Re: Matlab is too slow and can't keep up with the USRP


#1

Has anyone successfully used LabVIEW here? Is USRP interface code
available for either Matlab OR LabVIEW?

Phil


Message: 4
Date: Mon, 30 Apr 2007 13:18:18 +1200
From: Jonas removed_email_address@domain.invalid
Subject: [Discuss-gnuradio] Re: Discuss-gnuradio Digest, Vol 53, Issue
53
To: removed_email_address@domain.invalid
Message-ID: removed_email_address@domain.invalid
Content-Type: text/plain; format=flowed; charset=ISO-8859-1

On Thu, Apr 26, 2007 at 02:48:34PM -0400, Kevin Rudd (Contractor)
wrote:

Thanks for the suggestion. I dove into the TCP code to try and
figure out

what it was doing. I found my problem. I have to use the following
read

command…

data = pnet(con, ‘read’, 1000, ‘SINGLE’, ‘NATIVE’);

Now I notice that matlab seems to be lag behind the USRP. So, maybe
matlab

is not the best path to process the data in realtime. I am curious
what the

best data rate someone has sustained using a TCP connection.

TCP on GigE can do upwards of 100MB/s :wink:

If I get it up and going, I will post my code and a howto so others
can

stream data to and from MATLAB in real time.

Thanks.

Eric

Hi Kevin,

I’m never directly sending samples to and from Matlab (for my
application Matlab is too slow and can’t keep up with the USRP).
Instead, in Matlab I generate my waveform first, and then use Matlab’s
built in FTP to send the file to the USRP server computer. To generate
the file that I send, I use write_float_binary.m
(gnuradio-core\src\utils). The resulting “.dat” file can then be used
directly with a regular GNU Radio file source.

I wouldn’t be surprised if the USRP overflows that you are seeing are
attributed to Matlab’s inability to keep up (assuming you are going from

the USRP to Matlab).

Jonas