correlation - Autocorrelation in Matlab WITHOUT xcorr -

so i'm expected make autocorrelation of multiple signals (in order see difference of power spectrum it's irrelevant) note: don't want other way resolve exercise explanation problem :)

the signals periodic , data length multiple of period.

the problem don't same answer xcorr function of matlab. i'm using classic (i think) r_x(lag) = sum(k=-inf +inf) of x(k)*x(k-lag). when searching answer found other method 1 used using basically: ifft(fft().*conj(fft())) if understood correctly taking inverse fourier transform of periodigram equivalent power spectrum, fourier transform of autocorrelation. don't know why "classic" way doesn't work..

can please explain why/if code wrong?

if don't want run code can go directly % ! autocorrelation!

code:

% making prbs y, of period 2^size(a,1) -1 , span [-lambda,lambda] = [   1 0 0 0 0 1;...     1 0 0 0 0 0;...     0 1 0 0 0 0;...     0 0 1 0 0 0;...     0 0 0 1 0 0;...     0 0 0 0 1 0];   c = [   0 0 0 0 0 2*lambda];  x_0 = [ 1; 0; 1; 1; 1; 0]; % can 1 , 0 , @ least 1 one  m = 63; %found earlier n = 4*m-1;  x = zeros(6,n+1); y = zeros(1,n+1); x(:,1) = x_0;  = 1:n+1     z = a*x(:,i);     x(:,i+1) = mod(z,2);     y(i) = c*x(:,i) - lambda; end  % ! autocorrelation !   r_y = zeros(1,m);  % lambda going 0 m-1   tau = 0:m-1      iter = 1;  % periodicity      k = 0:n-1          if k == m-1  %when enters new period              iter = iter +1;          end          if k-tau+1 <= 0  %if lag tau takes y(i) out of boundary take y(i+period) in boundary              r_y (tau+1) = r_y(tau+1) + 1/n * y(k+1)*y(k-tau+iter*m+1);          else               r_y (tau+1) = r_y(tau+1) + 1/n * y(k+1)*y(k-tau+1);          end      end  end % compare  r_yy = xcorr(y,m); r_yy = r_yy(m+1:end-1)/n;   x_axis = [0:m-1]; figure plot(x_axis,r_y, x_axis,r_yy); grid on legend('mine','xcorr') 

i've tried adding comments explain reasoning.

thank lot answers!!

b.elling