I have a practical question: I have a received block of length K symbols. Now, which is more complex, to take the integration from kTs to (k+1)Ts for k=0,1,...,K-1, or take the K-point FFT for all symbols?
Integration, in its simplest form, is nothing more than addition, so all you need to implement an integrator is an accumulator. An FFT is not a simple integration.
Why would either one work for you? It's like saying "which is easier, driving to France or driving to China". You're not going to end up in the same place.
Integration, in its simplest form, is nothing more than addition, so all you need to implement an integrator is an accumulator. An FFT is not a simple integration.
Why would either one work for you? It's like saying "which is easier, driving to France or driving to China". You're not going to end up in the same place.
I am not trying to use either, but I want to compare in terms of hardware complexity and time consumption between my system with Np integrators, followed by one FFT/IFFT pair, and another system with one IFFT and Np FFT units. This has to do with different modulation schemes.
In both systems I use the same block size of N symbols, with oversampling factor of 1000, i.e., the block size after oversampling is K=1000*N. So, in each integration I need to sum over 1000 samples N times, a total of 1000*N*Np for all integrators. This is followed by one N-point FFT/IFFT units. On the other hand, for the other system one K-point IFFT and Np K-point FFT units are required. How many summations and multiplication in terms of K there are in each FFT unit?