USRP2 Maximum througput

Hello,

I am trying to test my maximum achievable throughput from our USRP2
to my host PC. The easiest way for me to do this is to invoke,
let say, usrp2_wfm_rcv.py. The ADC sample rate in usp2_wfm_rcv.py
is set to 100Ms/sec, and the decimation rate is set to 312 for a net
sample rate of 320 KS/sec. This configuration works fine, I have no
overruns. However when I change the decimation rate to 10, let’s say,
usrp2_wfm_rcv.py returns a number of errors. Additionally, when I
run usrp2_rx_cfile.py (to dump data to a file) and set the decimation
to 10 or below, although I dont get any errors, it does not work.
Setting
the decimation to 312 works just fine. Does anyone have any idea if
there is some limitation as to how little I can decimate and if so is
there a way to get around this. The USRP2 advertises 50MSamp/sec max
throughput, is there some other way I can validate this. Does anyone
know what the maximum throughput might be (is it 50Msamp/sec), assuming
my PC is not a bottleneck?

Thanks,
Sharif

On 02/10/2010 09:37 AM, Sharif S. wrote:

the decimation to 312 works just fine. Does anyone have any idea if
there is some limitation as to how little I can decimate and if so is
there a way to get around this. The USRP2 advertises 50MSamp/sec max
throughput, is there some other way I can validate this. Does anyone
know what the maximum throughput might be (is it 50Msamp/sec), assuming
my PC is not a bottleneck?
Thanks,
Sharif

The problem you are having is that the data from the USRP2 is being fed
into your soundcard after processing. So if you change the decimation
to 10, it means you are trying to feed many megasamples per second into
a soundcard and it won’t do that.

If you want to see how fast you can use your usrp, use usrp2_fft.py,
which shows the spectrum. It has a place to put in the decimation
ratio, so you can try 4 to 512.

Matt

Hi Matt,

Thanks so much for the response. I will try usrp2_fft
as you say. One other comment, I understand your point
about the sound card, but I would have expected audio overruns
or underruns, not that the usp2_wfm_rcv.py would not run.
I guess I am wrong on that?

Also, I dont believe usrp2_rx_cfile.py sends data anywhere
except a file. Regardless, I will try urp2_fft…thanks so much.

Sharif
----- Original Message -----
From: “Matt E.” [email protected]
To: “Sharif S.” [email protected]
Cc: [email protected]
Sent: Wednesday, February 10, 2010 1:11 PM
Subject: Re: [Discuss-gnuradio] USRP2 Maximum througput

On 02/10/2010 10:55 AM, Sharif S. wrote:

Hi Matt,

Sorry to bother again, but can I use a decimation value
of 2 to get 50MSamples/sec to my PC in usrp2_fft.py? I believe
you suggested 4 to 512, thus the question. Any information
would be greatly appreciated.

No. Gigabit ethernet is not fast enough for that. You would need to
switch to 8 bit samples instead of 16, and that isn’t implemented yet.

Matt

Hi Matt,

Sorry to bother again, but can I use a decimation value
of 2 to get 50MSamples/sec to my PC in usrp2_fft.py? I believe
you suggested 4 to 512, thus the question. Any information
would be greatly appreciated.

Thanks,
Sharif
----- Original Message -----
From: “Matt E.” [email protected]
To: “Sharif S.” [email protected]
Cc: [email protected]
Sent: Wednesday, February 10, 2010 1:11 PM
Subject: Re: [Discuss-gnuradio] USRP2 Maximum througput