Re: [Discuss-gnuradio] Using UHD, trying to tx and rx samples simultaneously

[Isaac Gerg]

Hi Josh,

Thanks for your responses.  I have looked all the example source code in the repo over the last few days.

You are correct in that when the thread "wakes" up, it sends a burst.  I have configured it to send one cycle of a complex exponential.  

Here is my function now.  I am setting the both burst flags to true and the time spec flag to false (I want it to send out the buffer immediatly)

{
    Int32 burstSizeSamps = 20;
    vector<complex<Float32> > vec(burstSizeSamps);

    m_pSdev->set_time_now(uhd::time_spec_t(0.0));

    // Set up tx params.  We wish to send a burst.
    // http://ettus-apps.sourcerepo.com/redmine/ettus/projects/uhd/repository/revisions/master/entry/host/examples/test_async_messages.cpp
    uhd::tx_metadata_t md;
    md.start_of_burst = true;
    md.end_of_burst   = true;
    md.has_time_spec  = false;
    //Float64 seconds_in_future = 1;
    //md.time_spec = uhd::time_spec_t(seconds_in_future);

    Float32 v = 0;
    for (Int32 k=0;k<vec.size();++k)
    {
        v = 2*M_PI*(Float32(k)/Float32(vec.size()));
        vec[k] = complex<Float32>(0.3*cos(v), 0.3*sin(v));
        //cout << vec[k] << " ";
    }
    cout << endl;

    // Send the entire contents of the buffer
    // This should be the only place where data is sent over the air!
    cout << "sending size [cpx samps]: " << vec.size() << endl;
    m_pDev->send(&vec.front(), vec.size(), md,uhd::io_type_t::COMPLEX_FLOAT32, uhd::device::SEND_MODE_FULL_BUFF);
}

However, this still does not work.  For burstSizeSamps > 10, I can only see ~10-20 samps that have amplitude on the rx end.  I tried doubling the samp rate of the rx part (for sanity checking) and then I can see twice as much.  I have modified burstSizeSamps and get the following number of samps with samplitude at the rx:
burstSizeSamps   numSamps at rx with abs()>0.01
1                         ~1-2
10                       ~10
20                       ~10-20
100                     ~10-20
1000                   ~10-20

Any ideas on why this isnt working?  

Thanks for your help so far,
Isaac

On Wed, Jan 26, 2011 at 11:35 PM, Josh Blum <address@hidden> wrote:

> If Im always just sending a burst message when I tx, is it safe to say that
> I should always have the EOB flag set?  My messages may be as large as a few

I believe so, see below

> thousand bytes, but I dont plan to do any continuous streaming.  Currently,
> the Tx thread has a queue that it just works off where each item in the
> queue is a message that could be as large as a few thousand samples.  The
> thread wakes up  every so many ms and then says Hey! theres messages to send
> and then one-by-one sends each message as a burst.
>

To get the terminology strait. It sounds like that each time the thread
"wakes" up to send, its sending a new burst.

The burst flags convey to the driver and the hardware that the samples
presented are part of one continuous stream. Suppose you have a
continuous stream of samples. The first call to send should have start
of burst set, and the last call should have end of burst set. If you
send all the samples in a single call to send, then set both start and
end of burst flags.

This code needlessly fragments a send operation into several smaller
ones as an example of how to use the flags:
http://code.ettus.com/redmine/ettus/projects/uhd/repository/revisions/master/entry/host/examples/tx_timed_samples.cpp#L94

-josh