Note on the phase/pathlength stability requirement for the receivers:

S.Guilloteau 31 May 2001

Problem:
	Because of sensitivity reasons, the basic phase calibration strategy =
for ALMA will use two different frequencies, one (called "C") for =
calibration and one (called "O") for observation. Hence its accuracy =
will rely on the relative stability of the phase and pathlength =
difference between the two "C" and "O" frequency channels.

	To determine the phase at the observing frequency, several parameters =
will play a role:
	1) the short-term reproducibility of the PHASE of either "C" or "O" =
when the LO system is commuted between "C" and "O" and back again.
	2) the medium-term stability of the PHASE difference between "C" and =
"O"
	3) the medium-term stability of the PATHLENGTH difference between "C" =
and "O"
	4) the atmospheric pathlength stability and/or
	5) the accuracy of WVR pathlength prediction=20

An ideal instrument would require the errors to be dominated by only =
non-controlable term,
i.e. the atmosphere. Fast switching and WVR can give an rms pathlength =
error of about 20 microns, although this is not easy. The specification =
in the hardware for the WVR have been to set a stability requirement =
corresponding to a rms pathlength of 10 microns.=20

I suggest that we put a similar specification for the short-term =
reproducibility of the PHASE
of the local oscillator system when switched back and forth between two =
receivers. To convert
such a pathlength specification into a phase specification requires to =
know the frequency.
In practice, it is the calibration frequency "C" which should be use for =
that, about 90 GHz.
(We can ignore the case when "O" is at a lower frequency than "C", =
because the 10 microns spec
is derived from the highest possible value of "O"). 10 microns =
correspond to about 1 degree
of phase at 90 GHz.=20

The medium-term stability must be such that, after removal of a drift on =
the calibration timescale, the residual be less than 10 microns of =
pathlength also. The relevant timescale is that of observations of a =
phase calibrator at both "C" and "O" frequencies. Its shortest
acceptable value is 0.5 to 1 hour.

Note that there is another technical solution, namely to have two LO =
loops working in parallel,
one controlling of calibration frequency "C", the other controlling the =
observing frequency
"O". This obviously is a more costly approach, but may be a fallback =
solution if the repeatability specification of 10 microns cannot be met.

In summary:
	Phase repeatability of the LO system: 1 degree (frequency /100 GHz) =
(i.e. 10 microns
	pathlength error)