I have two loops, how can i get rid of the inner loop "for j=1:nSym"?

for i=1:length(EsN0dB),
    for j=1:nSym
%        
%-----------------Transmitter--------------------
%---Generating a uniformly distributed random numbers in the set
%[0,1,2,..,M-1] i.e from [0 to 15]
info = ceil(M.*rand(1,Ntot))-1; % the reason we subtract '1' is because we want to go from '0' to '15 instead of 
% '1' to '16'.
%
% 16QAM constellation maping
x=RefPoint(info+1); % The reason we add 1 to info is because the array must be positive and since we have a '0'in info, we need add '1'.
%
%--- Reference Constellation for demodulation and Error rate computation-- 
refI = real(RefPoint);
refQ = imag(RefPoint);
%
%--------------Channel Modeling ----------------
%Adding noise with variance according to the required Es/N0
Esym=sum(abs(ofdm_signal).^2)/(length(ofdm_signal)); %Calculate actual symbol energy from generated samples
N0=Esym/(10.^(EsN0dB(i)/10)); %Find the noise spectral density
noiseSigma = sqrt(N0/2); %Standard deviation for AWGN Noise
n = noiseSigma*(randn(1,length(ofdm_signal))+1i*randn(1,length(ofdm_signal))); %Add noise in IQ plane.
%noise=1/sqrt(2)*(randn(1,length(ofdm_signal))+1i*randn(1,length(ofdm_signal)));
%r= sqrt((N+Ncp)/N)*ofdm_signal + n;
r= ofdm_signal + n;
%
 %-----------------Receiver----------------------
 %Removing cyclic prefix
 r_Parallel=r(Ncp+1:(N+Ncp));
%
 %FFT Block
 r_Time=1/sqrt(N)*(fft(r_Parallel,N));
 %Extracting the data carriers from the FFT output
 R_Freq=r_Time([(2:Ntot/2+1) (Ntot/2+13:Ntot+12)]);
%  
% %-------------I-Q Branching--------------- 
r_i = real(R_Freq);
r_q = imag(R_Freq);
%
 %Detection in the receiver - Euclidean distance Method
 [estimatedTxSymbols,dcap]= iqOptDetector(R_Freq,RefPoint);
%
 numErrors = sum((info~=dcap-1)); %Count number of errors
        %Accumulate bit errors for all symbols transmitted
 errors(i)=errors(i)+numErrors;
    end
end

Answer 1

[Walter Roberson]

The line

Esym=sum(abs(ofdm_signal).^2)/(length(ofdm_signal)); %Calculate actual symbol energy from generated samples

does not depend upon i or j or upon any data that is being changed in the nested loops. Move that line to before both loops.

The line

N0=Esym/(10.^(EsN0dB(i)/10)); %Find the noise spectral density

does not depend upon j, so it can be moved to before the "for j" loop. It could be further vetorized to

    all_N0 = Esym/(10.^(EsN0dB/10)); %Find the noise spectral density

and then in the "for i" loop,

N0 = all_N0(i);

However, N0 is only used in

noiseSigma = sqrt(N0/2); %Standard deviation for AWGN Noise

so instead that could be vectorized,

all_noiseSigma = sqrt(all_N0/2);

and then inside the the "for i" loop,

noiseSigma = all_noiseSigma(i);

... and so on. Keep picking at the code and finding places that do not change inside a loop and move those to before the loop. Places where you operate on one element at a time, move to before the loop and operate on as a vector and then (if necessary) index the vector result inside the loop.

[Ray]

Thanks. i will take a look and implement your comments.