Re: [Discuss-gnuradio] WBFM modulation. Bad sound and aUaU

[Marcus D. Leech]

On 03/05/2017 06:47 PM, Fernando wrote:
> Yes, this USRP has "very strange" sample rates, like 2.631579 MSps the 
> are far from rounded values, every single sample rate I try to use, it 
> change it to a slightly different one. 2.64Msps came from using a 
> sample rate over 2Msps, with a 44Ksps in audio and interpolating in 
> two stages one the FMTransmit (QuadratureRate/AudioRate =6 ) which 
> must be an integer and a second stage in the resampler (10), so 
> 44K*60=2.64M Using a UHD USRP source instead of osmocom source for the 
> USRPit works fine (it sound okay), altough some aUaU keep appearing 
> (maybe for mismatch between in sample rates?) regards 

Not that strange--a USRP N-series device has a master clock of 100MHz.  
Sample rates are constrained to be integer fractions of the
  master clock rate.  The sample rate you show is exactly 100e6/40, but 
is not a multiple of 44Ksps.

The osmocom block simply calls UHD, so in terms of sample-rate, there'll 
be no difference.

Since 44Ksps is not an integer fraction of 100Msps, you'll need to use 
an arbitrary resampler.   This is exceedingly common in DSP.
  Audio subsystems are sample-rate constrained, and SDRs are 
sample-rate constrained, and it's very rare that their constraints
  "mesh" in useful ways.   Doing fractional resampling in the software 
part of the flow is MUCH easier than having the hardware do
  arbitrary resampling.