spectral matching with different conventions in the headers

[kswang1029]

When exporting an image to a FITS file in CASA, there is an option to set the spectral convention as velocity in the header. So in principle we should see the original image and the exported image are perfectly matched with each others. However, we see an offset in the matched channel index.

test image

[veggiesaurus]

off by one channel?

[kswang1029]

By a lot. The two red lines in the screenshot should be at the same channel (the profiles are identical).

[veggiesaurus]

the spectral matching is done by AST based on the headers. Is it possible that the backend is passing incorrect headers for the CASA image?

edit, no wait. I just re-read the issue. It's just the channel matching that seems to be the issue

[veggiesaurus]

The issue is in getTransformedChannel (and I assume will be the same in getTransformedChannelList). When we set the FREQ based image to use the same matching type as the other image (VRAD), AST needs to choose a new unit as well. Since we normally display things as km/s, that's what it picks. But the VRAD image has its units in m/s, so the spectral matching is off by a factor of 1000.

The solution is to also set the correct units between both images

[kswang1029]

The issue is in getTransformedChannel (and I assume will be the same in getTransformedChannelList). When we set the FREQ based image to use the same matching type as the other image (VRAD), AST needs to choose a new unit as well. Since we normally display things as km/s, that's what it picks. But the VRAD image has its units in m/s, so the spectral matching is off by a factor of 1000.

The solution is to also set the correct units between both images

So I guess if the header is defined in wavelength we need to provide a correct unit too.

[veggiesaurus]

The issue is in getTransformedChannel (and I assume will be the same in getTransformedChannelList). When we set the FREQ based image to use the same matching type as the other image (VRAD), AST needs to choose a new unit as well. Since we normally display things as km/s, that's what it picks. But the VRAD image has its units in m/s, so the spectral matching is off by a factor of 1000.
The solution is to also set the correct units between both images

So I guess if the header is defined in wavelength we need to provide a correct unit too.

Luckily we have a map for default units already thanks to @duidae