Re: Scaling Database entries

[Andrew Johnson <anj@aps.anl.gov>]

Dirk Zimoch wrote:
> if you driver does not set ESLO and EOFF, you can set it in your db. 
> Calculate manually what to set. I did the same with StreamDevice.

...

> Don't forget to set LINR to "LINEAR", of course.

If you want to set ESLO and EOFF yourself, you should set LINR to 
"SLOPE" rather than "LINEAR" or there's a danger the device support will 
overwrite your values if someone modifies any of the conversion fields. 
 StreamDevice might not do so if it doesn't have a special_linconv 
routine, but then it's not following the full ai/ao standard.  Most 
analog device supports will overwrite ESLO and EOFF; by relying on non- 
standard behaviour you're making it harder for someone to replace 
hardware using a StreamDevice interface with some other device, because 
they now have to change more than just the DTYP and INP/OUT field of 
their database records when they move (and they might not realize that).

StreamDevice should have some way for the author of a protocol file to 
specify what the device's raw range actually is for analog records, and 
it should perform the calculation of ESLO and EOFF from that raw range 
and the EGUL+EGUF pair.  The other option you might want to support if 
you don't already would be to control whether any conversions get done 
at all - many instruments return engineering units anyway, and that's 
why the analog records allow the device layer to set the VAL field 
directly and return 2.  This is used a lot for GPIB instrumentation.

- Andrew
--
There is considerable overlap between the intelligence of the smartest
bears and the dumbest tourists -- Yosemite National Park Ranger