On the Performance of a Widely Linear

SC-FDE System Under Multiple Independent

Interferences
Bruno S. Chang1 , Carlos A. F. da Rocha2 , Hayfa Fhima3,4 ,
Rafik Zayani3 , Hmaied Shaiek4 and Daniel Roviras4
1
CPGEI/Electronics Department, Federal University of Technology - Paraná, Curitiba, Brazil
Email: [email protected]
2
Communications Research Group, Federal University of Santa Catarina, Florianópolis, Brazil
Email: [email protected]
3
Innov’COM Laboratory, Sup’Com, Carthage University, Tunisia
Email: {hayfa.fhima,rafik.zayani}@supcom.tn
4
CEDRIC/LAETITIA Laboratory, CNAM, Paris, France
Email: {hmaied.shaiek,daniel.roviras}@cnam.fr

Abstract—In this paper, a new widely linear equalizer QAM ones, since their envelope is constant or quasi-
for single carrier systems using frequency domain equal- constant [3]. These constellations are also known as
ization (SC-FDE) that takes into account up to K known improper, since the pseudoautocorrelation of symbols
interferers in the symbol estimation process is proposed. It
was found out that the error performance of the proposed from these constellations is non-zero. If the transmit-
widely linear equalizer is nearly immune to the effect of ted symbols come from improper constellations, widely
one interferer, regardless of its power; when compared to linear (WL) processing [4], [5], [6] can be used to take
its strictly linear version, the proposed structure also has advantage of this impropriety, by processing the signal
a better error performance for up to three interferers. together with its conjugate version to obtain a more
Index Terms—SC-FDE systems, widely linear processing,
interference cancellation.
precise estimate. Widely linear processing applied to
SC-FDE systems has been investigated in [7], [8], [9],
[10]; it was seen that WL processing improves the error
I. I NTRODUCTION performance of these systems and makes them more
Single carrier systems using frequency domain equal- robust to channel estimation errors.
ization (SC-FDE) [1] have a reduced peak to aver- There are scenarios in which a communications sys-
age power ratio (PAPR) when compared to Orthogonal tem will suffer strong co-channel interference, such as
Frequency Division Multiplexing (OFDM) systems [2], a system striving to obtain maximum spectral efficiency
while maintaining their convenient low complexity fre- and using a frequency reuse factor of one. In order to
quency domain equalization. This characteristic makes cancel the damaging effects of this interference in SC-
them more attractive compared to multicarrier systems FDMA systems, a time domain cochannel interference
for mobile uplink transmissions, which require energy suppression scheme was proposed in [11]; it was found
efficient and low cost power amplifiers in terminals. For out that the length of the filter used for interference sup-
these reasons the multiuser variant of SC-FDE systems, pression affects greatly the error performance. A time-
SC-FDMA (Single carrier systems with frequency do- domain system employing widely linear processing,
main multiple access) is used for the uplink in LTE such as the Single Antenna Interference Cancellation
systems. (SAIC) scheme proposed in [6], was found to be able
These systems normally transmit symbols from a to suppress the effect of one interferer. In [12], a SC-
complex Quadrature Amplitude Modulation (QAM) con- FDE system using a simplified widely linear equalizer
stellation. QAM symbols can be described as proper, that and operating under data-like cochannel interference was
is, their second-order statistics are completely described studied; in this work, the channel impulse response of the
by their autocovariance. The PAPR can be further re- interference is not known by the receiver. The authors
duced by the usage of transmitted symbols from real or in [13], [14] have investigated a scenario where more
offset constellations (such as Binary Phase Shift Keying than one interferer is considered in a system using a
(BPSK) and Offset QAM (OQAM) ones) instead of time domain widely linear equalizer, with the receiver
978-1-5386-3531-5/17/$31.00 c 2017 IEEE having knowledge of the channel impulse response of the

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SANTA CATARINA. Downloaded on November 02,2023 at 14:14:24 UTC from IEEE Xplore. Restrictions apply.
 ′
interferers. It was seen that the widely linear system will Since H is a circulant matrix, we can apply an eigen-
have a performance advantage with up to five interferers. decomposition to this matrix to obtain W∗ HW, where
In this work, we propose a widely linear equalizer for W is the normalized discrete Fourier transform (DFT)
SC-FDE systems that takes into account K interferer matrix of size N ×N and H is a N ×N diagonal matrix
signals in the equalization process, with the receiver with its (k,k)-th entry Hk,k corresponding to the k-th
having knowledge of the channel impulse response of the coefficient of the N -sized DFT of the channel impulse
interferers and using this knowledge for the interference response h̃.
cancellation process. It was seen that the proposed At the entry of the receiver, the time domain sig-
structure is immune to the effect of one interferer and nal r̃ = [r̃1 r̃2 . . . r̃N +O ]T with length N + O has
even with three interferers there is a performance gain its cyclic prefix removed and passes to the frequency
between the proposed widely linear equalizer and its domain through a Fast Fourier Transform (FFT), so
linear version. that equalization can be done in the frequency domain.
The remainder of the paper is organized as follows: Thus, the received signal in the frequency domain can
the system model of the WL SC-FDE system that be expressed as
takes into account K interferers and the derivation of K
the proposed WL equalizer is presented in Section II.
X
r = Hs + σi2 Hik sik + n, (1)
Simulation results comparing the WL equalizer with the k=1
strictly linear version for different signal to interference
ratios and number of interferers are shown and analyzed where s is the desired transmitted signal in the frequency
in Section III, and finally the concluding remarks are domain, σi2 is the total interference power, Hik expresses
stated in Section IV. the impulse response corresponding to the k-th inter-
Notation: vectors are represented by bold lowercase ferer channel in the frequency domain, sik represents
letters, while bold capital letters denote matrices. Time the frequency-domain transmitted signal from the k-
domain elements have a tilde appended to them. ⌊.⌋ th interferer and n is the noise, also in the frequency
represents the largest integer less than or equal to a domain. Thus, the signal to interference ratio SIR can
real number. The superscripts ()∗ , ()T and ()H denote, be expressed as
respectively, conjugate, transpose and Hermitian opera- σ2
SIR = s2 . (2)
tions. The i-th element of the vector v is given by vi σi
and the (i, j)-th element of the matrix M is given by Equalization and interference cancellation is per-
Mi,j . The identity matrix of size N × N is denoted by formed by filters based on the minimum mean square
IN . The trace of a matrix M is expressed by Tr {M}. error criterion. It is also assumed in this process that
Finally, the mathematical expectation is E {.}. the receiver has complete knowledge of the interferers’
channel impulse responses and of the total interference
II. S YSTEM M ODEL
power. However, since the equalizer is dealing with a
The system model for this work is presented on Figure signal from an improper constellation (which has non-
1. On the transmitter side, the block s̃ = [s̃1 s̃2 . . . s̃N ]T zero pseudocorrelation), it has to employ widely linear
of size N is composed by symbols s̃i belonging to a processing to use all the second-order statistics made
rectilinear improper constellation (in the case consid- available by the received signal. In order to do that, the
ered, M -PAM) with power σs2 . The transmitted signal original version of the received signal in the frequency
will pass through a channel with an impulse response domain and its conjugate version should be processed
h̃ = [h̃1 h̃2 . . . h̃Lh̃ ]T of size Lh̃ . Thus, the cyclic prefix by the equalizer [4], [6]. Thus, t combines the received
appended to the block s̃ before transmission must have signal and its conjugate version and is expressed as
a length O of at least Lh̃ − 1, resulting in s̃CP of length T
N + O. t = [r r∗ ] . (3)
This transmitted signal will be corrupted by K in- The cost function ǫWL used to derive the widely linear
dependent interference signals, which are considered to equalizer AWL based on the MMSE criterion is
be parallel cochannel improper transmissions following
the same procedure detailed above. Complex proper ǫWL = E[||AH t − s||2 ]
uncorrelated additional white Gaussian noise (AWGN) = Tr{AH Ctt A − AH Cts − Cst A + IN }. (4)
ñ with zero mean and variance σn2 also contaminates
the transmitted signal. where
Due to the cyclic prefix, the N × N channel matrix
′
H is a circulant one, with its first column containing the Ctt = E[ttH ]
    
channel impulse response h̃ appended by (N −Lh̃ ) zeros. r  H T
 Crr Crr
=E r r = ∗
r∗ Crr C∗rr

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SANTA CATARINA. Downloaded on November 02,2023 at 14:14:24 UTC from IEEE Xplore. Restrictions apply.
 Fig. 1. A SC-FDE system employing widely linear MMSE equalization under the effect of K interferers.

Crr = E rrH
 
After equalization and interference cancellation, an
K
X inverse fast Fourier transform (IFFT) is done on z so
= HHH + σi2 Hik HH 2
ik + σn IN , (5) that the symbol decision is realized in the time domain,
k=1 resulting in z̃ (with size N ). Due to the fact that
widely linear processing is employed in the equalizer,
K
X the estimated symbols z̃ at the output of the receiver
Crr = E rrT = HUHT + σi2 Hik UHT
 
ik , (6) will be purely real.
k=1
III. S IMULATION R ESULTS
with U = WWT expressed by Simulation results are presented in this section to
compare the error performance results from the widely
 
1 0 0 ... 0
 0 0 0 ... 1  linear equalizer proposed in this paper to its strictly
linear version, for different numbers of interferers and
U =  ... ... ... .. ..
 
, (7)
 
 . . signal to interference ratios. For these simulations, the
 0 0 1 ... 0  cyclic prefix size used O is the minimum sufficient to
0 1 0 ... 0 eliminate the interblock interference and the power loss
caused by the redundancy introduced by the cyclic prefix
rsH
    
H r H is taken into account in the SNR calculation. The symbol
Cts = E[ts ] = E s =E
r∗ r∗ sH blocks transmitted, both from the desired symbol and

H
 from the interferences, are from a BPSK constellation
= (8) and a sampling frequency of 10 MHz was used. To
H∗ U
calculate the final bit error performance in the Monte
and Carlo simulations, a minimum of 200 errors were taken
srH into account for each point. Channel fading is considered
    
H H r
Cst = E[st ] = E s =E to be quasistatic (time-invariant during each transmitted
r∗ s∗ rH
  block) and it is assumed that the receiver has perfect
H
= channel estimation, both from the desired channel and
UH∗
also from the interferer channels.
= Cts , (9) Figure 2 presents the error performance results for
with E[nnT ] = 0 (since the noise is proper), and N = 128, the ITU-T Vehicular A channel model and
WWH = IN . We obtain the optimal widely linear one interference, for SIR values of 20 and 16 dB only
equalizer AWL by differentiating ǫWL with respect to A for MMSE systems, 13 dB for both and finally 3 and 0
and equalling the result to zero, resulting in dB for the proposed WL-MMSE system. For reference
results when using the widely linear equalizer proposed
AWL = C−1
tt Cts . (10) in [8], which does not cancel the interference but treats it
as additional noise, are also presented. For this case, the
In comparison, its strictly linear version AL can be error performance degrades quickly with an increase of
expressed as the SIR, validating the approach proposed in this paper
AL = C−1 to filter together both interference and channel fading.
rr Crs , (11)
In the systems that are built to cancel the interference,
with it is possible to see that by employing widely linear
processing in the SC-FDE equalizer the system becomes
Crs = E[rsH ] = H. (12) nearly immune to the effect of this one interference,

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SANTA CATARINA. Downloaded on November 02,2023 at 14:14:24 UTC from IEEE Xplore. Restrictions apply.
 regardless of its power; with the regular strictly linear 3 interferences, N = 128, Vehicular A channel
100
equalizer, there is an error floor that becomes higher with MMSE, S/I = 20 dB
MMSE, S/I = 16 dB
an increase of the SIR. MMSE, S/I = 13 dB
WL-MMSE, S/I = 20 dB
WL-MMSE, S/I = 16 dB
1 interference, N = 128, Vehicular A channel 10−1
WL-MMSE, S/I = 13 dB
100
MMSE, S/I = 20 dB
MMSE, S/I = 16 dB
MMSE, S/I = 13 dB

BER
10−1 WL-MMSE from [8], S/I = 13 dB 10−2
WL-MMSE from [8], S/I = 0 dB
WL-MMSE, S/I = 13 dB
WL-MMSE, S/I = 3 dB
WL-MMSE, S/I = 0 dB
10−2
10−3
BER

10−3

10−4
0 2 4 6 8 10 12 14 16 18 20 22 24 26
Eb /N0 (dB)
10−4

Fig. 4. Error performance for N = 128, the Vehicular A channel

10−5
model and 3 interferences.
0 2 4 6 8 10 12 14 16 18 20 22 24 26
Eb /N0 (dB)
4 interferences, N = 128, Vehicular A channel
100
Fig. 2. Error performance for N = 128, the Vehicular A channel MMSE, S/I = 20 dB
model and 1 interference. MMSE, S/I = 16 dB
MMSE, S/I = 13 dB
WL-MMSE, S/I = 20 dB
Results (also for SIR values of 13, 16 and 20 dB) WL-MMSE, S/I = 16 dB
WL-MMSE, S/I = 13 dB
when increasing the number of interferers are presented 10−1

in Figures 3, 4 and 5, respectively, for 2, 3 and 4

interferers. In these first two cases the widely linear
BER

equalizer has an error performance which is now sen-

sitive to the interference effect; however, is still able 10−2

to compensate quite effectively the effect of the outside

interferers, leading to a performance gain and a lower
error floor with respect to the strictly linear equalizer.
These results are very interesting because they show 10−3
0 2 4 6 8 10 12 14 16 18 20 22 24 26
how a WL equalizer can perform better than its linear Eb /N0 (dB)

counterpart even in the presence of multiple interferers.

Finally, for the last considered case (4 interferers) both Fig. 5. Error performance for N = 128, the Vehicular A channel
model and 4 interferences.
equalizers have almost the same error performance and
high error floor, due to the fact that the WL equalizer
has reached its interference cancellation limit. IV. C ONCLUDING R EMARKS
2 interferences, N = 128, Vehicular A channel In this paper we have presented a SC-FDE system
100
MMSE, S/I = 20 dB with a widely linear MMSE equalizer designed to
MMSE, S/I = 16 dB
MMSE, S/I = 13 dB cope with K interferences. It was seen that its error
WL-MMSE, S/I = 20 dB
WL-MMSE, S/I = 16 dB
performance is almost immune to the effect of one
10−1
WL-MMSE, S/I = 13 dB interfererer, regardless of its power; it also has better
error performance and a lower error floor than its strictly
linear counterpart with up to 3 interferers. Suggestions
BER

10−2
for future work are the theoretical investigation of the
cancellation limits of the proposed widely linear equal-
izer, its expansion for quasi-rectilinear modulations and
10−3
the application of iterative equalization structures for this
task.
10−4
0 2 4 6 8 10 12 14 16 18 20 22 24 26 R EFERENCES
Eb /N0 (dB)

[1] D. Falconer, S. Ariyavisitakul, A. Benyamin-Seeyar, and B. Eid-

Fig. 3. Error performance for N = 128, the Vehicular A channel son, “Frequency domain equalization for single-carrier broad-
model and 2 interferences. band wireless systems,” IEEE Communications Magazine,
vol. 40, no. 4, pp. 58–66, Apr. 2002.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SANTA CATARINA. Downloaded on November 02,2023 at 14:14:24 UTC from IEEE Xplore. Restrictions apply.
 [2] H. G. Myung, J. Lim, and D. Goodman, “Single carrier fdma for no. 1, p. 124, 2014.
uplink wireless transmission,” Vehicular Technology Magazine, [9] B. S. Chang, C. A. da Rocha, L. S. Resende, and D. Le Ruyet,
IEEE, vol. 1, no. 3, pp. 30–38, 2006. “On the effect of channel estimation errors in sc-fde systems
[3] M. Luzio, R. Dinis, and P. Montezuma, “On the Use of Multiple using widely linear iterative equalization,” in 2015 Interna-
Grossly Nonlinear Amplifiers for an Efficient Amplification of tional Symposium on Wireless Communication Systems (ISWCS).
OQAM Signals with FDE Receivers,” in Vehicular Technology IEEE, 2015, pp. 526–530.
Conference (VTC Fall), 2011 IEEE, Sep. 2011. [10] B. S. Chang, H. I. Del Monego, A. Munaretto, M. S. Fonseca,
[4] B. Picinbono and P. Chevalier, “Widely Linear Estimation and R. D. Souza, “A simplified widely linear iterative equalizer
with Complex Data,” IEEE Transactions on Signal Processing, for sc-fde systems,” in Wireless Days, 2017. IEEE, 2017, pp.
vol. 43, no. 8, pp. 2030–2033, 1995. 159–162.
[5] P. Schreier, L. Scharf, and C. T. Mullis, “Detection and Estima- [11] U. L. Dang, W. Zhang, M. A. Ruder, and W. H. Gerstacker,
tion of Improper Complex Random Signals,” IEEE Transactions “On time domain co-channel interference suppression for sc-
on Information Theory, vol. 51, no. 1, pp. 306–312, Aug. 2005. fdma transmission,” in Wireless Communications and Networking
[6] P. Chevalier and F. Pipon, “New Insights Into Optimal Widely Conference (WCNC), 2011 IEEE. IEEE, 2011, pp. 1253–1258.
Linear Array Receivers for the Demodulation of BPSK, MSK, [12] M. Kim, J. H. Cho, and J. S. Lehnert, “Asymptotically op-
and GMSK Signals Corrupted by Noncircular Interferences - timal low-complexity sc-fde with noise prediction in data-like
Application to SAIC,” IEEE Transactions on Signal Processing, improper-complex interference,” IEEE Transactions on Wireless
vol. 54, no. 3, pp. 870–883, Mar. 2006. Communications, vol. 15, no. 3, pp. 2090–2103, 2016.
[7] B. S. Chang, C. A. da Rocha, D. L. Ruyet, and D. Roviras, [13] H. Fhima, R. Zayani, H. Shaiek, D. Roviras, B. S. Chang,
“Widely linear iterative equalizers for sc-fde systems,” in Wire- and R. Bouallegue, “Widely linear equalizer performance with
less Communication Systems (ISWCS 2013), Proceedings of the multiple independent interferences,” in ISCC 2017. IEEE, 2017.
Tenth International Symposium on. VDE, 2013, pp. 1–5. [14] ——, “Ber analysis for linear and widely linear equalization
[8] B. Chang, C. da Rocha, D. Ruyet, and D. Roviras, “Widely linear over frequency selective channels with multiple interferences,”
mmse precoding and equalization techniques for sc-fde systems,” submitted to IEEE Wireless Communications Letters, 2017.
EURASIP Journal on Advances in Signal Processing, vol. 2014,

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SANTA CATARINA. Downloaded on November 02,2023 at 14:14:24 UTC from IEEE Xplore. Restrictions apply.