For the JCMT-CSO Interferometer, IF signals from the receivers on the two telescopes are transmitted to the DAS to be correlated. In general, the astronomical signals arrive at one antenna before the other, and delay compensation is thus needed before correlation. Coarse delay compensation is realised by switching appropriate lengths of optical fibre into the signal paths. Finer corrections are made in the software during post-processing of the data. Numerically-controlled oscillators are used to generate offset frequencies which are introduced into the LO and IF chains to remove fringe rotation caused by the earth's rotation. The fringe rate can only be removed precisely for one frequency (chosen to be the centre of the IF band) and this sets a limit on the maximum integration time for individual records of around 10 seconds.
As for the pointing, irregularities of the azimuth track will contribute
to the displacement of the phase centre of the telescope from it's ideal
position, and this change in path length will cause a change in the phase of
the signal arriving at the correlator. Within a ten-second integration, this
phase change can be assumed to be constant (for any reasonable zenith angle).
For a monochromatic signal, as long as the change in path length is well known,
it does not matter if it is more than a complete turn of phase. For
a continuum signal, however, there will be a slope in phase across the
passband of the correlator. For a 1GHz bandwidth this limits the change
in path length to be less than 
30mm - two orders of
magnitude larger than the expected effects of the track, and so irrelevant
here.
The remaining effect then is that the post-processing software attempts
to perform fine phase correction when the individual 10 second records
are added together. In this case, errors in the phase correction
due to an uncertainty 
in the path length will cause a loss of
amplitude of:
A = A
( 1 - exp [ - 
] )
where 
= 
A 10% loss of signal will then occur when 
= 
. With
the appropriate definition of axes (z vertical, x forwards), and 
the
zenith angle:
= z cos(
) - x sin(
)
with an expectation value averaged over zenith angle:
Operation down to 450
m would therefore imply that the rms residual
uncertainty in 
due to track irregularities, after correction,
should be no more than 20
m .
As for the pointing, we can limit this specification to an
rms over some interval of azimuth/time. The relevant timescale is
hundreds of seconds, - i.e. rather small angular distances. However, if
we use the criterion that the phase correction between a calibrator and the
source be constant an appropriate
distance is indeed around 30
.
As for the pointing, if the track irregularities had a random
effect on the path length, we could simply characterise them by their
rms value. Since however we know that the errors are significantly affected
by the track joints, we add the criterion that no-where should
the peak-peak variations after correction exceed 100
m .
We have not as yet attempted to correct phase errors on the basis of measurements of the azimuth track, and models of the antenna A-frame. We assume that corrections similar to those made for the pointing will be possible, although these may be based e.g. on radial-arm inclinometry and an FEA model of the antenna.