[daip] Re: Smoothing in SPLIT ?

[Eric Greisen]

Nissim Kanekar writes:

 > I've got a bit of a strange problem in AIPS, namely that the frequency 
 > resolution in the FQ table doesn't seem to match that in the header of a 
 > file that has been SPLIT, with Hanning smoothing and resampling. The 
 > FQ table seems to show the original resolution, while the header of the 
 > file shows the (correct) smoothed resolution. The inputs to SPLIT were
 > a 128 channel MULTI-source data set and 
 > 
 > chinc 2
 > nchav 1
 > smooth 1 0
 > aparm 1 0
 > 
 > The header of the output file is below, and it shows the expected 0.25 MHz 
 > channel resolution (the original data had a resolution of 0.125 MHz).
 > 
 > AIPS 1: FREQ        64   3.2500000E+08       1.00  2.5000000E+05    0.00
 > 
 > The output of PRTAB on the FQ table is below, with the 0.125 MHz result 
 > and a claimed total bandwidth of 8 MHz.
 > 
 >     1         1        0.000000E+00      1.250E+05      8.000E+06    1

This is a subtle point.  The frequency increment becomes 0.25 MHz but
the width of the channel is 0.125 MHz and every other one was thrown
away (nchav 1; chinc 2).  This of course ignores the smoothing.

I am not sure what the task should do - I guess it has to get the
total bandwidth right and so may need to use the increment. Note that
if chinc = 1; nchav = 3 it would have also done something odd in the
FQ table.

This brings up a point - do you want to smooth your data befaore or
after applying the bandpass cal?  That depends on what you did when
you ran BPASS.  If you did not smooth in BPASS you want to apply
smoothing after applying the BPASS correction.  SMOOTH types 5,6,7,8
are done after bandpass.

Eric Greisen