Subject: Figures of merit for survey facilities

From: Tony Tyson

Submitted: Sat, 26 Jul 2003 18:28:14 -0400

Message number: 157 (previous: 156, next: 158 up: Index)

Nick,

In response to your lsst-general posting 137, here are my thoughts on
wide field imaging figures of merit and the comparison of PanSTARRS with
the 8.4m LSST. We should be consistent in comparing different
facilities with respect to details like filters and detectors. Figures
of merit for these facilities, to be meaningful, should assume identical
detectors and filters for each facility. The question these figures of
merit are meant to address is: based on a survey facility’s optical
throughput, seeing, sky background, and efficiency (on-sky time) what is
the relative power of a facility to carry out various types of science
programs (the three figures of merit). I could have used very wide
filters, but that would limit the range of science produced. So that is
why I use standard filters and assume identical 5e noise silicon
detector arrays. I agree that we should add an astrometry figure of
merit.

An advantage of high throughput for a survey facility is the ability to
pursue a wider range of science in the same survey. In fact, the
throughput of our 8.4m (6.9m eff) LSST design, which enables it to go
relatively faint wide and fast simultaneously, is just enough to pursue
the variety of science goals discussed in our last SWG meeting. For
example, the need for multi band color data over the whole visible sky
on a range of timescales is a requirement for many programs.

Once we have figures of merit for the key science drivers (I chose the
three discussed in the decadal survey) then it is easier to see how to
develop an optimal program. As throughput goes down below some
critical value it becomes advantageous to pursue a more limited survey
focused on one or two optimized modes. PanSTARRS and SNAP do this well
and, as I mentioned in my memo, represent significant leaps forward.
A sequence of surveys on such facilities then becomes attractive, and
one can imagine a linear algebra of these basic FOMs as an optimization
tool.

Regarding the PanSTARRS (PS) parameters, I have read the internal
document “Efficiency Notes.”
http://pan-starrs.ifa.hawaii.edu/project/people/kaiser/efficiency_notes.pdf
Your estimate of 30% obscuration for the 1.8 m PS telescopes updates my
estimate of maximum effective aperture. The resulting 1.78 sq.m
effective aperture of each of the PS telescopes is smaller than I had
assumed. For simplicity I also assumed sky noise limited exposures in
all those FOMs. For sky noise limited exposures with all four PS
telescopes in operation, the effective aperture of PS is four times the
above number: 7.12 sq.m. Using your 7 sq.deg FOV then leads to a PS
throughput (for sky noise limited exposures) of 49.8 sq.m sq.deg
compared with LSST’s 262 sq.m sq.deg. This is the factor of 5.3 to
which you refer.

While we are studying the advantages of thick Si p-i-n diode (hybrid
CMOS) arrays, we remain very interested in OTCCDs as a candidate for the
LSST focal plane. However, I do not see how OTCCDs can reduce the size
of the 0.3 arcsec PS pixel. So I do not understand your factor of 2
decrease in PSF. Such a gain could occur part of the time for PS if the
seeing were near 1” FWHM so that it was well sampled. But I am assuming
each ground based facility will de-weight poor seeing images. Under the
right conditions and for well sampled PSFs (like Jacoby, Tonry etal’s
implementation on WIYN) OTCCDs can improve the seeing statistics, and
they can compensate for a variety of high frequency telescope tracking
errors. OTCCDs have the potential of improving the PSF or introducing
errors into several moments of the PSF, depending on feedback details,
and I am eager to see what PSF shear improvements are obtained in
tests. In any case, since I assume the same detector for all
facilities, for the purposes of FOMs OTCCDs or any other assumed
detector would give the same advantage to all facilities.

There is some confusion over the values of the timescale tau I used in
the “time window” FOM. For the broad set of time domain programs
(moving objects to variable stellar objects) I take tau as the shortest
sky limited integration time. Instead I could have used a metric related
to the time resolution for variables or some measure of trailing in
moving objects, but this metric is different for each program and
becomes mired in exposure sequence strategy. In a survey for bright
flashes, like RAPTOR, sky noise and read noise are not as important.
Pace and FOV are more important. But for faint flashes or faint moving
objects one should use an expression for the FOMs which explicitly
include the read noise and sky noise, whose ratio depends in the
exposure time and throughput. Cutting the throughput in half yields a
double hit: exposure times to sky noise limit must be longer, and the
area surveyed to a given flux limit drops due to the decrease in pace.
For a multi-telescope survey facility like PS, each of the telescopes
must reach sky noise limit in the integration time. For each PS
telescope+camera the throughput is 12.5 sq.m sq.deg. I agree with Table
3 on page 3 of the PS “Efficiency Notes” where the PS exposure time for
sky noise limit is listed as 100 sec in V. Comparing PS with LSST, the
ratio of shortest exposure time (for sky limit) should be the ratio of
single telescope+camera throughputs: LSST/PS = 21. However, I did not
take LSST’s V-band tau as 5 sec because LSST is required to reach the
sky noise limit in B band in about 10 sec since we want to keep
integration times equal. While it is a worry for shorter exposure
times, there is no significant efficiency hit due to dome settling for
LSST’s mode of taking multiple exposures offsetting in altitude before
moving in azimuth. Telescope settling time can be made small via higher
power drives. While we have not yet decided if we will take quick pairs
of exposures any fraction of the time, note that it is not necessary to
take close pairs of exposures for cosmic ray splits since every patch
will be revisited often (this is not possible for low throughput). So I
keep LSST tau = 10 sec as the timescale on which variations can be
detected. Rather than use 210 sec for PS tau (as would be required for
a fair comparison using the single camera throughput ratio of 21) I will
update the PS tau to 100 sec (as listed in your document) in the
attached FOMs.

Of course you are not planning to use a V filter with PS. By using a
much wider filter PS can address some programs by trading color data for
shorter sky limit exposures, particularly slowly moving objects (KBOs
and some NEOs). While some NEO’s trail significantly in 15 sec, the
bigger effect will be in the pace: the rate of sky coverage is slower
for long exposures. So all-sky coverage with multiple exposures per
lunation becomes difficult. If PS focuses on a limited area near the
ecliptic this will be less of an issue. LSST will cover the visible sky
multiple times per lunation with standard filters.

LSST and PS are thus quite different facilities. In fact they are
complementary in other ways. They will undertake different science
programs largely, LSST will release all data live to the community, and
the funding sources of these two facilities do not overlap. I think both
projects should be done. In my judgment the single camera and larger
throughput of our current 8.4m design pose less engineering risk and
offers a range of unprecedented science possibilities. You have decided
to proceed with a four 1.8m (1.5m effective) telescope facility (PS). We
have decided to pursue engineering on the 8.4m (6.9m effective) LSST
design and see if we run into any problems. We have considered the
option of splitting LSST in half, building two telescopes and two
cameras, but the increased complexity and cost, and – significantly –
the loss of the capability to cover the entire visible sky multiple
times per lunation in standard color filters ruled out a sufficiently
wide range of science that it was rejected. Perhaps the SWG would like
to revisit this. Meanwhile, LSST Corp is proceeding with this phase of
engineering. So far it looks good.

At our last SWG meeting the need for FOMs was discussed. I was also
asked by the LSST Corp chair and two prospective funding sources to
estimate these FOMs for the three proposed facilities. These FOMs
indicate the relative areas of survey advantage for these facilities.
For the purposes of the SWG, a different set of FOMs are also relevant.
Each facility must choose to optimize parameters to address their own
suite of chosen science opportunities, and I think this is what you are
getting at. PS will take exposures as short as 30 sec by going to a
wide filter and thus focus on certain planetary problems (KBOs and
PHAs). So a set of FOMs like the “level playing field” ones, but with
these choices made, for each science program, would be useful. For
example, they would be a tool for addressing the question of what
fraction of that science program will be done by which facility and by
when. Getting those estimates will require simulations like those now
being carried out viz optical bursters of various time profiles and
frequencies, 90% of PHAs to 200m, weak lensing cosmology via all sky
cosmic shear and cluster counting vs redshift, all sky astrometric
standards of high density and related science, etc. One thing is clear
from the present relative FOMs: the 8.4 m LSST enables a unique range of
science using the same survey data.

Tony

LSST LSST LSST LSST LSST Mailing List Server LSST LSST LSST LSST LSST LSST
LSST
LSST This is message 157 in the lsst-general archive, URL
LSST http://www.astro.princeton.edu/~dss/LSST/lsst-general/msg.157.html
LSST http://www.astro.princeton.edu/cgi-bin/LSSTmailinglists.pl/show_subscription?list=lsst-general
LSST The index is at http://www.astro.princeton.edu/~dss/LSST/lsst-general/INDEX.html
LSST To join/leave the list, send mail to lsst-request@astro.princeton.edu
LSST To post a message, mail it to lsst-general@astro.princeton.edu
LSST
LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST