Attendees: Augusto, Blanco, Cecil (chair & minutes), Coulter,
Daggert, Kuhn, Moretto, S. Wolff. Presentations by: Goble & Pearson
(mirror supports), R. Wolff (Gemini TCS). The day before, Augusto, Baldwin,
Cecil, and Kuhn met to discuss science issues. This 2-day format worked
well. Notes updated to 2/5/97
Blanco, Cecil, Richard Green, and Kuhn will visit LNA in Sao Paulo March 3-5. Sessions are being organized by Horatio Dottori for us to present the issues under discussion, and seek Brazilian input on the final Science & Technical Requirements document for SOAR. Kuhn will also visit CTIO after this meeting, to talk to staff scientists. The S/TWG will submit our recommended configuration and trade study by mid-March. A draft version will be available for comment at the Web site just before the Brazil trip.
Blanco is optimizing the off-axis baseline (2 M2, parabolic/hyperbolic optics, f/10 & f/16 fold M3) configuration piece by piece. At the moment he has built variants of the OSS tower that supports M2, then used a feature of AutoCAD R13 to export the structures to NASTRAN for FEA. M2 is modeled as 3 lumped masses (750 lbs total). The whole structure is made from 3/8" steel plate, 7-8000 lbs.
He finds a 1mm static deflection at the top end. Various configurations
were examined, the prefered one (from the point of view of resonances) is
which has 2 main resonances near 20 Hz, the rest at >= 34 Hz. (These
#s will reduce once the top end is bolted to the mirror box-frame, but it is
already LESS stiff than the on-axis top-end, which has resonances that start
near 30 Hz.) A few other concepts, with
less favorable resonance structure:
The previous baseline optics were Gregorian. We now have parabolic/hyperbolic optics with a prime focus (discussed below.) Accessing prime makes the total assembly about the same length as before.
Blanco reported that L&F are comparing on- and off-axis designs, on alt-az and equatorial mounts. They are providing ROM quotes and the rationale, based on Blanco's design concepts. So far they have done only on-axis. Just tipping a beefed-up WIYN fork gives you a 45 deg blind spot at the N horizon. Cutting out a bit of the fork without weakening it gets you to within 25 deg. L&F has extended the fork to get a 20 deg blind spot. ROM costs increase from $2.75M (alt.az) to $3.25M (equatorial, 25 deg. blind spot), to $3.5M (20 deg. blind spot). The 1st step is an interesting (if preliminary) number. The $2.75M is consistent with the August Chapel Hill number when engineering costs are included (as they appear to be.)
L&F find that roller bearings will do, i.e. hydrostatics aren't required. Dan presented the following PRO/CON chart on equatorials to generate comments (got a few):
Everyone was amused to see the following figure, which shows the on-axis equatorial tucked neatly into the WIYN dome. This is because the WIYN dome is sized (as it would need to be if SOAR goes alt.-az) to provide floorspace for an elevator shaft to the coat facility.
.
The pier height is set to get out of the boundary layer. Baldwin will find out what the estimates are for the SOAR sites at CT and CP. Gemini is placing their M1 20m above ground, but this seems to be set by known values for MKO, which has different flow characteristics compared to CP. De Young told Baldwin that the sharp (70-deg??) dropoff angle at the Gemini-S site causes the wind to flow up and over the enclosure. In contrast, the less steep (45-deg??) angle at the proposed SOAR site near Gemini lets the air hug the slope, so presumably we don't need to go as high. Baldwin will report at the next S/WTG meeting in Feb.
Roy Coulter showed his concept first aired at the last meeting. Things will be different now that we have abandoned a Gregorian, in that the mirrors need to be removed to get to prime focus. He felt that the M2 flip was now better handled outside the hexapod tip/tilt unit.
Daggert gave the shockingly high price Gemini paid for their unit. Clearly, major cost drivers for them were the large throw angles and accelerations required to satisfy their 2 and 3 point chop requirement. Lockheed will quote us a lobotomized unit that preserves t/t, and another quote that adds in rapid focus piston (because any Gemini instrument will be able to diagnose the need for this using their On-Instrument Wavefront Sensor.) Gemini continues to debate the need for this; their simulations show a 0.06" Strehl improvement with rapid focus.
Coulter felt that magnetostrictive actuators would be fine, because we only need to move a few arcsecs for near-IR t/t and optical rapid guiding. These dissipate much less power, and have non-restrictive cabling (<20 volts) so electronics could be removed from the unit. The Gemini unit has large power amplifiers in the unit to drive the large chopper throw, with cooling lines to extract heat from or adjacent to the telescope beam. So the bottom line is that the Lockheed unit is likely to be grossly overbuilt for what we need.
To explore more suitable alternatives, Baldwin will get the frequency response of the now-operating Blanco tip/tilt unit (best value 0.2" to date). Daggert will also get info on the Physik Instrumente units that have been used at UKIRT and the UH 2.2m telescopes at MKO.
It was agreed that SOAR would baffle the near-IR and optical fields in an optimal fashion, with a deployable baffle if necessary. Baffling for thermal IR will be done within instruments. A subdiameter M2 is undesirable in any event because you can't be sure what will be in the adjacent sky!
Pearson reviewed his charge. Barr had agreed that even the 8"-thick concept in the baseline needed to be revisited. Barr's thermal study (soon to be archived at the Web site) showed that ``mirror seeing" will be an issue for the 8" blank a non-trivial time fraction unless surface heating is employed with a high duty cycle. It had been agreed that Pearson would pick up Goble's work (because Goble's work on the CLEAR OSS had removed him from SOAR/CLEAR M1 support studies). Cecil gave Pearson fairly sketchy performance criteria before Xmas to set the limits of his study. It was agreed that Pearson would be able to demonstrate an existence for the 4"-thick support by the end of the study. Some variant of this could be used on an equatorial and on an 8"-thick blank.
His solution is an evolution of Goble's. It currently uses 96 rigid actuators to stiffly couple the mirror to its cell. This reduces the effects of strong wind gusts to levels below those discussed (in a sketchy fashion) in the 8" design. Schwesinger lateral supports are not used along the edge of the mirror to ensure that the surface can be completely exposed for maximal air flushing and to maintain low edge stresses. Hydraulics are spurned to ensure low maintenance. Pearson does not use spreaders on the actuators or an airbag support, to minimize linkage-induced position uncertainties and uncertainties in the support weights. The airbag is higher maintenance because it is hard to seal, being perforated by support struts and by a possible M1 central hole. He said the Gemini concept was developed over an endless series of trades meetings and was therefore completely specific to their needs.
Blanco feels that the support on 96 pts & air bag would be exactly as stiff as the 96 pt support w/o airbag, only the bumps go away. In other words the air bag has nominally zero stiffness, so it neither adds to nor subtracts from the 96 pt stiffness. The bumps would show up again if the wind blows hard enough on the mirror - but Blanco doesn't think that will happen often enough to be a big concern. He was concerned about coupling the vibration modes of the cell to the mirror through this rigid multipoint support with no damping (unlike a hydraulic support which provides excellent damping).
With the actuators rigidly connecting M1 to its cell, the CTE of the cell vs. none in ULE becomes an issue. Attachment rods want to expand radially. Pearson is producing an ``athermal" cell -- aluminum and steel -- that produces a low amplitude, controlled bend to M1 when T varies by several degrees. He is on track to a solution by mid-Feb.
However, a few days before the meeting, Barr's technical description
of the baseline report resurfaced in Cecil's office, and Pearson realized
that his supports were going to produce surface bumps 3x those of Barr.
(Barr's model had 0.035 waves rms of wavefront error, vs. the Rayleigh
criterion of 0.075 waves rms.) There was a suspicion that, while better
than any mirror in current use, our M1 might be deficient w/ the 8m Gemini
mirrors in certain aspects of its performance relevant to scattered light,
away from zenith where polished out bumps reappear as dips. (Gemini avoids
dips with an airbag. However, an airbag means the mirror is less stiff
against wind gusts, requiring additional mechanisms.) By 60-deg zenith
angle, the dips are already at 70% of their full (horizon-pointing) magnitude.
As the scattered light maven, Kuhn has checked the effect on the PSF:
``The solid line shows a pure aperture diffractive calculation (no mirror roughness) while the dotted line shows Earl's current mirror. (Much of structure comes from the aliasing caused by the relatively coarse sampling [100x100] of the mirror.) Earl's mirror has a diffracted ring at about 0."35 radius, but otherwise near-identical scattered light. The scattered light from the mirror surface falls as the square of the roughness amplitude. I think this mirror is okay, but given more time a good goal would be to decrease surface bumps by at least a factor of 2".
Pearson believes that the bulk of this artifact comes from the periodic nature of the support printthrough, and that these surface errors can be greatly damped in a straightforward fashion by ``randomizing" the support points at a later point in the design. CTIO will also verify this. This should get us to the Rayleigh criterion even at 45-deg (if anyone cares, since noone would see this.) It is however, a refinement that will not effect his charge which is to prove a viable support for a 4"-thick mirror. Pearson is now working w/ Goble's 5-hr/wk input. Barr will also be asked to sit in with Goble, Pearson, and Cecil to provide input.It is also worth noting that the effects we are discussing are worst-cases. Within 45-deg of the zenith, the effect is <70% of the numbers above, close to the factor of 2 that is Kuhn's goal.
Quite apart from this issue (which emerged from discussions at and after the S/TWG meeting, and which is reported here as of 1/30) there was still uncertainty about whether the required resolution of the displacement sensors is adequate. Kuhn noted that we had had THAT conversation 3x before! It was decided to simulate the effect of 1-step errors on encoder readouts to see if we could live w/ this. [Pearson's subsquent simulations showed that a 1:2000 error in force sensing is completely swamped by wind gusts. i.e. if wind flow were steady, we could cruise on lookup tables forever. The dominant term that arises is astigmatism, with effects on the image that are smaller than the Airy disk at 0.5 micron wavelength. Therefore the dynamic range of these sensors is not the issue.]
[Pearson is also examining the effect of ``puck placement errors", when they are glued onto the M1 back. These would arise if M1 is not polished on its final surface. The obvious remedy is to require that the M1 polish support use the same pucks. This is usually always reflected in polish quotes, and costs extra.]
The central issue is in fact that the force sensors (load cells) can't tell whether the rods they are attached to are being compressed because wind is pushing on the front of the mirror or because the mirror cell is reacting on the back. An external wavefront measurement breaks this dichotomy. The final discussion therefore revolved around the desire of astronomers to work with a perfect open loop mirror control system (set once at the start of the night then coast on lookup tables) vs. the engineer's desire to close the active optics servo loop with whatever integration length one needs on the inevitable guide star. Wolff reflected on the many hours she spent debating such isssues in Gemini. WIYN resets the mirror several times a night, and the MMT routinely resets after each large pointing change. Pearson thinks that the only reason WIYN observers don't reset more often is because they are usually limited by KPNO seeing. Hopefully we will be doing better! Blanco confirmed that the expectation hence design driver for the WIYN active optics had been a high duty cycle of M1 tweaks, but that the system had proved surprisingly stable: virtually the same M1 force matrix could be used even after the mirror was removed from its cell, recoated, and reinstalled.
It is worth noting that telescopes are evolving rapidly to online control. The NTT will be in this state when it emerges from its VLT-induced ``Big Bang" upgrade this spring. It seems a conservative goal for SOAR in 5 yrs, and one that the project will probably implement no matter what anyone says at this point. We must ensure that the mirror control is not so ``jittery" that it fails without a closed loop. Pearson's passes that test. There is a performance hit back down to ``the best you can do today" if the star isn't there.
Drained by discussions of the mirror supports, we ajourned for a visit to Optics-R-Us (aka SOML) to see the slowly revolving 1wk-old 8.4m, annealing LBT blank and several 6.5m mirrors in various stages of polishing.
While we ate a pizza lunch (for the 2nd day in a row, ugh) various
were presented by Richard Wolff. He concentrated on the Gemini system, which would be available to SOAR free. It is distributed in all ways that one could imagine (geographically, processor heirachy, etc.) The effort has been toward careful design. Pat Wallace is doing: all astrometry, coordination of M1/M2, tip/tilt. The project is still struggling w/ the M1 support system. Everyone is overwhelmed with documents in this 1st serious effort to define a TCS before making parts.
Several modules (MCS - mount control system, CRCS - Cass. rotator control system, at least) must be customized if we go equatorial (and the latter even if we don't.) There are rigorous interfaces between modules to ensure no ``backdoor" tweaks. Each VME crate starts at $6K. for a bare box. The programming is EPICS (from CEBAF, RHIC, Fermilab = accelerator community), which talks to VX/WORKS, the real-time OS. Everything flows along a 100 Mbps data LAN, a control Ethernet, and video LAN. Telemetry is available in an EPICS database. Signals are verified (Blanco notes this is NOT the case at WIYN), but Richard was unsure if major tasks can be scheduled at absolute times. The emergency STOP overrides all software.
Cecil asked if adopting Gemini meant we were locked into several year-old hardware, desynchronized with planned upgrades of rapidly obsolete electronics. Richard agreed that several processor decisions already looked outmoded and all the hardware was not even assembled yet. However, he felt that the hardware and software was modular enough to allow piecemeal updates.
Staffing: 3 people at Gemini are working on the control software, 3 at NOAO are working on the observatory control system, DOA has several people working on among other things the archive tools (to interface w/ their data archive and distribution system), and there is a larger group in the UK doing everything else. The initial budget was $6M (no overhead allowed), a long time ago. Compare this to Wisconsin who got $1M from WIYN (but probably ended up spending twice that.)
Full maintenance documentation will be provided. Baldwin noted that Gemini will have programmers in La Serena for this; he doubts that the Blanco 4m can be integrated into the Gemini TCS, which somewhat defeats the concept of a ``two-telescope observatory". However, it will presumably be possible to duplicate the look-and-feel of the Gemini TCS.
Kuhn summarized what he had presented to the science reps the day before. He has a several page report that will soon be archived at the Web site for further perusal. He noted that theory and lab work on micro-roughness is well developed. Everything is in terms of BRDF, (Bidirectional Reflectance Distribution Function) = the scattered irradiance divided by incident flux. At fixed angle BRDF scales as sigma^2/lambda^4 , so scattering is no longer a problem in the IR. Kuhn argued that 10 Ang. microroughness was fine for angles out to 100" from a bright source. Baldwin will review these conclusions at CTIO, so that we can all agree on exactly what quantitative gains can be made in scattered light control. We agreed that for the trade study we would make the PSF (core and wings) preeminent. Kuhn is working to quantify an envelope of realistic performance goals at various radii.
These went out on 1/23, in botched form. HDOS is paying for the price quote, allowing our 10K to go entirely toward defining the performance spec. Kuhn added several criteria:
Contractor will:
Cecil and Daggert have also added active optics criteria (Basic Quality & Intrinsic Quality specs.) cribbed from the Galileo contract to Zeiss. Likely HDOS will implode when they read this request, so the list above is being further refined. We will also include a request that we be allowed to define polish support points. As of 2/5/97, HDOS still did not have our quote; they claim that they will be able to respond by the end of Feb.
We have the REOSC quote from the CLEAR study, which is recent and detailed enough (e.g. it splits off superpolish and mentions explicit test criteria) that it can serve as it stands for a ``reality check" on HDOS. Daggert was told by them that they would be ready to grind the SOAR blank by Oct. 97 (VLT and Gemini blanks will be done by then), and for polishing 1 yr later. (Does this mean that it would take 1 yr to grind?? Corning will rough grind.) These times are consistent with the project schedule, for off- or on-axis. But M1 is still the critical path. Daggert was told that Kodak was interested and that Tinsley was also currently working on large off-axis optics. He will investigate.
Cecil met with Buddy Martin from Steward Observatory Mirror Lab. Because SOML lost the SOFIA polish contract (to REOSC), there is less potential business for a new 4m polishing machine at Optical Sciences. SOML will therefore likely not be in a position to bid on the SOAR polish contract, but is glad that we are interested in their adaptive M2. Martin said that the first adaptive M2 will be all but aspherized by the end of the year. 2 prototypes have been built, of 50 cm diameter (the first adaptive M2 is 64 cm diameter, being built for the MMT upgrade.) This is convex, the other 2 under design are concave, for the LBT.
Baldwin reported that the building modifications to the Gemini-S enclosure to allow entry by a Pachon-based SOAR will cost only $10K. We could use the existing magnetons in the chamber (their height can be adjusted to bring them near enough to the SOAR mirror for an excellent coat, more uniform than can be attained in the Blanco tank.) We would need to provide our own handling cart.
Moretto showed the latest optical improvements for off-axis. Minor tweaks
to the current baseline have been made by powering M3 at f/16 to reduce
the elliptical field curvature to negligible levels. A weak asphere is
used. The f/10 M3 remains a fold mirror, but a 3-element refractive corrector
+ schematic ADC is used to open the fov to 12x12 arcmin with
acceptable spots at f/10. The corrector has a number of aspheric surfaces
(mostly concave), and he hopes to eliminate a couple. Cecil will post spot
diagrams etc. in a few days, once he translates the prescriptions from
CodeV to ZEMAX. The current configurations are:
Moretto is preparing for his ``final" task: to update his reflective field corrector to dispense once and for all with a 2nd M2 on the off-axis. This will be an important simplification of the whole design, at the possible cost of the prime focus and/or a 2 mirror-to-f./10 coronagraphic location, an outcome that is unacceptable to low-scattered light enthusiasts. Thus, this configuration will be presented as a trade, with cost/performance implications. Cecil will redirect Richardson from his current obsessions to completing the design for a refractive f/16->f/10 focal reducer+ADC. This will be the on-axis compliment to Moretto's efforts.
Cecil will attend the Feb. 10 AO workshop at KPNO to learn more about possible options for SOAR. Baldwin views low-order AO as the strongest candidate for the first few years of the post-first light instrument budget.
Daggert has begun to fill in the blanks on the cost chart, revisiting the August Chapel Hill #s. So far we have an on-axis baseline quote from L&F which agrees with the previous #, as well as the differential (discussed above) for the equatorial, a huge tip/tilt price via Gemini from Lockheed which we are working to clarify, and higher numbers from Corning on 8" and 4" (5% higher) blanks. Costs are still dominated by personnel, with attendant assumptions about the management style.
M1 remains the critical path. There is 18 months for polish. Daggert has added a year for site prep and another year for construction. These do not delay first light.
[Barr reported to Cecil a conversation w/ Bob Jones at Corning on 1/31. Jones says Corning needs from us a ``mirror sketch" of the blank, detailing radius of curvature, inner/outer diameters, & thickness; and a request for a formal price quote. Once this is accepted, they will need approx. 3 months to prepare for the blank fabrication. This time will be used to: hire their project manager, assemble the team from other efforts, cut boules, remeasure their CTE's, and rebuild the 4m oven. Obviously, before this begins we need to settle on a mirror thickness, and whether to go off- or on-axis (because this sets the radius of curvature.)]
Cecil presented simulations of the psf's expected for CP, based on the
measured r0 distribution at CT and the Cn^2 profile of Mauna Kea. (Nice
Observatory will measure the latter with balloons, which Baldwin said will
start in August.) Here is the officially sanctified r0 distribution from
the CT data (this now reflects a slight adjustment to median r0 measured
in each 8-minute long stellar data string. Median r0 is 16 cm vs. 21 at
MKO): 
The limited data on CP appear to be a little bit better, but span <100 nights which moreover were during the period of best seeing at CT. Therefore we agreed that a reasonable spec is 15% degradation of top third conditions at CT, r0 = 19 cm. or a tip/tilt stabilized image at 2 microns of 0."25 FWHM. (This is still better than median conditions.) This is a very tight spec, equal to Barr's entire 1992 telescope error budget! It is not obvious how we attain 0."29 FWHM best delivered image quality at 2 microns when the site and telescope are both 0."25 and both quantities add in quadrature. Barr is willing to admit that his error budget is ``attainable", which it has been argued is not usually the case. Pushed, he felt that one could be optimistic and shoot for a quadrature sum a little over 1/3" FWHM.
Model input psf's were computed by Cecil using the above r0 and the
Skylight package, with and without tip/tilt stabilization. (Skylight generates
``phase screens" at different altitudes, to simulate the effects of seeing.)
The sum of short-exposure psf's (i.e. speckle patterns) become source images.
These are fed as input images into ZEMAX and many rays traced to the focal
planes produced by Moretto's optics to assess Strehl reduction. The resulting
psf cores are measured to set 50, 75, and 85% encircled energy. The
purpose of this effort is to accurately assess the effect of site seeing
to ensure that the optics are not over-specified. Some of the preliminary
results are shown below, and will be finalized before the Feb. S/TWG meeting.
This shows the delivered PSF's from the current baseline off-axis optics,
bare f/16 port. 3 wavelengths are shown (back to front): 0.7, 1, & 2
microns. Left 2 columns: psf input into optics (r0=18.3 cm = top third
of the CT distribution, Cn^2 for MKO) and delivered image at the focal
plane. Right 3 columns: input atmosphere after the tip/tilt signal below
is removed, image delivered at field center, image at 3' radius. Each pixel
is 0."5 square; each profile is plotted in a box 2.9" square.
Strehl reductions due to residual optical aberrations are obvious at 2
microns. (Of course the optics have now changed yet again, so these are
no longer relevant. A side-by-side comparison between off-axis and
on- will be made in a few weeks.) The tip/tilt correction at 1 micron
is:
It was agreed among all the partners to base the image-specific trade issues on the PSF, core and wings. What increments require what $s? The plan is to freeze external contracts on Feb. 15, review engineering results then, and review science requirements on the 16th and 17th.
If you have comments or corrections, please email to cecil@physics.unc.edu