LMS and RLS Channel Estimation Algorithms For LTE-Advanced

JOURNAL OF COMPUTING, VOLUME 3, ISSUE 4, APRIL 2011, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 155
LMS and RLS Channel Estimation Algorithms

for LTE-Advanced
Saqib Saleem1, Qamar-ul-Islam
Abstract: For increased data rate and reduced latency for 4G radio communication standards, ITU made proposals for LTE-Advanced in 2009.
rd
To achieve Release-10 targets, made by 3 Generation Partnership Project (3GPP), channel state information at the transmitter is a pre-
requisite. In this paper, analysis of Least Mean Square (LMS) and Recursive Least Square (RLS) channel estimation techniques, using a priori
channel statistics, is drawn for different Channel Impulse Response samples and channel taps for LTE-Advanced system. The effect of different
parameters involved in these adaptive filters is also optimized. MATLAB simulations are used to compare their performance, in terms of Mean
Square Error and Symbol Error Rate, and the complexity in terms of computational time.
Keywords: LTE-Advanced, LMS, LSE, RLS, LMMSE, DM-RS, PUCCH, PUSCH
——————————  ——————————
knowledge of channel by processing a sounding
1. INTRODUCTION reference signal from UE. For DL, user specific
To lower the capital expenses (CAPEX) and Operating demodulation reference signals (DM‐RS) and cell‐
Expenses (OPEX) of the future broadband wireless specific reference signals are proposed. Channel State
network, 3GPP proposed to ITU that the capabilities of Information (CSI) can also be achieved through PUCCH
mobile systems in LTE Release‐10 and beyond should go and PUSCH for DL transmission. Adaptive equalization
beyond IMT‐2000, named as IMT‐Advanced [1]. In LTE‐ is required in case of time‐dispersive and multi‐path
Advanced the requirement of high data rate of 1Gbps fading channel for reliable communication [4]. For this
for DL and 500 Mbps for UL and high spectrum purpose reference signals are transmitted in place of the
efficiency of 30 bps/Hz for DL and 15 bps/Hz for UL can unknown transmitted data. Iterative receivers for 4G
be achieved by employing high order MIMO ( ), mobile standards performing joint detection and
wider transmission bandwidth, adaptive modulation decoding are proposed for high performance gain [5].
and coding (AMC), advanced channel coding, by using For these iterative receivers, adaptive filtering
LDPC convolutional codes, and multihop transmission techniques are most suitable as compared to LSE and
[2]. LMMSE [6]. LMS and RLS algorithms can be used for
At radio bearer, the key technologies used are OFDMA wiener‐based channel estimation, which may or may not
and SC‐FDMA, which are used in a hybrid form and require the second order statistics of the channel and
both utilize OFDM technology as a basis. The larger noise [7]. To the best knowledge of the authors, first
bandwidth requirement of 70 MHz can be achieved by time adaptive filters using channel statistics are
using both symmetric and asymmetric carrier investigated in this paper. Effect of varying step‐size on
aggregation as proposed in Release‐10 [3]. For high performance and complexity is also presented. We also
spectral efficiency, UL transmissions rely on channel show that how these adaptive algorithms can be
reciprocity, for which beam‐forming, SU‐MIMO and optimized by taking filter length and multi‐path channel
MU‐MIMO techniques are used. In this technique the taps into consideration.
transmitter, enhanced NodeB (eNB), gets the The rest of the paper is organized as: Section II describes
the physical layer aspects of LTE‐Advanced according to
Release‐10, in section III different channel estimation
————————————————
algorithms are discussed and their simulation results are
 Saqib Saleem is a student of MS Wireless Communication Engineering of
Department of Communication System Engineering at Institute of Space shown in Section IV and the last section draws the
Technology, Islamabad, Pakistan. conclusion.
 Dr. Qamar-ul-Islam is Head of Department (HoD) of the Department of
Communication System Engineering at Institute of Space Technology,
Islamabad, Pakistan.
2. SYSTEM DESCRIPTION
2.1 Frame Structure
According to Release‐10, in LTE‐Advanced two types of
radio frames are used for UL and DL transmissions:
Generic frame structure and Alternative frame structure.
Generic radio frame, which can be used both for TDD
and FDD transmissions, consist of 20 sub‐frames of
length while alternative radio frame,
which is applied to only TDD transmissions, consist of 2
half‐frames, each of length . Each
half‐frame consist of seven sub‐frames. In both cases
each sub‐frame consist of subcarriers
and the number of OFDM symbols, , are given in
Table 1.
TABLE 1
NUMBER OF OFDM SYMBOLS PER SUB-FRAME Fig.1. Mapping of Reference Signals for Generic Frame Structure
[2]
N symb
Configuration
Generic frame Alternative frame
structure structure TABLE 2
Normal cyclic prefix 7 9 OFDM PARAMETERS FOR FRAME STRUCTURE
Extended cyclic prefix 6 8 Generic Frame Structure Alternative Frame Structure
Cyclic prefix length Cyclic prefix length
Physical layer uses two types of DL signals: Reference Normal CP 160 for l  0
signals and Synchronization signals. One reference 224 for l  0,...,8
144 for l  1,...,6
signal is transmitted per antenna. DL transmit antennas Extended CP 512 for l  0,...,5 512 for l  0,...,7
can be 1,2 or 4. The 2‐dimensional reference signal
sequence is generated according to
2.3 OFDM System Model
, , , Fundamental steps carried out in an OFDM system are
shown in Figure 2. To make ISI free communication, a
Where , is 2‐dimensional orthogonal sequence and wideband signal of bandwidth is converted into flat‐
, is 2‐dimensional pseudo‐random sequence. The fading sub‐carriers. Then all these sub‐carriers are
mapping of DL reference signals for generic frame transmitted over one radio channel by performing IFFT
structure is shown in Figure 1 operation. To make sub‐carriers orthogonal to each
For LTE‐Advanced according to Release 9, two DL other CP of appropriate length is added. At receiver
reference signals are used: Reference signal for PDSCH side, first CP is discarded and then FFT operation is
demodulation and reference signal for Channel State performed on the received symbols, resulting in the
Information (CSI) estimation, which is cell specific, following form [8]
sparse in time/frequency and punctured into data of the
radio sub‐frame.

2.2 OFDM Signal Where subscript shows the corresponding sub‐carrier.
The continuous‐time signal in OFDM symbol is given by is ICI component given by [8]
/
,
, /

For ,
where the value of , is given in Table 2
Fig. 2. OFDM System Model

Where is total number of multipaths. shows ,

AWGN with zero mean and variance. ICI component

is more prominent in high mobility conditions.
Where , is the weighting factor, which is
commonly of the following exponential form
2.4 Channel Model
The channel impulse response can be described as [9]
,
,
Where is less than but should be close to 1.
The necessary steps carried out in RLS algorithm are
where is path gain that undergoes the multipath
Rayleigh fading and is described by the following 1‐ Correlation matrix is updated by
exponential function [10]

| |
2‐ Adaptation gain is given by
Where is multipath delay time and is the power
delay constant.
In frequency domain, Channel Frequency Response is
[11] 3‐ A priori error is given by

∆
,

Where ∆ is the spacing between sub‐carriers. 4‐ The conversion factor is given by

3. CHANNEL ESTIMATION
3.1 RLS Channel Estimation
To implement LSE as adaptive filtering, the desired 5‐ A posteriori error becomes
output and the past data values are required at each
iteration but in RLS algorithm, which is based on LS
estimate of co‐efficients at iteration , to
estimate the co‐efficients at iteration only new data 6‐ The updated co‐efficients are given by
values are required [12].
At iteration , the optimal filter co‐efficients results
in the minimization of the following function
After iteration , the estimated channel is given by

3‐ Error Vector

Where is the length of RLS filter.
4‐ Co‐efficient Updating
The gain vector is given by

And Where is the step‐size parameter.
5‐ Weight error vector is given by
The initialization parameters are

.
The performance of LMS algorithm is expressed in form
and
of Mean Square Error (MSE), defined as
.
Where
Where is found by LSE . is the expectation operator.
Where is the regularization parameter. In conventional
RLS algorithms, the whole reference signals are assigned In wireless communication, the step‐size parameter is of
equal forgetting factor value and the parameters are small value. It is proved in [14] that for the stability of
. and . . For comb‐pilot mode, two‐step LMS algorithm, the optimal adaptation constant value is
forgetting factor value is assigned such that the low
forgetting factor value is used for the first few reference ,
signals to make communication more dependent on Where
channel information while high forgetting factor value is
assigned to the remaining reference signals to make the
communication dependent on the statistical information
of the channel response [13]. is Doppler Spread of channel tap, P is number of
reference signals and is SNR value.
3.2 LMS Channel Estimation
To avoid the matrix inversion, involved in LSE and results in slow co‐efficient updating but better
LMMSE [6], LMS algorithm can be used to solve channel estimation while is fast channel tracking
Wiener‐Holf equation, which may or may not require algorithm with poor estimation because in this case
statistical a priori information of the channel and data. .
A summary of LMS algorithm is given as follows
3.3 Leaky-LMS Channel Estimation
1‐ First estimate the channel, by using LSE Under fast fading conditions, there may be the
technique. possibility of no convergence even for large which
2‐ Filtering gives results in the unstabilization of the LMS algorithms. To
force this un‐damped mode to zero, a leakage co‐
efficient is introduced which gives the following
adaptation [15]
Where

…
Where .
Where is the length of LMS filter.

3.4 Normalized LMS Channel Estimation complexity as shown in Table.4. LMS and RLS take
For minimum disturbance in adaptation of co‐efficients more computational time to converge due to the
the following technique is employed [15] requirement of the second order channel statistics used
in this paper, which are not required for simple LS.

TABLE 3
SYSTEM PARAMETERS
Parameters
Where is only used for the numerical problems. This Frame Structure Generic
approach results in time‐varying step‐size LMS Reference Signals CAZAC
algorithms, whose convergence rate is faster than LMS. Bandwidth 70 MHz
Carrier Frequency 2 GHz
FFT Size 2048
3.5 LMS Channel Estimation Modulation QPSK
For less complex hardware implementation, Power Spectral Density Jake’s Model
Multipath PDP EVA
function is utilized in LMS such that [15]
TABLE 4
COMPARISON OF LS, LMS AND RLS
5000 Simulations 1 OFDM 1 Bit
(mSec) Symbol (nSec)
(nSec)
3.6 Linearly Constrained LMS Channel Estimation LS 0.34 5.24 2.62
LMS 1.9 29.68 14.84
To make estimation technique more optimized, some RLS 1.8 28.12 14.06
constraints are taken into consideration, for which we
have [15] -300
Plot of MSE vs SNR for LS,LMS and RLS Estimator
10
LS
LMS
Where -302
10 RLS

-304
10
is a constant vector.
MSE
-306
3.7 Self-Correcting LMS Channel Estimation 10
The performance of LMS can be improved by
comparing the ideal channel with the estimated channel -308
10
that is closer and closer to the ideal channel. The
estimated channel at iteration is [15] -310
10
5 10 15 20 25
SNR (dB)
Fig.3. MSE v/s SNR for LS, LMS and RLS Estimators
This technique can be implemented by employing any

In LMS, the initial estimated channel can be taken either
LMS algorithms discussed above.
by LS or LMMSE. Figure 4 shows that at low SNR
values LMS, utilizing LMMSE as initial estimated
4. SIMULATION RESULTS channel, demonstrates better performance than LS‐LMS
In this paper, the performance and complexity of the
while at higher SNR, the performance remains almost
proposed algorithms is carried out by using MATLAB
same and only complexity increases in case of LMMSE‐
Monte‐Carlo Simulations in 64‐tap Rayleigh fading
LMS. MSE comparison between LS, LMS and its
channel. The remaining system parameters, according to
different variants is shown in Figure 5. LMS and Leaky‐
Release‐10, are given in Table.3.
LMS show almost same performance for all operating
Figure 3 shows the performance comparison of LS, LMS
SNR values. At low SNR values, the performance of self‐
and RLS. It is clear from Figure 3 that RLS outperforms
correcting LMS degrades but at high SNR it approaches
both LS and LMS for whole range of SNR values. RLS is
to LMS but has 15% more complexity. NLMS and ‐
better than LMS not only in performance but also in
LMS demonstrates better behavior at low SNR but
advantage in case of ‐LMS is in form of reduced 0.5 while further increase of step‐size has no effect on
complexity. No doubt the performance of constrained complexity. The effect of different values of step‐size for
LMS is degraded but the computational time remains LMS is shown in Figure 6. It is clear from Figure 6 that
same for all values of the step‐size. Table.5 shows that there is a wide gap of performance for step‐size 0.1 and
the effect of changing step‐size is most prominent in 0.5 while further increment in step‐size does not affect
case of LMS only and for other techniques the effect is the performance significantly. The computational time
not so much significant. for different step‐size values is shown in Table 6.
-105
Plot of MSE v/s SNR for LS-LMS and LMMSE-LMS Estimator
10 Plot of SNR v/s MSE for LMS Estimator
0
LS-LMS 10
-110
LMMSE-LMS
10
-115 -50
10 10
MSE
LS
MSE
-120 LMS mu=0.5
10
LMS mu=0.1
-100
10 LMS mu=0.9
-125
10
-130
10 -150
5 10
15 20 25 10
5 10 15 20 25
SNR (dB) SNR (dB)
Fig.4. MSE v/s SNR for LS-LMS and LMMSE-LMS Estimators Fig.6. MSE v/s SNR for LMS for different Step-Size Values
PLOT OF SNR V/S MSE FOR AN OFDM SYSTEM using LMS ESTIMATOR
10
5 TABLE 6
COMPUTATIONAL TIME OF LMS FOR DIFFERENT
STEP-SIZE VALUES
0
10 1000 Simulations
(mSec)
0.1 2.6
-5
10 0.5 2
0.9 1.8
MSE
-10
10 Symbol Error Rate for different LMS techniques is
LSE
LMS
Leaky LMS
demonstrated in Figure 7 and SER for different step‐size
-15
10 Sign LMS
NLMS
values is shown in Figure 8. MSE of LMS for different
Constrained LMS
Selfcorrecting LMS
channel taps is shown in Figure 9. As we go on
10
-20 increasing the number of channel taps, the performance
5 10 15 20 25 also improves but this comes at high computational
SNR (dB)
Fig.5. MSE v/s SNR for LMS Estimators complexity as shown in Table.7. From Table.7 it is clear
that by increasing channel taps from 4 to 10, complexity
TABLE 5 increases approximately 15% and for 20 channel taps,
COMPUTATIONAL TIME FOR DIFFERENT LMS complexity increases 32% for 1000 independent
ALGORITHMS
=0.1 =0.5 =0.9
simulations. The performance of LMS for different filter
(mSec) (mSec) (mSec) lengths is shown in Figure 10. Less channel impulse
LMS 9.8 9.7 9.6 response samples show better performance because of
Leaky-LMS 10.5 9.8 9.6
-LMS 10.5 9.8 9.8
high energy concentration while greater number of CIR
NLMS 19.1 19.2 19.2 samples not only degrades the performance but also
Norm-sign-LMS 19 19 19.1 increases the estimator’s complexity. Table.7 shows that
Constrained-LMS 13.2 13.2 13.2
Self-correcting LMS 11.3 10.2 10.1
complexity increase 33% by increasing CIR samples
In Leaky‐LMS, ‐LMS and self‐correcting LMS the from 2 to 5 and 10 CIR samples also results in same
complexity reduces when changing step‐size from 0.1 to increment of computational time.
MSE behavior of Leaky‐LMS for different leakage co‐ Plot of SER vs SNR of LMS Estimator for different mu
efficients is shown in Figure 11, which demonstrates 0.52

that by increasing the leakage co‐efficient the mu=0.5
performance degrades for low SNR while there is no 0.51 mu=0.1
effect on performance for high SNR values. Figure 12 mu=0.9
shows the effect of value of constant term added in
S y m bol E rror Rate

0.5
NLMS. By increasing this small constant to a certain
value the performance improves but by increasing 0.49
further, the performance degrades.
0.48
Plot of SER vs SNR for LMS Estimators
0.6
0.47
0.55
0.5 0.46
5 10 15
20 25 30
0.45 SNR (dB)
Sym bol Error Rate
NLMS Fig.8. SER v/s SNR for LMS for different Step-Size values
0.4
LMS MSE v/s SNR for different Channel Taps for LMS Estimator
-110
0.35 Leaky LMS 10
Constrained LMS Channel Taps=4
0.3
-120 Channel Taps=10
10
0.25 Channel Taps=20
0.2 -130
10
0.15
M SE
-140
0.1 10
5 10 15 20 25 30
SNR (dB)
-150
Fig.7. SER v/s MSE for different LMS Estimators 10
TABLE 7 -160
10
COMPLEXITY COMPARISON OF LMS
1000 1 OFDM 1 Bit
Simulations Symbol (nSec) 10
-170
(mSec) (nSec) 5 10 15 20 25
CIR Samples 2 1.66 25.93 13 SNR (dB)
5 2.2 34.37 17.2 Fig.9. MSE v/s SNR for LMS for different Channel Taps
10 2.32 36.25 18.2
4 1.92 30 15 PLOT OF SNR V/S MSE FOR AN OFDM SYSTEM using LMS ESTIMATOR
-40
Channel Taps 10 2.32 36.25 18.2 10
20 2.52 40 20
Table 8
-60
COMPLEXITY COMPARISON OF RLS 10
1000 Simulations 1 OFDM 1 Bit CIR=2
( Sec) Symbol (nSec) CIR=5
(pSec) CIR=10
CIR Samples 2 51.78 8.09 4.045
MSE
-80
10
5 52.78 8.25 4.125
10 55.65 8.7 4.35
4 468.97 7320 3660
Channel Taps 10 476.20 7500 3750 -100
10
20 538.1 8400 4200
-120
10
5 10 15 20 25
SNR (dB)
Fig.10. MSE v/s SNR for LMS for different CIR Samples
Plot of MSE vs SNR of Leaky-LMS Estimator for different Leakage Co-efficient MSE v/s CIR Samples for RLS Estimator
0.02
SNR =5 dB
Leakge Coefficient=0.1 0.018
SNR =10 dB
Leakge Coefficient=0.9
0.016 SNR =15 dB
Leakge Coefficient=0.5 SNR =20 dB
115 0.014
10 SNR =25 dB
0.012
MSE
0.01
MSE
0.008
110 0.006
10
0.004
0.002
0
1 2 3 4 5 6 7 8 9 10
105 CIR Samples
10 Fig.13. MSE v/s CIR Samples for RLS
10 1216 14
18 20 22 24
SNR (dB)
MSE v/s SNR v/s CIR Samples for RLS EStimator
Fig.11. MSE v/s SNR for Leaky-LMS for different Leakage Co-
efficients x 10
-3
Plot of MSE vs SNR of NLMS for different Constant Term Values 4
Constant Term=1
3
Constant Term=0.5
Constant Term=0.15
MSE 2
Constant Term=2
1
MSE
0
5
-5
10
10 0
15 20
20 40
60
25 80
SNR CIR Samples
6 8 10 12 14 16 18 20 22 24 Fig.14. MSE v/s SNR v/s CIR Samples for RLS
SNR (dB)
Fig.12. MSE v/s SNR for NLMS for different values
MSE v/s Channel Taps for RLS Estimator

For RLS, MSE as a function of CIR samples for different 0.08
SNR values is shown in Figure 13. Performance will be SNR =5 dB
better for less CIR samples and for high SNR values. The 0.07 SNR =10 dB
SNR =15 dB
effect of CIR samples is dominant only for low SNR SNR =20 dB
0.06
values as for high CIR samples, performance remains SNR =25 dB
same for all SNR values while complexity goes on 0.05
MSE
increasing, as shown in Table.8. 2% more time is

required for 5 CIR samples in place of 2 CIR samples 0.04
while it becomes 8% for 10 CIR samples. Overall effect
0.03
of SNR and CIR samples on MSE for RLS is shown in
Figure 14. In Figure 15, it is shown that performance 0.02
improves significantly for increasing channel taps up‐to
10 but further increment does not improve performance 0.01
0 10 20 30 40 50 60 70
and only increases computational time as shown in Channel Taps
Table.8. The combined effect of SNR and channel taps Fig.15. MSE v/s Channel Taps for RLS Estimator
on MSE for RLS is shown in Figure 16.
MSE v/s SNR v/s Channel Taps for RLS EStimator Journal of Electrical and Computer Sciences, Vol. 11 , No.02,
pp.6‐12, April 2011
0.08 [7] Yongming Liang, Hanwen Luo, Jianguo
Huan,”Adaptive RLS Channel Estimation in MIMO
OFDM Systems”, 0‐7803‐9538, IEEE 2005
0.06 [8] Soeg Geun Kang, Min Ha and Eon Kyeong Joo, “A
comparative Investigation on Channel Estimation
Algorithms for OFDM in Mobile Communication”, IEEE
MSE
0.04
Transactions on Broadcasting, Vol. 49, No. 2, June 2003.
[9] Hideo Lobayashi and Kazuo Mori, “ Proposal of OFDM
0.02
Channel Estimation Method using Discrete Cosine
Transform”, IEEE 2004
0 [10] Yen‐Hui Yeh, Sau‐Gee Chen,”DCT‐based Channel
0 Estimation for OFDM Systems”, IEEE Communication
10 Society 2004
0 [11] Andreas Ibing, Konstantinos Manolakis, “
20 20
40 Synchronization Tracking for Cooperative MIMO‐OFDM
30 60
80
with Propagation Delay Differences”, IEEE ISWCS 2008
SNR Channel Taps
[12] M.A.Mohammadi,M.Ardabilipour,”Performance
Fig.16. MSE v/s SNR v/s Channel Taps for RLS Estimator
Comparison of RLS and LMS Channel Estimation
Techniques with Optimum Training Sequences for
MIMO‐OFDM Systems”, 978‐1‐4244‐1980‐7, IEEE 2008
5. CONCLUSION [13] Yongming Liang,Hanwen Luo,”Adaptive RLS Channel
In this paper, two adaptive channel estimation Estimation in MIMO‐OFDM Systems”, Proceedings of
algorithms are optimized based on the channel filter ISCIT, 2005
length and number of multi‐path channel taps. Among [14] Dieter Schafhuber,Markus Rupp,”Adaptive
Identification and Tracking of Doubly Selective Fading
LMS techniques, Leaky‐LMS is proposed both for better
Channels for Wireless MIMO‐OFDM Systems”, 4th IEEE
performance and less complexity, by using small value Worshop on Signal Processing Advances in Wireless
of leakage co‐efficient. We note that the performance of Communications, 2003
RLS is better than conventional estimators, LS and [15] Alexandar D.Poularikas, “Adaptive Filtering Primer with
LMMSE, but its complexity is large due to the MATLAB” Taylor & Francis.
requirement of apriori knowledge of the channel. For
wireless communication such as LTE‐Advanced, a
compromise between performance and complexity can
be achieved by using a channel filter of length 2‐3 CIR
samples and 1‐6 channel taps for RLS but for LMS
channel taps should be more, for example 15‐20, for
better performance.

REFERENCES
[1] 3GPP TS 36.211 v.8.5.0, “Physical Channels and
Modulation”, 2008
[2] 3GPP TS 36.814 v.9.0.0, “Further Advancements for E‐
UTRA Physical Layer Aspects”, 2009
[3] 3GPP Release 10 v.0.0.6, “Overview of Release 10”, 2010
[4] Saqib Saleem, Qamar‐ul‐Islam, ”Optimization of LSE
and LMMSE Channel Estimation Algorithms based on
CIR Samples and Channel Taps”, IJCSI International
Journal of Computer Science Issues Vol. 8 Issue.1,pp.437‐443,
January 2011
[5] Jongsoo Choi, Martin Bouchard and Ter Hin
Yeap,”Adaptive Filtering‐Based Iterative Channel
Estimation for MIMO Wireless Communication”, 0‐7803‐
8834‐8, IEEE 2005
[6] Saqib Saleem, Qamar‐ul‐Islam, ”Performance and
Complexity Comparison of Channel Estimation
Algorithms for OFDM System”, IJECS International

LMS and RLS Channel Estimation Algorithms For LTE-Advanced

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

LMS and RLS Channel Estimation Algorithms For LTE-Advanced

Uploaded by

Copyright:

Available Formats

JOURNAL OF COMPUTING, VOLUME 3, ISSUE 4, APRIL 2011, ISSN 2151-9617

LMS and RLS Channel Estimation Algorithms

Keywords: LTE-Advanced, LMS, LSE, RLS, LMMSE, DM-RS, PUCCH, PUSCH

Fig. 2. OFDM System Model

Where is total number of multipaths. shows ,

This technique can be implemented by employing any

efficients is shown in Figure 11, which demonstrates 0.52

S y m bol E rror Rate

Plot of MSE vs SNR of NLMS for different Constant Term Values 4

MSE v/s Channel Taps for RLS Estimator

increasing, as shown in Table.8. 2% more time is

You might also like