Re: [Discuss-gnuradio] Disabling Decimation/I&Q Mixing in USRP Firmware

[Marcus D. Leech]

On 04/03/2011 2:46 PM, Jeffrey Lambert wrote:
> Okay,
>
> So I think I've answered my own question with regards to the disabling 
> the NCO/Mixer and the CIC decimator stages in the receive chain.  It 
> DID require modifying a patch that I found thanks to Google and 
> editing the source code so that Quartus would not give an error on 
> compiling.
>
> Without the decimation stage, there should not be any decimation.  
> Running the custom firmware suggests I was getting an approximate 
> sample rate of something like 25.334 MHz.  I repeated the experiment 
> using 8 bit samples and the number improve to a more respectable 32 
> MHz.  This seems to suggest that interleaving is still being done; the 
> o-scope display suggests that half of the data stream from the USRP is 
> still discarded (very possibly from buffer overflows?).  Why am I 
> getting these results?  Does disabling the NCO remove I/Q interleaving 
> and provide only real values over USB to the host?
>
> If anyone could answer I would greatly appreciate it; I am trying to 
> get the highest sample rate possible from the USRP into a PC for 
> capturing some very short (nanosecond scale) pulses.
9000ft view here.

The USRP1 ADC runs at 64Msps, and the highest data rate you can 
reasonably push over USB using the standard USRP "framing" is about 
32Mbytes/second,
  which for 8-bit samples tops out at 16Msps.  If you go to 4-bit 
samples, you can reasonably push 32Msps over the USB, and at 2-bit 
samples, you could get to
  64Msps, but it'll require some work on the FPGA image.