[daip] AIPS frustration-related issue

[Eric Greisen]

Bob Garwood wrote:
> 1) Does the reweight[2] value refer to the weight in the gridded image - 
> i.e. those cells with a final weight below this value are blanked - or 
> does it refer to the weight assigned to the data when it's loaded in 
> from the external file (i.e. prior to gridding).
> 
> 2) Can someone remind me what the available tools are for looking at the 
> ungridded data values and weights? (plots would be nicer than just 
> tabulated output)
> 
> 
> Thanks.
> 
> Bob
> 
> Glen Langston wrote:
>> Hi Eric,
>>
>> Thanks for the suggestion.
>>
>> Unfortunately we've been using a very small
>> number for reweight, (1e-12), because of a
>> problem of some single dish fits files
>> have very high weights (that are re-oobserved)
>> while other edges of the map may only get
>> a short obs.   With reweight closer to 1,
>> the short duration obs disappear, causing
>> the observer confusion.
>>
>> Also there is something up with the
>> sync/bessel function as the interpolation
>> should always be between the interpolated
>> values.  It does not seem to be the
>> data, as the other weighting functions do
>> not cause this problem.   We do want to
>> prefer the sync/bessel function as
>> it yields the best angular resolution, but
>> it seems (to me) that its likely there
>> is some numerical error in the sync/bessel
>> interpolation code.  Perhaps extreme weights could
>> be discarded in the convolution function(?)
>>
>> Thanks for your help.
>>
>> Glen
>>
>>   
>>> Glen Langston wrote:
>>>     
>>>> The imaging program SDGRD makes images with very noisy
>>>> edge pixels.  The range in these output images using
>>>> the sync and Bessel function (XTYPE = -16)
>>>> is strangely large.   Most of the other convolution
>>>> functions produce more stable output ranges, so
>>>> the map contains pixels with values only in
>>>> the range between the max input and spectra.
>>>> (Ie gridding the data with convolution function -12
>>>> yields reasonable data)
>>>>
>>>> For XTYPE = -16, the data in the center of the map
>>>> are fine, so to find the data range, use
>>>> TVWIN and IMSTAT to find the min and max values
>>>>
>>>> A simple to work around this problem use
>>>>
>>>> pixra = -1, 20  (for orion)
>>>> pixra = -20, 250 (for the moon)
>>>>
>>>> for these sources,  then
>>>>
>>>> TVLOD
>>>>
>>>> Glen
>>>>
>>>>
>>>>       
>>>>> Thanks.
>>>>>
>>>>> A few more random AIPS questions:
>>>>>
>>>>> I made an image and the full data range somewhere was quite large so
>>>>> that I had to fiddle with the TV color table and the image looked good,
>>>>> but it was all compressed into a tiny range.  Setting aside the issue
>>>>> as
>>>>> to what was different for me in getting to that image, how do I limit
>>>>> the TV to just consider a smaller data range?  I see that tvlod has a
>>>>> pixrange parameter, but setting that never seemed to make any
>>>>> differences in what was actually loaded.   So, how would you handle
>>>>> that?
>>>>>
>>>>> Are you still doing the baseline subtraction in AIPS?  Can you remind
>>>>> me
>>>>> what that task name was?
>>>>>
>>>>> Finally, for that "continuum" image.  How did you generate that in
>>>>> AIPS?  Are you using an AIPS task to average several channels together?
>>>>> Presumably that's all done before any baseline subtraction?  Or is that
>>>>> just one of those channels from that image prior to any baseline
>>>>> subtraction?
>>>>>
>>>>> Thanks.
>>>>>
>>>>> -Bob
>>>>>
>>>>> Glen Langston wrote:
>>>>>         
>>>>>> The image circulated was made with the sinc/bessel function
>>>>>> convolving function, (XTYPE=-16).
>>>>>>
>>>>>> Using the exponential convolving function allows producing a smoother
>>>>>> image, without occasional high spots (XTYPE=-12)
>>>>>>
>>>>>> Glad that last trick worked.
>>>>>>
>>>>>> Glen
>>>>>>
>>>>>> PS. Using different convolving functions can sacrifice angular
>>>>>> resolution for smoothness.
>>>>>>
>>>>>> Here's an image with 40 arc-second convolving function size,
>>>>>> and 16 arc-second convolving function FWHM.
>>>>>> (XTYPE=-12, XPARM=40,16,2,0)
>>>>>>
>>>>>>
>>>>>>
>>>>>>           
>>>>>>> Changing groups didn't help.  I verified that the group was actually
>>>>>>> changed by making a local file where the group was seen to be set to
>>>>>>> aipsuser.  I still can't make any changed to the /home/aips/RUN
>>>>>>> directory.
>>>>>>>
>>>>>>> However, the last trick worked.  Thanks.
>>>>>>>
>>>>>>> You made at least 3 images as seen in aips message log.  Which one is
>>>>>>> the image that ultimately led to the continuum image that was
>>>>>>> included
>>>>>>> in that e-mail?
>>>>>>>
>>>>>>> -Bob
>>>>>>>
>>>>>>> Glen Langston wrote:
>>>>>>>
>>>>>>>             
>>>>>>>> Sometimes you have to change groups to make the links
>>>>>>>>
>>>>>>>> newgrp aipsuser
>>>>>>>>
>>>>>>>> or something similar
>>>>>>>>
>>>>>>>> Alternatively you can put the runfile in the
>>>>>>>> directory you started aips in.
>>>>>>>>
>>>>>>>> If you have the file IDLTOSD.001, put it there,
>>>>>>>> Then you type
>>>>>>>>
>>>>>>>> VERSION = 'PWD'
>>>>>>>> input RUN
>>>>>>>> RUN IDLTOSD
>>>>>>>>
>>>>>>>> This should load the code
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>               
>>>>>>>>> Hi Glen,
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> How do I load the IDLTOSD run file?  Typing:
>>>>>>>>>
>>>>>>>>>  > run idltosd
>>>>>>>>>
>>>>>>>>> results in "TEXT FOR IDLTOSD UNAVAILABLE".  I can see the RUNFIL
>>>>>>>>> environment variable is defined and points at /home/aips/RUN which
>>>>>>>>> sure
>>>>>>>>> seems to have an IDLTOSD.001 in it, so I don't understand why this
>>>>>>>>> is
>>>>>>>>> failing for me.
>>>>>>>>>
>>>>>>>>> I also can't make a symbolic link for a version using my current
>>>>>>>>> AIPS
>>>>>>>>> user number there because I apparently lack the necessary
>>>>>>>>> permission
>>>>>>>>> to
>>>>>>>>> do that even though I am a member of the aipsuser group.  So, what
>>>>>>>>> am
>>>>>>>>> I
>>>>>>>>> doing wrong.
>>>>>>>>>
>>>>>>>>> -Bob
>>>>>>>>>                 
>>> The REWEIGHT(2) parameter is very helpful here - values like 0.1 or 0.3
>>> even might help in preventing excessive **extrapolation** of the data.
>>> It is places where the convolving weight is ~0 that cause the bad
>>> brightnesses at the edges of the data.
>>>
>>> Eric Greisen
>>>
>>>     
>>
>>   
> 
> _______________________________________________
> Daip mailing list
> Daip at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/daip

REWEIGHT(2) refers to the sum of the products of the data weights and
the convolution function weights at each cell scaled by the maximum such
sum.  Output cells with weights below REWEIGHT(2) * Max_convolved_weight
are blanked.  This is what the help file says.  The input data weights
are not examined, they are just used.  That is partly an error - it
should at least be looking for flagged data (weights <= 0) and not
including them.  Other interferometry tasks allow the user to alter the
data weights on input (e.g. use sqrt, fourth root, 1, ...)  If these
options are needed I can fix the code.

The available tools and how one uses them for single-dish work are
described in chapter 10 of the AIPS Cookbook.  The PostScript of that is
available at your site as $AIPS_ROOT/TEXT/PUBL/COOK10.PS or from the web
at http://www.aips.nrao.edu/cook.html.

Eric Greisen