[evlatests] Zeeman problem?

Barry Clark bclark at nrao.edu
Thu Sep 22 13:19:06 EDT 2011 

The noise correlation between baselines is an intensity interferometer,
so you get a different answer, depending on whether the source is
resolved or not.  The issues are a lot simpler to look at from the
phased array viewpoint, but you can view it as combining 27 independent
noises, instead of the 351 independent noises on a weak source.

Perhaps we need a separate section in our documentation for calibrating
strong sources.

Bob Sault wrote:
> Rick,
>  
> It would be great if Barry's analysis is the explanation (despite how
> silly I feel). It would tend to agree with my impression that Q and
> U were also corrupted, but at a fractionally lower level. Barry's
> explanation makes sense of this point.
>  
> With self-noise like this with a point source, because the "noise" is
> the same on all baselines, it will obey amplitude closure. So, if
> self-cal is used (as has been the case), the calibration procedure
> is likely to reduce the "noise" level below what a radiometer sensitivity
> equation (ie Barry's argument) will tell you - at least in RR and LL.
>  
> Calibration with these data is quite different to what would be done
> conventionally, because the line is so strong, and because most forms
> of conventional calibration does not have the sensitivity needed.
> When I was reducing the data, the approach I took was to a picked a
> range of channels (roughly the FWHM of the Stokes I line), and deduce
> antenna gain calibration of RR and LL (independently) integrated over
> this range of channels. This approach assumes the bandpass does not
> change significantly over this range of channels (given how
> narrow the line is, this was assumed to be a safe statement). It also
> assumes the Stokes V integrated over the range of channels is 0 (which
> is should be for a classic S curve). Generally I did not do polarimetric
> leakage calibration was not done (Stokes Q and U are not so large, so it
> does not matter).
>  
> The need to consider self-noise would be less of a issue for the ATCA
> observation I did on the line: the ATCA system has higher Tsys, and more
> integration time was used because of the smaller collecting area.
>  
> Best regards
> Bob
>  
> --
> ----------------------------------------------------------------------
> Phone:    +61-2-98721028
> Email:    rsault at nrao.edu <mailto:rsault at nrao.edu>
> Web:      http://www.ph.unimelb.edu.au/~rsault
>  
> -----Original Message-----
> From: Rick Perley [mailto:rperley at aoc.nrao.edu] 
> <mailto:[mailto:rperley at aoc.nrao.edu]>
> Sent: Thursday, 22 September 2011 12:40 AM
> To: Michael Rupen
> Cc: Barry Clark; rsault at unimelb.edu.au <mailto:rsault at unimelb.edu.au>; 
> evlatests at aoc.nrao.edu <mailto:evlatests at aoc.nrao.edu>; Bob Sault
> Subject: Re: [evlatests] Zeeman problem?
>  
>     Right.  The key question to me is why the spectra shown by Emmanuel
> from the VLA data did not seem to show the higher
> noise in the line, as Barry's (quite correct, in my opinion) argument
> suggests should be present.   Separating a noise origin from a gain
> instability origin should be relatively simple ... (says here ...)
>  
>   
>  
> Michael Rupen wrote:
>>  Hi Barry --
>> 
>>    boy, would I love this to be the explanation (regardless of how silly
>>  we'll all feel :)  It should be easy to confirm:
>> 
>>  1- This would have been true with the old correlator as well as the new,
>>    but I thought we were told that the old correlator did not see this
>>    sort of problem.  Emmanuel, can you confirm that was on much weaker
>>    lines?
>> 
>>  2- This argument suggests that the self-noise varies from channel to 
> channel
>>    directly proportional to the signal in that channel.
>>    - this should be seen in all Stokes parameters, so the noise in
>>      self-cal'd Stokes I should be similar.  Hard to check with recent A
>>      cfg data but should be obvious in older stuff, and in a new D cfg 
> image.
>>    - Stokes V spectrum should show noise prop.to the line in that channel.
>>    - Higher-resolution observations should show different noise levels
>>      depending on the shape of the spectral line.
>>    This might be tough to see but perhaps do-able.
>> 
>>    I believe we looked at some very strong lines in L band.  Did those show
>>    the increased noise?
>> 
>>  3- Self-noise should integrate down with time; my impression was that
>>    this did not.  Emmanuel, can you confirm?  It might be interesting to
>>    try stacking the data from several observations in this regard.
>> 
>>  Cheers,
>> 
>>                   Michael
>> 
>>  
>> > We've been thinking for a while about looking to see whether the
>> > Zeeman problem occurs with the phased array.  So I thought I'd
>> > try a thought experiment to see how a perfect instrument would
>> > behave.  To simplify further I made the thought experiment a
>> > filter bank, rather than a correlation spectrometer, which
>> > obviously must give the same result.  Looking at the output of
>> > this receiver, the system receiver noise is, in SEFD terms, about
>> > 12 Jy, the antenna SEFD divided by the number of antennas.
>> > 
>> > So, in the center frequency channel, the maser line _really_
>> > dominates the system, by a factor of 25 or so.  That being the
>> > case, the RMS on the channel will be the maser flux divided by
>> > the square root of BT.  For B = 2 kHz, T = 25m (roughtly the
>> > July 12 setup), the expected rms on the central channel is
>> > 300 Jy / sqrt(2000 * 1500) = 170 mJy.
>> > 
>> > In the end channels, off the line, the power is about 25 times lower,
>> > so the rms will be also, about 7 mJy.
>> > 
>> > It seems to me that this is not incompatible with what Emanuel
>> > sees in his reduction of the Zeeman observation.  The peak excursion
>> > on his Zeeman pattern is 290 mJy, about 1.7 sigma. something not
>> > unexpected.  That is, what we see is about (within a factor of two,
>> > anyway) what we would expect with a perfectly functioning correlator.
>> > 
>> > We may have a Zeeman problem, but I no longer regard the evidence
>> > for it as very convincing.
>> > _______________________________________________
>> > evlatests mailing list
>> > evlatests at listmgr.cv.nrao.edu <mailto:evlatests at listmgr.cv.nrao.edu>
>> > http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>> > 
>> >    
>>  _______________________________________________
>>  evlatests mailing list
>>  evlatests at listmgr.cv.nrao.edu <mailto:evlatests at listmgr.cv.nrao.edu>
>>  http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>>   
>  
> --
> -------------------------------------------------------
> 
> Phone: +61-2-98721028
> Email: rsault at nrao.edu
> Web: http://astro.ph.unimelb.edu.au/~rsault

More information about the evlatests mailing list