RE: Analog output device support design

[Angus Gratton <angus.gratton@anu.edu.au>]

Hi Mark & Ralph,

On Wed, 2010-07-28 at 11:38 -0500, Mark Rivers wrote:
> I generally agree with Ralph that if you are just supporting the ao
> record with your DAC that asyn may be overkill.
> 
> On the other hand, my asyn driver for the Industry Pack dac128V is only
> 161 lines of code total (the source is appended).  It uses the
> asynPortDriver C++ base class.  I think it may be a challenge to write
> device support for your new DAC in only 161 lines of code.

Thanks for the detailed explanations, and thanks for the source - that's
very useful.

I've actually now moved on to implementing support for our ADC
(XVME-500), where I'll probably use asyn, and when I'm done there I may
look into going back and refactoring the DAC support.

Our current requirements are not particularly onerous in terms of
frequency or performance. The non-EPICS system that I'm replacing
currently does very straightforward I/O, synchronous even for ADC
conversions (20uSec)(*), with no particularly tight timing constraints. 

So in this rewrite I'm mostly trying to maximise code simplicity and
readability over any other metrics (which it seems can be achieved with
asyn as well.)

- Angus

(*) So, actually, if I correctly followed the explanation in one of the
lecture videos, 20uSec ADC conversions are probably fast enough that I
could implement this device support synchronously as well, instead of
using asyn. Although I'm probably going to use asyn.