Comparisons of 16QAM Modulation Schemes Considering PAPR

for Single-Carrier FDMA RadioAccess in Evolved UTRA Uplink

Teruo Kawamura*, Yoshihisa Kishiyama*, Kenichi Higuchi*, and Mamoru Sawahashi**
*IP Radio Network Development Department, NTT DoCoMo, Inc.
3-5 Hikari-no-oka, Yokosuka-shi, Kanagawa, 239-8536 Japan
Tel: +81-46-840-3470, Fax: +81-46-840-3735
E-mail: kawamura@mlab.yrp.nttdocomo.co.jp
**Department of Electronics and Communication Engineering, Musashi Instituteof Technology
1-28-1 Tamazutsumi, Setagaya-ku, Tokyo, 158-8557 Japan
Abstract - This paper compares the achievable packet error rate
and throughput performance levels of 16QAM modulation
schemes considering the peak-to-average power ratio for single-
carrier frequency division multiple access (SC-FDMA) in the
Evolved UTRA uplink. Simulation results show that the effective
required average received signal energy per symbol-to-noise
power spectrum density ratio (Es/N0) including the cubic metric
(CM) of the (8, 8) star 16QAM modulation scheme is decreased
by approximately 0.8 dB compared to that of the square 16QAM
modulation scheme using Turbo coding with the coding rate of
R = 1/3 in the six-ray Typical Urban channel model, when the
amplitude ratio between the inner and outer ring, Rradius, is equal
to 1.2. The results also show that although the star 16QAM
modulation scheme reduces the effective required average
received Es/N0 including the CM compared to the square 16QAM
modulation scheme, this merit is concealed by the application of
low modulation schemes such as QPSK and 8PSK when adaptive
modulation and coding is applied. Therefore, we conclude that
the square 16QAM modulation scheme is an appropriate
candidate for SC-FDMA radio access in the Evolved UTRA uplink.
I. INTRODUCTION
The specification of the study item on the Evolved UTRA (UMTS
Terrestrial Radio Access) and UTRAN (UMTS Terrestrial Radio
Access Network) is anticipated to be accomplished in June this
year in the 3rd Generation Partnership Project (3GPP). The
Evolved UTRA and UTRAN will achieve full IP-based
functionalities with low latency and low cost employing the 3G
spectrum with the maximum channel bandwidth of 20 MHz [1].
In the downlink, orthogonal frequency division multiplexing
(OFDM) based radio access was adopted, since OFDM is robust
against severe multipath interference, i.e., time dispersion, which
occurs in broadband channels wider than, e.g., 5 MHz.
Meanwhile, the achievement of wide area coverage is one of the
most important requirements in a cellular system. Thus, radio
access and relevant techniques are necessary to achieve this
requirement particularly in the uplink where user equipment (UE)
transmits with restricted transmission power due to battery
usage. Accordingly, single-carrier frequency division multiple
access (SC-FDMA) was adopted as the uplink radio access
scheme in which its low peak-to-average power ratio (PAPR)
feature is achieved [2]. In SC-FDMAradio access, the PAPR can
be decreased compared to multicarrier based radio access such
as OFDM. This brings about a reduction in the transmission
back-off in the transmitter power amplifier. Therefore, assuming
the same peak power of the power amplifier, the coverage area is
extended. In addition to the usage of SC-FDMA radio access, an
efficient data modulation scheme with a low PAPR is necessary.
In the Evolved UTRA, the current working assumptions for the
data modulation schemes are QPSK, 8PSK, and 16QAM. In order
to reduce the PAPR, the application of the r/2-shifted BPSK
modulation scheme was proposed and adopted [3]. However,
the the degree of effectiveness in decreasing the PAPR is low
for the application of the r/4-shifted QPSK and offset QPSK
(OQPSK) modulation schemes. The PAPR reduction effect for
the same approach for the 16QAM modulation scheme is also
small. Therefore, different approaches that decrease the PAPR
are necessary for QPSK and 16QAM modulation schemes. This
paper focuses on the 16QAM modulation scheme. It is well
known that the square 16QAM modulation scheme provides
the least required received signal-to-noise power ratio (SNR) to
satisfy the required bit error rate or packet error rate (PER) among
the 16QAM modulation schemes [4]. However, the achievable
PER or throughput considering the PAPR has not yet been
reported to be best knowledge of the authors.
Therefore, this paper compares the achievable PER and
throughput performance levels of 16QAM modulation schemes
considering the PAPR for SC-FDMA radio access in the Evolved
UTRAuplink. In the paper, we use a cubic metric (CM), which is
an empirical criterion corresponding to the transmission back-
off of the transmitter power amplifier [5]. We investigate the PER
and throughput performance levels including the calculated CM
value into the required average received signal energy per
symbol-to-noise power ratio (E
s/N0). The rest of the paper is
organized as follows. First, Section II describes the 16QAM
modulation schemes assumed in the paper. Then, PAPR, CM,
and PER performance levels are investigated for the various
16QAM modulation schemes in Section III. Finally, Section IV
presents the overall PER and throughput performance levels
considering the CM as transmission back-off.
II. 16QAM MODULATION SCHEMES
The 16QAM modulation scheme is employed near a cell site
with a high received signal-to-interference plus noise power
ratio (SINR) condition. However, even near the cell site, an
efficient 16QAM modulation scheme is necessary, which
achieves a low received E
s/N0 value taking into account the
transmission back-off for the limited peak transmission power at
a UE. In the paper, we investigate the achievable throughput
performance for square 16QAM [4] and star 16QAM modulation
schemes [6],[7]. Figure 1(a) shows the constellation for the
square 16QAM modulation scheme, which is a current working
assumption as the uplink data modulation scheme in the Evolved
UTRA. Let d(t) be the narrowband modulated signal waveform
of the transmitted signal represented as

-
-
i
i T t u i j i g t d ) / ( )] ( exp[ ) ( ) ( o , (1)
332
2006 IEEE Ninth International Symposium on Spread Spectrum Techniques and Applications
0-7803-9780-0/06/$20.00 2006 IEEE
(a) Square 16QAM
Figure 1. Signal constellation of various 16QAM modulation schemes
(c) (8, 8) star 16QAM (d) 16PSK (b) (4, 12) star 16QAM
0001
0010
0011
0100
0101
0110
0111
0000 1000
1001
1010
1011
1100
1101
1110
1111
a
1
a
2
(a
2
/ a
1
= 3.0)
0001
0010
0011
0100
0101
0110
0111
0000 1000
1001
1010
1011
1100
1101
1110
1111
a
1
a
2
(a
2
/ a
1
= 3.0)
0001
0010
0011
0100
0101
0110
0111
0000
1000
1001
1010
1011
1100
1101
1110
1111
r
1
r
2
(R
radius
= r
2
/ r
1
)
0001
0010
0011
0100
0101
0110
0111
0000
1000
1001
1010
1011
1100
1101
1110
1111
r
1
r
2
(R
radius
= r
2
/ r
1
)
0001
0010
0011
0100
0101
0110
0111
0000
1000
1001
1010
1011
1100
1101
1110
1111
r
1
r
2
(R
radius
= r
2
/ r
1
)
0001
0100
0101
0010
0011
0110
0111
0000
1000
1001
1100
1101
1010 1011
1110
1111
r
1
r
2
(R
radius
= r
2
/ r
1
)
0001
0100
0101
0010
0011
0110
0111
0000
1000
1001
1100
1101
1010 1011
1110
1111
r
1
r
2
(R
radius
= r
2
/ r
1
)
0001
0100
0101
0110
0111
0010
0011
0000 1010
1011
1110
1111
1100
1101
1000
1001
0001
0100
0101
0110
0111
0010
0011
0000 1010
1011
1110
1111
1100
1101
1000
1001
where T is the symbol duration and u(t) is the unit function such
that u(t) = 1 for 1 0 < s t and otherwise, u(t) = 0. Furthermore,
g(i) and o (i) are the amplitude and phase components of the
narrowband data modulation represented in the following
equations for the square 16QAM modulation scheme.
'
`

'
'

- - - = 1 , 0 , 1 , 2 ;
5 . 2
1
) 1 2 (
x x
q q x ,
'
`

'
'

- - - = 1 , 0 , 1 , 2 ;
5 . 2
1
) 1 2 (
y y
q q y ,
2 2
) ( y x i g - ,
x
y
i
1
tan ) (
-
o . (2)
Moreover, we investigated the performance for two types of
star 16QAM modulation schemes as shown in Figs. 1(b) and
1(c), the (4, 12) and (8, 8) star 16QAM modulation schemes,
respectively. As shown in Figs. 1(b) and 1(c), the star 16QAM
modulation scheme employs multiple rings with different
amplitudes. In the paper, we assume two rings. In the figures,
R
radius indicates the ratio of the inner ring amplitude, r1, to the
outer ring amplitude, r2. The narrowband data modulation for
star 16QAM modulation schemes is represented in the following
equations. For the (4, 12) star 16QAM modulation scheme,
'
'

ring) (outer
ring) (inner 1
) ( 3 1
16
) (
2
radius radius
R r
r
R
r i g ,
{
'
'

=
ring) (outer 11 ..., , 1 , 0 , 3
ring) (inner 3 , 2 , 1 , 0 , 1
2 / ) (
q p
q p
p q i r o . (3)
For the (8, 8) star 16QAM modulation scheme,
'
'

ring) (outer
ring) (inner 1
) ( 1
8
) (
2
radius radius
R r
r
R
r i g ,
{ 7 ..., , 1 , 0 ; 4 / ) ( = q q i r o . (4)
The feature of the (4, 12) star 16QAM modulation scheme is
such that the inner ring is robust against noise and interference
owing to the long Euclidian distance compared to the (6, 10) and
(8, 8) star 16QAM modulation schemes. Meanwhile, robust signal
mapping is possible in the (8, 8) star 16QAM modulation scheme
by assigning three bits for phase difference and one bit for
amplitude difference independently. As we indicate in the
subsequent section, the PAPR in the star 16QAM modulation
scheme becomes lower than that for the square 16QAM
modulation scheme, since the number of transitions that cross
the zero point is reduced. Figure 1(d) shows the signal
constellation of the 16PSK modulation scheme, which we use as
a reference.
III. PERFORMANCE COMPARISONS OF 16QAM
MODULATION SCHEMES
A. PAPR and CM Performance Levels
We investigate the complementary cumulative distribution
function (CCDF) of the PAPR and CM for the respective 16QAM
modulation schemes employing a localized FDMA transmission
with the symbol rate of 4.096 Msps associated with the roll-off
factor of 0.14. The CM is defined in the next equation.
CM = ( 20 log10(vrms) 1.52 ) / 1.85, (5)
where vrms represents the root mean squared value of the
instantaneous cubic amplitude for each symbol which is
normalized by the average amplitude of the input signal [5]. In
Eq. 5, 1.52 and 1.85 are empirical factors. It was reported in
[5] that the CM could be directly used as transmission back-off
for the power amplifier. Tables 1(a) and 1(b) summarize the PAPR
at 99.99% CCDF and the CM values for the (4, 12) star 16QAM
and (8, 8) star 16QAM modulation schemes, respectively, with
the relative amplitude value, R
radius, as a parameter. The
performance levels of the square 16QAM and 16PSK modulation
schemes are also given in the respective tables. We see that the
16PSK modulation scheme achieves a lower PAPR and CM than
the other modulation schemes. We also find that the (4, 12) star
16QAM modulation scheme achieves a lower PAPR and CM
compared to the (8, 8) star 16QAM modulation scheme when
the same R
radius value is used. For instance, the PAPR at 99.99%
CCDF and the CM values for the (4, 12) star 16QAM modulation
scheme with Rradius = 1.5 are reduced by approximately 1.0 dB
and 0.7 dB, respectively, compared to those of the square 16QAM
modulation scheme. Therefore, the tables show that the star
16QAM modulation scheme is more advantageous than the
square 16QAM modulation scheme from the viewpoint of the
reduction in the PAPR and CM.
B. Average PER Performance
We also investigate the average PER performance of the
respective 16QAM modulation schemes without considering
the CM. Table 2 gives the radio parameters assumed in the paper.
333
Table 1. Comparison of PAPR and CM for various 16QAM
modulation schemes
(a) PAPR at 99.99% CCDF
(b) CM
Table 2. Simulation parameters
16PSK
(8, 8)
star
16QAM
(4, 12)
star
16QAM
Square
16QAM
R
radius
5.33
5.43
5.67
5.80
5.87
5.97
6.03
1.2
1.5
1.8
2.0
2.5
3.0
6.07
6.60
7.03
6.70
5.53
6.40
6.87
PAPR at 99.99% CCDF (dB)
16PSK
(8, 8)
star
16QAM
(4, 12)
star
16QAM
Square
16QAM
R
radius
5.33
5.43
5.67
5.80
5.87
5.97
6.03
1.2
1.5
1.8
2.0
2.5
3.0
6.07
6.60
7.03
6.70
5.53
6.40
6.87
PAPR at 99.99% CCDF (dB)
(Roll-off factor = 0.14)
16PSK
(8, 8)
star
16QAM
(4, 12)
star
16QAM
Square
16QAM
R
radius
0.66
0.73
0.91
1.06
1.13
1.26
1.33
1.2
1.5
1.8
2.0
2.5
3.0
1.15
1.70
2.28
1.61
0.78
1.51
2.06
Cubic metric (dB)
16PSK
(8, 8)
star
16QAM
(4, 12)
star
16QAM
Square
16QAM
R
radius
0.66
0.73
0.91
1.06
1.13
1.26
1.33
1.2
1.5
1.8
2.0
2.5
3.0
1.15
1.70
2.28
1.61
0.78
1.51
2.06
Cubic metric (dB)
(Roll-off factor = 0.14)
14 samples Cyclic prefix duration
6-ray Typical Urban, AWGN Channel model
Square 16QAM, 16PSK,
(4, 12), (8, 8) star 16QAM
Data modulation
256 samples Data block size
4.096 Msymbol / sec Maximum symbol rate
2-branch receiver reception,
LMMSE frequency domain equalizer
Turbo coding (K = 4, R = 1/3, 1/2)
0.5 msec
5 MHz
128 samples
2048 samples Number of samples per sub-frame
Pilot block size
Sub-frame length
0.244 sec Symbol (sample) duration
Channel coding
0.14 Roll-off factor
Localized FDMA Transmission scheme
Transmission bandwidth
Receiver
14 samples Cyclic prefix duration
6-ray Typical Urban, AWGN Channel model
Square 16QAM, 16PSK,
(4, 12), (8, 8) star 16QAM
Data modulation
256 samples Data block size
4.096 Msymbol / sec Maximum symbol rate
2-branch receiver reception,
LMMSE frequency domain equalizer
Turbo coding (K = 4, R = 1/3, 1/2)
0.5 msec
5 MHz
128 samples
2048 samples Number of samples per sub-frame
Pilot block size
Sub-frame length
0.244 sec Symbol (sample) duration
Channel coding
0.14 Roll-off factor
Localized FDMA Transmission scheme
Transmission bandwidth
Receiver
We assume a localized FDMAtransmission with the symbol rate
of 4.096 Msps using the roll-off factor of 0.14. A random sequence
is used for the pilot symbols (note that the comparison results
of the achievable PER performance do not depend on the pilot
sequence, since we assume a single-user condition). At the
receiver, we employ two-branch antenna diversity reception.
The received path timings are estimated from the generated power
delay profile using the time-multiplexed pilot channel with a
threshold detection method. The threshold value is set such
that it becomes six times higher than that of the noise power
level. The channel impulse response, i.e., channel gain, and noise
power are also estimated employing the pilot channel. The
channel gain of each sub-frame is derived by linear interpolation
or extrapolation of the channel estimates of two pilot blocks in
the time domain. The noise power is estimated as variance of the
received pilot signals. By employing the estimated channel gain
and noise power, frequency domain equalization using the linear
minimum mean square error (MMSE) algorithm is performed.
Figures 2(a) and 2(b) show the average PER performance for
the (4, 12) star 16QAM modulation scheme with the R
radius value
as a parameter using the channel coding rate of R = 1/3 and 1/2,
respectively, as a function of the average received Es/N0 per
receiver branch. The PER performance of the square 16QAM
and 16PSK modulation schemes are also given in the figures.
The figures show that according to the increase in the Rradius
value, the PER performance is improved since the Euclidian
distance between the signal constellations for the inner ring and
that for the outer ring becomes long. As anticipated, the figures
also show that the PER performance of the square 16QAM
modulation scheme is superior to those of the (4, 12) star 16QAM
and 16PSK modulation schemes without considering the CM.
Figures 3(a) and 3(b) show the average PER performance for
the (8, 8) star 16QAM modulation scheme with the R
radius value
as a parameter using the channel coding rate of R = 1/3 and 1/2,
respectively, as a function of the average received Es/N0 per
receiver branch. The figures show that the (8, 8) star 16QAM
modulation scheme achieves better PER performance than the
(4, 12) star 16QAM modulation scheme. This is explained as
follows. As shown in Fig. 1(c), three bits represent the phase
modulation, while one bit represents the amplitude modulation
in the (8, 8) star 16QAM modulation scheme. Thus, the
independent bit mapping in the amplitude and phase provides
greater robustness than in the (4, 12) star 16QAM modulation
scheme mapping for multipath Rayleigh fading. Therefore, we
find that the (8, 8) star 16QAM modulation scheme achieves
almost the same PER performance as the square 16QAM
modulation scheme when the optimum R
radius value is used in
both cases of R = 1/3 and 1/2.
IV. OVERALLAVERAGE PER PERFORMANCE COMPARISONS
CONSIDERING CM
A. Average PER Performance Considering CM
The results of the PER performance in Figs. 2 and 3 do not
consider the CM. Therefore, we compare the average PER
performance considering the CM in this section. Figure 4 plots
the required average received Es/N0 per receiver branch for the
average PER of 10
-1
considering the CM, as a function of the
Rradius value for R = 1/3 and 1/2. The vertical axis indicates the
effective required average received Es/N0 including the
transmission back-off. We assume an additive white Gaussian
noise (AWGN) channel in Fig. 4(a) since the channel variation
approaches the AWGN condition when channel-dependent
scheduling is applied. Figure 4(a) shows that in the case of R = 1/3,
both the (8, 8) and (4, 12) star 16QAM modulation schemes achieve
lower required received E
s/N0 values than the square 16QAM
modulation scheme with a low Rradius value owing to a smaller CM
feature. In particular, the required average received Es/N0 of the
(8, 8) star 16QAM modulation scheme with Rradius = 1.2 can be
decreased by approximately 0.8 dB compared to that of the square
16QAM modulation scheme. Meanwhile, in the case of R = 1/2,
the required average received Es/N0 of the (8, 8) star 16QAM
modulation scheme with Rradius =1.5 can be decreased by
approximately 0.2 dB compared to that of the square 16QAM
modulation scheme.
Figure 4 (b) plots the required average received Es/N0 per
receiver branch for the average PER of 10
-1
considering the CM
334
(a) R = 1/3
Figure 2. Average PER performance of (4, 12) star 16QAM
(b) R = 1/2
(a) R = 1/3
Figure 3. Average PER performance of (8, 8) star 16QAM
(b) R = 1/2
10
-3
10
-2
10
-1
10
0
2 4 6 8 10 12
A
v
e
r
a
g
e

P
E
R
Average received E
s
/N
0
per receiver branch (dB)
Square 16QAM
R
radius
= 3.0
R
radius
= 2.5
R
radius
= 2.0
R
radius
= 1.8
R
radius
= 1.5
R
radius
= 1.2
16PSK
(4, 12)
star
16QAM
6-ray TU
10
-3
10
-2
10
-1
10
0
4 6 8 10 12 14 16 18 20
A
v
e
r
a
g
e

P
E
R
Average received E
s
/N
0
per receiver branch (dB)
Square 16QAM
R
radius
= 3.0
R
radius
= 2.5
R
radius
= 2.0
R
radius
= 1.8
R
radius
= 1.5
16PSK
(4, 12)
star
16QAM
6-ray TU
10
-3
10
-2
10
-1
10
0
2 4 6 8 10 12
A
v
e
r
a
g
e

P
E
R
Average received E
s
/N
0
per receiver branch (dB)
Square 16QAM
R
radius
= 3.0
R
radius
= 2.5
R
radius
= 2.0
R
radius
= 1.8
R
radius
= 1.5
R
radius
= 1.2
16PSK
(8, 8)
star
16QAM
6-ray TU
10
-3
10
-2
10
-1
10
0
4 6 8 10 12 14 16 18 20
A
v
e
r
a
g
e

P
E
R
Average received E
s
/N
0
per receiver branch (dB)
Square 16QAM
R
radius
= 3.0
R
radius
= 2.5
R
radius
= 2.0
R
radius
= 1.8
R
radius
= 1.7
R
radius
= 1.6
R
radius
= 1.5
16PSK
(8, 8)
star
16QAM
6-ray TU
in the six-ray Typical Urban (TU) channel model [8] as a function
of the Rradius value for R = 1/3 and 1/2. Figure 4(b) shows that in
the case of R = 1/3, both the (8, 8) and (4, 12) star 16QAM
modulation schemes achieve lower required received Es/N0 values
than the square 16QAM modulation scheme with a low Rradius
value owing to the smaller CM feature. In particular, the required
average received Es/N0 of the (8, 8) star 16QAM modulation
scheme with Rradius = 1.2 can be decreased by approximately 0.8 dB
compared to that of the square 16QAM modulation scheme. As
shown in Fig. 3(a), the loss in the PER performance of the (8, 8)
star 16QAM modulation scheme compared to that of the square
16QAM modulation scheme is small without considering the CM
owing to the large channel coding gain. Accordingly, the CM
(PAPR) in the (8, 8) star 16QAM modulation scheme reduces the
effective required average received E
s/N0 compared to the square
16QAM modulation scheme. Meanwhile, in the case of R = 1/2,
the square 16QAM modulation scheme achieves better
performance than the (8, 8) and (4, 12) star 16QAM modulation
schemes.
B. Throughput Performance Considering CM Assuming AMC
In Section IV. A., we showed that the star 16QAM modulation
scheme reduces the effective required average received Es/N0
including the transmission back-off particularly with a low
channel coding rate. However, we should investigate the
applicability of the star 16QAM modulation scheme assuming
the adaptive modulation and coding (AMC) scheme. Here, we
assume QPSK, 8PSK, and 16QAM modulation schemes. Figure 5
shows the throughput performance for various 16QAM
modulation schemes as a function of the required average
received E
s/N0 including the transmission back-off corresponding
to the CM value. Figures 5(a) and 5(b) assume an AWGNchannel
and the six-ray TU channel model, respectively. We use the
optimum Rradius value for the (8, 8) and (4, 12) star 16QAM
modulation schemes. Figure 5(a) shows that the (8, 8) and (4, 12)
star 16QAM modulation schemes achieve a higher throughput
than the square 16QAM modulation scheme for R = 1/3. However,
as indicated in the figure, the QPSK modulation scheme with
R = 2/3 achieves higher throughput than that achieved by the
16QAM modulation scheme with R = 1/3. We also find that the
(8, 8) and (4, 12) star 16QAM modulation schemes with R = 1/2
achieve almost the same throughput as that achieved by the
8PSK modulation scheme with R = 2/3. This means that the
advantageous received E
s/N0 area of the (8, 8) and (4, 12) star
16QAM modulation schemes is covered by low modulation
schemes such as QPSK and 8PSK.
Similar to the results in the AWGN channel, Fig. 5(b) shows
that the (8, 8) and (4, 12) star 16QAM modulation schemes achieve
a higher throughput than the square 16QAM modulation scheme
for R = 1/3. However, as indicated in the figure, the QPSK
335
REFERENCES
[1] 3GPP, RP-04061, Proposed study item on Evolved UTRA and
UTRAN, Dec. 2004.
[2] 3GPP, RP-050758, LS on UTRAN LTE Multiple Access Selection,
Nov. 2005.
[3] 3GPP, R1-060293, Nokia, UL modulation scheme, Feb. 2006.
[4] J. G. Proakis, Digital Communications, 4th ed., New York: McGraw-
Hill, 2001.
[5] 3GPP, R4-040367, Motorola, Comparison of PAR and cubic metric
for power de-rating, May 2004.
[6] W. T. Webb, L. Hanzo, and R. Steel, Bandwidth efficient QAM
schemes for Rayleigh fading channels, IEE Proc. I. vol. 138, pp.
169-175, June 1991.
[7] F. Adachi and M. Sawahashi, Performance analysis of various 16
level modulation schemes under Rayleigh fading, Electron. lett.,
vol. 28, pp. 1579-1581, Nov. 1992.
[8] 3GPP, TS45.005 (V7.3.0), Radio transmission and reception, Nov.
2005.
Figure 5. Throughput performance as a fuction of average
received Es/N0 including transmission back-off
(a) AWGN channel
(b) Typical Urban channel
Figure 4. Required average received Es/N0 considering CM as a
function of Rradius
(a) AWGN channel
(b) Typical Urban channel
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
1.0 1.5 2.0 2.5 3.0
R
e
q
u
i
r
e
d

a
v
e
r
a
g
e

r
e
c
e
i
v
e
d

E
s
/
N
0
R
radius
p
e
r

r
e
c
e
i
v
e
r

b
r
a
n
c
h

f
o
r

a
v
e
r
a
g
e

P
E
R

=

1
0
-
1

+

C
M

(
d
B
)
AWGN
(4, 12) star 16QAM
(8, 8) star 16QAM
R = 1/3
R = 1/2
Square 16QAM
16PSK
6.0
7.0
8.0
9.0
10.0
11.0
12.0
13.0
14.0
1.0 1.5 2.0 2.5 3.0
R
e
q
u
i
r
e
d

a
v
e
r
a
g
e

r
e
c
e
i
v
e
d

E
s
/
N
0
R
radius
f
o
r

a
v
e
r
a
g
e

P
E
R

=

1
0
-
1

+

C
M

(
d
B
)
p
e
r

r
e
c
e
i
v
e
r

b
r
a
n
c
h

6-ray TU
(4, 12) star 16QAM
(8, 8) star 16QAM
R = 1/3
R = 1/2
Square 16QAM
16PSK
modulation scheme with R = 2/3 achieves higher throughput
than that achieved by the 16QAM modulation scheme with
R = 1/3. We also find that the (8, 8) and (4, 12) star 16QAM
modulation schemes with R = 1/2 achieves almost the same
throughput as that achieved by the 8PSK modulation scheme
with R = 2/3. Therefore, the advantage of the (8, 8) and (4, 12)
star 16QAM modulation schemes with a low channel coding
rate is concealed by the usage of the QPSK or 8PSK modulation
scheme with a high coding rate. Therefore, we conclude that
although the star 16QAM modulation scheme reduces the
effective required average received E
s/N0 including the CM
compared to the square 16QAM modulation scheme, the merit is
concealed by the application of low modulation schemes such
as QPSK and 8PSK.
V. CONCLUSION
This paper compared the achievable PER and throughput
performance levels of 16QAM modulation schemes considering
the PAPR for SC-FDMA radio access in the Evolved UTRA
uplink. The simulation results showed that the effective required
average received E
s/N0 including the CM of the (8, 8) and (4, 12)
star 16QAM modulation schemes when Rradius = 1.2 and 1.8, is
decreased by approximately 0.8 and 0.3 dB compared to that of
the square 16QAM modulation scheme using Turbo coding with
the coding rate of R = 1/3, respectively, in the six-ray TU channel
model. It was also clarified that although the star 16QAM
modulation scheme can reduce the effective required average
received E
s/N0 including the CM compared to the square 16QAM
modulation scheme, the merit was concealed by the application
of low modulation schemes such as QPSK and 8PSK when the
AMC was applied. Therefore, we conclude that the square
16QAM modulation scheme is an appropriate candidate for SC-
FDMA radio access in the Evolved UTRA uplink.
0
2
4
6
8
10
12
0 1 2 3 4 5 6 7 8
T
h
r
o
u
g
h
p
u
t

(
M
b
p
s
)
Average received E
s
/N
0
per receiver branch + CM (dB)
(4,12) star 16QAM
(8,8) star 16QAM
Square 16QAM
16PSK
QPSK, R = 2/3 ( )
16-level mod.
R = 1/3
16-level mod.
R = 1/2
AWGN
8PSK, R = 2/3 ( )
0
2
4
6
8
10
12
0 2 4 6 8 10 12 14 16
T
h
r
o
u
g
h
p
u
t

(
M
b
p
s
)
Average received E
s
/N
0
per receiver branch + CM (dB)
QPSK,
R = 2/3 ( )
16-level mod.
R = 1/3
16-level mod.
R = 1/2
8PSK, R = 2/3 ( )
6-ray TU (4,12) star 16QAM
(8,8) star 16QAM
Square 16QAM
16PSK
336