Here's a page where to list the CVEL tests done in light of the recent bug.
Adrienne's current approach to the problem is to re-run the post-calibration steps on the .LINCOP file, not the .SPLIT file. SPLIT is run at the end after UVLSF and CVEL:
- Run UVLSF on the LINCOP file with
FLAGVER 1 This is equivalent to applying the calibration and then subtracting the continuum. Note that this will subtract your phase and flux calibrators from themselves because they are considered "continuum". You cannot run this with SOURCES '[galaxy]'; it effectively SPLITs out the galaxy and creates a single source file with continuum subtracted. Not what we want. CVEL does not work correct on this file - I believe it only shifts the first scan because it only prints out average shifts for the first scan and not for any of the later ones (unlike when using the multi-source file).
- Run SETJY with the proper inputs on the LINCOP file:
SYSVEL [gal velocity in km/s]
RESTFREQ 1.420e9, 405752
APARM(1) = [ref channel] In our case, our velocity reference channel is in the middle of the cube (109). This tells AIPS that it should associate channel 109 with the galaxy velocity and calculate doppler shifts from the HI line. If you get huge errors about shifting large amounts in CVEL (the next step), you have probably done something wrong here.
- Run CVEL with the same inputs as before:
APARM(1) = [gal velocity in m/s] (Note the different from SETJY which is in km/s)
APARM(2) = [ref channel] (109 for VLA-ANGST)
APARM(3) = 1 (heliocentric velocity)
APARM(4) = 0 (optical velocity definition)
APARM(5) = 1.420e9 (HI rest freq part 1)
APARM(6) = 405752 (HI rest freq part 2)
APARM(7) = 1 (VLA Data)
- Run SPLIT to copy out GR8 and create a single source file after all processing steps.
- Run imaging script (COMBINER/COMBINA) normally to DBCON all test data sets together and image them.
I then checked the difference between the two dirty maps (one with CVEL before SPLIT and one with CVEL after SPLIT) by subtracting the dirty maps in miriad. Here's a link
to the difference image cube (normal reduction cube - test cube) in fits format. I am currently a bit confused because it seems like one of the maps has more flux than the other - we shouldn't be losing flux, we should just be redistributing it, so I imagine some channels should be more negative and some more positive. It appears that the normal cube has about 1-3 mJy/beam less flux on average where there is galaxy emission. Our noise is about 1.5 mJy/beam.
I believe the negative in one cube is due to the difference in beams between the normal and test cubes: here's a difference image of the beam at channel 109 (the reference channel): link
(again fits format)
- Convolve the two cubes to the same size and subtract. 20"? This is difficult because there is no clean beam information because they're dirty maps.
- Compare CLEANed cubes?
Also note that this is all for observations with a single FREQID! I am not sure yet if this will work in the cases were we have 3 FREQIDs.