Re: Signal Capture and Playback Noise

[Marcus D. Leech]

On 2022-01-28 21:29, nickrestivo@cornictech.com wrote:

Thanks again for the reply Marcus, I will try increasing the gain when I have access to the hardware again. I can imagine this being the LO leakage, but this is way higher than the specified 50dBc of LO suppression… could this be because the gain is so low? I feel like something else could be at play here because when transmitting a pure sine (not recording and playing back) the spectrum looks totally different.

 

Best,

Nick

    You can easily convince yourself that the TX chain is behaving
    "sanely" with one of the example apps, like "tx_waveforms".
    
    UHD TX Waveforms Allowed options:
      --help                    help message
      --args arg                single uhd device address args
      --spb arg (=0)            samples per buffer, 0 for default
      --nsamps arg (=0)         total number of samples to transmit
      --rate arg                rate of outgoing samples
      --freq arg                RF center frequency in Hz
      --lo-offset arg (=0)      Offset for frontend LO in Hz (optional)
      --ampl arg (=0.300000012) amplitude of the waveform [0 to 0.7]
      --gain arg                gain for the RF chain
      --ant arg                 antenna selection
      --subdev arg              subdevice specification
      --bw arg                  analog frontend filter bandwidth in Hz
      --wave-type arg (=CONST)  waveform type (CONST, SQUARE, RAMP,
    SINE)
      --wave-freq arg (=0)      waveform frequency in Hz
      --ref arg (=internal)     clock reference (internal, external,
    mimo, gpsdo)
      --pps arg                 PPS source (internal, external, mimo,
    gpsdo)
      --otw arg (=sc16)         specify the over-the-wire sample mode
      --channels arg (=0)       which channels to use (specify "0", "1",
    "0,1", 
                                etc)
      --int-n                   tune USRP with integer-N tuning
    
    
    I think the main issue is that in a record+playback situation, the
    magnitude of the baseband samples are typically fairly small.  So,
    when you combine that with
      a low TX gain setting, the mixed signal magnitude is very very
    small.

 

From: Marcus D. Leech <patchvonbraun@gmail.com>
Sent: Friday, January 28, 2022 11:21 AM
To: nickrestivo@cornictech.com; discuss-gnuradio@gnu.org
Cc: 'lists' <lists@lazygranch.com>
Subject: Re: Signal Capture and Playback Noise

 

On 2022-01-28 22:05, nickrestivo@cornictech.com wrote:

Hi Marcus and lists,

 

Thank you for the fast reply, and I appreciate your insight. I agree that the 1kHz offset of the pure TX waveform is within the ppm spec of the board, and the spectrum analyzer is a Keysight N9020B. I was only thinking that offset might have been relevant because it was the same offset as the spectral artifacts seen during the playback. I am feeding into the B200 the 500MHz signal at -30dBm, although I find it unlikely this was leading to non-linearities though because the input into GNU Radio was fairly clean and at the correct frequency when plotted during the input. I did as you suggested and plotted the waveform during output:

 

Seemingly no spectral artifacts present, and to ensure this wasn’t specific to the spectrum analyzer, I went ahead and plotted the B200 output on a different spectrum analyzer:

 

This was also recreated at 200MHz, again no noise in GNU Radio Frequency Sink during playback, but plenty more artifacts present on spectrum analyzer:

 

Am I doing something wrong here? Or is this amount of spectral artifacts expected? I’d love to know if someone could recreate this on another B series, I’d hate to have to purchase another B200 just to replicate this issue.

 

Thanks again for any more help,

Nick

 

Don't forget that when you're recording from the USRP SOURCE, then magnitude of the samples will be very tiny, typically.  Even for -30dBm input.  If those are then
  replayed as-is, the actual modulating component will be not a lot different than the LO leakage.  I'd place a multiply block between the recorded samples and actual
  transmission to the USRP sink.