[daip] AIPS and ALMA

[Eric Greisen]

Chris Carilli wrote:
> Eric,
> 
> Well, the good news is that I am happily calibrating and reducing ALMA
> data in AIPS. In fact, just about to submit a first very interesting paper
> where AIPS calibration and imaging is front-and-center, and the results
> are in fact much better than CASA.  I realize this isn't a competition,
> but thought you might be happy to hear.

I am happy to hear that - note that I just got some other ALMA data 
written by CASA and there were major problems.  The frequency increments 
were given as negative when they should have been positive and the 
antenna positions were given in an erroneous coordinate system.  That 
only matters if you run UVFIX which will get wrong answers.

It is not supposed to be a competition, but there are still management 
types that are acting as if it were...

> 
> now the bad news, and please read on, because it isn't really bad news,
> just annoying. i think these 2 issues will be easy to address:
> 
> 
> 1 .seems like something has changed in the new aips version.
> if i run imagr on ALMA data in AIPS using default version (''), i get:
> 
>> go
> IMAGR2: Task IMAGR  (release of 31DEC11) begins
> IMAGR2: Doing no flagging this time
> IMAGR2: Create J1058+015   .IMAGR .   1 (UV)  on disk  1  cno  122
> IMAGR2: Beginning channel    1 through    1 with  1 IFs
> IMAGR2: DATGET: ERROR  1 READING UV DATA
> IMAGR2: UVREAD: ERROR   1 READING UV DATA
> IMAGR2: UVREAD: ERROR READING Input UVdata
> IMAGR2: IMACPY: ERROR COPYING Input UVdata
> IMAGR2: TO UVdata work object
> IMAGR2: Deleting UV work file:
> IMAGR2: Destroyed UV image file: catno=    122 disk= 1
> IMAGR2: Purports to die of UNNATURAL causes

I note that your "TST" version is 31DEC11 rather than 31DEC12 - where 
are you?  Is this in fact a version current to the end of 31DEC11?  And 
I do not know what the logical "OLD" points to.

I tried your channel 1 data set and IMAGR works just fine in the 
following versions:
    31DEC12  64-bit  gnu compiler
    31dEC12  64-bit   Intel compiler
    31DEC11  64-bit  gnu
    31DEC10  64-bit  gnu
    31DEC12  32-bit  Intel
    31DEC11  32-bit  Intel
    31DEC10  32-bit  Intel
all on Linux-type boxes.  Are you on a Mac?

What are your inputs?

> 
> 
> If I then run imagr  with version 'old',  it runs just fine and produces a
> beautiful map.  so something has changed between the two?
> 
> 
> 2. this isn't relevant to ALMA data yet, but we have run into a problem
> when reducing data for an array with more than 50 antennas (PAPER). many
> of the tasks don't like arrays bigger than 50 antennas.  Dave Green dug a
> little and found that 50 seems to be hard-coded in some antenna arrays in
> POPS.   is there any way you could increase this to like 128 or 256 or
> something?
> 
the short answer is no.  The limit on antenna number in adverbs of 50 
comes from the assumption that no user will want to type 50 numbers and 
the adverb can be used as include these or exclude these which means 
that up to 100 antennas can actually be handled.  Inside AIPS the limit 
is currently set at 90 antennas and it is quite expensive to raise that 
since arrays come in as max antenna squared.  The UV data format limits 
the maximum number to 255.  Note that the 90 is in fact a parameter in 
an include file and can be changed easily.  A local version of AIPS 
could be compiled with any number you want up to the 255 limit which is 
much harder.

> 
> thanks, and hope all is well,
> see you in August,

I'll be here.

Cheers

Eric