FILLM and proper weights

[K. Desai]

Eric,

Michael and I explored my new FILLM weights scheme and it looks promising but
of course, needs more testing.

Your key test of comparing the nominal sensitivity and the Tsys or Tant
numbers is inconclusive.  There is a fudge factor that 'somebody' sticks
in that goes into the nominal sensitivity but is not included in the 
Tsys or Tant.  Grossly, the NS and Tsys track each other but the exact scalings
especially between different correlators [IFs or POLs] are clearly off
due to these fudge factors.

Michael's preliminary tests indicate that the weights do properly reflect
the scatter in the visibility data. Also, Ed, Michael, and I discussed it and
we think a sensible option for the weights would be to put them in units of
inverse squared millivisibilities .  Eg, with a noise of 5 mJy, the weights 
would be about 0.04 when the data is finally calibrated into Janskys.

It's not completely clear that I understand the etac factor mentioned in FILLM
and it looks like you introduced it so if you could find some notes on it and
explain it to me, that would be useful...  

[I'd like to be a priori sure that those little niggling factors don't represent
some weight scaling that I am leaving out...]


will this effort do to supplant my promised memo?

-ketan


PS, does anybody know how APCAL would react if confronted by only a GC table and
not a TY table?