Re: channel 'bounce' with ni_pcimio

From: Dan Christian <dchristian_at_mail.arc.nasa.gov> · Date: Tue, 3 Sep 2002 09:09:02 -0700

When I've seen this, it has either been that samples were dropped or 
extra data inserted.  Since this throws off the channel 
de-multiplexing, you get data moving between the various channels.

The data that you show actually looks different than this.  You show 
data switching channels, and then switching back.  It could be that 
data is being transferred out of order.

Does the channel swapping happen in the middle of a long acquistion 
(many K samples) or just at the end?

-Dan

On Friday 30 August 2002 17:18, Jack Culpepper wrote:
> Hi all,
>
> Using the ni_pcimio driver, I am seeing something very odd: if I
> sample from 8 channels, and put different looking signals on each
> channel, it is obvious that every 'once in a while' the data from one
> channel shows up on another and vv.  It looks something like this:
>
> 1 ================&&&========
> 2 ---------------------------
> 3 ++++++_at__at__at_++++++++++++++++++
> 4 ***************************
> 5 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
> 6 &&&&&&&&&&&&&&&&===&&&&&&&&
> 7 ###########################
> 8 _at__at__at__at__at__at_+++_at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at__at_
>
> I just realized this was happening.  It was probably happening all
> along (I've been using comedi for a couple of months now), but I just
> didn't notice b/c all my signal were so similar looking.  I am
> sampling at 4000 Hz, and using a large read buffer.
>
> Does anyone have an idea of what this could be?  It is almost
> certainly a software problem.  My code is almost identical to the
> code in demo/cmd.c, using the prepare_cmd() function with very few
> other changes..  After sampling, I call comedi_to_phys() to convert
> the data into volts (doubles), which are immediately copied to shared
> memory.  Thus, the code for the process sampling using comedi is
> extremely minimal..  I don't know what could be wrong.
>
> Thanks,
>
> Jack
>
> _______________________________________________
> comedi mailing list
> comedi_at_comedi.org
> https://cvs.comedi.org/cgi-bin/mailman/listinfo/comedi