pointing measurements at both 3mm and 1mm using cross and xpnt were made
in excellent weather between midnight and 3am on 11feb99 using 3c273. 
the main results are:

- pointing errors on all antennas were small (<0.2' peak) at both freqs,
	with most antennas showing errors < 0.1'.
- xpnt and cross give the same result within 0.1 arcmin at 86 ghz, and
	each routine takes about 3 minutes to run.
- cross found a fit for only a few scans at 86 ghz and none at 220 ghz,
	while xpnt converged to a good solution every time at both freqs.
	the problem with cross is due to the fitting algorithm or a bug;
	the data are good and can be easily fit by eye.
- pointing offsets between 3mm and 1mm were at most 0.1 arcmin according
	to xpnt, using the 3mm/1mm horn offsets derived earlier from total
	power sun pointing (spoint).
- the following lists the original and modified horn offsets from the xpnt
	data at 3 and 1mm.  differences from the nominal values of 7.1 and
	1.8 of the order 0.1 arcmin appear to be real, but are close to the
	uncertainty in the measurements.  since these data are of similar or
	better quality than the spoint data, the horns.ascii file has been
	modified to match the new values.

ant	  original	difference	   new
	daz	del	daz	del	daz	del
1	7.1	2.0	.1	0	7.2	2.0
2	7.0	1.9	0	-.1	7.0	1.8
3	
4	7.0	1.9	.1	0	7.0	2.0
5	7.1	1.9	0	-.1	7.1	1.8
6	7.0	1.8	0	0	7.0	1.8
7	7.0	1.8	0	0	7.1	1.8
8	7.1	1.9	0	0	7.1	1.9
9	7.1	1.8	0	0	7.1	1.8
10	7.0	2.0	0	-.2	7.0	1.8

in summary, nighttime pointing is as expected at both 3mm and 1mm, with
peak errors <0.2' for all antennas at both freqs.  the horn offsets
between 3 and 1mm are mostly within 0.1' of the geometrical values, with
the largest differences occurring for ants 1 and 4 (0.2').  cross runs
reliably and produces accurate data, but it's fitting algorithm needs
work.  xpnt for once managed to converge reliably at 1mm, but is
unlikely to in typical weather conditions and on sources other than
3c273.  however, since it works reasonably well at 3mm there is little
incentive to make it work at 1mm at our present sensitivity level. 
problems with xpnt which definitely need to be addressed include:
 - omitting shadowed or non-pointing (collision) antennas
 - ability to drop a listed antenna (when OFF'd) more gracefully
 - avoidance of the fitting algorithm's occasionally asking for user response
 - improvements in speed and ability to converge in marginal conditions would
	be nice, if they do not compromise accuracy or result in false
	convergence.

improvements in the on-line pointing model are also needed.  the main
problems are (1) the inability of the model to account for large
elevation errors near the zenith for ants 2 and 9, which both have
substantial collimation terms; (2) the presence of errors up to 0.5' due
to azimuth bearing irregularities; (3) the large day/night differences
noted recently, which can produce errors up to 0.7' on sunny days near
sunrise and sunset and a general daytime bias of 0.2-0.4' for most
antennas. 

jrf - 11feb99