[daip] AIPS limitations for large numbers of antennas
Eric Greisen
egreisen at nrao.edu
Thu Oct 16 13:46:07 EDT 2008
Lynn Matthews wrote:
>> Ther is a fundamental limit of ~250 antennas due to the method by
>> which the "baseline" is recorded (256 * IA1 + IA2).
>
> Could this be changed? I asked some MIRIAD types how they handle this,
> and was told that they now use a new formula to encode the baselines,
> allowing up to 2048 antennas.
Short answer - no.
The issue is bounded on two sides. Any new formula has to distinguish
old forms from new. Thus the lowest number in the new form must exceed
the highest number in the old form. Taking 30 as the maximum antenna
number (AIPS actually allows 90 and there are interferometers with close
to that for Solar work), the highest baseline number now extant in data
is 29 * 256 + 30 or 7454. Thus we would like to use 8192 in place of
the 256. Doing so, the highest number for 30 antennas becomes only
29*8192 + 30 = 237598. But to then allow 2048 antennas, this number
becomes about 2^24. To use the standard UV FITS format, this number
must be exactly represented in IEEE floating point and either this is
the largest number that can be accurately done or already too large.
If UV data were processed as tables, this could be larger in integer
form with no problem. AIPS does write an integer table that could have
that form - but only AIPS reads these tables - not CASA or MIRIAD. And
AIPS uses the baseline number internally in floating - so that would not
really help.
For simulations, you really do not need the baseline ID so long as
imaging is the goal. As soon as you want the full power of AIPS, then
you do not the baseline info as you have discovered. To implement the
proper solution, separate random parameters for antenna 1 and antenna 2
would be a lot of work. It could be done but there are a lot of more
pressing things than studies for telescopes that will not be build in
10s of years especially since no other software package would be able to
read the result.
Eric Greisen
More information about the Daip
mailing list