OFDM frequency sensitivity (MATLAB related)

[O.J.]

Hello everyone,

I am trying to simulate an OFDM system and part of what I want to do is investigate the Bit Error Rate with respect to frequency offset. I understand how the offset vs BER curve should be looking like the curve obtained in the code in this link http://www.dsplog.com/db-install/wp-content/uploads/2009/08/script_frequency_offset_ofdm.m which is available on this page:
http://www.dsplog.com/2009/08/08/effect-of-ici-in-ofdm/
All u have to do is run the code in MATLAB and you will see the correct offset vs error plot.

I am developing a code with which the offset vs BER plot seems incorrect. By that I mean, the error starts growing slowly from the 0 offset (no offset) and slowly grows (like a parabola) when the slope of the curve should be steepest (theoretically) just above or below zero offset and then should start to slow down approaching +/- 0.5 offset.

I cannot understand what I should modify in my code to address this issue. Here is the code I am using:
%% OFDM parameters
N = 2048; %number of carriers
Tu = 0.001; %symbol period for each carrier
T = Tu/N; %Symbol period of serial stream
R = 1/8; %guard time ratio (WiMAX standard) R*N should be integer
G = Tu*R; %guard time
fu = 1/Tu; %symbol rate of each datastream
Fs = fu*N; %sampling frequency
Q = 4; %upsampling factor



%% TRANSMITTER

%generate message
Mt = randint(1,N);
tn = 0:N-1;
stem(tn/Fs,Mt); xlabel('Time (seconds)'); ylabel('Binary Data');

%BPSK (mapping 0's to -1's)
M1=2*Mt-1;
stem (tn,M1); %plot the original message (BPSK)
xlabel('Time (seconds)'); ylabel('BPSK Data');

%S/P conversion
A = reshape(M1,N,1);

%modulate each subcarrier using IFFT
B = ifft(A);

%P/S conversion
C = reshape(B,1,N);

%guard time/cyclic prefixing
Guard = zeros(1,length(1:R*N));
C_gcp = [C Guard];
C_gcp(N+1:N+R*N) = C_gcp(1:R*N);

%oversample by a factor of Q (DAC ~)
D = ones(Q,1)*C_gcp;
D = conj(D(:)');

%plot the OFDM signal
t = 0:Q*(N+R*N)-1;
plot (t/(Q*Fs),D);xlabel ('Time (seconds)');ylabel ('OFDM Signal');title ('OFDM Symbol (with guard time & cyclic prefix) vs Time');SNR = 10;
BER = 0;
BER_t = 0;
N_Max = 50;
S = [];
f0 = 1/Tu;
offset = -0.5:0.1:0.5;
t = 0:Q*(N+N*R)-1;

for tr = 1:length(offset);
for n = 1:N_Max

%% RECEIVER

%noise
D_n = awgn(D,SNR,'measured');

%frequency offset
D_n_o = D_n.*exp(sqrt(-1)*2*pi*offset(tr)*f0*t/(Q*Fs));

%Downsampling
E = [];
for j =1:N+R*N
E(j) = D_n_o(j*Q);
end

%Removing CP
F = zeros(1,N);
F(1:N) = E(1:N);

%S/P
M2 = fft(F);

%Demapping
h = modem.pskdemod(2,pi);
Mr = demodulate(h,M2);

%Bit Error Computation
Error = length(find(Mt~=Mr));
Error_Rate = Error/length(Mt);
BER_t = BER_t + Error_Rate;
end
BER = BER_t/N_Max;
S = [S BER];
BER_t = 0;
end

plot(offset,S); xlabel('Normalized Offset'); ylabel('Bit Error Rate');
grid on

 

[O.J.]

bump...