Autumn 2009

EE359 Project

Nima Soltani

Comparison of Single-Carrier FDMA vs. OFDMA as 3GPP Long-Term Evolution Uplink

Friday December 4, 2009 EE359 Project Nima Soltani

1. Introduction
In order to transition from today's 3rd generation (3G) communications systems to meet the needs of 4th generation (4G) systems, the 3rd Generation Partnership Project (3GPP) has released the Long Term Evolution (LTE) specification. Among the numerous differences between these generations are changes in the physical layer (PHY), specifically in the modulation and multiple access schemes. While its parent generation relied on variations of Code Division Multiple Access (CDMA), LTE implements Orthogonal Frequency Division Multiplexing (OFDM) for its downlink and Single-Carrier Frequency-Division Multiple Access (SC-FDMA) for its uplink. The purpose of this project is to investigate the reasoning for this discord between uplink and downlink modulation schemes; specifically, why Orthogonal Frequency Division Multiple-Access (OFDMA) was not used as the uplink. OFDMA and SC-FDMA are the multiple-access versions of OFDM and a similar modulation scheme, Single-Carrier Frequency-Domain Equalization (SC-FDE). In order to compare the differences between the multiple-access methods, it is important to first cover the differences between their underlying modulation schemes. Section 2 covers these differences, building the foundation for a comparison between the multiple-access methods themselves in Section 3. The bases used for comparison will be capacity, outage probability and peak-to-average power ratio (PAPR). While the first two bases have always been traditionally used in analyses, PAPR is especially important for the uplink of mobile devices. Amplifiers used in circuits today have a linear region in which they must operate so as not to introduce signal distortion, and it is ideal to run with maximum amplification. However, if there is a high PAPR, the device is forced to run with lower amplification so the peak power does not lie in the non-linear gain region. The farther these amplifiers are operated from the peak, the less power efficient the devices become, leading to increased power consumption and while this might not be very important for a base station, it will reduce drain batteries on mobile devices more quickly. Therefore it is important to keep a low PAPR on the uplink.

1.1

Simulation setup

The simulations in this report are conducted with values similar to that in the LTE specifications. Some of the assumptions made are that the carrier is 1950MHz which is in Band I, the sample rate is 30.72MHz, and the FFTs/IFFTs will be 2048-point ones, with cyclic prefixes of 160 samples [1]. In the multi-usercase we assume that each user will use 12 subcarriers, requiring additional 12-point FFT/IFFTs as explained in Section 3.2. Also, 7 symbols will be sent at a time, corresponding to a single resource-block in the LTE specification [1]. The simulated channel follows the multipath propagation conditions specified by the 3GPP in [2]. The code used to model the channel is based on one used by the IEEE 1

Autumn 2009

EE359 Project

Nima Soltani

802.16 Broadband Wireless Access Working Group [3], but modified it to pertain to the LTE propagation conditions by changing the delays, gains and Doppler shifts, as well as the Doppler spectrum to be the classical one assuming a uniform scattering environment [4]. The channel corresponds to a 2 Rayleighfading tap delay filter, specified in Table 1. There will also be a 5.4Hz Doppler shift as well, as the user is assumed to be moving at 3km/h in this case [2]. In this case, the coherence time is in the order of 0.1s, which equates to over 1000 OFDM symbol times. Thus in the simulations 7 symbols can be sent sequentially with high certainty that channel will not change for the duration of the burst.
Table 1. Channel filter parameters for simulations

Relative Delay (ns) 0 976

Relative Delay (samples) 0 30

Average Attenuation (dB) 0 10

2. Modulation Schemes
When facing a frequency selective channel, one well-known method of improving performance is to use multicarrier modulation. By transmitting at multiple carrier frequencies, it is possible to increase capacity by adapting the data rate to be higher for carriers with higher received SNR and lower for carriers with lower received SNR. A common multicarrier modulation scheme is OFDM, and it is the modulation scheme that OFDMA is based upon. A modulation scheme that is based on OFDM, but uses only a single carrier is SC-FDE, and it is the modulation scheme SC-FDMA uses.

2.1

Orthogonal Frequency Division Multiplexing (OFDM)

OFDM takes advantage of frequency diversity to achieve higher data rates. Figure 1 shows a discrete-time version of an OFDM system [5]. The input binary data is mapped to constellations in frequency domain and is converted to a time domain signal using an IFFT. Afterwards, the last part of the signal is appended to the beginning of the signal, known as the cyclic prefix (CP), in part to fight inter-symbol interference (ISI) and also to convert the signals linear convolution with the channel impulse response to a circular convolution. One important feature of OFDM is its ability to equalize in frequency domain. There are many different ways of equalizing, two of which are the zero-forcing (ZF) equalizer and the minimum mean square error (MMSE) equalizer. While they both correctly recover the phase of the signal, the ZF equalizer causes noise enhancement and the MMSE equalizer causes time distortion in the signal. One benefit of equalization using any frequency domain technique, is that applying it is simple an elementwise multiplication in the frequency domain.
Modulator x[k] Mapping S/P IFFT Append CP P/S Channel

y[k]

Demapping Demodulator

P/S

Equalizer

FFT

Remove CP

S/P

w[n]

Figure 1. Discrete-time system model for OFDM

Autumn 2009

EE359 Project

Nima Soltani

2.1.1 Capacity By the orthogonality of the subcarriers, the capacity of OFDM is effectively N parallel channels with their own respective capacity. Assuming that the transmitter has no CSI and sends with constant power P and the receiver knows the channel transfer function H perfectly, it is shown in [6] that the instantaneous capacity (or more appropriately, instantaneous spectral efficiency) of the kth sub-carrier in the ith block is , = log 2 , 1+ 2
2

Thus the total instantaneous capacity of the ith block is

N1 N1

=
=0

, =
=0

log 2

, 1+ 2

And the ergodic capacity of the channel would thus be the expected value of the instantaneous capacities, . While there are simulations in [6], Section 2.3.1 has simulations done corresponding to the LTE system multipath propagation model [2]. This model works well in the LTE case, because the coherence time is typically larger than the OFDM symbol time by a few orders of magnitude, and therefore the transfer function of the system would change very little throughout the duration of the symbol. However, an important consideration that the paper fails to make is the effect of the cyclic prefix on the capacity. In [7], written about OFDMA which also has the a cyclic prefix appended, it is assumed at the effect of the cyclic prefix on capacity is only to scale it by /( + ), where TS is the OFDM symbol duration and is the cyclic prefix duration. 2.1.2 Outage Probability After having calculated the instantaneous capacity/spectral efficiency per OFDM, [6] goes on to define an outage to be when the capacity of the symbol drops below a certain threshold. Thus for the case when L OFDM symbols are transmitted at a time, it is possible to define the outage probability, , as the probability that the average capacity of the L symbols is below a rate R: 1 =
+1

<
=

Again there are simulations done in [6] showing the outage probability two different channels, but these simulations have been done again in Section 2.3 in the multipath propagation environment given in the LTE specifications. 2.1.3 PAPR OFDM signals are mapped to constellations in the frequency domain and then using an IFFT they become time domain signals. Although the mapping of the constellations is deterministic, it is difficult to compute the values of the PAPR of the time domain signal analytically for a given set of constellations. Thus a simpler way is to assume that the constellations map to complex Gaussian random variables with variance 2 in the time domain [4]. This results in the complementary cumulative distribution function (CCDF) of the PAPR to be 3

Autumn 2009

EE359 Project 0 = 1 1 0

Nima Soltani

Where P0= 02/2 and N is the number points in the IFFT. Simulations for this scenario can be seen in Section 2.3.2.

2.2

Single-Carrier Frequency-Domain Equalization(SC-FDE)

A variation of OFDM is SC-FDE, a modulation scheme that contains all the same blocks but moves the IFFT from the transmitter to the receiver. Figure 2 shows a discrete-time version of this system [5]. The difference between this system and OFDM is that the constellation mapping takes place in time domain, and the CP is the last time samples appended to the beginning of the signal. Equivalently, an SC-FDE symbol is a group of N+ modulated (MQAM, MPSK, etc.) sub-symbols, where N is the length of the FFT/IFFT used in the receiver, and is the length of the cyclic prefix. This makes for a very simple transmitter. By moving the IFFT to the demodulator, equalization is done with a simple multiplication in frequency domain like in an OFDM demodulator.
Modulator x[n] Mapping S/P Append CP P/S Channel

y[n]

Demapping Demodulator

P/S

IFFT

Equalizer

FFT

Remove CP

S/P

w[n]

Figure 2. Discrete-time system model for SC-FDE

2.2.1 Capacity The capacity of an SC-FDE system becomes much more difficult to calculate than OFDM because in SCFDE for every sub-symbol sent (individual constellation), the many frequencies end up being used, while in OFDM every sub-symbol (constellation) modulates its own subcarrier. This means that to find the maximum possible throughput for a given channel transfer function, it is necessary to consider the effects of the equalizer to find the SNR of each SC-FDE symbol. In [6], two equalizers are explored: the zero 1 forcing (ZF) equalizer, with , = , for each symbol i and subcarrier k; and the minimum mean
squared error (MMSE) equalizer, with , = , / , 2

. The paper then finds their

corresponding instantaneous capacities to be

= log 2 1 +

1 1
1 =0

Autumn 2009

EE359 Project

Nima Soltani

= log 2

1 1
1 =0

, / 2 + 1

The capacity of the system using a ZF equalizer is less than that of OFDM, and it suffers from noise enhancement as well. For MMSE, there is no noise enhancement, but there is distortion introduced within in the SC-FDE symbol, and in the formula it is assumed considers the ISI between the SC-FDE subsymbols to be white noise [6]. Section 2.3.1 considers simulations with appropriate conditions for LTE. In the simulations in Section 2.3.1, outage probability is taken into consideration in the same manner described in Section 2.1.2, but using the capacity formulae above. 2.2.2 PAPR Since the mapping of SC-FDE sub-symbols is done in time domain, and the only processing done on an SC-FDE symbol is to attach a cyclic prefix, the PAPR for SC-FDE is the average PAPR of the modulation scheme for the sub-symbols (e.g. MQAM, MPSK, etc.). It is easy to assume that the PAPR is the ratio of the power of the constellation point farthest from the origin to the average power of the signal, but this assumes that the pulse shape is rectangular pulse shape used is a rectangular pulse [8]. This pulse shape has infinite frequency support, however, and it practice it is not used, but filters such as raised cosine filters are used instead, which is simulated in Section 2.3.2

2.3

Simulations

For each simulation, 10000 independent trials were run in MATLAB with multipath coefficients following the LTE multipath propagation model [2]. However, since LTE uses multi-user versions of these modulation schemes, we assume that all 2048 points of the FFT/IFFTs are used for a single user. 2.3.1 Capacity and Outage Probability Using the method described in [6], the plot shown below in Figure 3 was generated. The modulation scheme used mapping used was 16-QAM. It is clear that the capacity of OFDM is consistently higher than SC-FDE for all SNRs. This is mainly due to the fact that when the system experiences frequency selective fading, the bits that are transmitted on the subcarriers in deep fade result in lower instantaneous capacities, while the other subcarriers experience higher instantaneous capacity as they are not in a fade. SC-FDE however does not have this luxury, because the bits are mapped in time and transmitted across all subcarriers. It is also evident that at SC-FDE using an MMSE equalizer has consistently higher capacity than SC-FDE using a ZF equalizer. This difference, however, is quite minimal throughout the range of SNRs used, being less than 0.2 bps/Hz.

Autumn 2009

EE359 Project

Nima Soltani

Capacity of OFDM and SC-FDE Systems 10 9 8 7

Capacity (bps/Hz)

OFDM SC-FDE w/ZF Equalizer SC-FDE w/MMSE Equalizer

6 5 4 3 2 1 0

10

15 SNR(dB)

20

25

30

Figure 3. Capacity of OFDM and SC-FDE systems

The simulation results for outage probability can be seen below in Figure 4. It is clear that OFDM performs better than SC-FDE. Again, this is due to the fact that frequency selective fading only affects the bits being sent on faded frequencies in OFDM, while in SC-FDE it affects all bits in the symbol since the bits are mapped in time and thus spread in frequency. What is interesting to note from this graph, is that for the multipath propagation model used, the performance of SC-FDE with MMSE equalizer performs only ~2dB poorer than OFDM for the same outage probability. SC-FDE with ZF equalizer, however, performs much poorer than the SC-FDE with MMSE and OFDM for outage probabilities less than ~10%. This can be attributed to the noise enhancement that occurs when the signal is in a deep fade and the channel transfer function is inverted.

Autumn 2009

EE359 Project

Nima Soltani

Poutage of OFDM and SC-FDE Systems 0 -0.5 -1 -1.5 -2 -2.5 -3 -3.5 -4 OFDM SC-FDE w/ZF Equalizer SC-FDE w/MMSE Equalizer

log10(Poutage)

10

15 SNR(dB)

20

25

30

Figure 4. Outage probability distribution of OFDM and SC-FDE systems

2.3.2 PAPR To compare the PAPRs of OFDM and SC-FDE, a MATLAB simulation was done in which 10000 random symbols were generated and their PAPRs were measured, in a similar manner as in [9] for SCFDMA and OFDMA. To measure the effect of pulse shaping in SC-FDE, the signal was up-sampled by Nupsampled = 8 and convolved with a raised-cosine FIR filter for one case and a rectangular pulse of duration Nupsampled for another. The raised cosine filter has a roll-off factor of 0.5, and duration of 6Nupsampled+1 samples. For each set of 7 OFDM and pulse-shaped SC-FDE symbols, the maximum value of the signals during that time is divided by the mean value, to get the PAPR for that set of symbols. Having calculated all the PAPRs, a complementary cumulative distribution function (CCDF) was empirically measured for the test samples and can be seen in Figure 5. It is clear that between SC-FDE with rectangular pulse shaping and raised-cosine pulse shaping, there is 3dB difference for CCDF<10-1/2, and for both pulse shapes, the PAPR is consistently lower for SC-FDE than in OFDM. For SC-FDE, this allows circuit designers to operate their amplifiers higher up in their linear regions, with confidence that the probability that the signal falls into the non-linear region of the amplifier is very low, minimizing signal distortion compared to OFDM.

Autumn 2009

EE359 Project

Nima Soltani

CCDF of PAPR for OFDM and Pulse-Shaped SC-FDE 1 OFDM SC-FDE w/Rectangular SC-FDE w/Raised Cosine

log10(Prob(PAPR > PAPR0))

-1

-2

-3

-4

-5

-6 0

10 12 PAPR0 (dB)

14

16

18

20

Figure 5. CCDF of PAPR for OFDM and SC-FDE

3. Multiple Access Methods

In order to support more than one simultaneous transmission, multiple access techniques are employed. Both of the techniques that will be analyzed in this project achieve are forms of Frequency Division Multiple Access (FDMA). The basic premise of converting modulation schemes such as OFDM and SCFDE to their FDMA counterparts, OFDMA and SC-FDMA is assigning each user a subset of the available subcarriers and having the users transmit only on at these frequencies. In this section capacity and PAPR will be investigated.

3.1

Orthogonal Frequency Division Multiple-Access (OFDMA)

OFDMA is very similar to OFDM in function, with the main difference being that instead of being allocated all of the available subcarriers, the base station allocates a subset of carriers to each user in order to accommodate multiple transmissions simultaneously. A block diagram of this multiple access method can be seen in Figure 6 [9].
Modulator x[n] Symbol Mapping S/P Subcarrier Mapping M-point IFFT Append CP P/S Channel

y[n]

Symbol Demapping Demodulator

P/S

Subcarrier Demapping/ Equalizer

M-point FFT

Remove CP

S/P

w[n]

Figure 6. Discrete-time system model of OFDMA

Autumn 2009

EE359 Project

Nima Soltani

A problem with OFDMA as a compared to OFDM is its high sensitivity to frequency offset [9]. In OFDM since all the frequencies of the subcarriers were generated by one transmitter, maintaining orthogonality of the subcarriers is relatively easy. In OFDMA, since many users transmit simultaneously, each with their own estimates of the subcarrier frequencies, a frequency offset is inevitable and multiple access interference occurs as users power leaks into subcarriers bands. 3.1.1 Capacity The following expression for the capacity of the ith OFDMA symbol can be found in [7]

=
=1

log 2 1 +

+ 2

Where there are K users transmitting simultaneously, each with signal power P, noise power 2 and a fraction of the total bandwidth, and the ratio of other users power leaked into user j's spectrum to user j's power, as f. To simplify, the model [6] used in Section 2.1.1 will be modified to find the capacity of an OFDM channel. We have KN available subcarriers to use, and if there was no frequency offset at all between carriers, the capacity would be

=
=1

, =
=1

log 2

, 1+ 2

To incorporate frequency offset, we add the same factor f used in the previous equation [7]. This leads to
2

=
=1

, =
=1

log 2 1 +

, , + 2
2

=
=1

log 2 1 +

1 2

Simulations for this capacity estimate are conducted in Section 3.3.1. 3.1.2 PAPR Like in the OFDM case, it is difficult to come up with an analytical solution for the PAPR of OFDMA, because the sub-symbol constellation mapping is done in frequency domain. Thus the assumption of the time domain signal being a complex Gaussian random variable still holds, and the equation given in Section 2.1.3, is still valid. The numerical simulations in Section 3.3.2 will give a better indication as to its distribution.

3.2

Single-Carrier Frequency-Division Multiple-Access (SC-FDMA)

SC-FDMA is the multiple access version of Single-Carrier Frequency-Domain Equalization (SC-FDE), which is similar to OFDM, in that they both perform channel estimation and equalization in the frequency domain. The system model for this access method can be seen in Figure 7 [9]. Multiple access is achieved in frequency domain in SC-FDMA. Thus to transition from SC-FDE to SC-FDMA requires division frequency amongst frequencies. This is achieved by the first performing an N-point FFT on the output of the symbol mapper, and mapping the output of the FFT to N of M subcarriers, with M = QN where Q is an integer called the bandwidth expansion factor of the symbol, which is effectively the number of 9

Autumn 2009

EE359 Project

Nima Soltani

simultaneous users that the system supports [9]. Afterwards the M-point IFFT converts the signal back into time domain. Likewise in the demodulator, only N of the M frequencies of the output of the FFT correspond to this user, so those N values are equalized and an N-point IFFT is performed to get the original signal back.
Modulator x[n] Symbol Mapping S/P N-point FFT Subcarrier Mapping M-point IFFT Append CP P/S Channel

y[n]

Symbol Demapping Demodulator

P/S

N-point IFFT

Subcarrier Demapping/ Equalizer

M-point FFT

Remove CP

S/P

w[n]

Figure 7. Discrete-time system model for SC-FDMA

An important consideration to be made is how the N points of the signal are mapped to the M subcarriers of the system. Two main strategies exist: use N adjacent subcarriers, called localized FDMA (LFDMA) or distribute the N values across the M subcarriers using every Qth subcarrier, called interleaved FDMA (IFDMA). The LTE specification follows LFDMA as it assigns each users a set of 12 adjacent subcarriers. As explained in [9], a disadvantage of IFDMA is that it is very sensitive to frequency offset like OFDMA since different users take adjacent subcarriers. LFDMA, however, is much less affected by this because it only has at most two adjacent users. 3.2.1 Capacity In [7], the capacity of SC-FDMA is given as

=
=1

log 2 1 +

+ 2

1
SC-FDMA 10 10 )

Where K, P, , and f, are the same as the ones defined in Section 3.1.1, and LSC-FDMA is an empirically determined value representing the link loss relative to OFDMA in dB. It is said that typical values of L are 0dB for QPSK and about 1dB for 16-QAM and 64-QAM. In order to get a better idea of the difference, the capacities found in 2.2.1 will be applied. Since LFDMA is being used in LTE, frequency offset affects each user minimally since it only has two neighbouring cells. It is therefore not a far stretch to assume that there is no frequency offset or overlap between the different users. With this assumption, the capacity with K users becomes K times the capacity of a single SC-FDE user, which are the ZF and MMSE equalizer capacities found in Section 2.2.1. This scenario is simulated in Section 3.3.1.

3.2.2 PAPR The PAPR of SC-FDMA signals is analyzed in [8] for both the LFDMA and IFDMA cases. In the case of IFDMA, the output of the M-point IFFT of Figure 7 is just the original N-point signal repeated in time M/N times, thus making the PAPR the same as in the SC-FDE case. LTE, uses for LFDMA, the output of 10

Autumn 2009

EE359 Project

Nima Soltani

the M-point IFFT is the original N-point signal with M/N-1 samples in between that are a complexweighted sum of all the N samples, thus increasing the PAPR compared to IFDMA. Section 3.3.2 simulates and compares the PAPR of LFDMA which is used in LTE with that of OFDMA.

3.3

Simulations

For each parameter, 10000 independent trials were run in MATLAB with multipath coefficients following the LTE multipath propagation model [2] using 16-QAM constellation mapping. It is assumed that each user takes 12 subcarriers of the 2048 available in the FFT (at most a total of 600 subcarriers are actually used in the standard). 3.3.1 Capacity For capacity simulations we will further assume that 10 users transmit simultaneously, using 120 subcarriers. The methods used to simulate are covered in Sections 3.1.2 and 3.2.2. For OFDMA, two values of f are considered: 0, corresponding to 0 frequency offset, and 0.001 corresponding to frequency offset causing 0.1% of a signals power to leak into adjacent subcarrier. The results can be seen below in Figure 8. Although the methods used for OFDMA and SC-FDMA calculations were approximate, it shows some noteworthy trends. Firstly, it verifies the LSC-FDMA link loss factor referred to in [7], since the OFDMA with f = 0 case is about 1dB off from the SC-FDMA signals. Also, OFDMAs sensitivity to frequency offset is shown by the fact that for f = 0.01, increasing SNR causes more power to be leaked into adjacent subcarriers decreasing potential capacity. The MMSE and ZF equalizers also make less of a difference on the capacity of the channel in SC-FDMA than in the SC-FDE.
Capacity of OFDMA and SC-FDMA Systems 100 90 80 70
Capacity (bps/Hz)

OFDMA w/f=0 OFDMA w/f=0.001 SC-FDE w/ZF Equalizer SC-FDE w/MMSE Equalizer

60 50 40 30 20 10 0

10

15 SNR(dB)

20

25

30

Figure 8. Capacity of OFDMA and SC-FDMA systems

3.3.2 PAPR To obtain an empirical measure of the PAPR, the same setup as in Section 2.3.2 is used. This time it is an SC-FDMA signal that is passed through a raised-cosine pulse shaping filters, and rectangular pulse shaping filters. The OFDMA signal is not pulse shaped. The measured CCDF of the PAPR experiment can be seen below in Figure 9. Compared to SC-FDMA, OFDMA is more likely to have the higher PAPR 11

Autumn 2009

EE359 Project

Nima Soltani

ratios. It is also noteworthy that considering the pulse shaping filter does not seem to affect the CCDF of the PAPR.
CCDF of PAPR for OFDMA and SC-FDMA 1 OFDMA SC-FDMA w/Rectangular SC-FDMA w/Raised Cosine

0
log10(Prob(PAPR > PAPR0))

-1

-2

-3

-4

-5

8 10 PAPR0 (dB)

12

14

16

Figure 9. CCDF of the PAPR of OFDMA and SC-FDMA (LFDMA)

4. Conclusion
Power efficiency is an important criterion to consider when selecting a modulation scheme in mobile uplinks because battery usage is an important concern. In this project it became clear that SC-FDE and SC-FDMA are better choices in this regard. Their PAPRs achieve lower values on average, mainly due to the fact that they map their input bits to time symbols, as opposed to OFDM and OFDMA which map them directly to frequency symbols. The capacity of SC-FDE is inherently lower than that of OFDM, and in the estimates for SC-FDMA and OFDMA it was shown that their capacities were roughly equal without frequency offset. OFDMA is also sensitive to frequency offsets, causing it to achieve lower capacities for high SNRs. Although there are many variations of OFDM/OFDMA and SC-FDE/SCFDMA, comparing the basic models of each type, one can conclude that SC-FDMA is the better option for a cellular uplink in LTE, because of its higher efficiency due to low PAPR, its lower sensitivity to frequency offset because it has at most only two adjacent users, and its capacity is about the same as OFDMA.

12

Autumn 2009

EE359 Project

Nima Soltani

References
[1] 3rd Generation Partnership Project (3GPP). (2009, Sep.) Evolved Universal Terrestrial Radio Access (E-UTRA); Physical channels and modulation. [Online]. http://www.3gpp.org/ftp/Specs/htmlinfo/36211.htm [2] 3rd Generation Partnership Project (3GPP). (2009, Oct.) User Equipment (UE) conformance specification; Radio transmission and reception (FDD); Part 1: Conformance specification. [Online]. http://www.3gpp.org/ftp/Specs/html-info/34121-1.htm [3] IEEE 802.16 Broadband Wireless Access Working Group. (2001, Jul.) Channel Models for Fixed Wireless Applications. [Online]. http://www.ieee802.org/16/tg3/contrib/802163c-01_29r4.pdf [4] A. Goldsmith, Wireless Communications. New York, USA: Cambridge University Press, 2005. [5] J. Li, Y. Du, and Y. Liu, "Comparison of Spectral Efficiency for OFDM and SC-FDE under IEEE 802.16 Scenario," Proceedings of the 11th IEEE Symposium on Computers and Communications (ISCC'06), 2006. [6] T. Shi, S. Zhou, and Y. Yao, "Capacity of single carrier systems with frequency-domain equalization," IEEE 6th CAS Symp. on Emerging Technologies: Mobile and Wireless Comm., pp. 429-432, May 2004. [7] F. Khan, LTE for 4G Mobile Broadband: Air Interface Technologies and Performance. New York, USA: Cambridge University Press, 2009. [8] H. G. Myung, J. Lim, and J. Goodman, "Peak-to-Average Power Ratio of Single Carrier FDMA Signals with Pulse Shaping," The 17th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC'06), pp. 1-5, Sep. 2006. [9] H. G. Myung, J. Lim, and D. J. Goodman, "Single Carrier FDMA for Uplink Wireless Transmission," IEEE Vehicular Technology Magazine, vol. 1, no. 3, pp. 30-38, Sep. 2006.

13