Date: Tue, 20 Apr 2004 21:07:09 -0400 (EDT)
From: David Schlegel
Subject: Re: Spectro focus once again [was RE: 2.5m obslog 20040419 (MJD
Pete,
As per spectro collimation procedures: I would suggest we
go with the procedures as suggested some time ago, e.g.
  sdss-obs/3743
That basically says that we try to start each plate "in the
black" for XSIGMA/WSIGMA, and then there should never be any
loss of data.
The lost time that we've had has always been when plates are started
with the spectrographs just "on the edge" of being inside specs.
My guess as to the numbers are something like:
  |COLLIMATE offset|  0.7 pix : In the red according to SOS
(The above numbers could be measured, but I don't think ever have;
the guesses above are made from just taking some quadrature differences.)
Thus, if you collimate and put the shifts all within 0.15 pix
(esp. at zenith), then it's very much in collimation.  Moving
the telescope around is at worst a 0.2 pix shift peak-to-peak
(cf.  sdss-obs/3618 and refs),
which isn't a big deal.  Temperature shifts are probably the
biggest culprit for changing things.
If we knew the shifts w.r.t. temperature, one could also pro-actively
change collimation at the beginning of each plate.  But even if that
we done, we should *still* have the procedures as stated above.
Didn't someone look at the temp shifts at some point -- I can't
seem to dig that up.
-David