more efficient throughput in terms of joules per bit and or megabytes per watt hour both at the e node B and the user device (to extend the user data duty cycle) and in small cells and large cells and for stationary users and highly mobile users (potentially up to 500 kph and certainly 350 kph). We have stated that scheduling at the physical layer in small cells can be in half millisecond increments and or across different OFDMA sub carriers in the frequency domain and or across spatial channels. Scheduler signaling load is localized to the e node B.
133 Physical layer scheduling can be either max CQI , proportional fair or round robin. Max CQI scheduling allocates LTE resource blocks on the basis of measured channel quality. This is the most efficient scheduler from an operator/network perspective in terms of throughout against available bandwidth and power. However edge of cell users would suffer very poor service. This would result inconsistent user experience metrics which could trigger high product return and churn.

Round robin services all users and devices in turn. It delivers the best user experience and hence lowest product return and churn however it is theoretically the least efficient option from an operator/network perspective (throughput against bandwidth and power). Proportional fair scheduling is a compromise point between these two extremes and can be set to meet specific user experience expectations. Irrespective of where the scheduling compromise point is set, there is a directly beneficial relationship between user equipment performance and scheduler efficiency gain. Base band interference cancellation and advanced receiver techniques will be particularly effective in interference limited conditions; improved 134 RF selectivity will also help. Improved RF sensitivity will be particularly helpful in noise limited conditions. Scheduling is also performed at packet level in order to differentiate best effort traffic from streamed or interactive or conversational traffic and to respond to resource availability which in turn is a function of physical layer scheduling efficiency. Traditional handover to other channels and or other bands is also supported providing opportunities to deliver macro scheduling gain (see earlier analysis). In terms of e node B energy consumption, increased sensitivity in a user device operating in a noise limited environment will result in a higher reported downlink CQI which will translate into a higher allocation of downlink time and frequency sub carrier resource blocks. A better link budget also translates into being able to support higher order modulation and lighter error protection/correction overheads which means that the ratio of information bits to error protection bits improves. Thus any improvement in sensitivity will result in greater information throughput for a given amount of input energy. Note that channel coding and other adaptive physical layer mechanisms mitigate the impact of poor user equipment sensitivity or selectivity but the bandwidth and power cost is effectively transferred to other users. The relationship of user equipment performance to overall network QOS and QOE is therefore direct and inescapable. Given that the base station is sharing a finite amount of transmit power, for example 20 watts in a macro or large micro site, then this means that more power is available to serve other user downlinks or put another way, the same amount of power is being used more efficiently across all users being serviced from the site. A similar argument applies for devices with increased receive selectivity operating in an interference limited environment. On the uplink, user device TX efficiency will influence the user data duty cycle but TX linearity is equally important. The base station has to demodulate a signal modulated with noise and phase sensitive higher order modulation. Much effort is expended in mitigating the impact of channel distortions on this complex signal but if the signal is distorted before it leaves the antenna of the user device there is not much that the base station can do with the received signal energy. This is the reason for the nearly 5x more strict EVM specifications compared to WCDMA. As with RX sensitivity, error vector magnitude is a specified performance metric described precisely in the conformance standard and user devices would generally be designed and cost engineered to exceed this standard by an acceptable but not unnecessarily large margin. A lower EVM with additional margin over

133 134

Channel Quality Indicators – With thanks to Icera, technology update briefing August 13 2010

Geoff Varrall

Page 61 LTE UE efficiency and network value Final

15/09/2010

and above the conformance requirement would however translate into higher uplink data rates and more uplink capacity at the base station. Improving the RF performance in the front end of the user’s device reduces baseband coding overheads which reduces baseband DC power drain. Applications will also run faster which further reduces power drain and adds to user experience value. The benefit will accrue irrespective of where a user is in the cell though the maximum opportunity cost is likely to be incurred in edge of cell conditions. This can be said to generally apply in all macro, micro, pico and femto cell topologies and also applies irrespective of whether the user is stationary or highly mobile. LTE QOE and compression Some dramatic improvements are being made in compression techniques particularly in digital TV where the DVB T 2 standard is now complete and close to implementation. However digital TV is not power constrained and is essentially a one way delivery medium. It also has a relatively generous first order and second order latency budget (first order is the amount of latency, second order is the variation between worst and best latency) In a mobile broadband network, indeed in any two-way radio network, an increase in compression ratio means that more information can be loaded on to each bit sent. This means that the energy per bit relative to the noise floor improves. However each bit is now carrying more information so if the bit is received in error proportionately more information is lost. Even with all of the adaptive process in an LTE radio network, the radio channel is still an error prone channel when compared to broadcast networks and the errors have a habit of occurring in bursts that can translate at higher layers of the protocol stack in to packet errors and retries which are disruptive to compression techniques that rely on memory - the result is error extension. Additionally, higher order compression is computationally complex, power hungry and time hungry – the extra clock cycles needed to compress a signal in digital TV can be easily accommodated in the power and time budget both at the transmitter and the receiver (which is normally connected to the mains). This is not the case in a mobile broadband network. Doing higher order compression on the user equipment TX path for instance will introduce compression delay, absorb energy and require performance parameters such as error vector magnitude to be more closely controlled (which in turn will have an impact on the user’s energy budget). Compression techniques have generally been developed either to maximise broadcast transmission efficiency (DVB T as an example) or memory bandwidth efficiency (MP3 and MP4). These efficiency gains do not translate comfortably into a mobile broadband channel. Improvements in compression techniques may have some beneficial impact on future spectral efficiency but these benefits are likely to be marginal rather than fundamental and less useful than RF performance and baseband performance gain. MOS and QOS Measuring the quality of voice codecs is done on a well-established mean opinion score basis (MOS) and similar processes are being agreed for video perceived quality benchmarking. Note that if actual video quality stays the same but audio quality improves the perceived video quality will improve. Similarly, user experience opinion scores will generally be higher when the delivered quality is relatively constant and will be higher than a service which is sometimes good and sometimes bad. This has an impact on scheduling implementation but it can be generally stated that any improvements in user equipment RF and baseband performance will translate directly into a better and more consistent user experience MOS. The three companies below have web sites with information on voice and video MOS comparisons. Some additional standards work on application performance benchmarking may be useful in the longer term. Opticom Psytechnics Radvision The cost of store and forward (SMS QOS) and buffering

Geoff Varrall

Page 62 LTE UE efficiency and network value Final

15/09/2010

Present and future mobile broadband networks have a mix of store and forward processes whose purpose is to reduce the cost of delivery both in terms of bandwidth and energy used. An example is SMS where in extreme circumstances several hours might elapse between a message being sent and received. At the physical layer best effort data will be buffered in the e node B or user’s device to smooth out traffic loading (which reduces the offered traffic noise) and or to take advantage of CQI-based scheduling gain. There is an alternative argument that says short term (buffering) or longer term (store and forward) has an associated memory cost that needs to be factored in to the efficiency equation – fast memory in particular incurs a capital and operation (energy) cost. In the British Post Office, an organization not generally known for its efficiency, a second class letter costs more to send than a first class letter as it spends more time and occupies more space in the sorting office. Applying this analogy to mobile broadband networks suggests it may be more cost efficient to minimize store and forward and buffer occupancy. This can be achieved by improving user equipment RF and baseband performance. In appendix 2 we referenced a case study that established a plausible and probably under-stated relationship between receive sensitivity and data throughput in which it could be shown that an extra 3 dB of sensitivity translated into a doubling of the downlink data rate in a noise limited environment This could also be equated to a doubling of the user data duty cycle or a halving of the user power budget for the same amount of data. Similarly it can be argued that a 3dB improvement in selectivity would translate into a doubling of the downlink data rate in an interference limited environment Some RX RF performance improvements can deliver a double benefit. Achieving better/lower phase noise in a front end PLL and LNA improves sensitivity but also improves the accuracy with which antenna adaptive matching can be realized. The impact of a link budget gain in a mobile broadband network is similar but different to the impact of a link budget gain in a legacy voice and text network. Improving the link budget in a voice and text network improves capacity and coverage depending on whether any particular link at any particular time is noise or interference limited. The user experiences better voice quality (though the operator can trade this against capacity gain by using a lower rate codec), improved geographic coverage and in building penetration better, less blocked calls, less dropped calls and a longer duty cycle (minutes of use between battery recharge). Improving the link budget in a mobile broadband network increases coverage and capacity and average per user data throughout rates and improves end to end latency. This translates into reduced application latency for the user, lower blocked session rates, lower dropped session rates and a longer data duty cycle (megabytes of use between battery recharge). Assuming a mobile broadband network will also be supporting voice and assuming voice traffic is on an IP virtual circuit then the link budget will need to be sufficient to ensure users have the same voice quality, 135 coverage and talk time that they have on present devices. The link budget can be improved by increasing network density, improving base station efficiency up to an interference threshold, or by improving user equipment efficiency or a mix of all three. Increasing network density has a capital and operational cost implication in terms of site acquisition, site rental, site energy costs and backhaul. The increased burstiness 136 of offered traffic in a mobile broadband network means that backhaul needs to be over provisioned to cope with higher peaks versus the traffic average. Backhaul costs are therefore becoming more important over time. 137 Improving the link budget by improving user equipment performance does not have these associated capital or operational costs [. Additionally improving TX and RX efficiency in user equipment can be shown to deliver non linear gains in scheduler efficiency (micro scheduling gain).

As stated in other sections IP voice has some additional processing overheads that are inherent in processing a narrow-band information stream, tyically12 kbps, from a relatively broadband channel (20 MHz) The user will expect IP voice to be either free or significantly cheaper than existing voice service offerings and of at least equivalent quality. 136 A series of RTT technology topics addressed this topic in 2003 – for one example follow this link. 137 Femto cells are part of the solution to backhaul costs and take some users off outdoor cells but have limited impact on wide area access economics.

135

Geoff Varrall

Page 63 LTE UE efficiency and network value Final

15/09/2010

If RF and baseband performance improvement in user equipment can be combined with extended band flexibility then additional gain can be achieved by implementing more comprehensive inter band inter operator handover. However every additional band increases cost and decreases sensitivity and selectivity. Additionally inter band measurements are slow and absorb network signaling bandwidth and power. Ideally user equipment would be capable of accessing all available bandwidth in all countries with a parallel capability to scan the available bandwidth for the best available connectivity. To be efficient this would need a separate receive function in the user’s device (effectively a dual receiver). Such a device does not presently exist. Summary The traffic forecasts in this study indicate a 30 fold increase in data volume over the next five years (from 3 to 90 exabytes). Over the same timescale based on present tariff trends, income will have increased by at most a factor of three. Using a combination of averaging, adaptation and opportunistic scheduling including in certain channel conditions spatial multiplexing and multi user diversity, LTE delivers spectral efficiency benefits over and above HSPA Release 6 by roughly an order of three which is impressive but short of the increase needed to balance this traffic/income disparity. Appendix 5 Operator cost/benefit analysis including energy costs

So for the time being at least, short term performance gains are more likely to achieved from RF and baseband efficiency improvements (micro scheduling gain) rather than extended multi band capability (macro scheduling gain). Looking at this positively, it could mean that for the first five years of LTE a one dB gain per year over and above the conformance specification could be achieved from improving RF and baseband efficiency. Adding a band per year could probably be achieved in parallel without compromising this target. From year 5 onwards an ability to support extended multi band if combined with a dual receiver capability would open up the possibility of realizing macro scheduling gain. This could probably deliver another five years of year on year performance improvement. This brings us to the final reckoning as to the whether RF and baseband performance gain in the user device will be commercially worthwhile for the operator community and overall for the industry supply chain. Note that if it did not make sense to all parts of the supply chain then it is unlikely to happen. The start point is that if something is technically efficient it should also be possible to make it commercially efficient. It is far harder to take something that is technically inefficient and make it commercially efficient. It is our contention that the mobile broadband physical layer is presently technically and commercially inefficient irrespective of the technology used. This is due to a global regulatory policy that has created a spectral auction process that has brought more spectrum to market than can be presently absorbed given present RF technical and commercial constraints in the user device. Our thesis is that a transition to LTE if combined with more aggressive and properly rewarded RF and baseband innovation in user devices could resolve this. Looking at this from a positive perspective, let us look at what would happen fiscally if it could be ensured that the RX sensitivity and selectivity and TX accuracy of LTE user devices could be progressively improved to deliver a dB of uplink and downlink performance gain per year from 2010 to 2015 over and above the present conformance standard. As stated earlier this closely mirrors the GSM model from 1992 to 1997. We would suggest this scale of improvement could be reasonably easily accomplished over a three to five year period in a single band phone without any significant RF component cost increase. The real challenge is to replicate this performance gain in a multi band phone given that in each subsequent year it may be necessary to add an additional band over and above the present supported band plan in each successive year. Given that an additional band at present introduces approximately one dB of performance loss then this represents a more significant challenge and almost certainly additional cost. This additional cost should be off set to some extent by inventory management costs (the ability to ship a standard multi band phone to most markets) but an increase in the RF BOM would seem more or less certain. This increase has to be off set against the increase in operator value which in turn will depend on the additional value that a multi band LTE device would deliver to roaming and national users.

Geoff Varrall

Page 64 LTE UE efficiency and network value Final

15/09/2010

If both international and national roaming was implemented in a significant number of markets within this time scale, and this seems increasingly likely, then the ability to hand off high opportunity cost subscribers to other operators in other bands would transform the delivery economics of mobile broadband connectivity both in terms of direct cost and indirect cost. We have said that user devices can rather crudely be divided into four groups, basic phones handling voice and text, smart phones including iPhones, tablets and slates of which the iPad is an early example and lap tops with dongle or embedded connectivity. It could be argued that the lower frequency bands, for example the 700 and 800 MHz bands are likely to be predominantly larger form factor devices which would make some of the RF design issues rather easier. However it seems rash to base a 700/800 MHz business model on a product mix that excludes smaller form factor smart phones and other smaller devices so all four device types are included in the model. Note that delivering traffic over other operator’s infrastructure, which is one of the potential benefits of band flexibility, is particularly beneficial to EBITDA performance. International roaming provides an example. An analysis of operator results in Europe suggests that revenue from international roaming is approximately 138 3%. However roaming revenue does not have a capital cost associated with it. This means that every 250 million euros of roaming revenue provides an EBITDA contribution that is equivalent to one billion euros of revenue if spectral and network cost amortization is included. This of course also makes a compelling economic case for national roaming and highlights the potential network cost benefits of load shedding to other operators working on other bands. The operator that off loads the traffic enjoys an improved EBITDA, the operator that delivers the traffic enjoys an improved EBITDA. As the basis for an initial economic analysis we have made a number of additional assumptions that are stated in the EBITDA model at the end of the study. Energy Costs Network energy costs represent a small percentage of national energy consumption but are still significant in terms of network operational cost. In the UK, 3 G networks have an approximate energy consumption of 139 300 GWh/year, 2 G networks have approximate energy consumption of 1000 GWh/year. Electricity costs over the seven years are detailed below Year Large user price per kWh in pence 2.8 pence 4.8 pence 6.7 pence Dollar equivalent (x 1.5) 4.2 cents 7.2 cents 10.05 cents Very large user price per kWh in pence 2.4 pence 4.0 pence 5.6 pence Dollar equivalent

2001 2007 2008/9

3.6 cents 6 cents 8.4 cents

At large use rates this suggests GSM networks are costing £64 million per year ($96 million) or £1600 ($2400) per site per year and 3 G networks are costing £20 million ($30 million) per year or £1000 ($1500) dollars per site. While this is less than site cost rental and associated real estate costs it still represents an expense that can be reduced. We have not scaled this globally but self evidently this represents a useful margin contribution. The report also states that a phone charged for three hours a day will consume 7.5 kWh/ year. While individually relatively insignificant, the total for five billion users will be in the order of 37.5 million megawatts per year so a saving here might be useful too. The energy impact of moving and storing > 80 exabytes of

Statistics and analysis from The Mobile World. These figures are taken from the report commissioned by Ofcom in the UK in April 2009 and researched by Plextek – energy costs are taken from UK government data for grid power energy costs.
139

138

A full copy of the report which includes an analysis of base station efficiency can be downloaded from the Ofcom site http://stakeholders.ofcom.org.uk/binaries/research/technology-research/environ.pdf. The report was undertaken by Plextek

Geoff Varrall

Page 65 LTE UE efficiency and network value Final

15/09/2010

data by 2015 will not be insignificant 140 . Improving user equipment RF efficiency at least addresses the radio network part of this equation. User energy budgets and battery density In section 1 we drew attention to battery performance limitations. The table below compares present battery densities in term of watt hours per kg and Wh per litre.

Wh/kg Wh/litre

Ni-Cads 60 175

NiMh 90 235

Lithium Ion 120 280

Zinc Air 210 210

Lithium Thin Film 600 1800

Lithium ion batteries using a liquid electrolyte deliver better energy density (120 Wh/kg) but also have a relatively high self discharge rate. Lithium metal batteries, using a manganese compound, deliver about 140 Wh/kg and a low self discharge rate, about 2% per month compared to 8% per month for lithium ion and 20% per month for lithium polymer. Lithium thin film promises very high energy density by volume (1800 Wh/liter).However delivering good through life performance remains a non trivial task. These very high density batteries like to hold on to their power, they have high internal resistance. This is not inherently well suited to devices that have to produce 141 energy in bursts on demand. Appendix 6 Network infrastructure cost benefit analysis

Over the past ten years all of the major infrastructure vendors have decoupled their network hardware and software R and D development and market and business activity from their handset activities which are now run as separate businesses. The concept of adding cost to user equipment to reduce network density and network hardware and software investment is therefore not apparently appealing. However the infrastructure vendor community is dependent on operator spending which in turn is dependent on operator EBITDA Improved operator EBITDA combined with compelling LTE user experience benefits (thoughput, duty cycle and application and task latency) and delivery cost reduction could potentially increase the transition rate presently anticipated by the LTE vendor community. The overall impact of this should be an increase in the volume and value of LTE network hardware and software shipments. The EBITDA benefits are shown in the model. Appendix 7 User equipment vendor cost benefit analysis

Similarly a more rapid transition to LTE should be good news for user equipment vendors with strong capability in LTE user device design. Recent press comments about iPhone RF performance have highlighted that a strong user value proposition can be compromised by a poor or inconsistent connectivity experience. Conversely it should be true that a strong user value proposition can be enhanced by better than average and or better than expected connectivity. If this is combined with a longer data duty cycle (a longer time between recharge) then the value proposition is further increased. However we have said that this can only be achieved at least in multi band devices by increasing the RF BOM cost. This should however be off set by savings in inventory management costs and potentially in higher realised prices. There are other subtler reasons why improving multi band flexibility would be useful for user device vendors. We have said that the market will be made up of some mix (choose your own numbers) of standard phones, smart phones, tablets and slates and lap tops. The industry has become accustomed to volumes that are predicated on replacing user devices every 18 months or less in some markets. This is great for volume but introduces costs both in terms of
One informed industry observer has calculated that by 2020 server storage will be absorbing more energy than the aviation industry. The calculation does not factor in the associated connectivity energy needed to access that data. 141 Additional resources are available on this subject. If you would like these sent to you please email geoff@rttonline.com
140

Geoff Varrall

Page 66 LTE UE efficiency and network value Final

15/09/2010

administration and in some cases subsidy expense. Most of us prefer to keep our lap tops rather longer, preferably between three and five years. If connectivity is via an LTE dongle then a device vendor has the opportunity too supply an upgrade, a faster modem or additional band support for example. If embedded, the user is tied into a connectivity solution which gets out of date and the vendor loses an update sale opportunity. The modem as stated in an earlier section is becoming more software configurable over time and in a three to five year time horizon it can be reasonably assumed that upgrades could be downloaded much in the same way as most current ADSL modems. However it is not presently possible to add bands through the lifetime of the device. A software reconfigurable RF front end presently remains an ambition rather than a market reality but would completely change the economic model for this part of the industry supply chain.

The EBITDA model
The model has been prepared for this study by The Mobile World. We have made some assumptions which are described below. If you would like to input your own assumptions or have the model run for specific markets or would like to comment on the model please contact; John Tysoe john@themobileworld.com Assumptions The market growth and market mix assumptions are an estimate based on past trends analysed from The Mobile World statistical data base. The following assumptions have been made; Vendor estimates suggest a dongle in a lap top or a lap top with embedded connectivity presently generates about 2 GB a month of data traffic, an iPhone about 165 MB 142 , a smart phone about 115 MB per month and a standard phone about 30 MB per month including voice. A tablet or slate will presumably be somewhere between a lap top and iPhone but presently it is too early to tell. The table below puts this in perspective but also highlights that the assumptions we have used (the lower numbers for Smart phones and iPhones) are likely to be conservative. The more recent numbers also suggests ‘ordinary’ smart phones are catching up with iPhone functionality. Mobile Broadband user equipment offered traffic per month Form factor As reported February 2010 Standard phone 30 MB Smart phone 115 MB iPhone 165 MB Tablet ? Lap top 2 GB As reported August 143 2010 30 MB 300-500 MB 300-500 MB ? 2 GB

The product mix assumptions are based on past trends. For tablet devices there is not a huge amount of trend information to go on so the actual may be different from the projection. On this basis our calculations show an annual data volume of 87 million terabytes (more or less equivalent to 87 exabytes) per year by 2015, nearly four times more than the NSN estimates but similar to some other forecasts. All of these numbers would be significant under estimates based on present growth trends. Some recent tariff product and service offerings integrating 3 G connectivity with multiple Wi Fi devices suggest the 2 GB per month figure for lap top traffic may also be conservative. The table below shows the scale of the increase in data volume that this implies.
Mobile Broadband data scale and industry volume projections 8 bits = 1 byte Options – multiply by 1000 or 1024 or 1048 (IBM and Microsoft) http://www.whatsabyte.com/ Kilobyte Megabyte Gigabyte Terabyte Petabyte 1000 1000 1000 1000 1000 bytes kilobytes megabytes gigabytes terabytes
142 143

Exabyte 1000 petabytes

Zetabyte 1000 exabytes

Yottabyte 1000 zetabytes

Brontabyte 1000 yottaytes

Geopbyte 1000 brontabytes

See earlier footnote - Operators are suggesting this may be nearer 300 to 500 Mb per month rather than 165 Mb – reported via NSN inputs August 2010

NSN reports from operator customers August 2010

Geoff Varrall

Page 67 LTE UE efficiency and network value Final

15/09/2010

Paragraph of text

CD ROM 600 megabytes

10 yards of books

1000 copies of the Encyclopedia Britannica

20 million four draw cabinets of text

all the words ever spoken by mankind

The entire world wide web

Mobile broadband data volume 2010 3 exabytes NSN projection for 2015 23 exabytes TMW projection for 2015 87 exabytes Cisco projection for 2015 90 exabytes All of the above would be an underestimate based on present growth trends. See also this web site http://www.iec.ch/zone/si/si_bytes.htm for definitions of the size of a megabyte.

Operators can of course influence product mix by subsidy and or by tariff product and service bundling (follow link) and or by choking demand by under investing in radio and or network bandwidth. The ideal user equipment product and service mix would maximize average margin per user (AMPU) and or average margin per device (AMPD) and or average margin per service (AMPS) or average margin per application (AMPA) which of course means that operators need visibility to device specific delivery cost which is a composite of delivery and memory bandwidth cost and customer support costs expressed in terms of dollar costs and fractional dollar income per byte (dollar bytes). 144 But ignoring this caveat let’s go with our projections based on past volume metrics and what we feel could be realistic data volumes if an improved investment return is used to provide year on year improvements in the user experience. We have factored in an additional dollar per year of RF BOM cost for each additional band assuming this band flexibility is achieved with a 1dB per year gain in sensitivity and selectivity over and above the conformance standard. An assumed 1% incremental benefit to ARPU from each additional band on a year on year basis is then added to the calculation .The financial effects are then shown with an incremental gain on the overall investment of 5.6% in year one, 9.4% in year two, 13.6% in year 3, 18.4% in year 4, 24.6% in year 5 and 33.1% in year 6. On this basis and on these assumptions it is clear that a year on year improvement in user equipment efficiency combined with a year on year improvement in band flexibility provides a positive EBITDA return even when additional BOM costs are included. Operators are faced with demand side projections that suggest additional spectrum will be needed to meet future mobile broadband user needs and expectations. However before lobbying or bidding for this spectrum, the returns from present spectral and network investment including LTE 700 and planned 800 and 2600 MHz deployments need to be improved. There is effectively a safe absorption rate that governs how fast operators can invest in spectrum and network hardware and software and generate an adequate return. User equipment performance has always been and continues to be a major factor in this ROI equation, the lack of performance competitive GSM mobiles in 1992 being a notable past example. In terms of LTE single band, multi band and extended multi band user equipment, the challenges over the next three to five years will be range availability, RF and baseband performance and efficiency. The direct coupling between user equipment performance, band flexibility, quality of service, quality of the user experience and operator and supply chain EBITDA suggests that any or all improvements that can be delivered will result in a positive return. The linkage between additional bands and increased EBITDA is open to debate. Our argument rests on the improved user experience that should be realisable from inter band handover. Philosophically, a technical challenge should translate into a commercial opportunity. A commercial challenge should translate into a technical opportunity. Mobile broadband appears to meet both criteria.

Mobile broadband radio access economics - the Model

144

Dollar bytes are a way of describing and comparing delivery value and delivery cost in mobile broadband networks.

Geoff Varrall

Page 68 LTE UE efficiency and network value Final

15/09/2010

2010 ASSUMPTIONS Net Additions to Global Subscriber Base (m) Phone Upgrades (m) Global Handset Shipments (m) Proportion of New Devices Laptop Devices iPAD & Other Handheld Smartphones iPhones Other Smartphones Ordinary Phones New Devices as a Proportion of Net New Connections Laptop Devices iPAD & Other Handheld Smartphones iPhones Other Smartphones Ordinary Phones Increase in Data Usage per Month (MBytes) Laptop Devices iPAD & Other Handheld iPhones Smartphones Ordinary Phones 683 700 1,383

2011 600 800 1,400

2012 550 880 1,430

2013 500 968 1,468

2014 500 1,065 1,565

2015 450 1,171 1,621

6.0% 1.1% 15.0% 3.0% 12.0% 77.9% 49.4% 8.0% 1.5% 20.0% 4.0% 16.0% 70.5%

6.4% 1.6% 18.4% 4.0% 14.4% 73.6% 42.9% 8.0% 2.0% 23.0% 5.0% 18.0% 67.0%

7.7% 2.6% 22.1% 6.0% 16.2% 67.7% 38.5% 9.0% 3.0% 26.0% 7.0% 19.0% 62.0%

8.5% 3.4% 24.7% 6.8% 17.9% 63.5% 34.1% 10.0% 4.0% 29.0% 8.0% 21.0% 57.0%

9.4% 4.7% 27.2% 7.7% 19.6% 58.8% 32.0% 11.0% 5.5% 32.0% 9.0% 23.0% 51.5%

10.2% 6.0% 29.8% 8.5% 21.3% 54.1% 27.8% 12.0% 7.0% 35.0% 10.0% 25.0% 46.0%

20.0% 40.0% 35.0% 30.0% 10.0%

24.0% 50.0% 40.0% 35.0% 10.0%

28.8% 80.0% 45.0% 40.0% 10.0%

34.6% 100.0% 50.0% 45.0% 10.0%

41.5% 80.0% 55.0% 50.0% 10.0%

49.8% 50.0% 55.0% 50.0% 10.0%

DATA VOLUME CALCULATIONS Net New Connections (m ) Of which: Laptop Devices Of which: iPAD & Other Handheld Of which: Smartphones Of which: iPhones Of which: Other Smartphones Of which: Ordinary Phones Total Connections (m ) Of which: Laptop Devices Of which: iPAD & Other Handheld Of which: Smartphones Of which: iPhones Of which: Other Smartphones Of which: Ordinary Phones Average Global Device Population (m ) Laptop Devices iPAD & Other Handheld Smartphones iPhones Other Smartphones Ordinary Phones

2010 55 10 137 27 109 482 5,332 148 22 555 97 458 4,608

2011 48 12 138 30 108 402 5,932 196 34 693 127 566 5,010

2012 50 17 143 39 105 341 6,482 245 50 836 166 670 5,351

2013 50 20 145 40 105 285 6,982 295 70 981 206 775 5,636

2014 55 28 160 45 115 258 7,482 350 98 1,141 251 890 5,893

2015 54 32 158 45 113 207 7,932 404 129 1,299 296 1,003 6,100

120 17 487 83 403 4,367 4,991

172 28 624 112 512 4,809 5,632

220 42 765 146 618 5,180 6,207

270 60 909 186 723 5,493 6,732

323 84 1,061 228 833 5,764 7,232

377 114 1,220 273 947 5,997 7,707

Data Usage per Month (GBytes) Laptop Devices iPAD & Other Handheld Smartphones iPhones Ordinary Phones Data Usage per Month (Gbytes) Laptop Devices iPAD & Other Handheld Smartphones iPhones Ordinary Phones GigaBytes per Month TeraBytes per Month TeraBytes per Year

2010 2.458 0.560 0.223 0.150 0.033

2011 3.047 0.840 0.312 0.202 0.036

2012 3.925 1.512 0.452 0.283 0.040

2013 5.282 3.024 0.678 0.410 0.044

2014 7.472 5.443 1.051 0.615 0.048

2015 11.191 8.165 1.630 0.922 0.053

295,649,280 9,377,200 108,414,653 12,467,553 13,309,395 439,218,080 428,924 5,147,087

522,998,907 23,408,700 194,597,519 22,615,500 18,583,967 782,204,592 763,872 9,166,460

864,970,345 63,681,660 345,698,043 41,339,209 24,684,926 1,340,374,183 1,308,959 15,707,510

1,426,663,223 182,551,320 616,218,485 76,022,765 31,754,352 2,333,210,145 2,278,526 27,342,306

2,410,607,509 457,868,376 1,115,465,609 140,152,825 40,244,471 4,164,338,790 4,066,737 48,800,845

4,220,162,499 927,664,164 1,987,663,182 251,711,844 50,314,370 7,437,516,059 7,263,199 87,158,391

Geoff Varrall

Page 69 LTE UE efficiency and network value Final

15/09/2010

RF ECONOMICS CALCULATION Additional Mobile Subscribers (m) CapEx per Incremental GB Capital Expenditure ($m) Cumulative Capital Expenditure ($m) New Mobile Devices Laptop Devices iPAD & Other Handheld Smartphones iPhones Other Smartphones Ordinary Phones

2010 683 0.044 154,923 154,923

2011 600 0.035 141,482 296,405

2012 550 0.021 138,147 434,552

2013 500 0.011 122,863 557,415

2014 500 0.004 90,641 648,056

2015 450 0.001 40,506 688,561

83 16 207 41 166 1,077 1,383

90 22 258 56 202 1,030 1,400

109 36 316 85 231 968 1,430

125 50 362 100 262 931 1,468

146 73 426 120 306 920 1,565

165 96 482 138 345 877 1,621

Average Handset Price ($) Laptop Devices iPAD & Other Handheld Smartphones iPhones Ordinary Phones Weighted Average Handset Price Incremental Cost of Additional Bands Weighted Average Handset Price, including additional bands Handset Costs ($m ) Incremental Handset Costs ($m) Cumulative Incremental Handset Costs ($m) Handset Costs, including new bands ($m) Additional Handset Cost (%)

30.00 500.00 200.00 200.00 30.00 90.79 1.00 91.79 126,942 1,383 1,383 128,325 1.1%

29.10 475.00 184.00 190.00 29.10 99.07 2.00 101.07 141,501 2,800 4,183 144,301 2.0%

28.23 451.25 169.28 180.50 28.23 108.93 3.00 111.93 160,064 4,290 8,473 164,354 2.7%

27.38 428.69 155.74 171.48 27.38 113.19 4.00 117.19 172,041 5,872 14,345 177,913 3.4%

26.56 407.25 143.28 162.90 26.56 118.08 5.00 123.08 192,593 7,824 22,169 200,417 4.1%

25.76 386.89 131.82 154.76 25.76 121.92 6.00 127.92 207,387 9,728 31,897 217,114 4.7%

2010 Global Average ARPU ($) Of which Voice Of which Non-Voice Trend in ARPU Voice Non-Voice Global Service Revenue ($m ) Of which Voice Of which Non-Voice Global EBITDA Margin Global Service EBITDA ($m) Global Operating Costs Capital Intensity Mobile Device Revenues ($m ) Laptop Devices iPAD & Other Handheld Smartphones iPhones Other Smartphones Ordinary Phones 17.00 13.60 3.40

2011 16.15 12.65 3.50 -5.0% -7.0% 3.0% 1,091,482 854,802 236,679 35.0% 382,019 709,463 13.0%

2012 15.37 11.76 3.61 -4.8% -7.0% 3.0% 1,144,797 876,128 268,668 35.0% 400,679 744,118 12.1%

2013 14.65 10.94 3.72 -4.7% -7.0% 3.0% 1,183,851 883,717 300,135 35.0% 414,348 769,503 10.4%

2014 14.00 10.17 3.83 -4.5% -7.0% 3.0% 1,214,997 882,898 332,099 35.0% 425,249 789,748 7.5%

2015 13.40 9.46 3.94 -4.3% -7.0% 3.0% 1,239,553 875,025 364,529 35.0% 433,844 805,710 3.3%

1,018,062 814,450 203,612 35.0% 356,322 661,740 15.2%

2,489 7,779 41,490 8,298 33,192 32,310 125,559

2,607 10,640 47,734 10,640 37,094 29,985 138,701

3,088 16,455 54,452 15,358 39,094 27,327 155,774

3,417 21,397 57,926 17,117 40,809 25,503 166,169

3,886 29,792 63,332 19,500 43,832 24,426 184,769

4,260 37,322 66,740 21,327 45,414 22,596 197,659

2010 Additional Bill of Materials per Band ($) Additional Bands Cumulative New Bands Increm ental Benefit to ARPU of Additional Bands Voice Non-Voice Average Voice ($) Non-Voice ($) Average ($) Average, Cumulative ($) Financial Effects Incremental Service Revenue ($m) Cumulative Incremental Service Revenue ($m) Incremental EBITDA ($m) Cumulative Incremental EBITDA ($m) Return on Incremental Investment in Handset Incremental Return on Total Investment 1.00 1 1

2011 1.00 1 2

2012 1.00 1 3

2013 1.00 1 4

2014 1.00 1 5

2015 1.00 1 6

1.0% 1.0% 1.0% 0.14 0.03 0.17 0.17

1.0% 1.0% 1.0% 0.13 0.04 0.16 0.33

1.0% 1.0% 1.0% 0.12 0.04 0.15 0.49

1.0% 1.0% 1.0% 0.11 0.04 0.15 0.63

1.0% 1.0% 1.0% 0.10 0.04 0.14 0.77

1.0% 1.0% 1.0% 0.09 0.04 0.13 0.91

10,181 10,181 8,798 8,798 636.1% 5.6%

22,404 32,585 19,604 28,402 679.0% 9.4%

36,139 68,724 31,849 60,251 711.1% 13.6%

51,035 119,759 45,163 105,414 734.8% 18.4%

66,975 186,734 59,151 164,565 742.3% 24.6%

83,770 270,503 74,042 238,607 748.1% 33.1%

Comments on the model to John Tysoe

john@themobileworld.com

Geoff Varrall

Page 70 LTE UE efficiency and network value Final

15/09/2010

Study sponsor profiles
Peregrine Semiconductor
Peregrine Semiconductor is a global leader of high-performance RF CMOS solutions with a patented Ultra CMOS™ process technology – enabled by silicon on sapphire substrates. These devices achieve high levels of monolithic integration through a broad portfolio of mixed-signal RF ICs which can be used to build efficient and flexible and innovative RF front end architectures that combine band flexibility with exceptional transmitter and receiver efficiency. This translates directly into improved data and voice duty cycles with a range of device solutions directly targeted at the LTE user equipment market. Contact Jim Cable jcable@psemi.com 001 858 731 9400

Ethertronics
Ethertronics is a leading developer and manufacturer of high performance embedded antennas for Cellular, WiFi/WiMax, GPS, Bluetooth and other wireless devices. Ethertronics patented, Isolated Magnetic Dipole (IMD™) technology resists antenna detuning by confining the electromagnetic currents on the antenna element to deliver higher isolation. This higher isolation delivers improved network coverage and capacity even when using compact antennas. These benefits significantly reduce time-to-market and design risks for user equipment vendors and improve connectivity, battery life, peak and average data throughput and application and task latency, improving the user experience and the user value realizable from present and future mobile broadband spectral and network investment. Contact Barry Matsumori bmatsumori@ethertronics.com 001 858 550 3845

Research Partners 
  The IWPC study group A detailed technical study on tuneable components and architectures is being undertaken by a working group within the IWPC. The objective of this group is to identify where tuning will be generically useful in the RF front end. Cases being investigated for a report due in September include Reducing antenna size while supporting lower bands by using active components for tuning Improving MIMO and SIMO performance for the second antenna Improving efficiency in the RF chain via impedance matching –on/near the antenna or antenna switch module Improving efficiency of the PA by adjusting the load line impedance, PA bias and/or low-loss switches to use different PA stages/sections The technology necessary to make tunable filters both technically feasible and economically affordable Contact larry.morrell@cavendish-kinetics.com 001 408-240-7384 National Microelectronics Institute The National Microelectronics Institute (NMI) is a trade association representing the semiconductor industry in the UK and Ireland. Its objective is to help build and support a strong micro and nano-electronics community by acting as a catalyst and facilitator for both commercial and technological development. A notfor-profit organization funded by its members, the NMI has a membership that spans the supply chain and includes fabless semiconductor manufacturers, IDMs, foundries, design services, IP providers, business associates, research and academic institutions. Contact info@nmi.org.uk 01506 401210

Geoff Varrall

Page 71 LTE UE efficiency and network value Final

15/09/2010

RTT Contacts
Lead author for the study Geoff Varrall geoff@rttonline.com 00 44 208 744 3163 www.rttonline.com RTT Programmes Limited Twickenham, UK TW1 2BG Additional resources relevant to this study are available from the linked web site www.makingtelecomswork.com Monthly (free) updates on mobile broadband technology, engineering, market and business issues www.rttonline.com/techtopics.html Index of technology topics (130 archived over the past ten years) www.rttonline.com/sitemap.html This study issued September 1st 2010.

Geoff Varrall

Page 72 LTE UE efficiency and network value Final

15/09/2010

Sign up to vote on this title
UsefulNot useful