March 3-4/97, LNA San Paulo
Baldwin, Blanco, Cecil (chair & minutes), Green, Kuhn, Smith, 35 Brazilian astronomers
[To maximize the utility of these minutes, I have included in [] SWG activities that occured ``off-line" from the meeting but that are germane to issues raised there. Hopefully this device will satisfy ``the lawyers". I taped most of the discussion.]
Day 1:
Words of welcome from S. Viegas, stressing necessity of SOAR, ditto from R. Green for NOAO, De Medieris (President of Brazilian Astronomical Society) restated firm support of SOAR.
Cecil's presentation on the telescope included Moretto's latest off-axis optical designs which now depower M3 at f/10 to allow multiple instruments there. [He has also now moved the field corrector before the fold mirror, so all f/10 ports can attain 15-arcmin field.] Still 2xM2! Several Brazilian astronomers argued that tip/tilt was insufficient for a 4m that will enter operation in 5 years. Baldwin asked Cecil to include wavelength dependence in the system error budget (e.g. remove ``corrector" term for f/16 port.) Smith asked Cecil to include low-order AO (a la CFHT) on the error budget. STATUS: latter now done, revised plot of expected performance is:
[Note that partial AO will alleviate certain ``telescope ills", leading to smaller image degradation. I have attempted to account for some of these in the plot above. I reviewed Barr's error plot and set several more sources of degradation to ``epsilon" = 0."05 FWHM, his estimate for any type of ``measurement error". All this is unofficial of course, but the net effect of these preliminaries is to drop degradation from 0."18 to 0."13 Strehl width. CFHT seems to do slightly better, so I guess we too will approach the bottom blue line. This will be worked on over the next few weeks.]
Kuhn presented the goals for MSU astronomers. They have 100+ nights/yr on telescopes already, also buy time on Wyoming IR telescope. They want a novel observing capability, viz. one that emphasizes control of scattered light. Ideal configuration would be an off-axis. [If the project manager comes back and says s/he can't achieve the science program with identified funds, then 1) reduce the requirements that depend on an on-axis design, 2) reduce the requirements that depend on an on-axis equatorial (to avoid time-dependent spider rotation), 3) eliminate the requirement of AO shortly after first light.]
Smith presented the CTIO position: Blanco & SOAR should be viewed as complementary telescopes with time-trades to access long-term instruments on both. The main emphases are: tight core PSF, Q-scheduled, rugged platform for state-of-art (including Gemini) instruments. He asked other SOAR partners to discuss whether we should sacrifice part of the instrument budget for a first-light low-order AO system.
Cecil outlined UNC goals:
Dottori asked if spectropolarimetry & coronagraphy could be done on the on-axis? Cecil: yes, a Cass. focus can be incorporated for a light (couple hundred kg) instrument.
Don't want to do off-axis if: too expensive, too risky, no science benefits. Arguments in favor of off-axis:
Kuhn then summarized the contributors to the PSF: Atmospheric anisoplanism, light scatter from off-field, instrument scatter (he felt from the limited data he had worked with that e.g. the WIYN ADC doesn't scatter light), atmosphere aerosols. Wavelength dependence: longward of 1 micron, diffracted energy from the spiders and aperture are comparable, mirror scatter is not important. 0.45 micron: aperture diffraction dominates at small radii, mirror scatter dominates >3" when no spiders; eliminating these effects (the latter by a coronagraphic mask) will decrease the scattered light between 1-100" by 10x.
Spiders are always important: they take energy out of PSF core.
Mirror roughness scatter dominates in blue (several % at short wavelengths),
2cm-thick spider [2x too thick!] gives <0.5%. But it's the angular dependence
of spider diffraction that is important (dust is distributed uniformly,
spider light isn't.) Net effect of dust is to remove light from core & distribute
it uniformly in angle. Dust doesn't dominate scattered light until you
get to 1000" radius. Still important to SOAR because it will
dominate emissivity, and reduces throughput. Kuhn showed the spatial influence
of the rotating spider during a 1-hr exposure. The spiders rotate through
at least 10% of the 90-deg azimuth angle between diffraction spikes over
most of the sky. His plot below gives the angular fraction of 90-deg that
is covered during this exposure:
Bottom line is that our goal for scattered light should be to reduce the total scattered power by a factor of 100x from the HST mirror. (This # comes from the measured surface roughness of just M1, not including WFPC2 etc.)
Did not have time in SWG study to optimize 4 telescopes. So, made all mounts as similar as possible (``versatile mount"). Studied most obvious differences between designs. Optimize selected (by cost & science benefit) design.
WIYN is the on-axis baseline because: performance at alt.az. is good, and cost/performance can be convincingly scaled to 4m on-axis alt.az. Projection is reasonable to off-axis alt.az., less convincing projection possible for equatorial. E.g. L&F study confirmed that bearing loads and swing clearance are v/ different for the equatorial. Off-axis equatorial was considered by L&F ``the worst possible choice" for a mount. [Indeed, COAST Steel advocates a horseshoe equatorial mount for maximum stiffness. Blanco feels that supporting clusters of heavy instruments is the real challenge to the equatorial.]
WIYN features that are applicable to SOAR (regardless of SOAR's configuration): AO cannot correct for mirror & dome seeing close to telescope (small eddy currents). WIYN M1 is actively by blowing T-controlled on it (hold to 0.5 C during evening, better usually), actively ventilate telescope (air drawn from all heat sources into base cone; heated air is exhausted from the bldg.) Active collimation of M2 (as it tracks through the gravity vector).
WIYN has 3 instrument ports, 4-reflection Cass. (using a fold mirror
below M1 for polarimetry), M3 articulated for quick change. The most ``hair
raising" part of the design is folding M3 out of the beam this structure
hangs over M1!) Green: does M3 moving out of beam introduce hysteresis
in the registration of M3? Yes, this limits our ability to generate
lookup tables and gives WIYN a slight performance hit. This is a 1.2m
flat for WIYN, but will only be 20cm for M3 on SOAR.
Blanco kept his finite element models simple (unrealistically optimistic), ran lots of models to try different ideas, optimized for high resonant frequencies (smaller gravity deflection & less wind shake). The most plausible equatorial (``the world's largest fork mounted optical telescope" [but not by much, the ESO 3.6m is on a fork]) would still have a 20 deg N-horizon blind spot. Natural frequencies fall as you go from on- to off-axis but off-axis truss may achieve comparable performance to on-axis (in resonant frequency) with further optimization. For 90% of the operating conditions, there is no difference in the performance of the 3 different telescope structures considered. L&F did FE analysis of the mount, including the weight of a Gemini instrument at Nasmyth focus.
Blanco also discussed WIYN-based enclosures, which are light and well
insulated. Air temperatures in and out are matched within 20 min. Dome
is covered with a poor radiating tape to prevent it from supercooling below
nighttime ambient T. Equatorial fits into a slightly smaller dome than
alt. az. L&F estimates $350K to go from on-axis alt.az. ($2.75M)
-> off-axis ($3.1M). On-axis equatorial will cost $3.5M. Breakdowns
for structure will be made by L&F to identify possible Brazilian subcontracts.
Showed what studies had been commissioned. Reviewed the following in detail:
HDOS & REOSC can both superpolish the mirrors. Additional cost is confused at present. [HDOS seems to be saying that on-axis w/o superpolish is $3M, superpolish is 10% extra, f/1 parent is 10% extra, off-axis is 10% extra, 5% to combine everything, i.e. $1M extra to polish off-axis or $0.7M if we agree that superpolish would also be done on the on-axis optics. This will be clarified in a few weeks. The HDOS on-axis price is currently higher than REOSC, who can also do superpolish and off-axis. Here is what we know about prices to date, BUT THESE ARE NOT FINAL FIGURES!]
Baldwin showed from John Filhaber's (CTIO) work that the on-axis when tilted/decentered introduces coma, which is easily diagnosed by the image analyzer. However, the off-axis produces complicated spots which include astigmatism. M1 also produces astigmatism, so there is ambiguity about which optic is out of spec that can be resolved by measuring spots across the field. The off-axis is 2.5x more sensitive to alignment errors than the on-axis. Still within range of active collimation. Bottom line: it will be harder but doable to keep an off-axis telescope in alignment.
Baldwin outlined: the motivation from Racine's study at CFHT for minimizing mirror seeing (stay within 0.5 C), Barr's thermal analysis of how to control the temperature of a 10cm-thick mirror, and Pearson's position-actuator solution. There are several possible ways to support a 10cm-thick mirror, arguably the least plausible being to scale the Gemini force-actuated solution for SOAR (it would require assembly and service by ``vertically challenged" individuals.)
Pearson's solution would ideally use an image analyzer outside the science
field to constantly monitor the M1 figure. Force sensors and actuators
would come from Keck. If no star is available, would use strain gauges
to make open-loop adjustments until you could get to a star later in the
night, as things are done now. Thermal control would occur by blowing air
through the cell. Pearson's approach would work for either on- or off-axis,
alt. az. or equatorial mount. Pearson has now identified hardware, and
has produced this Russell Porteresque sketch of his athermal cell:
[A few comments about this design (by Pearson to Cecil, after the SWG meeting): the athermal cell introduces <10 microns of distortion into the mirror surface due to T & gravity changes. Various tripod structures link the different levels together. Proceeding from mirror surface to bottom of the cell (and with reference to the figure above), we find: a small tripod that connects the mirror to the top of the displacement-style actuator. This is a tripod to produce a design that works for an equatorial mount. If SOAR chooses an alt.az. then the tripod is replaced with a bipod, a somewhat stiffer structure. This is the only difference (at this early stage in the design) between the two different mounts. The extra linkage in 2d rather than 1 produces a slightly larger uncertainty in the rod position, which translates into a small performance hit on our knowledge of the mirror surface. The tripod or bipod acts both as a spreader, to reduce the amplitude of bumps on the mirror surface, and to introduce the counteracting moment required because we are not attaching to the center of mass of the mirror. Then there is the steel cuniform, which attaches the steel plate (hence cell) to the rest of the telescope at 4 points. Then each attachment point to the steel plate is surrounded by 3 points that are are apices of the tripods connecting the steel to aluminum plates. Triangles on triangles.]
As mentioned above, this is one example of a support system for a 4"-thick blank. Together they prove existence. At present the SWG has passed the buck to the soon-to-be-hired project manager. This person will presumably then turn around and ask the scientists to choose, very soon after taking control! Acceptable image degradation from mirror seeing has been specified as a science driver (essentially the Gemini spec.)
Baldwin then summarized the properties of the proposed site on Pachon, and plans for verifying the height of the boundary layer by using wind speed monitors on a tower at there. 5 wind gauges at 5 heights show that wind direction at 30m is well correlated w/ flow at 20m and 10m heights but not at 6m. So boundary layer at Pachon is somewhere between 6-10m. CTIO's tower at SOAR's site will do the same experiment. After several months we should know how high to set M1. Have no equipment or personnel to do differential-image motion monitoring.
Cecil noted that ``lofting" by the dome would raise the effective height of the layer. Pearson has suggested that the dome be on stilts [like his mirror supports!], to minimize the obstruction of the airflow. The Coast Steel clamshell dome has somewhat this character, as does the WIYN-derived Magellan dome (although this has more structure in the way).
SWG selected a group of problems that would push the capabilities of the telescope, to see if we can distinguish between the results of imaging with ideal (= no instrument degradation) on- and off-axis configurations. Kuhn & Baldwin presented the results, on behalf of groups at MSU and CTIO. (The data are available at the Web site for others to work on.)
Both telescope configs. were assumed to have optical polish that decreases the total scattered power by 1) the same as HST or 2) 1% of HST (may be achieved by ensuring that edge polish is ok, or may require a global improvement. TBD w/ HDOS study.) This is reasonable because the SWG can certainly say at this stage that we will specify a smooth polish on the optics, regardless of which configuration we settle on in April.
Kuhn simulated 4 configurations
on several different problems. At the minimum, an off-axis telescope will increase Strehl. It never damages the image.
Jupiter detection 5 AU from a 0-mag * at 6 pc distance. Jupiter would have a surface brightness of 15-mag/arcsec^2, well above sky background. But starlight at same distance is 5 mag/arcsec^2. [Ratio is similar for a less luminous star.] We clearly need to minimize scattered light!
First, a coronagraph will remove most of the aperture diffraction, allowing us to reduce the difference in surface brightness between planet and star to 5.5 mag/arcsec^2, nearly a 1% signal. An obstructed 8m will need much more of the pupil masked to attain comparable performance, reducing its efficiency. The larger 8m aperture doesn't help as much as you'd think: 8m obs flux/4m unobs. flux = 3.2 not 4 because you have to over-occult the 8m pupil to block ringing from the spiders & M2. In contrast 4m unobs/4m obs = 1.25.
Second, detection is better on an obscured or unobscured 4m SOAR than on 8m at wavelengths longer than 1 micron because D/r0 is optimized for SOAR as long as both telescopes are operating at the same level of AO. [BUT: if light contrast between a true Jupiter and the star is optimized by going into the thermal-IR, the 8m wins because SOAR but not the 8m will be diffraction-limited at e.g. 2 microns. Moreover at whatever wavelength, if the 8m has higher-order AO than SOAR, it wins. So this is a transient advantage for SOAR in the IR, somewhat less transient in the visible. Note that this advantage does not depend on SOAR being unobscured.]
Kuhn's conclusion: assuming coronagraphs on both telescopes, an unobscured 4m must detect delta-6.5m signal (0.25%) vs. obscured 8m must detect a delta-8.5m signal (0.04%). So, the relative efficiency obstructed-8m/unobstructed-4m = 3.2x0.04/(1.25x0.25) = 0.41, i.e. an unobstructed SOAR is >2x more efficient than an obstructed Gemini. The advantage reduces to slightly better than Gemini for an obstructed SOAR.
Kuhn then compared gaussian galaxies + foreground *s, does one see diffuse light that really is just scattered light? Yes, at the level of 1/1000 of the sky.
Baldwin: compared DAOPHOT analysis of a very crowded cluster field ranging over 8-16th mag. for on- and off-axis telescopes w/ 0."8 and 0."4 FWHM images using 0."1 pixels. Analyzed by 2 groups of skilled photometrists. Spiders and higher Strehl for the offaxis are the differences. However, there is no difference in the measured photometry compared to the input. What counts is not the particular telescope you have but rather how tight you can make the central core (errors blow up from 0.5-mag to 5 mag as you degrade from 0."4 and 0."8). Strehl improvement of off-axis is not enough to make the internal errors smaller. No significant difference in detection rate of *s. (Baldwin & Kuhn were both disconcerted to note that the DAOPHOT errors were reported to be much smaller than the true errors!)
For faint objects in an ``empty" field, made 2048^2 image w/ 0."1 pixels, 5000 1" diameter circ. galaxies, using galaxy luminosity function of Koo & Kron, 500 *s down to 20th mag distributed in intensity range 1e3-1e7 counts, dark sky brightness for CTIO. Making a frame of this size was a mistake: it took Kuhn forever to generate the image, and forever more for Baldwin to dribble the image down to CTIO. As before see Strehl differences and spiders as the only differences between 2 frames. How much do spiders mess up the ability to search for faint objects in ``blank" areas? Spiders appear when you look at 1% of sky background. In 1" seeing real data from 4m show spikes on 15.3 and 16 mag *s, not @ 18. So stay away from these if you want to find faint objects. In simulation see spiders around 15 & 16 mag *s, so the simulations appear to have the spiders correctly normalized. Also, a large fraction of the area is available between diffraction spikes at a level of 1% of the night sky: viz. 5-deg away from diffraction spikes, at 1" you are down by a factor of 100, at 10" you are down by a factor of 1000. 45-deg away it is down even more.
Haven't yet done point object on extended object.
Baldwin noted that the effects of spiders should be worse when you go into the IR. But in fact you don't see spider spikes in the IR because modern IR instruments have pupil masks which mask them out (on equatorial telescopes! Not alt. az.) Can we put a rotating pupil mask at the Lyot spot into at least a new IR instrument? (No room in COB for this.) Kuhn pointed out that such a mask would lead to significant throughput losses because of ringing. [After the SWG meeting, Green & Cecil talked to Fred Gillette (Gemini) about rotating masks: Gemini has no optimized coronagraph yet. Spider vanes are 1cm thick steel; their emissivity is only 10% that of the mirror coating. Suggested look into composite spiders, might be able to cut thickness to 0.5cm.]
Cecil covered scheduling, emphasizing the need soon for $ to form the project team. Also discussed need for integrated M1+cell assembly (like WIYN) and factory test (possibly in Brazil) of the steel structure. Blanco felt it was unrealistic to checkout the TCS in a factory vs. an optical shop. Telescope is also not mounted on a stable pier.
Day 1 ended with Cecil noting that we had experienced an abnormally coherent SWG meeting because mirror supports had NOT been discussed! We were drained nonetheless, so it was time for food & drink.
Day 2:
SWG was presented with ``sneak previews" of the soon-to-be-delivered Brazilian science programs:
Steiner (synthesized Brazilian interests): Galaxies & Large-scale structure, AGNs & starbursts, star formation, stellar populations & chemical evolution, stellar pulsations, binary systems, ISM, planets (solar system dynamics). Need to improve maturity of Brazil. community, SOAR must support graduate programs which means a broad community now and even broader in the future. Should not specialize SOAR too much, but Brazil recognizes the interests of other partners to build a unique telescope.
Basically Brazil wants: image quality, flexibility + time resolution variability, suggests seasonal coatings to optimize UV (impacts corrector design), most extragalactic want to work in near-IR, stellar needs UV. Need AO asap, if possible at first light. Propose to consider it as instrumentation. IR spectroscopy is a major new interest in the Brazilian community.
Smith stressed that it is unclear given manpower how one can run queue scheduling and rapid changes on SOAR. The current operating model he is exploring with the SOAR Operations Working Group does NOT have queue scheduling on the Blanco 4m. This has implications for time trades. He noted that a lot of UV interest of Brazil is monitoring which probably could not be switched over to the Blanco. Operations budget for Gemini is $5M/yr for each, CFHT is currently $7M/yr, SOAR will be <$1M/yr. He is committed to developing a 20 yr operating scenario, but ``the partnership won't like the implications."
Cecil responded to Brazilian science requirements. He asked the group when they wrote up their research proposals to consider what they would do when not in top quartile seeing. The difference in FWHM between bottom and top quartile conditions is not large in the near-IR but is in the optical. The main benefits for AO are going to be at <1 micron, although Strehl will be increased in the near-IR, important for detections. Also noted in the project schedule that things will move rapidly once the project manager is hired in a month or two, i.e. we'll need real money then!
Brazil, CTIO, and MSU want low-order AO near first light. UNC is interested as well, but not at the expense of restricting our choice of instruments. There was discussion of the possibility of removing up to $2M from the $5M instrument line, to be identified as funds available to build a low-order, several arc-minute field, near-IR/red AO imager. The AO unit might also provide a deaberrated beam for an integral field spectrograph.
$1M could also be transfered from instruments to telescope contingency, leaving $2M available to build one instrument (an optical spectrograph being the most likely candidate.) If NOAO contributes COB (Cryogenic Optical Bench) and a CCD mosaic at no cost to SOAR, we have the imagers covered (although COB would be shared w/ Gemini during at least their commissioning.)
A clone of the Gemini-N IRS (GNIRS) could be provided by NOAO at cost of increased NOAO observing share (TBD, perhaps equivalent to a $1M capital contribution) if no other source of funds materializes from the partners to pay for an IR instrument. While this 8m instrument would most certainly not be optimized for the 4m SOAR, it would provide IR spectroscopy when not in use on Gemini-S (approx. half the time), with a choice of 0.1" and 0.3"/pixel and full coverage with multiple dispersions from 1 to 5 microns. It is the scientific choice of CTIO for the SOAR spectrograph. Other SOAR partners might prefer a dedicated IR spectrograph, optimized for 4m and their science interests.
Rev 02 of the Science & Technical Requirements document is now available for perusal at the Web site. It includes comments on AO for SOAR. Blanco is integrating the CFHT PUEO AO-bench design into the on- and off-axis configurations for SOAR. (This work is not reflected in rev02.)
Moretto will leave for BC in a week or so to work with Harvey Richardson on optics. One hopes that Richardson will finally generate a focal reducer/field corrector for the on-axis.
Cecil, Kuhn, and Daggert will travel to HDOS in late March to be briefed on their polish study for SOAR. Costs for the off-axis superpolish will also be ``scrubbed" at that meeting. HDOS will then produce a document outlining how they would test and qualify the optics. This document would be suitable to solicit bids from rival polishers (e.g. REOSC, Contraves). SWG will get a quote from at least REOSC using the HDOS specs.
Cecil, Daggert, & Blanco met with Bob Jones (Corning) on 3/11. We told him we expect to be in a position to decide on M1 thickness, radius of curvature, and on/off-axis in 2 months. Good, because there is an ARPA 4m emerging to compete with us for Corning resources ... a space-based laser project. Jones says this is good for us because Corning now need not charge SOAR the full cost of refurbishing the 4m furnace (updating it to reflect lessons learned from the 8m castings.) It has not been used for nearly 10 years.
Jones says that mirrors are now very low priority at Corning and that it is a continual challenge to keep people from being raided by the booming fused silica and fiber optics divisions. For this reason, even if our soon-to-be-hired project manager flips a coin and says 10cm, Corning won't actually start forming the blank until Sept.-Oct. Before this, they will need 3 months or so to hire their project manager. During this time, they want to remeasure CTE & revisit the boule layout. (This has nothing directly to do with the discolorations identified at the hex boundaries of the Gemini-S blank. We now know a bit about that.)
Jones will get to us in a week or two Corning's estimate of a realistic production schedule, assuming that the other 4m blank materializes (he believes that it was funded this year.) Daggert will fold this into the SOAR schedule, but it is clear at this stage that M1 is back on the critical path. Sigh.
The final meeting of the SWG will be held in Tucson April 11-13. This is 1 wk later than originally planned so that Horatio Dottori can attend from Brazil. Augusto Damineli will also come from JILA, appropriate because Brazil has been under-represented at several earlier SWG meetings. Friday will be an engineering closeout, then a weekend science review will culminate in a recommended telescope configuration that satisfies partner science goals. The SOAR Interim Board has been told, and has approved, that a draft of the SWG final report will be available for perusal by consortium members at the Web site shortly before this meeting, which will include the criteria to be used to make the recommendation. The SWG will work hard to ensure that the final report is delivered to the Interim Board by April 15th. The Interim Board chair (Wolff) has told Cecil not to cost cap our science goals, but that it is appropriate to include costs when we order the rank-order these goals.