Re: [Discuss-gnuradio] usrp over/underrun problems

[Bob Vincent]

At 12:23 PM 2/15/2005, Matt Ettus wrote:
> Quoting Bob Vincent
> <address@hidden>:
>
> > We've got a couple of USRP boards running on "embedded"
> ebx boards.
>
> Great!
>
> > Unfortunately, we are seeing usb overruns/underruns when we run the
> fsk_tx
> > / rx example. I removed the fft from the signal chain, and we still
> see
> > regular overruns on the rx side. Also cut the cordic freq. to 2e6,
> with the
> > same results. I thought this might change the sampling rate. Is that
> so?
>
> CORDIC freq will not affect sample rate.  You will get the lowest
> sample rate by
> setting the decimator rate to 256.  This will give a sample rate of
> 64e6/256 =
> 250 ksamples/sec.

So in fsk_*, on the rx side it's a function of data rate, while on the tx
side the same calculation goes to the interpolator. Got it.
Unfortunately, I get problems when I set data_rate for most of the values
I've tried. It appears that both calculations have to wind up in the
range 1-256. 
Do they need to be powers of 2? I'm not much with fpga code. 


> What code are you
> running?

Latest. I'm trying to get fsk to play nice. I've also put in socket for
file source and sink. We're going to try to run our ad-hoc routing
protocols on top of the "radio".
I noticed that when you stop the app, the usrp recirculates. I presume
that's what will happen when we source data from a socket as well. 
Do you know of any way to "stop" the usrp, apart from muxing in
zeroes? I want to avoid as much coding as possible.



> > Finally,
> setting the data rate lower seems to make things much worse.
> > Perhaps that is because the correlator runs longer?
>
> What correlator?
>
> > We're using fedora 3 on a 1G pentium 3, approximately. I suspect the
> usb
> > isn't fast enough, but that's the hardware we're stuck with. So
> turning
> > down the a/d clock would be good.
>
> Are you certain that this board has USB2?  Most P3's did not have
> USB2.
>
> Matt