[daip] Ringing from BPASS

[Ingyin Zaw]

On Thu, 7 May 2009, Eric Greisen wrote:

> Ingyin Zaw wrote:
>>  Hello,
>>
>>  I'm running into a problem where the bandpass solution from BPASS
>>  introduces ringing in spectral line data. Furthermore, the problem is
>>  excerbated when (8) spectral channels are averaged with AVSPC. Since we
>>  are looking for narrow water maser lines, this ringing poses a serious
>>  problem.
>>
>>  The details are as follows:
>>
>>  AIPS version: 31DEC09, 64 bit (problem also seen in 32 bit, 31DEC09)
>>
>>  Computer system:
>>  Linux rglinux9.cfa.harvard.edu 2.6.18-128.1.1.el5 #1 SMP Mon Jan 26
>>  13:58:24 EST 2009 x86_64 x86_64 x86_64 GNU/Linux
>>
>>  Dataset: BK114A, taken Nov. 28, 2004, 256 Channels, 8 IFs
>>  (NB: similar problem seen in BB242D, taken Jan. 21, 2008, 512 channels,
>>  8IFs)
>>
>>  Details on ringing:
>>  - Ringing in data (amplitude) when viewed with POSSM and BP is applied
>>  (doband 1).
>>     - Mainly in baselines to Ant 12, the phased VLA array.
>>     - Behavior is different (more severe) on data averaged with AVSPC
>>       (avoption 'SUBS'; channel 8)
>>     - Behavior also differs when the file contains only one source or
>>       when it contains several sources even though BPASS and POSSM are run
>>       only on one source.
>>  - Ringing is not seen when BP is not applied (POSSM, doband -1)
>>  - Ringing is less noticeable when looking at the BP table itself
>>     (POSSM, aparm(8)=2)
>>
>>  I've made 3 files available for testing via ftp from:
>>  cfa-ftp.harvard.edu
>>  Name: anonymous
>>  Password: <youremailaddress>
>>  Folder: incoming/lincoln/
>>  Files:
>>  (1) BK114A_2145_256.FITS
>>  (2) BK114A_2145_32.FITS
>>  (3) BK114A_2145FILE.FITS
>>
>>  (1) Contains only the source, 2145+067, on which I'm running BPASS (loaded
>>  with FITLD with source selection).
>>  (2) Is the file with AVSPC (avoption 'SUBS'; channel 8) ran on file (1).
>>  (3) Contains other sources, in addition to 2145+067, all from one file
>>  loaded with FITLD.
>>
>>  Each file contains one BP table, output of BPASS with identical inputs.
>>  The parameters different from default are:
>>
>>  calsour '2145+067'
>>  solint -1
>>  soltype 'L1'
>>  refant 6 (LA)
>>
>>  (docalib -1; doband -1; bpassprm 0)
>>
>>  Please let me know how the ringing can be fixed.
>
> I have taken a look at these data.  If one changes the AN table (as a test 
> only) to make it not be the vlba, then the BP correction makes the end 
> channels (one or both) some 1000's of times larger than the rest.
> Therefore the BP solution is bad at the ends (not surprising but worse than 
> usual).  You could try UVCOP of the data omitting a few channels at each end, 
> then do BPASS and POSSM.
I tried UVCOP of channels 11 to 246 and rerunning BPASS. There is still 
ringing at the edges. Also, it is puzzling that the BP solution itself 
does not show strong ringing and the data shows no ringing but when the BP 
table is applied to data, it shows strong ringing. Also, why is the 
ringing worse for the AVSPC file?

> A VLBI knowledgable observer queried why you are doing BPASS and also was 
> very perplexed by your using FRING instead of CALIB when asking it basically 
> not to search in delay space at all.  If I understood him correctly, FRING is 
> normally used an IF at a time for single IF delays and then a multi-band 
> delay is fit on AVSPC'd data.
Ideally, I want to run FRING with at least 100 ns and 100 MHz windows. The 
smallness of my dparm parameters was to demonstrate that it fails even 
with those unreasonably small numbers.

The full sequence of steps I'm running is: (1) FRING on 1 src 
(aparm(5)=0), (2) BPASS on that source to remove additional, non-linear 
delays, then (3) FRING on all sources with (aparm(5)=2) for the full 
multi-band delay. The aparm(5)=2 setting runs without memory problems on 
the AVSPC file. However, the rining from BPASS is much worse for the AVSPC 
file, which is why I was trying to run it on the full set of channels.

Thanks,
Ingyin