Wednesday, July 9, 2008

Two More Ways to Identify Bad Data

Crystal also said a good way to identify RFI is to look at Stokes V, either in TVFLG or UVPLT or whatever. RFI is usually polarized and will pop out as unusually high baselines/channels/times.

Stokes V will usually mimic your amplitude structure (as a function of baseline length). So, for example, if your source has high amplitudes at short baselines, Stokes V should also be a bit higher on these short baselines (so don't get confused and flag them).

And another way to find that bad data-- Data weights. Plot your data in UVPLT with BPARM = 0 13. The weights for all the data should cluster-- if there are any points that are anomalously low or high, flag them! Crystal says she likes to use WIPER. Obviously, be more concerned about data with really high weights that really low weights-- because this data will affect your overall data set more, as it is up-weighted!

1 comment:

Anonymous said...

I am re-reducing 1667 MHz OH spectral line data from project AC319 (JUN-92,
polarizations RR, LL, RL, LR). I chose one spectral channel from one day,
and I plotted the visibility amplitudes with both TVFLG and UVPLT. The
data are split, and they are for the program source W22.

I plotted the visibility amplitudes in UVPLT (stokes 'rr'), and I don't see
any amplitudes over 20 Jy for baseline lengths 4 - 9 kilolambda. However,
when I plot the same data in TVFLG (stokes 'rr', sorted by length), I see a
number of amplitudes higher than 20 Jy in the same baseline range.

Why would the higher amplitudes be displayed by TVFLG but not by UVPLT? In
both cases, I chose flagver =-1, so no flagging should have been applied
when I ran either task.