The Hamm windowed complex data forms some sort of a quantized pattern whereas in the same code if tukeywin(0.25) is applied normal noise is observed. Any idea why ?enter image description here

  • $\begingroup$ I will throw in a guess that you are seeing what is equivalent to Intersymbol Interference (ISI) based on a bandlimited channel with a consistent impulse response that is dependent on the previous symbols transmitted. If you rotate the pattern 45°, you can see that there are 3 levels as to where the real and imaginary portions of the waveform are, at the point where the signal is sampled, meaning 3 possible path projections on I and Q, depending on prior symbols sent. Please provide more details of your specific waveform generation and processing $\endgroup$ – Dan Boschen Dec 21 '18 at 13:41
  • $\begingroup$ ....(channel, noise addition, filtering, windowing) up to the point of symbol sampling and that theory or alternate theories will be clearer. $\endgroup$ – Dan Boschen Dec 21 '18 at 13:42
  • $\begingroup$ This is only Hamming windowed and no filter has been added at this point. No noise has been added and directly demodulated after windowing. Does this help? $\endgroup$ – samz12 Jan 8 '19 at 17:16

Assuming we're looking at plots of the time domain I&Q, I'll guess that your multiplication operation is of limited bit width and is causing the quantization.

If you look at the Hamming Window in the time domain, only the central sample(s) are weighted at 1.0, with all the other weights down along a relatively smooth curve. https://en.wikipedia.org/wiki/File:Window_function_and_frequency_response_-Hamming(alpha_%3D_0.53836).svg

If you look at the Tukey Window most of the samples are weighted at 1.0, with some samples on the edges of the window weighted along a curve. https://en.wikipedia.org/wiki/File:Window_function_and_frequency_response_-Tukey(alpha_%3D_0.5).svg

| improve this answer | |
  • $\begingroup$ I am sorry I am trying to wrap my head around this. Kinda new to DSP but I was wondering if there could be a fix to it $\endgroup$ – samz12 Jan 8 '19 at 17:15
  • $\begingroup$ There is no advice I can give you without more information. As @Dan_Boschen mentioned "Please provide more details of your specific waveform generation and processing [...]". You said "[...] directly demodulated after windowing," but gave no details on how you are demodulating. $\endgroup$ – Andy Walls Jan 8 '19 at 17:23
  • $\begingroup$ 1. Remove CP from the 2. FFT the above streams 3. Get the central 1200 sub carriers out of 2048 4. Demodulate using " qamdemod(fft_stream, M(QPSK), 'bin') " 5. Convert above to binary But wht I have scatter plot is only the FFT'ed data. Does this info help? $\endgroup$ – samz12 Jan 10 '19 at 17:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.