Date: Fri, 14 May 1999 11:11:47 -0700
From: Russell E Owen
Subject: comment on guiding
I helped out some with guiding last night, but the decision was made
to switch to imaging before we had it fully debugged.
We do know that the instrument data is still not fully correct. The
telescope can hold a position, but it cannot adjust itself to center
the stars up in the probes. Until it can do that, I don't think we do
a reasonable job with spectroscopy (though one can get close by
manually centering things up).
We realized yesterday that the guide probe orientation was being
expressed with respect to the instrument, not the rotator (a 90
degree error), which would prohibit any guiding from working.
However, we did not realize until about 11:00 last night that the
guide probe positions on the plate were ALSO being expressed with
respect to the instrument, not the rotator. This would not affect
guiding in az and alt, but would mess up guiding the rotation and
possibly scale.
This known problem does not, as far as I can see, explain why Dan had
trouble using TRACK/PTERR to center a star on a probe. That is quite
disturbing. It suggests that orientations are still not known
adequately or that there is some subtle bug in the TCC which doesn't
affect the engineering camera or the 3.5m off-axis guider, but does
affect guide probes.
This ongoing confusion between rotator and instrument coordinates
should, I think, be more directly addressed. Any of the following
proposals would work fine:
1) Side-step the problem by claiming that the instrument frame IS the
rotator frame. To do this set rot_inst_ang = 0 in the instrument data
file, then rotate +/- 90 to use the plate.
Advantages: the rotator and instrument frames become identical,
completely eliminating the confusion that has plagued us. Translating
from plate x,y to probe positions that the TCC needs becomes quite
simple. The original instrument data files will work just fine.
Disadvantages: this has the effect of changing the instrument
coordinate system convention. Practically speaking I'm not sure it
matters -- the same code used in the same way can still be used to
drill plates. But it sounds as if it could cause confusion somewhere.
It also violates the TCC convention that instrument y = north when
you rotate 0 object.
2) Finish confronting the problem. Rotate the guide probe positions
from the instrument frame to the rotator frame. Do this for all
cartridges all the time. It's a 90 degree rotation -- darn near
trivial -- but could be an ongoing source of confusion.
On the whole, I feel pretty comfortable with either solution. We
don't have to even pick one right now. Scott has proposed trying both
next time we guide.
I will try to check in tonight and tomorrow, but cannot offer any
extended help until next week.
Suggested tests:
- Try to figure out why pointing error corrections do not work. I'll
see what I can learn about that today, when I'm not helping with
collimation. One question: have they ever worked, e.g. with the
incorrection probe orientation data? I see this is a pretty important
problem to solve, but it may have to wait until I can work nights
(i.e. next week).
- Fix the guide probe positions and see if multi-object guiding is
capable of centering up (guide on/nref=n). I'd be a bit surprised if
this works unless pointing error corrections work, but although they
use similar math, they aren't entirely the same code.
-- Russell