Re: [alma-sw-ssr] Paris meeting agenda

[Bryan Butler <bbutler@aoc.nrao.edu>]

steve wrote:

> 
> 
> Here are some topics that I hope we can address at our meeting. It 
> includes several items that we have touched on but not explored in 
> depth.
> 

...

> 8) Correlator data processing
>   a) What lag spectrum window functions do we support?
>   b) Decimation after windowing and FFT?
>   c) If you truncate the lag spectrum (with a rectangular or other
>      window function), it is the same as smoothing in the spectral
>      domain. Allowing variable width window functions can tailor the
>      resolution to the science requirements. Do we want to do this?
>      Or perhaps other smoothing implemented directly in the spectral 
>      domain?
>      

shouldn't this all be done in post-processing?  just save the lag
spectrum, then you can do any windowing that you want after the
fact, prior to FFTing into the spectrum.  this is how it seems to 
me that it should be done anyway.

i guess the argument is in data rate though...

if you *are* going to do the lag-spectrum windowing in the on-line
system, then shouldn't it be absolutely controllable by the user?
for example, have a table of taper/window value vs. lag which can
be input.  of course for those who don't want to mess with it, then
whatever prepares the "control files" for the system should have
some obvious preset types (no tapering, cosine or cosine squared,
kaiser-bessel, blackman-harris, e.g.).  for those who want to really
control the taper, they can create their own taper table (say, for
instance, you wanted to do the prolate spheroidals optimized for a 
particular angular scale size and they weren't supported in the 
"standard" taper types)...

	-bryan