Re: [Discuss-gnuradio] determining throughput

[Eric Blossom]

On Mon, Jun 27, 2005 at 04:42:30PM -0400, John M Daniel wrote:
> > I not sure I'm following your question.  Are you trying to benchmark
> > the throughput of a particular graph in gnuradio?  What are the input
> > and output sources?  Files?  USRP?
> 
> I am using a PCI-DAS4020, and want to know if and when any
> discontinuities might arise while collecting data for an extended period
> of time.  What I had in mind was outputting to a file then using a c++
> program to check for any gaps in the data, however I believe there's a
> less painful way of doing it.  What "stuff" would I use to check this in
> the example that you provided?

The gr-mc4020 class mc4020_source.{h,cc} keeps track of this.
There's currently no way to fish the value out, but it's trivial to
add a method that returns d_total_lost.

If the code is writing "O" to stderr, you're having overruns.
You shouldn't be getting any.

> Also on another tangent, I am interested in doing some spectroscopy.
> More specifically after doing a FFT I want to add the square of the real
> and imaginary components so that any noise is smoothed out and anything
> hidden shows up.

The FFT sink plots 20*log10(abs(fft_output)).  This is already
squaring and adding the real and imaginary components (and taking
the square root).  To smooth it out, enable averaging.  Right click in
the fftsink for a menu, or you can provide an argument to the
constructor.  See fftsink.py

> Would I need to write up a new block that does this or
> can I already get it done in python?  In addition could I use fftsink to
> plot my results or would I need to use another one.  Many thanks.

No need for a new block, however you will need to convert the two's
complement short output of mc4020.source to float using
gr.short_to_float() before connecting it to the fft.

The existing gnuradio-examples/python/mc4020/mc4020_fft.py already
does all of this for you.

Eric