[daip] VLBABPSS

[Walter Brisken]

I think I now know what was triggering the problem and may understand why 
changing BPASSPRM(5) allowed things to work.  GBT was included in this 
experiment and when it first started observing, it took approximately 1/2 
of the bandpass calibrator's scan to stabilize in amplitude.  I suspect 
BPASS sees the variance on a per-channel basis as evidence of not 
converging, and the divide by channel 0 effectly counters that changing 
amplitude on a timescale shorter than the BPASS solution interval, perhaps 
even on a per-integration-time timescale.

-Walter

On Fri, 3 Aug 2018, Walter Brisken wrote:

>
> I read the help from a distance, so must have missed the 'NOT'.  Anyway, 
> changig the state of this to the non-default value (from VLBABPSS 
> perspective) allowed solutions to pass that otherwise would not have. There 
> is still a puzzle remaining here...
>
> -W
>
> On Fri, 3 Aug 2018, Eric Greisen wrote:
>
>>  On 08/03/2018 10:35 AM, Walter Brisken wrote:
>> > 
>> >   Hi DAIP (probably Amy?)
>> > 
>> >   Tom and I are looking at some of his data that he is processing using 
>> >   the
>> >   VLBA scripts in AIPS.  After running VLBABPSS it is apparent that 
>> >   nearly
>> >   all of the LCP channels are getting flagged by the BPASS task.  If we 
>> >   tget
>> >   BPASS to see the parameters that are used we find BPASSPRM(5) is set to 
>> >   1,
>> >   indicating desire to "divide by channel 0".  I worry that this is both
>> >   unnecessary and potentially problematic.  Indeed, setting this 
>> >   parameter
>> >   to 0 and running BPASS in its naked form results in sensible 
>> >   bandpasses.
>> > 
>> >   Is this a known issue?  Are there reasons that "divide by channel 0" 
>> >   is
>> >   actually the right thing to do?  Seems like any normalization that 
>> >   this
>> >   helps accomplish will be undone first by the normalization specified by
>> >   BPASSPRM(10), and later by ACSCL.  Maybe I'm missing something subtle?
>> > 
>> >        -Walter
>> > 
>>
>>  You did not read the help file.  BPASSP(5)=1 means do NOT divide by
>>  channel 0 and do the normalization after the fit based on the value of
>>  BPASSP(10). Setting it 0 causes the normalization to be done record by
>>  record.
>>
>>  Eric
>> 
>
>

-- 
-------------------------
Walter Brisken
Director
Long Baseline Observatory
(575)-835-7133 (office)
(505)-234-5912 (cell)