[daip] Re: Request for help on CLCOR (fwd)

[Leonia Kogan]

Mark,

you wrote:

>My long experience with VLBI software leads me to believe that
>sign conventions cannot be determined by pure thought. 

The sign convention was not determined by pure thought.

I have never use the "pure thought" in my conclusion of my AIPS tasks!!!!!!!!!

Yesterday I carried out several tests with input file having diferent sign 
at the vertical delay column

Today I have carried out another test which simulates the correlator 
correlation with 'shifted' atmosphere.

The following is the description of the test:

1. VPLOT of the data (phase) with DOCALIB =-1 (no correction)

2. SPLAT the data with DOCALIB=1 GAINUSE=4;
   CL=4 was created by CLCOR, opcode 'atmo' The vertical delay column 
   at the input file is positive.
   So the SPLAT's output data simulate observation with 'shifted'
   atmosphere.

3. CLCOR to correct the virgin CL table (CL=2 = copy of CL1) of the 
   SPLAT's output data.
   opcode 'atmo'; 
   The vertical delay column at the input file is negative

4. VPLOT of the SPLAT's output data.
   DOCALIB=1; GAINUSE=2;
   The plots are coincide (as expected) with the VPLOT's plots 
   of the origin data (DOCALIB=-1) (item #1)

This test proves the correct correction/application of the CL table
created by CLCOR opcode 'ATMO'

So enjoy.

I am very confident that CLCOR option 'ATMO' works properly with possible 
minor errors which can be found during your work with the new CLCOR.

You can use Tony's data to make additional test if you want.

Please "start using the new CLCOR"


Leonia

----------------------------------------------------------------------------
Leonia,

  Regarding testing the sign convention of atmospheric
corrections in the new CLCOR...

On Mon, 2 Dec 2002, Leonia Kogan wrote:

I first said...
> >It seems to me that the best way to test such a program
> >requires correlating some data twice: once with normal
> >atmospheric parameters in the correlator and once with
> >a significant change in the atmospheric parameters.
> 

You replied...
> Having spoken with analysts and John Benson I realized that it is very
> dificult to change the atmospheric parameters for the second
> correlation.
> 
> Right now I have checked the sign of the vertical delay given at the 
>> input file and I can report that the given value recalculated 
> to the given ZA is added to the correction and subtracted 
> from the column ATMOS at the Cl table.

   My long experience with VLBI software leads me to believe that
sign conventions cannot be determined by pure thought.  One old-time
VLBIer (Dave Schaffer) said that 
  "the probability of getting a sign convention correct is 1/2^n", 
where "n" is the number of times you think about it.  My experience
is that this is close to being right.

   There is, in my view, no option other than to run a test against 
the correlator.  If NRAO is developing software to support the 
analysis of telescope data, it must find a way to test the software.
This also should include proper documentation of how the test was
conducted and what the results were.  Simply saying that it is 
"hard" to do such a test is admitting defeat.  It was hard to build 
the VLBA; it was hard to write AIPS; but that is the nature of the 
enterprise we are involved in.

   In this case, it turns out that I remember an old data set
that Tony Beasley gave me a while ago.  It had, I believe, some
data processed with the atmosphere "turned off" for one(?) station.
I've put this data set on our anonymous FTP site.  You may want
to retreive it for testing purposes.

There are three files:

  147398400   16900001.FITS-CODE
  147225600   16900001.FITS-TJVB
         20   REIDME               (Tony has a sense of humor)

Here is how to access them:

  % ftp cfa-ftp.harvard.edu

    Login:       anonymous
    Password:    emailaddress

  ftp> cd outgoing
  ftp> cd reid
  ftp> mget *
  ftp> quit
=============


> I think you can test/use the new option with your data.
     Unfortunately, with corrections of order 3 cm in 
vertical path this is hard.  We could test the calibrator data, 
by adjusting it, then re-fringing, and re-solving for
vertical delays.  However, if we want to test the
conventions used to adjust interferometer phase (to fix
phase referenced data), the changes are too small for one to be 
confident that everything is correct.  

Mark

-------------------------------------------------------------
Mark J. Reid                 Phone: 617-495-7470 
Harvard-Smithsonian CfA      Fax  : 617-495-7345
60 Garden Street             Email: reid at cfa.harvard.edu
Cambridge, MA 02138, USA     Web  : cfa-www.harvard.edu/~reid                  
-------------------------------------------------------------