problems with scan record

['Dale L. Brewe' <brewe@pnc.aps.anl.gov>]

We've been having a problem w/ scans using the scan record of epics
v3.13b11, version 3.2 of the scaler record, and the XFD 3.13b7 release of
their scan database (v3.1). I'd like to know if there is anybody else who
has experienced similar problems or who has a suggestion about what might
be the cause/solution to the problem, or how/where to look further.

   The symptom of the problem seems to be that detector values are
sometimes received by the scan record before the detector data is valid:
scaler counts appear to be recorded before the scalers are finished
counting, so that if the desired time is 1.0sec, the recorded time might be
0.2sec, and the # of recorded counts are down from the expected values
accordingly (i.e. the elapsed time appears correct). We've only seen this
problem during energy scans when we also scan our ID along with the mono.
I believe the scaler itself isn't quitting prematurely. This follows from
the fact that a calculation whose result we record, and which processes
after the scaler finishes, is not updated when the problem occurs (i.e. the
value from the previous scan point is recorded).  I haven't had a chance to
try and confirm this.
   
To scan the ID we treat it as a positioner (ID20:ScanEnergy.VAL=>drive,
ID20:Energy.VAL=>readback). We add ID20:Busy.VAL to the 'all done' records
in all_com_xx.db. For energy scans we generally use 'Table' scans - we
haven't had a chance to see if the problem occurs in the Linear mode.  The
problem seems to have a threshold in the step size: for example, a scan
with 1.0eV steps might not show the problem, whereas a scan with 0.5eV
steps will; the value of the threshold step size seems consistent, but
energy dependent (since the angular step for a given energy step is energy
dependent, it's possible the threshold in angular units is constant). 

I might also add that the energy 'positioner' drive value is a calculation
that converts the energy to an angle to drive the mono.  The readback is
another calculation that converts back from angle to energy. We also use
the counting time as a positioner to get variable integration times during
the scan.

It seems to me that the wait records involved in the scan aren't working
properly, but I don't know how to determine this for sure. 

Dale Brewe  





~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 Dale L. Brewe     PNC-CAT/U. of Washington
 ph: 630 252 0582         fax: 630 252 0580
 Mail:
 APS/Argonne National Laboratory
 Bldg 435E sector 20
 9700 S. Cass Ave.
 Argonne, IL 60439
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~