[evlatests] Referenced Pointing Tests

[Barry Clark]

OK, somebody could mine the data, and see if median is detectably
different from first good.

Bryan Butler wrote:
> 
> i've always thought the median is the simplest good filter - i suspect 
> we'll find no better.  should i ask keith to implement that for testing?
> 
>     -bryan
> 
> 
> Claire Chandler wrote, On 11/18/10 13:41 PM:
>> The trouble is that telcal might be able to get a solution but it 
>> might be
>> rubbish.  I was thinking we should perhaps use the median of all 
>> subbands,
>> but there's clearly some work to be done to establish what algorithm is
>> best.
>>
>> Claire
>>
>> On Thu, 18 Nov 2010, Barry Clark wrote:
>>
>>> How about letting telcal try another subband if it can't get a solution
>>> on its first choice?
>>>
>>> Bryan Butler wrote:
>>>> yes, all the solutions are written into the pointing file.  i don't 
>>>> know
>>>> that we will ever have a truly robust way of avoiding RFI, other than
>>>> trying to avoid where we know it exists.  we had been thinking at one
>>>> point of at least allowing the user to specify which subband was used -
>>>> if we do that is it enough?  maybe.  certainly somebody should look 
>>>> into
>>>> whether trying to take the results from all the subbands is better than
>>>> one, but maybe we should just go ahead now and change the 
>>>> recommendation
>>>> for RSRO/ECSO observers so that they can use the same setup for
>>>> reference pointing as for observing.  i suspect that the problems
>>>> encountered from that will be less than the problems that arise from
>>>> changing configurations, but maybe i'm missing something.
>>>>
>>>>     -bryan
>>>>
>>>>
>>>> Claire Chandler wrote, On 11/18/10 12:18 PM:
>>>>> Hi Rick,
>>>>>
>>>>> You need to define what you mean by "works for wideband".  Recall 
>>>>> that at
>>>>> present the default algorithm in TelCal is to take the second 
>>>>> subband from
>>>>> the bottom of each baseband for the solution that is fed back to the
>>>>> online system.  Therefore it relies on that subband being RFI-free.  I
>>>>> believe that solutions are being written for all subbands for testing
>>>>> purposes at the moment, and what we really need to do is to 
>>>>> establish how
>>>>> to combine the results from all those subbands to obtain a solution 
>>>>> that
>>>>> is not affected by RFI, and can potentially be used to enable 
>>>>> pointing on
>>>>> weaker sources.
>>>>>
>>>>> Claire
>>>>>
>>>>> On Thu, 18 Nov 2010, Rick Perley wrote:
>>>>>
>>>>>>      I ran a test of referenced pointing this morning.    There 
>>>>>> were two
>>>>>> goals:
>>>>>>
>>>>>>      1) Does 'wide-band' referenced pointing work as well and 
>>>>>> reliably as
>>>>>> 'narrow-band'?
>>>>>>
>>>>>>      2) Does a 15-second cadence work as well as 20-second?
>>>>>> (Explanation:  In the past, the system used 10 seconds to move 
>>>>>> from one
>>>>>> position to the other, and 10 seconds to integrate.  But from 
>>>>>> holography
>>>>>> we know that the move-and-settle time is less than 5 seconds, so a
>>>>>> 15-second (5+10) regimen should work).
>>>>>>
>>>>>>      To test these, I used 3C147 (~5 Jy) to measure pointing 
>>>>>> offsets at
>>>>>> X-band, with four combinations:
>>>>>>
>>>>>>      1) Narrow-band with 20 seconds.
>>>>>>      2) Narrow-band with 15 seconds.
>>>>>>      3) Wide-band with 20 seconds
>>>>>>      4) Wide-band with 15 seconds.
>>>>>>
>>>>>>      The intention was to observe each mode sufficiently long for 3
>>>>>> consecutive pointing solutions.  Due to a blunder on my part, the 
>>>>>> 'fast'
>>>>>> tests were observed long enough for only 2 solutions.  But that is 
>>>>>> still
>>>>>> sufficient for the following conclusions:
>>>>>>
>>>>>>      A) Wide-band mode works as well as narrow-band.
>>>>>>
>>>>>>      B) Fast mode appears to be as good as 'slow'.
>>>>>>
>>>>>>      There were a few failed solutions for all four trials -- the
>>>>>> statistics are not good enough to claim that any one of these four is
>>>>>> better or worse than the others.  Comparison of the actual 
>>>>>> solutions has
>>>>>> been done for only one antenna (so far) -- there is no significant
>>>>>> differences, other than an expected change in elevation offset as the
>>>>>> sun rose.
>>>>>>
>>>>>>      By looking at the detailed log, the sequence of operations is 
>>>>>> easily
>>>>>> discerned.  (Ken claims this may be unreliable, and he intends to 
>>>>>> peer
>>>>>> deeper after lunch).  But, be this as it may, I deduce that the
>>>>>> *absolute minimum* time needed for a single referenced pointing
>>>>>> measurement is (presuming the antennas are on source) is:
>>>>>>
>>>>>>      124 seconds for 'slow' mode,
>>>>>>      94 seconds for 'fast'.
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> evlatests mailing list
>>>>>> evlatests at listmgr.cv.nrao.edu
>>>>>> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>>>>>>
>>>>> _______________________________________________
>>>>> evlatests mailing list
>>>>> evlatests at listmgr.cv.nrao.edu
>>>>> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>>>> _______________________________________________
>>>> evlatests mailing list
>>>> evlatests at listmgr.cv.nrao.edu
>>>> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>>> _______________________________________________
>>> evlatests mailing list
>>> evlatests at listmgr.cv.nrao.edu
>>> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>>>
>> _______________________________________________
>> evlatests mailing list
>> evlatests at listmgr.cv.nrao.edu
>> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests