[daip] autoboxing
Eric Greisen
egreisen at nrao.edu
Wed Jun 24 15:18:29 EDT 2009
Lynn D. Matthews wrote:
> Eric,
>
> I agree that having the option to write out a series of OBOX files would
> be useful. Personally I would also like a record of each channel's boxes
> in the history file, although it would make it long. Previously I had
> done the equivalent of all this through a POPS script I wrote to read a
> series of pre-made box files, clean one plane at a time, record the
> boxes in the HI file using HINOTE, insert the plane into a cube using
> MCUBE, etc.
For my hand reduction of NGC6503 - my script used a input BOXFILE from
channel N-1 for channel N and then wrote a new OBOXFILE. I used the TV
to adjust the boxes as needed. I did not put all in the HI file but did
save the many boxfiles. Eventually I used MCUBE to put the cube
together. I need to think about the cube reduction issue more deeply.
> Meanwhile I have been putting the autoboxing option to the test using
> some VLBA maser data. I found I needed to set IM2PARM(1) to a very low
> number to prevent initial boxing of sidelobes, so I chose IM2PARM(1)=1
> and left the rest of the array as zeros. I wasn't clear as to what
> IM2PARM(5) was doing when left to the default, but 0 worked, so I didn't
> alter it.
IM2PARM(4) is meant to help with sidelobes so that IM2PARM(1) could be
larger.
>
> The results are very encouraging. When cleaned to the same limiting flux
> as my hand-boxed image, the autoboxed image has slightly higher noise,
> but manages to avoid the production of annoying artifacts around the
> bright, spatially extended masers. IMAGR also finished 20% faster with
> the autoboxing. I compared with an image that Bill Cotton made using
> OBIT's autoboxing some time ago, and the AIPS/OBIT results are virtually
> identical. Some screen shots are attached for two test cases I tried if
> you're interested. E and K refer to two observational epochs.
>
> I did however run into one problem. When I had IMAGRPRM(8), (9), and
> (10) set for the second test, IMAGR crashed after filtering with the
> messages below. This did not happen during the first test.
>
I suspect the failure below is a consequence of problems handling
filtered output and only accidentally related to the auto-boxing. The
filtered data were so like the non-filtered data that there was nothing
left to Clean. Somehow this did not get passed properly and so it tried
an illegal subtraction and everything fell apart. I will have to look
at this - it may come from my adding a continue Cleaning after the
"final" filtering in case the filtering mattered. When it doesn't -
which is what one hopes for a "final" it should then complete gracefully.
>
> wrest > IMAGR1: Filter abs value below 0.001000 Jy in 13-pixel diameter
> area
> wrest > IMAGR1: Field 1 filtered 6814 components of 643.163 Jy
> wrest > IMAGR1: to 6811 components of 643.176 Jy
> wrest > IMAGR1: All fields totaled 6814 components of 643.163 Jy
> wrest > IMAGR1: now total 6811 components of 643.176 Jy
> wrest > IMAGR1: VISDFT: Begin DFT component subtraction
> wrest > IMAGR1: VISDFT: fields 1 - 1 chns 1 - 1 in 1 CC models
> wrest > IMAGR1: I Polarization model processed
> wrest > IMAGR1: Field 1 min = -165.5 MilliJy,max = 192.2 MilliJy
> wrest > IMAGR1: Resuming Clean with 6811 filtered components
> wrest > IMAGR1: CLRHIS: NO FIELDS THIS PASS ABOVE MINIMUM FLUX
> wrest > IMAGR1: VISDFT: Begin DFT component subtraction
> wrest > IMAGR1: VISDFT: fields 1 - 1 chns 1 - 1 in 1 CC models
> wrest > IMAGR1: VISDFT: ERROR 1 NO POINT COMPONENTS FOUND
> wrest > IMAGR1: OUMSUB: ERROR SUBTRACTING CLEAN field number 0001
> wrest > IMAGR1: OUMSUB: FROM UVdata work object
> wrest > IMAGR1: UVSUBM: ERROR SUBTRACTING CLEAN field number 0001
> wrest > IMAGR1: UVSUBM: FROM UVdata work object
> wrest > IMAGR1: CLNUV1 : ERROR CLEANING CLEAN process object
> wrest > IMAGR1: CLNUV : ERROR CLEANING CLEAN process object
> wrest > IMAGR1: Deleting UV work file:
> wrest > IMAGR1: Destroyed 1 extension files of type AN
> wrest > IMAGR1: Destroyed 1 extension files of type FQ
> wrest > IMAGR1: Destroyed UV image file: catno= 319 disk= 3
> wrest > IMAGR1: Purports to die of UNNATURAL causes
> wrest > IMAGR1: wrest 31DEC09 TST: Cpu= 1307.2 Real= 1346 IO= 94614
Eric Greisen
More information about the Daip
mailing list