[daip] AIPS and BPASS

[Eric Greisen]

Yuri Y. Kovalev wrote:
> Dear Eric,
> 
> Thank you very much for your comments.
> 
>> How stable are the phases within each scan?  If they are stable,
>> bpassp(5) should be 1.  If amplitude is stable, but phase is not,
> 
> I am sorry, but I do not understand, what kind of stability do you mean? 
> I did not understand this in the AIPS help to bpass, I still do not 
> understand it getting now in an e-mail from you.
> Stability in time?
> Stability (being constant) versus sp channels?
> The same question goes for amplitudes stability.
> 
> What does the 'stability' mean?
> 
> All the sources are different. And FRING solutions are different.
> 
>> bpassp(5) = -1.  With 0, the data of each record are normalized in amp
>> and phase which can introduce a Ricean bias since the expeted amplitude
>> in the case of no signal is not zero.  VLBI often uses all channels so
>> bpassp(10) = 1 or 4 rather than 2.
> 
> What is the differnece between bpassp(5) = -1 and bpassp(5) = +1 ?
> In a sence of getting bpass solutions and calibrating phases and 
> amplitudes by bpass calibration application?
> Is there anywhere a memo discussing it in details? Would appretiate it 
> very much. After your first comments I see that I trully need to 
> understand how does it work in AIPS. Thanks!
> 
> Yuri

By stability we usually mean with time.  If your data are very variable 
in phase with time then averaging will reduce the amplitude very much 
and therefore destroy the coherence of the bandpass.  In that case, 
correcting at least for average phase with BPASSP(5) = -1 is needed.
If the phases are good so that there is no loss of coherence when 
averaging the scan, the BPASSP(5) = +1 is better to avoid the somewhat 
noisier correction.

BPASSP(5) = 0 also does amplitude which is affected by Ricean bias (all 
amps are > 0 so not Gaussian in noise) so there can be a slight amp bias 
even if each time sample has good signal to noise.  Delaying the 
amplitude normalization until after the solutions are found (BPASSP(10)) 
is therefore better when possible.  Of course, if your amplitudes are 
also very variable with time, then it may be better to use BPASSP(5)=0 
and get a good averaged normalized bandpass - and then deal with that 
variability later (in self-cal).

Eric Greisen