Professional Documents
Culture Documents
Ebffiledoc 985
Ebffiledoc 985
https://textbookfull.com/product/from-internet-of-things-to-
smart-cities-enabling-technologies-1st-edition-hongjian-sun/
https://textbookfull.com/product/tfet-integrated-circuits-from-
perspective-towards-reality-navneet-gupta/
https://textbookfull.com/product/the-internet-of-things-and-big-
data-analytics-integrated-platforms-and-industry-use-cases-1st-
edition-pethuru-raj-editor/
https://textbookfull.com/product/enabling-the-internet-of-things-
fundamentals-design-and-applications-1st-edition-muhammad-azhar-
iqbal/
Ageing of Integrated Circuits: Causes, Effects and
Mitigation Techniques Basel Halak
https://textbookfull.com/product/ageing-of-integrated-circuits-
causes-effects-and-mitigation-techniques-basel-halak/
https://textbookfull.com/product/design-of-analog-cmos-
integrated-circuits-2nd-eddition-2nd-edition-behzad-razavi/
https://textbookfull.com/product/automated-hierarchical-
synthesis-of-radio-frequency-integrated-circuits-and-systems-a-
systematic-and-multilevel-approach-fabio-passos/
https://textbookfull.com/product/applications-of-organic-and-
printed-electronics-a-technology-enabled-revolution-integrated-
circuits-and-systems-2013th-edition-eugenio-cantatore/
https://textbookfull.com/product/circadian-rhythms-for-future-
resilient-electronic-systems-accelerated-active-self-healing-for-
integrated-circuits-xinfei-guo/
Massimo Alioto Editor
Enabling the
Internet of
Things
From Integrated Circuits to Integrated Systems
Enabling the Internet of Things
Massimo Alioto
Editor
And to everyone who has inspired my curiosity and love for life,
including Giusi, Rodolfo, Alfio, Dora, Gaetano, Annamaria
and Santina. And thankfully many, many others.
Preface
vii
viii Preface
call the “Internet of Things” (IoT). The IoT can evidently create a huge value
and bring unprecedented benefits to the society. To set this on a trend
perspective, we can extrapolate Hick’s law to artificial intelligence and
cloud computing: more physical data will enable us to take more automated
decisions with an effort that is only logarithmic in the space of decision
choice. The IoT is ultimately a powerful enabler to share on a larger scale,
make technology more human centric and real time, and decouple socioeco-
nomic progress from intensive use of resources. And, interestingly, IoT
silicon technology becomes so small that the user is immersed in it (there
is no more “user experience,” in a sense), with interesting implications in
terms of market and perceived value.
In spite of the daily IoT-related claims in the chip design community, the
tiny sensing nodes of the IoT at its edge (the “IoT nodes”) are still in their
technological infancy. Several challenges need to be tackled, such as energy
efficiency and related lifetime, cost, security, and interoperability, among
others. Such challenges need to be tackled in a holistic manner, developing
both an understanding of the different parts of IoT nodes and an insight into
the big picture and the strong linkage to applications and related
requirements.
To the best of our knowledge, this is the first book on integrated circuit
and system design for the Internet of Things. This book develops in both the
“vertical” and the “horizontal” dimension. Vertically, it provides a compre-
hensive view on the challenges and the solutions to successfully design chips
for IoT nodes as systems (from circuits to packages), a broad analysis of how
chip design needs to evolve to meet those challenges, and a fresh perspective
grounded on historical and recent trends. Horizontally, the book covers in
one place the very diverse domain-specific expertise of the subareas involved
in the design of IoT nodes, which was previously scattered across a large
number of talks, journals, and conferences.
This book provides a design-centric perspective, providing an understand-
ing of what the IoT really means from a design point of view. Typical
specifications of commercial IoT nodes are discussed, and constraints
imposed by IoT applications are translated into design constraints that chip
designers are used to deal with. Design guidelines to meet them are system-
atically discussed in every chapter.
This book started in the form of talks at various venues, such as VLSI
Symposium, HotChips, and ISCAS, where I had very interesting
conversations with several other speakers. Those talks were motivated by
the lack of a cohesive and detailed source of accessible knowledge on the
design of IoT nodes. The idea to write this book came exactly from those
conversations, which later continued throughout the interaction with chapter
authors. They really made this book possible, providing their deep insights
and invaluable expertise. I deeply thank all outstanding researchers and
designers who contributed to the chapters of this book, sharing their expertise
in an accessible and concise manner for the benefit of our community.
Preface ix
xi
xii Contents
xiii
xiv About the Editor
This chapter opens the book and provides a sum- its pervasive networking, cloud and related
mary of the challenges and the opportunities that advances (e.g., big data), the physical world
are offered by the Internet of Things (IoT), with through distributed sensing and people’s
emphasis on the aspects that are relevant to activities, in the unprecedented form of mostly
integrated circuit and system design from circuits real-time fine-grain and aggregated data from the
to packaging for IoT nodes. The chapter is knowledge coming from environments, goods,
organized along a chronological perspective, resources, tools, infrastructures, among the others.
first reviewing technology historical trends So far, the IoT has been defined in several
beyond mere Moore’s law, and summarizing different ways, and its meaning has become so
recent past achievements and capabilities that broad that it oftentimes includes any object on
are making the IoT possible. Then, present earth that is connected to the Internet, such as
challenges are described, as pathway to connected cars, drones, smartphones, smart
up-coming advances and developments in the appliances, industrial tools, and so on. Under
design of IoT nodes. Finally, mega-trends are such generic definition based on pure Internet
examined to unearth clues on longer-term evolu- connectivity, the IoT has been already realized
tion of the IoT and the implications on integrated as the number of computing devices connected to
system design. the Internet surpassed the worldwide population
back in 2008–2009 (Evans et al. 2011).
This book focuses on the IoT as pervasive,
1.1 The Internet of Things: Context unobtrusive, systematic and coordinated intro-
and Overview duction of sense-, compute-, communication-
ability and sensemaking of physical data in a
The concept of the IoT seems to first appear in very large number of objects on earth. This is
Kevin Ashton in a presentation delivered at enabled by the introduction of extremely
Procter & Gamble in 1999 (Ashton 2009), which miniaturized integrated systems (“IoT nodes”)
was then described as a large-scale network of with very long lifetime (e.g., decades) that are
smart RFIDs. On a broad perspective, the IoT autonomous in many respects, from functional-
lies at the intersection of the Internet realm with ity, to energy, to the way they interact with the
physical world and the network infrastructure.
M. Alioto (*) From this perspective, the IoT pushes such
National University of Singapore, Singapore 117583, capabilities beyond personal devices (e.g.,
Singapore smartphones), embedding them in everyday
e-mail: malioto@ieee.org
objects and living environments. This book predict a somewhat slower growth (Nordrum
addresses the challenges involved in the creation 2016). The IoT market size is expected to have a
of IoT nodes in the form of integrated circuits, global economic impact of 2.5–11.1 T$ by
covering the different areas involved in this pro- 2022–2025 (Dobbs et al. 2015; Jankowski et al.
cess including architecture, circuit building 2014).
blocks, design methodologies, packaging and As shown in the simplified architecture in
system demonstrations. Being the IoT an exten- Fig. 1.1, the IoT is structured into three tiers of
sive topic, the scope of this book purposely devices. At the bottom, IoT nodes perform sensing
excludes the challenges related to the integration and interact with the physical world. To assure
of IoT nodes into a cohesive and scalable net- scalability and ubiquitous network access, gate-
work comprising inter-operable and heteroge- ways and concentrators collect, protect (under
neous nodes, and related communication users’ control) and route data from several and
protocol and software layers. physically proximal IoT nodes, and route it to
A commonly agreed target of the IoT is to servers. The latter perform data aggregation and
expand the number of connected devices per per- knowledge extraction, and deliver physically-
son to the order of a thousand, thus reaching an enhanced cloud services. Some additional inter-
unprecedented scale of trillions of connected mediate levels of aggregation might be needed,
devices (Gaudin 2015). The number of depending on the amount of data generated, the
connected devices is expected to grow to 30–50 area covered by a sub-network, and the density of
billion devices by 2020, with an expected market IoT nodes, among the others. For example,
CAGR growth of 15–35% (Markets and Markets; concentrators might actually be a sub-set of the
https://newsroom.cisco.com/press-release-content network below an Internet hub/gateway, which is
?type¼press-release&articleId¼1771211; Ericss here omitted as this would be simply part of the
on Mobility Report; http://www.gartner.com/ existing Internet infrastructure.
newsroom/id/3165317; Greenough and Camhi The hardware requirements of the devices in
2015; Worldwide Internet of Things Forecast the three tiers in Fig. 1.1 are very different, by
2015; TechNavio 2015; Machina Research 2015; virtue of their significantly different number and
Bauer et al. 2014; Jankowski et al. 2014; Dobbs level of pervasiveness. The number of IoT nodes
et al. 2015; Digital Universe of Opportunities is expected to be approximately two orders of
2014), [IoT Analytics, Oct. 2014], (http://www. magnitude larger than the number of con-
postscapes.com/internet-of-things-market-size/). centrators, which in turn is plausibly higher than
Some forecasts question such fast growth and the number of server blades by another two orders
# DEVICES WORLDWIDE
size
cost/item USERS compute- communication- sense-
power ability ability ability status
100 Millions
illions 100-1,000
meterss GFLOPs well deployed
CLOUD 10,000 $
1,000-10,000 keeps expanding
1,000 W
10 Billions
ions 100-1,000
INTERNET GATEWAYS/ 10 cm MFLOPS well deployed
OF THINGS CONCENTRATORS 10+ $ keeps expanding
1 -10 W
1 Trillion
on 1 -100 commercial nodes
1 -10 mm MFLOPS still far from IoT
IoT NODES
1$ target
000 mW
0.1-1,000
PHYSICAL
WORLD
of RFID identifiers (Atzori et al. 2010). Finally, buildings and nations, toys, worksites, smart
the Internet as we know it today was mostly infrastructures, energy, lifestyle/entertainment,
designed for non-real-time sharing of documents among the others.
and data, with resiliency being the main concern As opposed to previous technological waves
(Greenemeier et al. 2013). Due to the generation in the semiconductor history, the IoT is the first
of large amounts of real-time data, the IoT pushes one that is so pervasive that it becomes invisible
the Internet towards its limit and hence needs to to the users, with several implications on the
be structured in a more decentralized manner to value capturing in the semiconductor industry.
assure sustainable scalability. For example, only 5–10% of the IoT technology
The above issues related to the IoT data del- spending is expected to fuel the semiconductor
uge are drastically mitigated by moving intelli- industry market (Dobbs et al. 2015), whereas
gence from the cloud to the concentrators and more value (15–20% each) will be captured by
most importantly to the IoT nodes in Fig. 1.1, software and integration services. Plausibly,
i.e. making the IoT nodes “smarter” most of the value of the IoT will come from the
(or “cognitive”, if intelligence means ability to data aggregation and the real-time response
detect and classify patterns) than they are today. (or actuation) of cloud services, as well as the
Indeed, pre-processing in the IoT nodes and more demand prediction for new proactive services
distributed intelligence reduce the data volume, and tasks that no longer need us to “push a
as only partially aggregated data needs to be sent button” (or click a mouse) to be executed. To
over the network, as opposed to raw data. capture more value from the large market volume
and by delivering integration services (e.g., from
IoT nodes to software for data aggregation and
1.2 Brief Review of IoT sensemaking), semiconductor companies will
Applications likely become more vertically integrated through
acquisitions, close partnerships and industrial
1.2.1 Considerations on the IoT consortia. As further benefit, this trend will also
Market Volume favor IoT node inter-operability and
standardization.
The IoT as a whole is inherently a general-
purpose technology, similarly to computers and
mobile devices in the past decades. Like any 1.2.2 Summary of Current
other general-purpose technology, it can boost and Prospective Applications
true productivity and create a value that is sub- of the IoT
stantially higher than its market size, as it can
serve as catalyst for bigger change (Brynjolfsson The IoT is a very fragmented application sce-
and Hitt 1998). Indeed, the IoT can further nario (Vermesan and Friess 2014), and
improve efficiency, economy of scale, ability to encompasses a wide range of applications, some
react to and predict demand in capex, labor and of which are summarized in the following.
energy. Also, the IoT is expected to enable better In the agriculture sector, the IoT infrastructure
coordination and usage monitoring of buildings, can monitor the quality, the actual usage and the
machinery, manufacturing processes, factories, availability of resources, for better and predictive
supply chain and resources. The IoT will impact management (e.g., irrigation) and storage (e.g.,
a very wide diversity of applications, from agri- avoid waste of feed and fertilizing). Monitoring
culture to consumer products, automotive, the environmental conditions permits to support
healthcare, retail, manufacturing and supply the growth of animals and plants (e.g., aquacul-
chains (e.g., Industry 4.0), telecommunications, ture), optimally time the next course of action,
logistics, public sector, financial, transportation and ultimately assure quality (e.g., wine) and
and shipping, smart environments from homes to raise the efficiency in the production process.
1 IoT: Bird’s Eye View, Megatrends and Perspectives 5
In automotive, the IoT enables the monitoring objects for personal care and hygiene can be
of the state of a vehicle down to its critical used to remind of regular but infrequent care
components, from initial shipping to usage, to activities based on dentist’s suggestions shared
assess their correct utilization (e.g., detecting in the cloud, and motivate positive behavior in
bumps, vibration) and maintenance (e.g., open- children. Smart clothing can remind of periodic
ing of containers, wearing parts). Based on actual cleaning based on actual usage. Smart toys can
usage, predictive maintenance can be performed be selectively enabled only upon the occurrence
to lengthen the vehicle lifetime, and lower the of desired conditions to create positive habits
upkeep. Such capabilities enabled by the IoT are (e.g., only at certain times or lighting conditions),
also very useful in fleet management and car and prevent danger by disabling them under the
sharing services. Also, distributed sensing and presence of others (e.g., toddlers). Smart jewelry
global sensemaking enables traffic control can be used to unobtrusively track activity, mea-
through differentiated and personalized road sure exposition to solar light and other environ-
pricing to encourage virtuous behavior and pri- mental conditions, and make emergency calls.
oritize tasks for commercial (e.g., car pooling Energy management at different scales can be
with multiple passengers sharing cost) and pri- made more effective by the IoT. At the city scale,
vate vehicles (e.g., fast delivery for critical the smart grid offers several opportunities to
goods), through virtual/dynamic city area leverage the sensing and sensemaking
boundaries. capabilities of the IoT to optimize the energy
In public transportation, the occupancy and usage across many users, a better coordinated
utilization can be monitored to assure an ade- usage and planning of alternative energy sources,
quate quality of service, detect potential danger ultimately reducing the overall energy and the
(e.g., potential collision between vehicles and currently large gap between the peak and the
pedestrians), and predict short term demand average consumption.
based on crowd monitoring in strategic locations. Health care is another important application
On the road side, excessive congestion and pol- area in which the IoT promises to fundamentally
lution can be managed with real-time demand- contribute to. As few examples, the miniaturiza-
response schemes where the road pricing is tion and long lifetime of IoT nodes provides an
dynamically adjusted through real-time obser- unobtrusive mean to constantly monitor vital
vations and utilization prediction, based on signs and other related parameters (e.g., behav-
previous history and real-time data in strategic ioral) and develop deeper understanding of the
locations. Also, the transportation of dangerous patient’s health evolution. In addition, the avail-
goods and the circulation of slow (or frequently ability of big data from a large number of patients
stationary) vehicles can be optimally coordinated offers an unprecedented opportunity to explore
with the ordinary traffic to minimize their nega- correlations, build models and tools for predic-
tive impact. Again, the IoT offers unprecedented tive diagnosis, early treatment and make drug
opportunities to share resources efficiently, while discovery more efficient and effective. Similar
preserving their running condition. considerations hold for the elderly and the
Consumer electronics substantially benefits disabled, as constant non-obtrusive monitoring
from the IoT, as its pervasiveness permits to allows for better and highly responsive/predic-
track smart goods (e.g., positioning systems for tive care, while preserving individual’s indepen-
object retrieval) and detect their exposure to dency and offloading hospitals. Remote
anomaly conditions (e.g., overheating, physical supervision also enhances the ability to share
shocks). IoT sensing can signal spatial professionals across a larger number of
co-presence of objects and specific people (e.g., individuals and patients, thus driving the care
kids) to signal potential danger, or to recommend cost down.
activities to complete when all necessary objects Industrial processes and logistics can also
tools are available in the same space. Smart highly benefit from the IoT, as it can enable
6 M. Alioto
ubiquitous sensing of operating conditions, real- Smart homes promise to automatically order
time tracking of semifinished products, detection supplies before exhaustion and upon the avail-
of events that slow down the process throughput ability of online offers, as a major step forward
and potential safety issues. The data generated in compared to the today’s Amazon Dash wireless
the production line can be intelligently shared button that simply orders goods online when the
with the quality assurance process and across button is pushed (Amazon Dash). At the same
different sites, to raise the yield and reduce time, waste management will be made more effi-
cost. On the warehouse side, product location cient by sensing the actual demand, and pricing
and storage conditions can be tracked for more based on actual consumption habits, thus encour-
efficient product delivery and distribution. Simi- aging virtuous behavior. The IoT can also make
larly, sharing real-time and historical data on residential compost recycling easier and
parts with the procurement process makes automated, through the monitoring of humidity
restocking more efficient, and reduces the inven- and temperature trends.
tory cost through more strategic purchasing Smart buildings can leverage the IoT to be
strategies. The presence of IoT within machines more adaptive to the actual demand and needs
can enable early and self-diagnosis, predictive of the occupants, while ensuring the highest
(rather than reactive or pre-scheduled) mainte- safety and comfort standards. Indeed, air quality
nance, pre-emptive vendor support to prevent and thermal/acoustic/visual comfort can be mon-
known failures. Again, the IoT enables better itored and controlled for the first time with a
economy of scale, efficiency and makes pro- granularity that goes down to the single room,
cesses leaner. with obvious advantages in terms of comfort
In the area of retail, smart malls can provide assurance and energy cost. Beyond normal build-
real-time shopping recommendations, matching ing operation, the real-time capability of the IoT
available offers with individual customers, dis- enables the ability to respond to critical events
card products for potential customers with (e.g., fire) quickly, minimizing the human and
allergy issues and provide other personalized material losses in case of emergencies.
services (e.g., for customer fidelization). The Through the IoT, smart cities can manage
tight coupling between the individual and collec- resources more efficiently, be made much more
tive customers’ behavior, the store setting and resilient to temporary malfunctions and disasters,
the warehouse permits to streamline the inven- and encourage virtuous behavior. Smart and
tory management, offer better shopping experi- weather-adapting lighting, water/gas leakage
ence, dynamically adjust in-store display based monitoring, smart parking with dynamic pricing
on the predicted demand, and cut inventory costs. and area allocation, no physical boundaries and
Through the IoT, smart homes can manage automated parking advice are just a few
utilities more efficiently by controlling individ- examples of how to use the IoT to solve today’s
ual appliances based on actual utilization and urban challenges. Ubiquitous vision can enable
needs, and purchasing electricity when cost is an unprecedented level of safety and security,
lowest within the day in demand-response energy detecting potential danger and provide crucial
pricing schemes. Unprecedented levels of secu- information on crowd behavior and citizens’
rity (e.g., perimeter access control) are achiev- needs (e.g., for adaptive and predictive transpor-
able thanks to the pervasiveness of IoT nodes and tation management, real-time digital signage
sensemaking ability. Occupant recognition recommendations to prevent immediate danger).
permits to adjust lighting, sound, air condition- Other than enabling ubiquitous and augmented
ing/heating based on individual preferences. This surveillance, vision in IoT offers physical aug-
can be done in a predictive manner, so that mentation to social media and recommendation
occupants do not need to “push any button”, systems (e.g., venue recommendation based on
leveraging the fine-grain knowledge of crowdedness, and crow sentiment), and human
occupants’ habits and the ability of the cloud to activity monitoring to achieve better match
generalize and extract trends and predictions. between demand and supply of services
1 IoT: Bird’s Eye View, Megatrends and Perspectives 7
1E+12
1E+10
10 years
1E+8
1 year
lifeme (s)
1E+6 1 week
1E+4 1 hour
1E+2 1 minute
1E+0
1E-2
nW mW mW
1E-9 1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0
power (W)
smartwatch battery button cell battery thin-film battery
Fig. 1.3 Lifetime vs. average power consumption for different batteries
Larger Pavg mandate the addition of an energy to deliver the peak power, if the former does not
harvester, whose size is generally proportional to have adequate instantaneous power capability, as
Pavg. Figure 1.4 shows the harvester size required dictated by its size.
for a given Pavg for various energy sources. From As opposed to purely battery-powered systems,
this figure, millimeter-sized photovoltaic energy harvested IoT nodes can operate nearly-
(indoor), thermo-electric (on-body patch) and perpetually, as long as the harvester power
airflow (indoor) harvesters can indefinitely sus- exceeds Pavg (i.e., the harvester size is large
tain Pavg in the order of μWs (Alioto 2015). Tens enough), and can hence indefinitely sustain the
of μWs are sustainable under more abundant power required by the IoT node. On the other
energy sources, such as photovoltaic (outdoor), hand, an increase in the targeted lifetime tlifetime
thermo-electric (industrial machines) and body ¼ Ebattery =Pavg for a given Pavg requires the adop-
vibration (e.g., walking) harvesting (Alioto tion of proportionally larger batteries. Hence,
2015). GSM radio-frequency energy harvesting energy harvesters are invariably more compact
can instead sustain only tens to very few than batteries for long enough lifetime targets.
hundreds of nWs. From Fig. 1.3, printed and Figure 1.4 shows the breakeven lifetime at which
solid-state batteries (see Chap. 15) enable harvester and battery have the same size, assum-
aggressive miniaturization at the cost of much ing a battery with energy density equal to typical
shorter lifetime, whose extension requires the alkaline button cell batteries (e.g., LR44). From
addition of an energy harvester in all practical this figure, harvesting is always more compact for
cases. As a third energy source option, the battery all practical lifetime targets under abundant
can be suppressed altogether by pairing the energy sources, such as photovoltaic (outdoor),
energy harvester with a small energy source thermo-electric (industrial machines) and body
(e.g., off-chip supercapacitor, on-chip capacitor) vibration (e.g., walking) (Alioto 2015). On the
1 IoT: Bird’s Eye View, Megatrends and Perspectives 9
IoT node
1
size target
0.1
0.1 1 10 100 1000 10000
avg power Pavg (mW)
photovoltaic (indoor) photovoltaic (outoor)
thermal (human) thermal (industrial)
vibration/motion (human) air flow (indoor)
electromagnetic (GSM)
other hand, harvesters become more compact than other living environments and infrastructures,
batteries for targeted lifetimes of 2–3 years and this translates into a lifetime of several decades.
longer, and hence in most of IoT applications. Industrial applications, transportation and
GSM radio-frequency harvesters are instead shipping might require a shorter lifetime,
always larger sized than the battery counterpart. although still in the order of a decade. The life-
Regardless of the specific lifetime target and time requirement can be further relaxed in other
energy source architecture, the volume of IoT applications such as retail, worksites, lifestyle/
nodes is certainly dominated by off-chip entertainment. Hence, the above considerations
components, and in particular by the energy on the power budget of IoT nodes apply almost
source, as the antenna can be made very thin. In unmodified in a very wide range of applications.
other words, the size of IoT nodes is essentially Meeting power budgets of few μWs or below is
set by their power consumption, which hence feasible only if the IoT node actively performs
become a very stringent and crucial requirement tasks (e.g., sensing, processing) only infrequently.
in any IoT node design. In other words, power needs to be aggressively
reduced by duty cycling the IoT node operation,
alternating active tasks and long sleep periods as
depicted in Fig. 1.5, with periodicity set by the
1.3.2 Interaction with the External
wake-up cycle Twkup. From an architectural stand-
World
point, this means that IoT nodes are organized into
an always-on (ALWON) sub-system that manages
With reference to Fig. 1.2, the interaction of IoT
the periodicity of the wake-up cycle and stores
nodes with the external world needs to last at
information across active tasks, and a duty-cycled
least the lifespan of the object/environment they
(DCYC) sub-system that periodically performs the
are embedded in, as battery replacement is not an
active task (Alioto 2012). Hence, the average IoT
option due to the large number or the inaccessi-
node power can be written as the sum of the
bility of nodes. When deployed in buildings or
10 M. Alioto
Twkup
ALWON power PALWON and the DCYC energy designed resolution (and hence higher cost and
EDCYC (Alioto 2012): power). From Fig. 1.6, most of IoT applications
require a minimum resolution that is below
EDCYC
Pavg ¼ PALWON þ : ð1:1Þ 12 bits, and 8 bits are sufficient for a rather
T wkup wide range of practical cases. On the secondary
From Eq. (1.1), the power reduction can be y axis, the figure also reports the energy per
reduced by reducing the power (energy) of the conversion, assuming an energy per conversion
always-on (duty-cycled) sub-system. In other step of 30 fJ, which is relatively optimistic espe-
words, nearly-minimum power design needs to cially for larger resolutions (Murmann 1997).
be pursued in the always-on sub-system, while The datarate range of the above sensors is plot-
nearly-minimum energy design is the objective ted in Fig. 1.7. This figure shows that most of the
in the duty-cycled one (see Chap. 4). Of course, sensors require only thousands of bits per second
larger Twkup and hence more infrequent active when operating continuously, whereas tasks
operation mitigates power, although Twkup is related to vision and audio processing need orders
upper bounded by the application, depending on of magnitude higher datarates (up to 10 Mbps in
how frequently data needs to be updated. Such the case of compressed VGA video streaming).
system-level tradeoffs are discussed in Chap. 2, From the above considerations, the specifications
whereas approaches to further reduce the leakage of IoT node sensing interfaces are actually quite
power cost of storing information across tasks is relaxed, thus cost and power consumption are far
discussed in Chaps. 5–7. more important aspects than pure performance.
Both challenges are well addressed by tailoring
such circuits around the specific application. The
power consumption of the ADC is proportional to
1.3.3 On-Board Capabilities of IoT the datarate in Fig. 1.7 and the energy per conver-
Nodes sion in Fig. 1.6, and is plotted in Fig. 1.8. From this
figure, the power consumption of ADCs for IoT
IoT nodes need to have sensing, computation, nodes spans a very wide range, mostly because of
and wireless communication capabilities. In IoT the wide energy per conversion range in Fig. 1.6,
nodes design for a specific purpose, sensing can as dictated by the exponential relationship between
be typically made more inexpensive by tailoring resolution and energy (Freyman et al. 2014). This
the MEMS design and the analog interface confirms that tailoring the ADC to the specific
around the specific application. This permits to application is crucial in IoT, and the same consid-
substantially reduce the complexity that is expe- eration applies to most of the other building blocks
rienced by general-purpose platforms, and hence and sub-systems.
the cost. As simple example, Fig. 1.6 shows that Let us now consider the case where the raw
sensors for IoT applications cover a wide range sensor data is transferred directly to
of resolutions, hence using the appropriate ADC concentrators and cloud. Assuming a best-in-
resolution (see Chap. 13) is necessary to avoid class radio consuming 5 nJ/bit (ISSCC 2016),
using general-purpose platforms with over- the resulting power to wirelessly transmit such
1 IoT: Bird’s Eye View, Megatrends and Perspectives 11
1E+6 5E-3
1E+4 5E-5
1 mW
1E+2 5E-7
1E+0
sensor
heart rate
humidity (capacitive)
battery monitor
temperature
accelerometer
magnetometer
altimeter/pressure
imager (VGA, RGB)
imager (MP4 compressed)
infrared proximity
gyroscope
microphone
CO2
light
strain
ultra-violet
Fig. 1.7 Datarate range required by various sensors, and wireless power required to continuously transmit data
(energy/bit assumed to be 5 nJ/bit (ISSCC 2016))
12 M. Alioto
raw data is reported on the secondary y axis in communication. For example, the IoT node can
Fig. 1.7. From this figure, most of sensors cer- be proactive and monitor for critical or important
tainly exceed 1 μW and hence the range of prac- events (e.g., the crossing of a threshold, or an
tical IoT node power targets mentioned above. increase rate larger than a pre-set value), and
Hence, mere computation offloading to transmit data only upon their occurrence. From
concentrators and cloud through raw data trans- Fig. 1.7, this is particularly crucial in applications
mission is not an option for IoT nodes operating involving large datarates, such as continuous
continuously. vision and audio sensing. In such applications,
For some environmental sensors (e.g., temper- more intelligence needs to be embedded in the
ature, CO2, light, UV), duty cycling discussed in IoT node, such as the ability to perform pattern
the previous subsection is applicable since recognition and classification. Other options to
measurements do not need to be taken continu- trade off computation and communication are in
ously, as the related phenomena exhibit slower the choice of the data representation and sampling
time constants. For such sensors, a duty cycle of approach (e.g., compressive sensing, including
percentage points reduces the average datarate computation in the compressive sensing domain
down to hundreds of tens of bits/second, and the (Shoaib et al. 2015)), as well as signal dimension-
power down to tens of nWs. Often times, the other ality reduction (e.g., in-node feature extraction,
sensors cannot be duty cycled as the dynamics of which is equivalent to compression, with the fur-
the related phenomenon does not really allow it ther advantage that it is often a necessary task to
(e.g., accelerometers, gyroscopes, imaging, be performed anyway in many algorithms).
audio). In these cases, further power reduction From the above considerations, the wireless
can be achieved by leveraging the well-known power is always an issue in IoT nodes, and hence
computation-communication tradeoff (Min et al. requires the choice of appropriate communica-
2001), moving computation onto the IoT node tion standard for the intended range and datarate,
to reduce the volume of wireless data as will be discussed in the next section.
1 IoT: Bird’s Eye View, Megatrends and Perspectives 13
Fig. 1.9 Scaling laws of computer market size and system volume
of mobile computing, the next technological concept of “learning curve” for the silicon
wave is expectedly composed of systems with a manufacturing, as plotted in Fig. 1.11. In general,
volume of hundreds of mm3 or lower, i.e. the the learning curve comes from the observation
IoT. Essentially, the last few decades have seen that the doubling in the large-scale cumulative
computers and integrated systems move from production typically leads to a fixed reduced cost
desks to pockets, and then anywhere. per unit at any point in time (Jaber 2011). The
The energy efficiency trend for computers and semiconductor industry is only one of the many
Digital Signal Processors is well captured by the examples of steady (exponential) cost reduction
Koomey’s (Koomey et al. 2011) and Gene’s law that comes from the accumulated knowledge and
(Frantz et al. 2000; Karam et al. 2009), which are improvements in the overall design/production
representative of control-heavy and data-heavy process. In particular, Fig. 1.11 shows that the
architectures. As summarized in Fig. 1.10a–b, transistor manufacturing cost fits a 55% learning
both classes of computing have equally benefited curve, as doubling the overall number of
from technology advances, which can be manufactured transistors reduces the cost to 55%
expected to hold true in the future as well. Histor- of the original cost. Such relentless cost reduction
ically, most of 100 energy reduction achieved is at the basis of the Moore’s law (Jovanovic and
every 10 years has been obtained through tech- Rousseau 2002), which has driven the learning
nology scaling (90% according to (International curve of transistor cost down relying on a steady
Technology Roadmap for Semiconductors (exponential) doubling of transistors per die at
2015)). In the IoT domain, energy efficiency each CMOS generation, assuming the cost reduc-
improvements will not come from technology tion comes from shrinking (Hutcheson 2009)
scaling, as advanced CMOS technologies are (which is no longer true at 20 nm and below due
simply too expensive for the very low cost target to higher lithography cost, thus making those
of IoT nodes, as discussed below. Hence, keeping CMOS generations unsuitable for low-cost IoT
the same pace in the energy reduction requires nodes, as discussed in Sect. 1.8). The cost per
major innovation at system level (Chap. 2), in key transistor has been constantly reduced through
building blocks (Chaps. 3–7, 10–14), through shrinking in the past decades, although this is
innovative design methodologies (Chap. 9), and clearly not the only way to do it. In other words,
suitable batteries (Chap. 15). the learning curve in Fig. 1.11 actually transcends
Regarding the cost of IoT nodes, and in partic- Moore’s law, and is expected to continue for a
ular the cost per transistor, it is useful to recall the long time in spite of the end of the latter (Rhines
1 IoT: Bird’s Eye View, Megatrends and Perspectives 15
(a)
100
(mW/MIPS=nJ/instrucon)
energy per sample
10
0.1
0.01
1970 1980 1990 2000 2010 2020
year
(b)
16 M. Alioto
1E-3
1E-6
1E-9
1E+09 1E+12 1E+15 1E+18 1E+21
cumulative no. of manufactured transistors
2015), leveraging on the accumulated knowledge replenishment. Circuits to monitor the state of
of semiconductor industry, and various other the battery are included to adapt the power man-
approaches described in Sect. 1.8. agement strategy to the actual energy availability
More considerations on trends in software and (e.g., brown-out detector). Specialized circuitry
communication infrastructure will be made in is embedded for the generation of proper multi-
Sect. 1.8. ple voltages from the single (and not perfectly
stable) voltage coming from the battery. Voltage
booster and battery recharger are inserted
1.5 Looking at the Present: Typical between the energy harvesters and the battery.
Specifications of IoT Nodes A Power Management Unit (PMU) coordinates
the interaction between the energy source, the
1.5.1 Architecture of IoT Nodes harvester and the power mode of the node.
The task performed in active mode (see
A relatively general architecture of IoT nodes is Sect. 1.3.2) is started by a periodic trigger
depicted in Fig. 1.12 (covered in Chap. 2), which (time-driven) or specific events (event-driven),
details its main sub-systems, including depending on the power mode and the specific
processing and security assurance (Chaps. 3–9), application. Time-driven control is determined
power conversion and delivery (Chaps. 10 and by deriving a proper clock from the system
11), analog interfaces (Chaps. 12 and 13), radios clock, whose frequency is typically 32,768 Hz
(Chap. 14), energy sources (Chap. 15), system to serve as Real-Time Clock and hence perform
integration and assembly (Chap. 16). Two accurate timestamping and inter-node synchroni-
examples of detailed architectures of IoT nodes zation. Event-driven control is achieved by con-
are provided in Chaps. 17 and 18. In this section, stantly monitoring digital and analog signals of
a review of the main features of IoT node interest, respectively through digital transition
architectures is provided. detectors and analog comparators that detect the
As in Fig. 1.12, sensors are connected to ana- crossing of a meaningful threshold to signal an
log interfaces that include an amplifier with pro- event occurrence. When more sophisticated com-
grammable gain (and sometimes analog filters), parison is needed for signals coming from
and are multiplexed to share a single ADC for all sensors, the ADC and processing are kept on to
analog channels. Analog voltages are generated acquire samples and process them until the event
by a DAC for actuation. Energy interfaces of interest is detected.
involve an energy storage element (e.g., battery, In Fig. 1.12, memories are an important part
supercapacitor), and energy harvesters for energy of IoT node processing. RAM is needed for the
1 IoT: Bird’s Eye View, Megatrends and Perspectives 17
digital interface
for configuration
SPI
settings
Non-Volatile
Memory (Flash)
analog triggers
EVENT
Power
Management
Unit (PMU) WIRELESS
COMMUNICATION
other supply DC-DC radio-frequency
voltages conversion RX/TX
ENERGY
system BUS
INTERFACES battery monitor
& brown-out
detection
general-purpose
booster for digital IOs
harvesting &
energy
harvester battery charger
I, V ANALOG
references INTERFACES
general-purpose
DAC analog IOs
sensors ADC
Fig. 1.12 Relatively general architecture of IoT nodes with detailed sub-systems
Language: English
A N D T H E I R D AY S
BY DR. DORAN
AUTHOR OF “LIVES OF THE QUEENS OF ENGLAND OF THE HOUSE OF HANOVER,”
“TABLE TRAITS,” “HABITS AND MEN,” ETC.
REDFIELD
3 4 B E E K M A N S T R E E T, N E W Y O R K
1856
TO
Is inscribed
BY HIS VERY SINCERE FRIEND,
THE AUTHOR.
CONTENTS.
PAGE
A FRAGMENTARY PROLOGUE
9
THE TRAINING OF PAGES 30
KNIGHTS AT HOME 36
LOVE IN CHEVALIERS, AND CHEVALIERS IN LOVE 51
DUELLING, DEATH, AND BURIAL 65
THE KNIGHTS WHO “GREW TIRED OF IT” 78
FEMALE KNIGHTS AND JEANNE DARC 104
THE CHAMPIONS OF CHRISTENDOM 113
SIR GUY OF WARWICK, AND WHAT BEFELL HIM 133
GARTERIANA 148
FOREIGN KNIGHTS OF THE GARTER 170
THE POOR KNIGHTS OF WINDSOR, AND THEIR DOINGS 184
THE KNIGHTS OF THE SAINTE AMPOULE 194
THE ORDER OF THE HOLY GHOST 200
JACQUES DE LELAING 208
THE FORTUNES OF A KNIGHTLY FAMILY 228
THE RECORD OF RAMBOUILLET 263
SIR JOHN FALSTAFF 276
STAGE KNIGHTS 295
STAGE LADIES, AND THE ROMANCE OF HISTORY 312
THE KINGS OF ENGLAND AS KNIGHTS; FROM THE NORMANS TO
THE
STUARTS 329
“THE INSTITUTION OF A GENTLEMAN” 351
THE KINGS OF ENGLAND AS KNIGHTS; THE STUARTS 358
THE SPANISH MATCH 364
THE KINGS OF ENGLAND AS KNIGHTS; FROM STUART TO
375
BRUNSWICK
RECIPIENTS OF KNIGHTHOOD 388
RICHARD CARR, PAGE, AND GUY FAUX, ESQUIRE 410
ULRICH VON HUTTEN 420
SHAM KNIGHTS 439
PIECES OF ARMOR 455
THE
Dr. Lingard, when adverting to the sons of Henry II., and their
knightly practices, remarks that although chivalry was considered the
school of honor and probity, there was not overmuch of those or of
any other virtues to be found among the members of the chivalrous
orders. He names the vices that were more common, as he thinks,
and probably with some justice. Hallam, on the other hand, looks on
the institution of chivalry as the best school of moral discipline in the
Middle Ages: and as the great and influential source of human
improvement. “It preserved,” he says, “an exquisite sense of honor,
which in its results worked as great effects as either of the powerful
spirits of liberty and religion, which have given a predominant
impulse to the moral sentiments and energies of mankind.”
The custom of receiving arms at the age of manhood is supposed,
by the same author, to have been established among the nations
that overthrew the Roman Empire; and he cites the familiar passage
from Tacitus, descriptive of this custom among the Germans. At first,
little but bodily strength seems to have been required on the part of
the candidate. The qualifications and the forms of investiture
changed or improved with the times.
In a general sense, chivalry, according to Hallam, may be referred to
the age of Charlemagne, when the Caballarii, or horsemen, became
the distinctive appellation of those feudal tenants and allodial
proprietors who were bound to serve on horseback. When these
were equipped and formally appointed to their martial duties, they
were, in point of fact, knights, with so far more incentives to
distinction than modern soldiers, that each man depended on
himself, and not on the general body. Except in certain cases, the
individual has now but few chances of distinction; and knighthood, in
its solitary aspect, may be said to have been blown up by
gunpowder.
As examples of the true knightly spirit in ancient times, Mr. Hallam
cites Achilles, who had a supreme indifference for the question of
what side he fought upon, had a strong affection for a friend, and
looked at death calmly. I think Mr. Hallam over-rates the bully Greek
considerably. His instance of the Cid Ruy Diaz, as a perfect
specimen of what the modern knight ought to have been, is less to
be gainsaid.
In old times, as in later days, there were knights who acquired the
appellation by favor rather than service; or by a compelled rather
than a voluntary service. The old landholders, the Caballarii, or
Milites, as they came to be called, were landholders who followed
their lord to the field, by feudal obligation: paying their rent, or part of
it, by such service. The voluntary knights were those “younger
brothers,” perhaps, who sought to amend their indifferent fortunes by
joining the banner of some lord. These were not legally knights, but
they might win the honor by their prowess; and thus in arms, dress,
and title, the younger brother became the equal of the wealthy
landholders. He became even their superior, in one sense, for as Mr.
Hallam adds:—“The territorial knights became by degrees ashamed
of assuming a title which the others had won by merit, till they
themselves could challenge it by real desert.”
The connection of knighthood with feudal tenure was much
loosened, if it did not altogether disappear, by the Crusades. There
the knights were chiefly volunteers who served for pay: all feudal
service there was out of the question. Its connection with religion
was, on the other hand, much increased, particularly among the
Norman knights who had not hitherto, like the Anglo-Saxons, looked
upon chivalric investiture as necessarily a religious ceremony. The
crusaders made religious professors, at least, of all knights, and
never was one of these present at the reading of the gospel, without
holding the point of his sword toward the book, in testimony of his
desire to uphold what it taught by force of arms. From this time the
passage into knighthood was a solemn ceremony; the candidate was
belted, white-robed, and absolved after due confession, when his
sword was blessed, and Heaven was supposed to be its director.
With the love of God was combined love for the ladies. What was
implied was that the knight should display courtesy, gallantry, and
readiness to defend, wherever those services were required by
defenceless women. Where such was bounden duty—but many
knights did not so understand it—there was an increase of
refinement in society; and probably there is nothing overcharged in
the old ballad which tells us of a feast at Perceforest, where eight
hundred knights sat at a feast, each of them with a lady at his side,
eating off the same plate; the then fashionable sign of a refined
friendship, mingled with a spirit of gallantry. That the husbands
occasionally looked with uneasiness upon this arrangement, is
illustrated in the unreasonably jealous husband in the romance of
“Lancelot du Lac;” but, as the lady tells him, he had little right to cavil
at all, for it was an age since any knight had eaten with her off the
same plate.
Among the Romans the word virtue implied both virtue and valor—as
if bravery in a man were the same thing as virtue in a woman. It
certainly did not signify among Roman knights that a brave man was
necessarily virtuous. In more recent times the word gallantry has
been made also to take a double meaning, implying not only courage
in man, but his courtesy toward woman. Both in ancient and modern
times, however, the words, or their meanings, have been much
abused. At a more recent period, perhaps, gallantry was never better
illustrated than when in an encounter by hostile squadrons near
Cherbourg, the adverse factions stood still, on a knight, wearing the
colors of his mistress, advancing from the ranks of one party, and
challenging to single combat the cavalier in the opposite ranks who
was the most deeply in love with his mistress. There was no lack of
adversaries, and the amorous knights fell on one another with a fury
little akin to love.
A knight thus slain for his love was duly honored by his lady and
contemporaries. Thus we read in the history of Gyron le Courtois,
that the chivalric king so named, with his royal cousin Melyadus, a
knight, by way of equerry, and a maiden, went together in search of
the body of a chevalier who had fallen pour les beaux yeux of that