From: Greg Aldering (aldering@panisse.lbl.gov)
Date: Sat Nov 02 2002 - 00:25:07 PST
Hi Isobel,
> We can certainly specify different conditions for brighter and fainter
> targets though.
Ok, let's do that then.
>Re Nodd & Shuffle - we will observe the source in both nod positions so
>the overhead is not much larger than if you offset along the slit between
>exposures in the conventional way. The difference is that we will nod much
>more frequently and will not read out the detector in between nids - the
>more frequent nodding is what adds the 25% overhead.
Am I right that with nod & shuffle one does pixel-by-pixel subtraction
of sky observed at the SN location for each nod position? Subtracting
"sky" from "object+sky" on a pixel-by-pixel basis gives a sqrt(2) hit
in S/N (for "object" << "sky"), unless "sky" is obtained by including
sky observations from several additional images.
What was tried when chopping along the slit with LRIS was to make a sky
residual image by medianing the sky-subtracted 2D spectra taken over
the span of about an hour surrounding the time of a given spectrum, and
then subtracting a version of that residual image which was smoothed on
a scale such that most fringe residuals survived. Unfortunately, due to
flexure, one could not median coadd observations covering a window of
more than about an hour. If GMOS has sufficiently small flexure, then
one could use this exact same approach - maybe even with a longer
averaging window in time - in order to decrease the S/N hit. To do
this, the SN has to be placed at different locations along the slit
even across targets so that sky is almost always being observed at the
location where a SN had once been placed during the night.
An approach to avoid the sqrt(2) hit for part of the spectrum is to do
pixel-by-pixel sky subtraction only where there are strong sky lines
and/or strong fringing, and do interpolated sky subtraction where the
fringing and OH lines aren't bad.
Since the nod & shuffle data can be reduced in a classical way - doing
interpolated sky subtraction - if desired, it certainly makes sense to
take the data in this mode for those faint candidates which we expect
to have their main features redshifted into the OH forest.
>For the faint ones we'll specify better IQ (~0.6?) with nod & shuffle
>and for the not-so-faint ones 0.8" with no nod & shuffle.
This sounds good: <0.6" for faint targets and <0.8" for brighter targets.
Now we have to define "faint" and "not-so-faint"! I would guess that the
transition is somewhere around i' ~ 24. What do you think?
>Then we can mention that we'd like to be observing on 8th if possible.
>(do we want to hold back some time for that night? It is possible that
>we could have used all our 10.3 hrs before that if things go well.)
If we stick to using our time on nights with sufficiently good seeing -
using the faint and bright IQ limits specified above - then we needn't
hold back time for the 8th.
Cheers,
Greg
This archive was generated by hypermail 2.1.4 : Sat Nov 02 2002 - 00:25:24 PST