Re: Analog input record offsets

[mrk@aps.anl.gov (Marty Kraimer)]

> From npr@jach.hawaii.edu Mon Nov 28 20:22 CST 1994
> Date: Mon, 28 Nov 94 16:23:38 HST
> From: npr@jach.hawaii.edu (Nick Rees)
> To: tech-talk@aps.anl.gov
> Subject: Analog input record offsets
> Content-Type> : > text> 
> Content-Length: 2155
> 
> 
> I have a rotary encoder with a range of travel less than 360 degrees
> that I want to read into an analog input record. (Yes, its the
> telescope mount encoder).  Unfortunately I have no guarantee where the
> encoder zero point is. The zero point adjustment has to be added to the
> raw value in encoder units, and then the value circularly normalised so
> the value from the encoder always lies between two accepted values.
> 
> I have to write analog input device support for the hardware, so I was
> intending to do this fiddling in device support. Looking at the record
> support, it would seem that the ROFF field might be a field I could use
> - it is in digitizer units (the most convenient in this case) and is
> added to the raw value just after it is returned from device support.
> Device support would then:
> 
> a. Read the value.
> b. Add ROFF
> c. Circularly normalize between the maximum and minimum possible values.
> d. Subtract ROFF (so that it can be added once again in record support).
> 
> A few questions:
> 
> 1. How do I set ROFF? The documentation says `it is the responsibility
>    of the device support routines to use EGUF and EGUL to compute ESLO
>    and ROFF'. All very well but none of the device support routines I
>    have reference ROFF. (grep roff ~epics/share/src/dev/*.c returns nothing).
>    What gives? Am I missing a trick somewhere?
> 

In gtacs, the precurser of epics, the field roff did not appear. The
raw value was always ajjusted in the drivers so that it was a positive number
with 0 representing egul. Thus roff was not needed. For epics roff was added
for generality. Note that for gtacs the raw value was not necessarily the value
read from hardware. I guess that all new device support has followed this model.

> 2. Can I use EGUF and EGUL in ways that were not intended? (i.e. basically
>    ignore them - they don't have sufficient precision). I presume so.
> 
> 3. If ROFF isn't used by device support, can I just set it at
>    initialization - or even in dct (this would require a change to
>    aiRecord.ascii) to a specific value without breaking things?
> 

You could use aoff. It is setable via dct or via Channel Access.

> I admit it is a bit ugly, but the alternative is to write a new analog
> input record for the special case of a circular encoder - and still
> have to write device support for it. If people think this is a better
> way (or have any other suggestions), do tell me.

Note that you always have the option of doing all conversions in
device support and returning a status of 2 which tells record support not
to do any conversions.

Marty Kraimer