[prev in list] [next in list] [prev in thread] [next in thread] 

List:       comedi
Subject:    Re: continuous aquisition using standard linux
From:       "yahoo" <schousek () yahoo ! com>
Date:       2000-10-19 13:51:51
[Download RAW message or body]

    What kind of data rates are we talking here? That seems to be the most
important question. If you are trying to monitor exact points on the 50
hertz cycle, then you probably need to go to the dma level. If, however, you
only need to monitor every cycle or ten then you could use a card that can
automatically trigger an A/D conversion based off an internal counter then
cause an interrupt on successful conversion. 20 milliseconds shouldn't
create too much hardship on a modern computer (gut feel) and 200
milliseconds will obviously be even better. I know that this ability in
cards is not really rocket science, but can't point you towards an exact one
nor say whether your nt atmio is capable. Of course I work for an
electronics company that does design consulting so I could always be
convinced to find out and even develop the software :-)

brian
----- Original Message -----
From: "Jim Skelton" <skelton@oak.u-net.com>
To: <comedi@stm.lbl.gov>
Sent: Wednesday, October 18, 2000 1:40 PM
Subject: continuous aquisition using standard linux


> Hi all
>

> I work for a company that builds data acquisition systems  and test rigs
> for the
> British energy generation industry. We often have a requirement to
> collect large quantities of data on a continuous basis. We currently
> use a hotch potch of equipment, varying from SONY data recorders to
> PCs with ADC cards running crude DOS programmes.
>
> I'd like to start using Linux for many of these applications, and want
> build some solid tools to replace our aging DOS applications
>
> I have experimented with reading single values (comedi_data_read()
> function) but would like some advice on the best way to collect data
> continuously at an accurate sampling rate. I would prefer not to use RT
> Linux at present because it is an extra complication and It would be
> easier to convince my bosses if the application was build on standard
> Linux.
>
> Now my first Idea was to use DMA and some kind of ring buffer but on
> reading the readme for the NT atmio card (one card I have access to)
> using DMA for this card, is not recommended: I know that the atmio
> cards have limited internal buffers can I transfer to this and then
> pull out at slower (asynchronous) rate? and still get glitch free
>
> In summary my questions are
> Q1 what is the best way to collect continuous data using standard Linux
> Q2 if I have to use DMA which cards offer the best support under
> comedi
> q3 if I use the atmio cards what is the best way to collect continous
> data from this?
>
> TIA Jim Skelton
>
>
> _______________________________________________
> comedi mailing list
> comedi@stm.lbl.gov
> http://stm.lbl.gov/cgi-bin/mailman/listinfo/comedi




[prev in list] [next in list] [prev in thread] [next in thread]