1
$\begingroup$

So, I tried out what i am learning in digital communication using GNU radio.

The plan is to modulate two baseband signals(sine and cosine) 1kHz with no modulation protocol(don't know any(yet)) to a carrier of 9kHz. I added gaussian noise of 10 amplitude to the passband. Which is followed by BPF, demodulation and LPF. The result(I and Q) and the raw gaussian noise(reference signal) is saved as binary. Flowchart of the system in GNU Radio

I want to do baseband processing like reducing the noise using RLS filtering but i couldn't get this done. The filter output is exactly the same as its inputs. output data(Q) from the GNU radio Filtered output from RLS

Is this because of the filtering(LPF and BPF)?

All the files used are here.

Thank you.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

Is this because the analog filtering(LPF and BPF)

nothing here is analog! This is all digital signal processing

makes the signal less correlated with the noise(reference signal)?

since signal and noise are uncorrelated to begin with, you cannot make them less correlated, so that makes no sense.

Does Filtering changes the noise characteristics in a signal?

Yes, of course. That's the point of a filter! It has a response, and you'll see exactly that, applied to the signal just as well as to the noise.

General note: you converting your complex signal to a pair of real signals makes your processing much harder to understand; stay with the complex numbers! That way, you only need to multiply with a complex sinusoid instead of multiplying with a sine and a minus cosine, you need to apply only one filter instead of a pair of filters and so on. Complex baseband is the abstraction used in modern communications engineering, anyways.

Improve this answer
$\endgroup$
4
  • $\begingroup$ Thanks for the reply. Why cant RLS filtering work in this scenario? About the correlation i got confused, a while ago i did some digging and found the autocorrelation is no longer zero for delay values other than zero after filtering, why does it affect adaptive filtering? $\endgroup$
    – user71499
    Commented Mar 12 at 11:26
  • $\begingroup$ It can work, nothing here says it can't. You have pretty low SNR, so maybe you're starting with a problem that's too hard for a beginning. Yes, the autocorrelation changes, the noise becomes more correlated, not less, with itself. $\endgroup$
    – Marcus Müller
    Commented Mar 12 at 11:32
  • $\begingroup$ Here's something I don't get it, if it becomes more correlated with itself doesn't it mean that it is more predictable?, if my filter is able to find the pattern isn't it just simple subtraction to get what i want regardless of my SNR? $\endgroup$
    – user71499
    Commented Mar 12 at 12:07
  • $\begingroup$ it becomes more predictable. But that doesn't mean any specific predictor designed to predict something else than the noise will be more happy about its presence. And you can't "simply subtract" noise, that's still random. $\endgroup$
    – Marcus Müller
    Commented Mar 12 at 12:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.