hassiweb / otfs-chan-est-and-eq Goto Github PK

Hello sir, do you have the code for evaluating the normalized mean square error (NMSE) according to your paper?

  % The FDF is not a causal filter, especially if the delay is less than 1.
  
  % In this case, the FDF cannot keep the energy of an input signal such as a discontinuous signal (e.g., delta function).
  
  % To keep the same energy between discontinuous signals and continous signals, operations below convert parameters to express fractional delays.
  
  this.pathOrigRayleighGains = this.pathRayleighGains;
  
  this.pathOrigDopplers = this.pathDopplers;
  
  [this.pathRayleighGains, this.pathDopplers, this.pathDelays, this.pathOffsets] = fdfpathpars(this.pathRayleighGains, this.pathDopplers, this.delayProfile, this.pathOffsets, this.SamplingRate);
  
  this.pathDelaysInSample = this.pathDelays*this.SamplingRate;

Why update the path parameters through the fractional filter, I see that the number of paths has increased after the update, can the basis of the operation here provide some reference? ^_^

Sir kindly help in this regard
getf1f2' is used in Parallel Concatenated Convolutional Coding: Turbo Codes.

Error in LteChannelCoder (line 88)
[f1, f2] = getf1f2(Kplus); % defined in TS36.212 section 5.1.3.2.3 Table 5.1.3-3

Error in OTFS (line 119)
hChannelCoder = LteChannelCoder(

if not(isempty(denominator(denominator<threshold)))
Pt = sqrt(MN)0.9472; % for 16QAM
Pr = mean(abs(Ydd_dft(:)));
Ph = (Pr - noiseVarsqrt(MN));
Ph_est = mean(abs(Hdd_dft(:)));
est_err = (Ph_estPt - Ph)/sqrt(MN)/2;
denominator = denominator + est_err;
end
What do Pt, Pr, Ph and Ph_est stand for?
Hope you can tell me, thank you

You had written about the reason why using the stopping criteria.
But I am confused about the "if " in the repeat loop.
Why stop when finding the current path has a higher gain than the last?

Hi,

I am very happy to find your code. But when I want to reproduce your results by running your code, one issue happened. The hChannelCoder.decoder outputs the whole decoded code, so its dimension will not meet with txBits's dimension. Besides, the BLERs of all SNRs are 1.00 when I run this code. I think that's because of the decoder problem. Can you fix it? Thanks.

best,
Xiang

P.S. I use MATLAB 2019 to run your code on WINDOWS. I don't know if my problem comes from a different platform.

Hello!

Thank you for sharing these interesting results and making the codes available.

I would like to know if you have any implementation done for OTFS Pilot-Based channel estimation in a channel that does not consider fractional doppler.

I started working on DNN-based channel estimation for OTFS with pilot insertion and channel estimation being done in the frequency domain, taking advantage of its properties of OTFS compared to OFDM and without increasing the complexity of going to DD domain estimation. I think that comparing the estimators I am proposing to your proposal will be a great indicator of the next steps, but I'm facing problems in changing the functions that implement the fading to design a channel without this fractional effect. That is crucial to have identical scenarios for a fair comparison between our estimators.

Again, thank you so much for your valuable work and help.