Re: [Discuss-gnuradio] comedi in grc

[Yann Le Coq]

Hello,

In the process of writing the xml files for gr-comedi in grc and testing 
the result, I believe I am getting in an gr-comedi issue.

The only comedi board I have access to is a USB-DUX Sigma, which outputs 
ADC data in the comedi type "lsampl_t" (32 bits per sample). The 
gr-comedi module, however, is hard coded for the comedi type "sampl_t" 
(16 bits per sample only).

The only way I see to make this work for the USB-DUX Sigma (or any other 
>16bits ADC equipped board) is to modify the gr-comedi module C source 
itself, so as to accomodate 32 bits ADC devices.

If I were to do this, I would, however, be tempted to change the output 
type entirely. The way it is done now, the comedi-device is expected to 
return "sampl_t" type raw data (unsigned 16 bits), which are then 
shifted down by -32767 and the resulting "signed and hopefully 
zero-centered" data is output as a short. There might be a good reason 
for that, but it seems more likely to be a quick hack, that has the 
drawback of making the whole module interface too device-dependant, IMHO.

The comedilib standard way would rather be to use the comedilib function 
"comedi_to_phys()", which takes raw data and makes it a float that 
actually represents a physical value (a voltage for example) based on 
the channel/board configuration (unipolar/bipolar, ADC range, etc.). 
This function helps a lot providing device-independence and I don't see 
good reason why gnu-radio shouldn't be using it. In fact, I don't see 
why one would want to have access to the raw data itself from gnuradio, 
especially if it means changing the output type depending on the device 
being used... Getting a physically meaningful float output seems to make 
ways more sense.

Doing this modification would, however, break backward compatibility 
with pre-existing gr-comedi-using top_blocks, so I am a bit hesitant and 
would like some feedback before investing the time that is necessary to 
make this change.

Best regards,

-Yann

Le 01/02/2016 18:19, Marcus Müller a écrit :
> Hello Yann,
>
> no, you're not missing something; I don't remember there ever being GRC
> XML files for gr-comedi, but if there where pre-3.7, they might have
> gone away during the restructuration for 3.7; and, as there seemed to be
> very little interest in getting gr-comedi to work from within GRC....
>
> At any rate, since the sink and source have no own method and their
> ctor/make functions only take sample rate and device address string as
> arguments, copy&pasting together a valid XML file should be done very
> quickly. However, I (used to think that I) don't know anyone who has
> uses comed to talk to a data aquisition or signal generation device –
> which would be useful for testing.
>
> So, would you be willing to contribute such XML?
>
> Best regards,
> Marcus
>
> On 01.02.2016 18:00, Yann Le Coq wrote:
> > Hello,
> >
> > Unless I missed a point in compilation from source, there doesn't seem
> > to be a grc blocks for the comedi driver gnuradio interface in a
> > standard installation (origin/maint commit
> > 11973c64437683cc99c48eae9eb4db8234f1ac42). Indeed, in the source code,
> > there is no /grc sub-folder in the gr-comedi folder to define the .xml
> > files.
> >
> > Am I missing something, or is there a place somewhere where this can
> > be found ?
> >
> > Thank you for your help.
> >
> > -Yann
> >
> >
> >
> > _______________________________________________
> > Discuss-gnuradio mailing list
> > address@hidden
> > https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
>
> _______________________________________________
> Discuss-gnuradio mailing list
> address@hidden
> https://lists.gnu.org/mailman/listinfo/discuss-gnuradio