Reported velocities offset from actual values

[taylorr5@…]

Hi, I am a PhD student at Cardiff University. I'm trying to use Duchamp as a source extractor on an HI data cube. As there are several sources of interference I run Duchamp over different subsections of the cube to avoid them. It seems to be working well over relatively nearby regions (0 - 6000 km/s), but at greater distances the velocities reported by Duchamp and the actual values (as viewed in kvis and checked with the miriad task mbspect) start to diverge. There is a clear trend that this gets worse at higher velocities, over 500 km/s difference at 18000 km/s. The positional matches are very good and the spectra look as expected, it's just that the reported velocities do not match. Below are shown the relevant parameters I'm using in case I've made a silly mistake somewhere. Many thanks for any help, Rhys Taylor

flagTrim false flagSubsection true subsection [16:296,11:66,175:730] flagsmooth false flagAdjacent false flagBaseline false flagGrowth false snrCut 3.5 minpix 3 threshSpatial 3 threshVelocity 5 minChannels 3

[MatthewWhiting]

Hi Rhys,

Sorry for not replying sooner -- the trac notification seems to have been playing up a bit.

Sounds like an odd problem, and perhaps due to the way Duchamp is interpreting the WCS headers -- perhaps not doing the conversion to velocity correctly. It's hard, though, to see exactly what the problem is without seeing the FITS header or the data...

Does the same thing happen if you don't use subsections? (Or do you get too much crud from the interference to tell?)

I can take a look at your data if you like, if it is not too big a file to send.

Matthew Whiting.

[MatthewWhiting]

Hi Rhys,

I have looked at this problem and think I have it working. The problem was that the code was transposing the spectral type into a different scaling system to what the FITS file had, meaning that the quoted velocities were calculated in a different way. I've stopped it doing this, so that it uses whatever spectral system the FITS file says to use.

The next release of Duchamp (1.1.2 -- which should come out soon) will have the fix in it, but I have put a "beta" version of that release on the website. If you can have a go with it and make sure the velocities come out to what you expect, that would be good. If they don't, let me know!

Matthew Whiting.

[MatthewWhiting]

As this problem appears to be fixed, I'm closing this ticket.

MW