discuss-gnuradio
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Discuss-gnuradio] Buffer Question for Streaming with USRP


From: Marcus D. Leech
Subject: Re: [Discuss-gnuradio] Buffer Question for Streaming with USRP
Date: Mon, 04 Apr 2011 13:11:22 -0400
User-agent: Mozilla/5.0 (Windows; U; Windows NT 6.1; en-US; rv:1.9.2.15) Gecko/20110303 Thunderbird/3.1.9

On 04/04/2011 11:10 AM, Tom Hendrick wrote:
Hi Martin,

Thanks,  for sure the C++ program is giving output in doubles.  I was told by Josh that you can use the gr file block with vector length 8 to read it in as a double with GRC, but none of the other blocks use double which is why I would need a double to float converter.

The C++ program I'm working with has a function to convert from double to float, but I don't think it handles the buffer correctly.  I didn't write it.

For sure the output looks good when I use the C++ program and the converter from double to float and read it in with GRC as float.  I checked with an oscilloscope, the problem is when I get the underruns.  I'm not exactly sure how to handle the clock, I'm using a camera, then the C++ modulator, then the C++ converter (double to float), and that gets fed into GRC.  In GRC I have the gr file block, a resampler (to match the sampling frequency of the signal to the USRP), a float to complex block, then the USRP sink block.

At this point even a simple custom block from double to float in GRC is hard for me to make myself due to my lack of programming experience. 
Thanks for the advice
The basic issue is that *some* part of your signal chain isn't keeping up with the USRP demand for data.  Looking at your flow-graph
  last night, that "demand" is 500Ksps.

It doesn't matter how much buffering you have, if it sometimes takes longer than 1 sample time to compute the sample, you'll underrun
  the USRP hardware, since there'll be no data available when it needs it some part of the time.

How much is your processor loaded when you run your signal processing chain?

How much processing does openRTSP do?  Is it providing samples in real-time, and at what rate?





--- On Mon, 4/4/11, Martin Braun <address@hidden> wrote:

From: Martin Braun <address@hidden>
Subject: Re: [Discuss-gnuradio] Buffer Question for Streaming with USRP
To: address@hidden
Date: Monday, April 4, 2011, 8:10 AM

On Sun, Apr 03, 2011 at 07:12:58PM -0700, Tom Hendrick wrote:
> Hi all,
>
> I'm using a C++ program that modulates a signal that I feed to GRC via an
> mkfifo (named pipe) file.  The problem is that I get underruns from the USRP.
>
> There is a function in the C++ program that converts from double to float and
> then dumps it to stdout which I then feed to an mkfifo file in the command
> line.  With the buffer size of 32 bytes in the stdout line in C++  I get
> consistent underruns.  With the buffer size to 4096 bytes, I get an initial
> underrun and then none for about 1-2 minutes and then I start getting them more
> consistently, and about 500msec of no signal for each underrun.  With the
> buffer size set to the size of a 2 second packet (at 500 kHz) I get an underrun
> every 4 seconds with  a 1-2 seconds of no signal randomly in between my
> modulated signal.
>
> Does anyone have any ideas if I need to use a certain buffer size?  I tried
> this on my netbook with an intel atom and also laptop with a duo core and got
> the same thing.
>
> When I run the C++ program and output to a regular file, and read in with the
> GRC script at the same time, I don't get any underruns at all, but of course
> this isn't real time.
>
> My other idea is to find a way to do the conversion from double to float in GRC
> with a custom block.  I also wonder if going from stdout to a pipe file and
> then reading the pipe file from GRC is also a problem.

Hi Tom,

I wonder... how do you get double-values in the first place? Literally
no GNU Radio block in the main tree uses double. Perhaps it's the Monday
morning, but the only way I can think of to get doubles into your flow graph
without writing lots of code yourself is by reading them from a file,
which you could convert to float beforehand.

Just checking here... are you sure you have double? After your custom converter,
is the data valid?
Your underruns might also be caused by something else. Do you have a
hardware clock on both ends of the flow graph (e.g. a USRP and a sound
card)?

However, writing simple blocks like a double->float converter is
super-easy. I reckon it would take you as long as it took to set up the
custom converter and the pipes. Use create-gnuradio-out-of-tree-project
to get started and then add a sync block.

MB

--
Karlsruhe Institute of Technology (KIT)
Communications Engineering Lab (CEL)

Dipl.-Ing. Martin Braun
Research Associate

Kaiserstraße 12
Building 05.01
76131 Karlsruhe

Phone: +49 721 608-43790
Fax: +49 721 608-46071
www.cel.kit.edu

KIT -- University of the State of Baden-Württemberg and
National Laboratory of the Helmholtz Association


-----Inline Attachment Follows-----

_______________________________________________
Discuss-gnuradio mailing list
address@hidden
http://lists.gnu.org/mailman/listinfo/discuss-gnuradio
_______________________________________________ Discuss-gnuradio mailing list address@hidden http://lists.gnu.org/mailman/listinfo/discuss-gnuradio


reply via email to

[Prev in Thread] Current Thread [Next in Thread]