[mmaimcal] Re: WVR memo

[Richard Hills]

Dear Larry,

Thanks for your comments.  I note that this seems to have been
copied to the receiver people, which is good.  In case there are
some who haven't seen the original note, see

   http://www.mrao.cam.ac.uk/~alma/wvr/ALMA_Radiometers.ps
   
   or

   http://www.mrao.cam.ac.uk/~alma/wvr/ALMA_Radiometers.doc

I am afraid it is rather long and turgid and it would be easier
to follow if we had been able to include some diagrams, but the
usual problems of running out of time and effort applied.

We'll try to take most of the points on board in an improved
version, but here are some quick responses on the main issues:

> 1.   the right word is "correction."

Agreed

> 2.                                            .............  If the
>     astronomical calibration is not to be degraded by use of the WVR,
>     then the systematic error in the WVR's estimate of the atmospheric
>     phase must change by much less than the spec over the interval
>     between astronomical calibrations.  Thus, the spec is relevant.

I agree that we need two specifications:
(i) the short term phase fluctuations that result, for example,
from noise in the radiometer and would produce de-correlation,
and 
(ii) the allowed drift or other systematic error over the period
of a calibration cycle (or better a few cycles, e.g. 5 minutes).

What I was trying to propose (although I accept that this was
not explained clearly) was that both these figures should be set
at 10(1+wv) microns of path.  I think this should be possible
but tough.

If you are suggesting that the figure for the drift should be
8.4 femtoseconds of delay, which corresponds to only 2.5 microns
of path, then I would be interested to hear the origin of that
requirement.  This is after all only 1 degree of phase at 345
GHz.  As best I can understand it the equivalent value in the
antenna specification is 15 microns.

I would also resist the idea that the calibration might be
"degraded" by the use of the WVR.  Obviously if conditions are
very good you simply turn off the correction.  (I would have
thought that times when the atmosphere was stable to 2.5 microns
of path would be rare indeed, but assuming the final result is
more like the figure I am suggesting this will presumably be
true quite often on short baselines.)  Assuming the correction
is done post-facto (which sounds as if it is not going to be a
serious problem for the data processing) then I would have
thought that it would be straightforward to test how much the
WVR is helping from the residuals on the calibrators, etc.

 > 3.  LO locking.

I think that if the 2 GHz and 125 Mhz can be provided then this
would be fine. (Plus the 20 Hz of course.) Good point about
phase ambiguity. The frequency step would probably want to be of
order 125 MHz which would move any spike by ~200 km/s but not
change the WVR performance too much. This should be fairly easy
to arrange internally to the WVR by changing a divide ratio
somewhere.

> 4.  Keeping open options for using the astronomical receivers.

I was mostly raising these issues so that we could discuss them
and with luck close them off.  It would be helpful if that is a
conclusion reached by the Science Group meeting tommorrow.
 
> 5.  Cooling.

I am encouraged by your statement that commercial mixers are
likely to be coolable.  How much improvement in noise
performance do you think one could expect?

I agree that a separate dewar and cooler is undesireable.

> 6.  Under "Form of switching," par 3, you say that for a 2-mixer
>     correlating radiometer "with appropriate switching we can
>     ... separate the sidebands."  I don't think so, since this is just
>     a radiometer, not an interferometer.  Do you know of some sideband
>     separating technique of which I'm unaware?

I have not been through this in any detail, but here is my
thought process.
In this section I was thinking of a device where one take the
signal from the sky, S, and that from a load, L, to the opposite
ports of a 180 degree hybrid (a magic tee or whatever) and feed
the sum and difference to two identical receivers.  The two
voltage outputs, A and B, are therefore S+L and S-L.  You then
form the product A times B to get S^2 - L^2, which has the usual
advantages of reduced sensitivity to gain fluctuations, etc. 
This is, as far as I can see, analogous to an interferometer
with a source S in a positive lobe and a source L in a negative
one.  If the receivers are heterodyne and double sideband, the
usual tricks involving phase switching the LO's apply:  180
degrees flips the sign of the measurement and 90 degrees enables
you to separate sidebands.  You do need a complex correlator
which would not strictly be necessary otherwise if you could
phase up the DSB system properly.

>7. Also under "Form of switching," you discuss the optics for
>   implementing the pointing correction feature.  To cover only half
>   the subreflector requires an f/16 beam (as opposed to the f/8 beam
>   for the whole subreflector), and consequently double the aperture
>   diameter in the focal plane (to about 110 mm at 5w).  It's very hard
>   to fit this in, especially if it needs to be close to the other
>   feeds so as to minimize the beam offset.  When this is combined
>   with the increased sensitivity requirement and with the need for
>   multiple-temperature calibration loads, the technical difficulties
>   of this receiver are getting substantial.  In view of the fact
>   that the pointing correction scheme is completely unproven, and
>   that the main points of the receiver design need to be frozen very
>   soon, I don't see how we can plan to support this feature.  Unless
>   it can be established within the next month or two that it is both
>   essential and likely to work, I expect it to be designed out.

I agree that doing the pointing correction is not going to be
easy, but ALMA is not an easy project. I don't think the work
has been done in terms of the analysis of the site data to see
how much we would loose by not doing this.  The area of the
focal plane that we have to play with is at least 500mm in
diameter and I find it hard to believe that there is no place to
put a 110mm diameter mirror.  I cannot therefore accept that it
should be designed-out now. We need to do the analysis and if it
does look necessary we should try to organise practical tests.

I hope you appreciate that my underlying approach is in sympathy
with yours:  we need to get the specifications fixed and the
concepts defined.
Unfortunately this area is lagging behind others.

Best Richard