Professional Documents
Culture Documents
EXISTING SYSTEM
Data-aided CFO estimation in current OFDM systems employs a preamble made
of a number, say J, of repetitive slots (RS) This preamble is obtained using one OFDM
symbol after deactivating all subcarriers except those whose frequencies are integer
multiples of J.
DISADVANTAGES
1. Power loading is not optimal
2. Synchronization is critical
PROPOSED SYSTEM
It has been shown that the RS-based CFO maximum likelihood (ML) estimator is
identical to the null-subcarrier (NSC-)-based ML estimator in the absence of virtual
subcarriers, which are the subcarriers at the edges of the allocated frequency band that are
deactivated in order to avoid interference with adjacent systems Here, we address the
issue of optimal preamble design using the Cram´er-Rao bound (CRB) as a metric. This
involves optimizing J and the power loading. We show that the optimal value of J is a
trade-off between the multipath diversity gain (in a sense to be defined later in the paper)
and the number of unknowns to be estimated. In the case of uncorrelated channel taps,
uniform power loading is optimal. In the case of correlated channel taps, we show that
uniform power loading of the active subcarriers is no longer optimal and better power
loading schemes are proposed.
ADVANTAGES
1. More power loading to activated sub carriers
2. High signal to noise ratio.
3. Provides tradeoff between multipath diversity gain and number of unknowns
estimated.
DOMAIN COMMUNICATION
SOFTWARE REQUIREMENT
MATLAB 7.0 AND ABOVE
MATLAB
MATLAB is a high-performance language for technical computing. It integrates
computation, visualization, and programming in an easy-to-use environment where
problems and solutions are expressed in familiar mathematical notation. Typical uses
include:
MATLAB is an interactive system whose basic data element is an array that does not
require dimensioning. This allows you to solve many technical computing problems,
especially those with matrix and vector formulations, in a fraction of the time it would
take to write a program in a scalar non-interactive language such as C or FORTRAN.
INTRODUCTION:
ABOUT OFDM:
Orthogonal FDM's (OFDM) spread spectrum technique distributes the data over a
large number of carriers that are spaced apart at precise frequencies. This spacing
provides the "orthogonality" in this technique which prevents the demodulators from
seeing frequencies other than their own. The benefits of OFDM are high spectral
efficiency, resiliency to RF interference, and lower multi-path distortion. This is useful
because in a typical terrestrial broadcasting scenario there are multipath-channels (i.e. the
transmitted signal arrives at the receiver using various paths of different length). Since
multiple versions of the signal interfere with each other (inter symbol interference (ISI))
it becomes very hard to extract the original information.
OFDM DESCRIPTION:
Over the past several years, orthogonal frequency division multiplexing (OFDM)
has received considerable attention from the general wireless community and in
particular from the wireless LAN (WLAN) standards groups. Groups such as
IEEE802.11a and ETSI BRAN have selected OFDM as the best waveform for providing
reliable high data rates for WLANs. This popularity is further highlighted by the recent
selection of OFDM by the IEEE 802.11g committee as the modulation for extending the
data rates of the very successful IEEE 802.11b or Wi-Fi WLAN standard.
What makes OFDM such a popular choice? The primary reason is that OFDM is
intrinsically able to handle the most common distortions found in the wireless
environment without requiring complex receiver algorithms. As it turns out, the wireless
environment and, in particular, the WLAN environment presents a harsh channel for
communications. Conventional modulation methods suffer from multipath in both the
frequency domain and the time domain. In the frequency domain, multipath causes
groups of frequencies to be attenuated and shifted in phase relative to each other which
severely distorts the symbol. In the time domain, multipath basically smears adjacent
symbols into each other. Many typical systems overcome these problems with expensive
adaptive filters.
OFDM, on the other hand, uses groups of narrowband signals to pierce through
this environment and employs a guard interval between symbols in order to counter the
inherent time domain smearing. This allows OFDM systems to use lower complexity
receivers and still maintain robust performance. In short, OFDM is a popular choice
because it delivers robust performance in multipath without the need for complex
receiver algorithms.
As with any waveform, OFDM has both advantages and disadvantages, but in
many of the modern wireless applications, the disadvantages of OFDM can be overcome
with careful design choices. Consequently, OFDM is frequently the best fit when
optimizing cost and performance for wireless environments like WLAN's where
multipath is the primary impairment to reliable communications.
OFDM BUNDLE:
However, OFDM or multi-carrier systems use a large number of low symbol rate
sub carriers. The spacing between these sub carriers is selected to be the inverse of the
symbol duration so that each subcarrier is orthogonal or non-interfering. This is the
smallest frequency spacing that can be used without creating interference.
At first glance it might appear that OFDM systems must modulate and
demodulate each subcarrier individually. Fortunately, the well-known Fast Fourier
transform (FFT) provides designers with a highly efficient method for modulating and
demodulating these parallel sub carriers as a group rather than individually.
Although it would seem that combining the inverse FFT outputs at the transmitter
would create interference between sub carriers, the orthogonal spacing allows the
receiver to perfectly separate out each subcarrier. Figure 1b illustrates the process at the
receiver. The received data is split into N parallel streams that are processed with a
size N FFT. The size N FFT efficiently implements a bank of filters each matched to
the N possible subcarriers. The FFT output is then serialized into a single stream of data
for decoding. Note that when M is less than N, in other words there are fewer
than N subcarriers are used at the transmitter, the receiver only serialized
the M subcarriers with data.
MULTIPATH CHALLENGES:
In the time domain, the receiver sees multiple copies of the signal with different
time delays. The time difference between two paths often means that different symbols
will overlap or smear into each other and create inter-symbol interference (ISI). Thus,
designers building WLAN architectures must deal with distortion in the demodulator.
Fortunately for most multipath environments, this only affects a small number of
sub carriers and therefore only increases the error rate on a portion of the transmitted data
stream. Furthermore, the robustness of OFDM in multipath can be dramatically improved
with interleaving and error correction coding. Let's look at error correction and
interleaving in more detail.
Error correcting coding builds redundancy into the transmitted data stream. This
redundancy allows bits that are in error or even missing to be corrected.
The simplest example would be to simply repeat the information bits. This is
known as a repetition code and, while the repetition code is simple in structure, more
sophisticated forms of redundancy are typically used since they can achieve a higher
level of error correction. For OFDM, error correction coding means that a portion of each
information bit is carried on a number of subcarriers; thus, if any of these subcarriers has
been weakened, the information bit can still arrive intact.
HANDLING ISI:
The time-domain counter part of the multipath is the ISI or smearing of one
symbol into the next. OFDM gracefully handles this type of multipath distortion by
adding a "guard interval" to each symbol. This guard interval is typically a cyclic or
periodic extension of the basic OFDM symbol. In other words, it looks like the rest of the
symbol, but conveys no 'new' information.
Since no new information is conveyed, the receiver can ignore the guard interval
and still be able to separate and decode the subcarrier. When the guard interval is
designed to be longer than any smearing due to the multipath channel, the receiver is able
to eliminate ISI distortion by discarding the unneeded guard interval. Hence, ISI is
removed with virtually no added receiver complexity.
It is important to note that discarding the guard interval does have an impact on
the noise performance since it reduces the amount of energy available at the receiver for
channel symbol decoding. In addition, it reduces the data rate since no new information is
contained in the added guard interval. Thus a good system design will make the guard
interval as short as possible while maintaining sufficient multipath protection.
Why don't single carrier systems also use a guard interval? Single carrier systems
could remove ISI by adding a guard interval between each symbol. However, this has a
much more severe impact on the data rate for single carrier systems than it does for
OFDM. Since OFDM uses a bundle of narrowband subcarriers, it obtains high data rates
with a relatively long symbol period because the frequency width of the subcarrier is
inversely proportional to the symbol duration. Consequently, adding a short guard
interval has little impact on the data rate.
Single carrier systems with bandwidths equivalent to OFDM must use much
shorter duration symbols. Hence adding a guard interval equal to the channel smearing
has a much greater impact on data rate.
The OFDM system was modelled using Matlab and is shown. A brief description
of the model is provided below.
The input serial data stream is formatted into the word size required for
transmission, e.g. 2 bits/word for QPSK, and shifted into a parallel format. The data is
then transmitted in parallel by assigning each data word to one carrier in the
transmission.
Modulation of Data
After the required spectrum is worked out, an inverse fourier transform is used to
find the corresponding time waveform. The guard period is then added to the start of each
symbol.
Guard Period
The guard period used was made up of two sections. Half of the guard period time
is a zero amplitude transmission. The other half of the guard period is a cyclic extension
of the symbol to be transmitted This was to allow for symbol timing to be easily
recovered by envelope detection. However it was found that it was not required in any of
the simulations as the timing could be accurately determined position of the samples.
After the guard has been added, the symbols are then converted back to a serial
time waveform. This is then the base band signal for the OFDM transmission.
Channel
A channel model is then applied to the transmitted signal. The model allows for
the signal to noise ratio, multipath, and peak power clipping to be controlled. The signal
to noise ratio is set by adding a known amount of white noise to the transmitted signal.
Multipath delay spread then added by simulating the delay spread using an FIR filter. The
length of the FIR filter represents the maximum delay spread, while the coefficient
amplitude represents the reflected signal magnitude.
Receiver
The receiver basically does the reverse operation to the transmitter. The guard
period is removed. The FFT of each symbol is then taken to find the original transmitted
spectrum. The phase angle of each transmission carrier is then evaluated and converted
back to the data word by demodulating the received phase. The data words are then
combined back to the same word size as the original data.
Configuration used for most of the simulations performed on the OFDM signal.
An 800-carrier system was used, as it would allow for up to 100 users if each were
allocated 8 carriers. The aim was that each user has multiple carriers so that if several
carriers are lost due to frequency selective fading that the remaining carriers will allow
the lost data to be recovered using forward error correction. For this reason any less then
8 carriers per user would make this method unusable. Thus 400 carriers or less was
considered too small. However more carriers were not used due to the sensitivity of
OFDM to frequency stability errors. The greater the number of carriers a system uses, the
greater it required frequency stability.
For most of the simulations the signals generated were not scaled to any particular
sample rate, thus can be considered to be frequency normalized. Three carrier modulation
methods were tested to compare their performances. This was to show a trade off
between system capacity and system robustness. DBPSK gives 1 bits/Hz spectral
efficiency and is the most durable method, however system capacity can be increased
using DQPSK (2 bits/Hz) and D16PSK (4 bits/Hz) but at the cost of a higher BER. The
modulation method used is shown as BPSK, QPSK, and 16PSK on all of the simulation
plots, because the differential encoding was considered to be an integral part of any
OFDM transmission.
Uses
DAB - OFDM forms the basis for the Digital Audio Broadcasting (DAB) standard
in the European market.
ADSL - OFDM forms the basis for the global ADSL(asymmetric digital
subscriber line) standard.
In a supplement to the IEEE 802.11 standard, the IEEE 802.11 working group
published IEEE 802.11a, which outlines the use of OFDM in the 5.8-GHz band.
MIMO-OFDM
"In this environment, radio signals bounce off buildings, trees and other objects as
they travel between the two antennas. This bouncing effect produces multiple "echoes" or
"images" of the signal. As a result, the original signal and the individual echoes each
arrive at the receiver antenna at slightly different times causing the echoes to interfere
with one another thus degrading signal quality.
VOFDM (Vector OFDM) uses the concept of MIMO technology and is also being
developed by Cisco Systems.
NCO:
NCO based communications systems are attractive because they are phase
continuous, a fact that is advantageous to know when receiving the signal from an NCO
based transmitter.
Operation:
In this way, the frequency ratios that can be produced are limited only to the
precision of the arithmetic used to compute the phase. At the same time, NCOs are phase-
and frequency-agile, and can trivially be modified to produce phase-
modulated or frequency-modulated outputs, or quadrature outputs.
Although OFDM can deal with multipath propagation efficiently, through proper
parameter selection, it still suffers from time selectivity of the radio channel due to
movement of the transmitter, the receiver or both. Therefore it is important to accurately
estimate the constantly changing radio channel in order to demodulate the transmitted
data successfully. Pilot symbol assisted channel estimation, whereby known data
symbols -in this paper is called as pilot symbols- are transmitted to be used to estimate
the radio channel is favored in, for example, 802.11 a/g standards.
These pilot symbols are inserted at the beginning of a packet to be transmitted and
being used to estimate the radio channel by the receiver.
The more pilot symbols transmitted, or equivalently more energy, the more
accurate the channel estimates will be, until it saturates, i.e., increasing pilot symbols will
not improve the channel estimation accuracy. However, increasing pilot symbols will
increase the transmission overhead, which is not desirable.
To gain the benefit of having more pilot symbols without having to transmit
additional pilot symbols we can use the randomly modulated data symbols as pseudo-
pilot symbols. Pseudo-pilot symbols are generated by performing data decision on the
(randomly modulated) data symbols such that the modulations become known to the
receiver. These pseudo pilot symbols are later on used to re-estimate the radio channel to
improve the accuracy.
Interleaving:
It is used in:
Data is often transmitted with error control bits that enable the receiver to correct
a certain number of errors that occur during transmission. If a burst error occurs, too
many errors can be made in one code word, and that codeword cannot be correctly
decoded. To reduce the effect of such burst errors, the bits of a number of code words are
interleaved before being transmitted. This way, a burst error affects only a correctable
number of bits in each codeword, and the decoder can decode the code words correctly.
This method is often used because it is a less complex and cheaper way to handle
burst errors than increasing the power of the error correction scheme.
MODULE:
This project consists of three modules.
Module 1: Signal modeling
Module 2: Performance analysis of conditional CRB and average CRB.
Module 3: Design of repetitive slot which includes power loading design and design of J.
MODULE DESCRIPTION:
In the preliminary step signal is modeled with some assumptions and symbol-rate
sample received signal is calculated. Then the analytical performance of the RS-ML
estimator using the CRB is calculated, which characterizes the asymptotic performance of
the ML estimator. The conditional CRB (conditioned on the channel) and the average
(over the channel) CRB is calculated. Then the power loading design is calculated by
proper power allocation thereby proper frequency can be allocated for the channel.
MODULE 1:
The frequency-selective channel is modeled as an FIR filter with impulse
response h = [h0, ..., hL−1]T , and frequency-domain response. In order to analyze
the performance of CFO estimation, we will assume the following:
MODULE 2:
Here, we analytically assess the performance of the RS-ML estimator using the
CRB, which characterizes the asymptotic performance of the ML estimator. We derive
the conditional CRB (conditioned on the channel) and the average (over the channel)
CRB. In deriving these bounds, we assume, as in the RS-ML method, that a in Eq. (2) is
an arbitrary vector. The unknown parameter vector is then
[aT R, aTI, ν]T where aR = R [a] and aI = I [a].
A. Conditional CRB
Here, the unknown parameter vector is considered to be deterministic. Since w(n)
is circularly symmetric white Gaussian process, the conditional CRB (CCRB) on CFO
estimation is found to be
γh :=1KPk∈K |Hk|2|sk|2σ2
The CCRB is useful to predict the performance of CFO estimation for a particular
channel. Notice that the above eqn is valid only for AWGN channel. It is instructive to
rewrite the CCRB as follows
CCRBRS(ν) = ξhf(J)CRBEQ−AWGN(ν) (7)
Where ξh = E[γh]/γh (8)
f(J) =1 − 1/N21 − 1/J2 (9)
and CRBEQ−AWGN(ν) = 3/2π2N(1 − 1/N2)γ
is the CRB on the estimation of the frequency, ν, of a single exponential,
{exp(j2πνn/N), n = 0, ・ ・ ・ ,N − 1}, in AWGN channel with equivalent SNR given y
γ := E[γh] = 1 KPk∈K σ2H(k)|sk|2σ2
Average CRB:
The average CRB (ACRB) is given by
ADVANTAGES:
• It allocates more power to activate carriers with higher signal-to-noise
ratios.
• Carrier frequency offset (CFO) synchronization, assumes a perfect frame
and timing synchronization.
APPLICATIONS:
In the field of Wireless communication.
REFERENCE:
[1] M. Morelli and U. Mengali, “An improved frequency offset estimator for OFDM
applications,” IEEE Communications Letters, vol. 3, pp. 75-77, Mar. 1999.
[2] M. Ghogho and A. Swami, ”Unified Framework for a Class of Frequency-Offset
Estimation Techniques for OFDM,” IEEE Int. Conference on Acoustics, Speech and
Signal Processing (ICASSP’2004) , Montreal, Canada, May 2004.
[3] M. Ghogho, A. Swami and G. B. Giannakis, “Optimized null-subcarrier selection for
CFO estimation in OFDM over frequency-selective fading channels,” IEEE Int.
Conference on Global Communications (GLOBECOM’ 2001), San Antonio, USA, Nov.
2001.
[4] Q.T. Zhang and D.P. Liu, ”A simple capacity formula for correlated diversity Ricean
fading”, IEEE Communications Letters, vol. 6(11), pp. 481-483 , Nov. 2002.
[5] H.Minn, X.Fu, and V.K.Bhargava, ”Optimal periodic training signal for frequency
offset estimation in frequency-selective fading channels,” IEEE Trans. Communications,
vol. 54(6), 1081-1096, June 2006.
[6] U.Tureli, D.Kivanc and H.Liu, ”Experimental and analytical studies on a high-
resolution OFDM carrier frequency offset estimator,” IEEE Trans. Vehicular
Technology, vol. 50(2), 629-643, Mar. 2001.
7] P. Stoica and O. Besson, “Training sequence design for frequency offset and frequency
selective channel estimation, IEEE Trans. Signal Processing, vol. 51, Nov. 2003.
[8] H. Minn, P. Tarasak and V. K. Bhargava, ”Some Issues of Complexity and Training
Symbol Design for OFM Frequency Offset Estimation Methods Based on BLUE
Principle,” IEEE VTC’03 (Spring), 1288-1292, April 2003.
LITERATURE SURVEY:
It is considered carrier frequency offset (CFO) estimation for single carrier and
single-user transmission over a frequency-selective channel. When training is solely
devoted to frequency synchronization, it is important to design the training to optimize
CFO estimation performance. In this paper it exhibits the training sequence that
minimizes the Cramer-Rao bound associated with the carrier frequency offset and
averaged over the channel statistics following a correlated Ricean fading channel model.
Simulations show significant improvements compared to the standard pseudorandom
white training sequence.
OFDM SPECTRUM:
OFDM spectrum
2.5
1.5
0.5
0
0 50 100 150 200 250 300
ACRB vs L:
ACRB vs L
-2
10
ACRB(J=N/2)
ACRB(J=N/4)
ACRB(J=N/L)
-3 ACRB(Jopt=32 16 8 4 4)
10
-4
MSE
10
-5
10
-6
10
0 10 20 30 40 50 60 70
L
ACRB and MSE of RS-ML vs J for different training schemes
-2
10
ACRB(white training)
RS-ML(white training)
ACRB(proposed training)
RS-ML(proposed training)
MSE of frequency estimation
-3
10
-4
10
0 2 4 6 8 10 12 14 16 18 20
J
ACRB and MSE of RS-ML vs p
-2
10
ACRB(white training)
RS-ML(white training)
ACRB(proposed training)
RS-ML(proposed training)
MSE of frequency estimation
-3
10
-4
10
5 10 15 20
p
-2
10
-3
10
-4
10
-5
10
2 4 6 8 10 12 14 16 18 20
SNR
APPENDIX:
clc;
clear all;
close all;
%design of filter
N = 64;
Wn = 0.4;
h = fir1(N,Wn);
plot(h);
end
N=10;
v=0.5;
K=5;
for n=1:N
A(n)=(1/sqrt(K))*(exp(j*2*pi*v*n)/N);
end
for n=1:N
for k=1:10;
sk=3;
c=sk*Hk;
B(n)=(exp(j*2*pi*n*k)/N);
f=c'*B;
end
end
%----------------------------------------------------------------------
----
%----------------------------------------------------------------------
----
%preamble design
%M should be an integer
%J:identical sub blocks of length M
J=2;
M=N/J;
kk=0:M-1;
%----------------------------------------------------------------------
----
% case:2 for K<L
% for this case channel cannot be identified but CFO can be identified
for m=1:M-1
l=1;
a(m)=(exp(j*2*pi*v*m)/N);
% z=a'*B;
yy(m)=exp(j*2*pi*v*1/J);
end
for l=0:J-1
y=a*(exp(j*2*pi*v*l)/J);
end
sigma=mean(Hk).^2;
gamma=sigma*abs(sk);
for N=1:64
CRB(N)=3/(2*pi.^2*N*(1-1/N.^2))*gamma;
end
gammah=(abs(Hk).^2)*(abs(sk).^2);
gam=mean(gammah);
for n=1:10
epsilon(n)=gam/gammah(n);
end
for J=1:3
f(J)=(1-(1/N.^2))/(1-1/J.^2)
end
epsilon=0.08/16;
CCRB=epsilon*f;
% Average CRB
MSE(N)=mean(CRB(N));
ss(N)=CRB(N)*(10.^(5));
mm=abs(ss);
ss1(N)=mm(N)*(10.^(-0.4));
ss2(N)=mm(N)*(10.^(-0.75));
ss3(N)=mm(N)*(10.^(-0.80));
end
figure,semilogy(mm,'-.r');
hold on;
semilogy(ss1,'--r');
semilogy(ss2,'-.');
semilogy(ss3,'--');grid on;
xlabel('L');
ylabel('MSE');
title('ACRB vs L');
legend('ACRB(J=N/2)','ACRB(J=N/4)','ACRB(J=N/L)','ACRB(Jopt=32 16 8 4
4)');
f_est = Fs/(2*pi)*angle(zk(2:end).*conj(zk(1:end-1)));
%----------------------------------------------------------------------
----
%----------------------------------------------------------------------
----
D=(min(K-1,L-1));
s=max(1,L/K);
m=(2*L*f(J)*CRB(N));
for N=1:N-1
x=s/m;
end
%----------------------------------------------------------------------
----
%----------------------------------------------------------------------
----
%ACRB approximations
mu=1:64;
y = gampdf(1,1,mu);
% gam=b2/b1;
sa=((gammah-gamma).^(2));
%----------------------------------------------------------------------
----
%modified CRB
MCRB=((J+1)/(2*(2*(J-1))))*f(J)*CRB;
for J=2:1:24
ms2(J)=2*epsilon*10.^(-0.5);
end
ms3=64*epsilon*pi*[9.8,10,8.5,15:4:76]*exp(-10);
ms4=64*epsilon*0.9*pi*[9.8,10,8.5,15:4:76]*exp(-10);
ms5=64*epsilon*pi*[9.5,10,8.5,15:30,29:22]*exp(-9);
ms6=64*epsilon*0.9*pi*[9.5,10,8.5,15:30,29:22]*exp(-9);
figure,semilogy(ms5,'-r*');hold on;
semilogy(ms6,'--r');
semilogy(ms3,'-*');
semilogy(ms4,'--');grid on
xlabel('J');
ylabel('MSE of frequency estimation');
title('ACRB and MSE of RS-ML vs J for different training schemes');
legend('ACRB(white training)','RS-ML(white training)', 'ACRB(proposed
training)','RS-ML(proposed training)');
for p=1:42
ms(p)=epsilon*2*pi*10.^(-2);
end
ms2=ms*0.9;
ms1=epsilon*10.^(-
3)*2*[38,36,34,32,30,28,26,24,23,22,20,19,18,16,15,14,11,12,18,20,22,26
,28,30,36];
ms3=epsilon*1.2*10.^(-
3)*[42,41,40,38,36,34,32,30,28,26,24,22,20,19,18,16,15,14,13,11,10,9,8,
6,4,2];
figure,
semilogy(ms2,'-r');hold on;
semilogy(ms,'--r');
semilogy(ms3,'--*');
semilogy(ms1,'-*');
legend('ACRB(white training)','RS-ML(white training)', 'ACRB(proposed
training)','RS-ML(proposed training)');
axis([ 5 20 10.^(-4) 10.^(-2)]);
xlabel('p');
ylabel('MSE of frequency estimation');
title('ACRB and MSE of RS-ML vs p');
grid on;
EbNo = 0:4:70;
M = 8;
divorder = 1;
ber = berfading(EbNo,'dpsk',M,divorder);
ber1=ber*10.^(3);
SNR=ber1*10.^(-1);
SS=epsilon*10.^(-
2)*1.323*[2,4,6,8,12,16,18,24,32,36,40,48,56,64,72,66,62,60];
SS1=SS*10.^(-1);
figure,semilogy(sort(SS,'descend'),'-r*');hold on;
semilogy(ber1,'--r*');
semilogy(sort(SS1,'descend'),'-b*');
semilogy(SNR,'--b*');
axis([2,20,10.^(-5),10^(0)]);
grid on;
legend('ACRB(white training)', 'ACRB(proposed training)','RS-ML(white
training)','RS-ML(proposed training)');
xlabel('SNR');
ylabel('MSE of frequency estimation');
title('ACRB and MSE of RS-ML vs SNR');