Part of the problem: miriad tries to keep velicities consistent. CASA tries to keep frequencies consistent.
Some explanations, mostly from George:
Are you assuming that the miriad images are labelled with frequencies and
velocities that are in the same frame and related to each other simply by
the rest freq? I wonder if this is actually true. I know that for
AIPS
(and possibly also miriad?), the
frequencies in data and images are not
well-accounted as regards frame (LSR, BARY, etc.). All of the effort is
spent on assigning correct
velocity start/step header parameters as
alternate axis parameters (necessarily linear, I think). The
frequencies (which, for VLA at least, are TOPO at some
difficult-to-ascertain [especially after split] timestamp), such as they
are, just come along for the ride.
To be clear, CASA is nominally attempting (with some difficulty in clean,
at the moment, granted) to maintain
frequency accountability throughout,
including through frame changes, so that velocities may be calculated
flexibly (e.g., for different species) and, if needed, non-linearly over
the full bandwidth (i.e., not as alternate
linear axis parameters, the
inherent errors in which will be more noticable for the wider fractional
bandwidths of the new instruments). If the
frequency labels are always
correct---and I think we should be able to achieve this in both data and
images---revisable renderings in velocity will generally be more
straightforward and accessible, and without having to refer to any
external information.
'>
- in casa in the ms (is there a difference between importuvfits and
'>
carmafiller? hope not)
I wouldn't be very surprised if there is a difference. I would tend to
want to believe carmafiller more than importuvfits, but I don't know
anything about carmafiller. Note that it is precisely the point I made in
the last email that makes it very difficult to ensure robust frequency
accountabilty in CASA for data originating in UVFITS. We just don't know
when those (TOPO) frequency labels were actually correct (and there may be
convention differences in miriad and AIPS when writing UVFITS). Our best
efforts in importuvifts involve a stationary frame (the Doppler-tracked
frame, if any) "back-calculation" according to the alternate velocity
parameters. Basically, velocity parameters imply Doppler-tracking, and
assuming the frequencies are stationary in the Doppler tracked frame (that
is the point, after all) enables us to assign a reasonably secure
fixed-frequency spectral window in CASA.
If the data in question were taken without Doppler tracking---i.e., with
fixed frequencies throughout---I would be a little more confident. But if
the alt vel parameters are present, importuvfits probably assumes that it
was Doppler tracked (when, in fact, they were only Doppler
set and left
fixed), that the frequency labelling is therefore unreliable, and do the
back-calculation anyway. Basically, UVFITS simply cannot describe
variations on these subtleties well enough to permit CASA to proceed with
reliable frequency accountability.
'>
But actually what you say below may explain something Michelle was
'>
telling me about this morning, that in some of her conversion
'>
experiments between miriad and CASA, the vels came out right but not the
'>
freqs.
YES! This is to be expected, I think, given AIPS/miriad/UVFITS
conventions (or lack thereof?) about
frequency accountability.
CASA will make an attempt at discerning the correct frequencies from the
velocity info, and the MS will be different than the UVFITS. If there is
no velocity info in the header, and the frequencies are TOPO and constant
in time, then (I thinK), the frame will be TOPO in CASA and the
frequencies will be correct. Then, when imaging in CASA, the
frequencies should (if all is working) come out correct, too (in
whatever frame you image to), but they will still differ from what
AIPS/miriad produce because they never bothered to convert the frequencies
to the output frame. If CASA's frequencies are correct, then its
velocities will be correct, too, and if AIPS/miriad are calculating
the velocities correctly, they'll match.