[daip] CVEL in AIPS

[Eric Greisen]

Ross Burns wrote:
> Dear AIPS helpdeskers
> 
> My name is Ross, I'm at Kagoshima university and am a 1st yr PhD student.
> Here at our lab we use AIPS to reduce VLBI data from the VERA 
> interferometer (it's like a small VLBA). 
> 
> We usually get along fine with all of the AIPS tasks that are used in 
> our data reductions, however it seems that the latest version has some 
> modifications that has caused us some trouble. The problematic change 
> seems to be in either SETJY or CVEL or both.
> 
> During past reductions it became apparent, long ago, that CVEL seemed to 
> work better when the array name was set to 'VLBA' and then change back 
> to 'VERA' after CVEL. Here is the code we would use as standard:
> 
> 
> *******
> *TABED
> *******
> tget tabed
> default
> indisk 2; getn 1; keyword 'arrnam'; keystrng 'vlba'; optype 'key'; 
> aparm (4)=3; inext 'AN';inver 1; outver 1; bcount 1; ecount 0; clro; 
> go; wait 
> *Change array name to VLBA
> 
> *******
> *SETJY
> *******
> tget setjy
> default
> indisk 2; getn 1;source 'X_Hya 
> restfreq 2.223e10, 5080000; 
> veltyp 'lsr'; veldef 'radio';
> bif 0; eif 0; 
> aparm (1)=187; 
> sysvel 23.66; 
> go; wait
> 
> *******
> *CVEL 
> ******
> tget cvel
> default
> indisk 2; getn 1; outdisk 2;outclass 'CVEL''
> doband -1; gainuse 1; source 'X_Hya
> aparm 187 0; 
> go; wait; pcat
> * Creating a CVEL file. 
> 
> 
> *******
> *TABED
> *******
> tget tabed
> default
> indisk 2; getn 2; keyword 'arrnam'; keystrng 'VERA'; optype 'key'; 
> aparm (4)=3; inext 'AN';inver 1; outver 1; bcount 1; ecount 0; clro; 
> getn 1; go; wait 
> getn 2; go; wait 
> * Change array name back to 'VERA'
> 
> 
> Where 187 is the frequency channel of the emission peak 
> and 23.66 is the velocity of the emission peak
> 
> However. In the latest version, 31DEC13 we found that this procedure 
> wouldnt work. The velocity offsets were far too big and often seemed to 
> fail altogether.
> We realised that CVEL was always setting the system velocity to zero. 
> 
> So we changed our procedure by first locating the velocity and frequency 
> channel of the emission peak, and then calculating the frequency 
> channel corresponding to the zero velocity. This worked quite well, 
> though it may someday become problematic if the bandwidth does not 
> include the zero velocity, then we would calculate negative values for 
> the corresponding frequency channel, and thats no good. 
> 
> The other problem is that this method only allows us to allocate a 
> system velocity that is a multiple of the velocity resolution (0 km/s +- 
> some number of channels) this gives us less control on the values we use 
> to define the peak of the emission. 
> 
> Generally, it is a pain for any new students who cant get hold of an 
> 'old' version of AIPS. For anyone in our lab running 31DEC12 or earlier 
> this problem does not occur, though I recently re-installed the current 
> 31DEC12 using binary and an internet connection and found the CVEL 
> problem, so perhaps it was included as part of a recent patch? 
> My old (Downloaded around Jan2012) 31DEC12 did not have this issue.
> 
> I think this problem could be something to do with the name change to 
> 'VLBA' though I am not sure. I cant imagine such a fundamental change 
> would have been made to CVEL itself.
> 
> 
> 
> 
> If you could shed some light on what might be going on here it would be 
> a great help for me and all my fellow lab members! Sorry about the long 
> message but i though I should be thorough in the explanation of whats 
> troubling us.

Let me ask a question - the 31DEC13 version that you are using - when 
did you last update it (with the MNJ)?  A bug was put in the code March 
12, 2012 and corrected after October 13, 2013 (correction put in both 
31DEC12 and 31DEC13 then).  This bug was tricky since it affected 32-bit 
Linux only (binary installations) - the very real error in the software 
did not cause the 64-bit Intel compiler or those on Macs or GNU to do 
anything bad.  So any 31DEC12 version taken between these dates or 
31DEC13 version taken before October 21 will have an error.  If this 
does not affect you please let me know - but try the MNJ to update both 
versions and then retry the tasks.

Note - CVEL assumes that the "normal" thing has been done with VLBA data 
and they have been shifted wrt to the first or second antenna in the 
baseline.  The VLBA is different and shifts everything to the center of 
the Earth.  It sounds like VERA does the same - I will add VERA to the 
test in CVEL.

Eric Greisen