1

I read about Time recovery (as an example, the Mueller&Muller algorithm); it is linked to when you should sample your signal.

I am currently working on a project with GNU Radio and there is some blocks that perform time recovery.

I don't get how those blocks actually perform a time recovery in the digital world because they can't change the sample that the USRP produces... I don't understand how such algorithm can be implemented on the digital side with out affecting the USRP itself which is the device sampling the signal.

Marcus Müller
  • 30,525
  • 4
  • 34
  • 58
MathieuL
  • 765
  • 1
  • 8
  • 12
  • There are two kinds of sampling going on. The USRP turns a continuous signal into a discrete signal. After matched filtering (implemented in digital backend), the signal is sampled again at specific time instants, to get the transmitted bits. This second sampling is what M&M is about. – MBaz May 27 '16 at 16:18
  • @MBaz so the samp_rate only affect how you sample the already sampled stream of data that come from the USRP? So the samp_rate is a downsampling? – MathieuL May 27 '16 at 17:14
  • Depends on what you mean by samp_rate. If it is the number you give the USRP, then no, the USRP's sample rate is the rate at which the RF continuous signal is discretized. – MBaz May 27 '16 at 18:32
  • Well you are saying that the USRP sampling rate is constant what ever the samp_rate variable is set @MBaz – MathieuL May 27 '16 at 19:19
  • samp_rate is a variable. You can use it to specify anything. You can use it to specify the sampling rate of the USPR, or you can use it for something else. You have not specified what that variable is. – MBaz May 28 '16 at 01:01
  • @MathieuL are you aware of what pulse shaping does? – Marcus Müller May 28 '16 at 09:27
  • @MarcusMüller mhm is that the effect cause by the non ideal sampling that we use rectangular window instead of dirac to sample the signal? – MathieuL May 30 '16 at 13:04
  • @MBaz sorry if I was unclear, I refer to the typical variable create by default in a flow chart of GNU Radio – MathieuL May 30 '16 at 13:05
  • A window, but in useful modulators, that window is not rectangular! – Marcus Müller May 30 '16 at 13:22
  • 1
    You must really understand what a pulse shaping filter does, @MathieuL; that's very important for understanding what timing recovery needs to do! – Marcus Müller May 30 '16 at 13:23
  • @MarcusMüller I just read my reference about pulse shaping, a nutshell, it is to respect Nyquist criterion in bandwidth limited channel. So we shape our signal with a raised cosine to meet the Nyquist's criterion. Time recovering is to sample at the peak of the cosine therefore where the eye is the most open. – MathieuL May 30 '16 at 13:45
  • Exactly! Also, note that the conditions for pulse shape filters are that their impulse response = 0 for every full symbol period (except for the t=0 one), which means that at exactly these "peaks" it's ok to demodulate, because the value there only depends on the current symbol. go a little back or forth in time – and the pulse shape of the adjacent symbols are part of what you observe! – Marcus Müller May 30 '16 at 13:49
  • Let us continue this discussion in chat. – MathieuL May 30 '16 at 15:08

0 Answers0