Attending: Baldwin, Sam Barden (NOAO), Cecil (Chair & minutes), Diaz, Elston, Simkin. Dottori & McMahan were absent.

Barden began by reviewing the recent Tenerife conference on Optical Fibers in Astronomy, of which he was one of the organizers. The biggest excitement was for the presentation by Gerhard Schottz (sp?) from Hydrasil (sp?) who talked about new glass silica fibers instigated by Dan Fabriccant (SAO).  Ion implantation, normally used to remove OH for red transmission is now also used to remove the chlorine which causes most of the UV scattering.  This produces a broad-band fiber which goes  down to 350 nm with good throughput. There is still a little tuning to be done if you want highest throughput in the UV (at the expense of the red).  Hydrasil (sp?) is now making 2 different fiber types, which can be drawn and clad into fibers by the fiber companies. Hydra-CTIO will use these new UV fibers.

Sam noted that the cladding thickness on any fiber should be at least 10x the operating wavelength. Data were shown at the conference that supported this in the IR: fibers with 5 micron thick cladding showed a monotonic decline in transmission as the fiber length was increased. Jacques Berand (sp?) talked about data that showed the presence of modal noise. Multi-mode fibers are required for the highest throughput, but there is noise associated with each and this is an additional quadratic term in the overall noise that becomes apparent in high S/N data on the pixel/pixel level. This is not fully understood, but does appear to scale as expected with the mode number.

There was not much information on IR fibers, other than that people are using flouride fibers to get to K. SMIRFS-2, a multi-object IFU being build at Durham for UKIRT, is using flouride fibers for K and fused silica for J & H, both using warm 1.2m-long fibers. Sam suspected that both types performed well, with the flouride fibers working down to 0.6 microns, roughly 1/3 less efficient that glass (over a 1.2 m run) at J & H but can work cold. The fibers themselves don't contribute significantly to the background. Zirconium fibers don't transmit well below H. A fiber spectrometer for SOAR would probably need two different types of fibers to do J, H, and K.

Next, the meeting discussed IFU's, including science. The Spanish reported on INTEGRAL at the WHT. There was some discussion about the minimum fiber diameter that did NOT show focal-ratio degradation (FRD) [whereby light exits the fibers spanning a larger solid angle than it had when it went in.] The situation below 50 micron fiber diameter is controversial: some groups see FRD because strong diffraction now would cause the fiber to act like a waveguide, others (Taylor & Watson at the AAO) have gone down to 30 microns and regard the FRD as acceptable. Barden felt that over a couple of meters run those fibers would be fine; he has certainly seen a noticeable degradation with 100-micron diameter fibers over 20-30 meter lengths! Cecil noted that even a 15-micron diameter fiber has 10's of waves across it, so why would one expect to be in the regime of strong diffraction?

In previous conferences people were using bare fibers. Now everyone is going to lens arrays to feed fibers. There were both macro- and micro-lens approaches discussed. For FUEGOS, the Saclay concept is to use 7 macrolenses & fibers to cover the target, to act as an image slicer and to maximize spectral resolution by using a small slit. They get 62% efficiency in coupling light into the fiber/lenset array. (Sam was unclear if this was theoretical or measured efficiency.) The AAO also uses macro-lenses, which requires fore-optics to magnify the telescope focus to f/100. Macro-lenses are advantageous because they are large enough to be ground from optical glass rather than cast in a resin mold. This reduces the scattered light. Telecentricity constraints are also relaxed because the surface curvature of a macrolens is smaller than that of a microlens. On the other hand, the Durham group quotes only 40% efficiency for their best epoxy fiber/microlens IFU but part of this is because they are using microlenses on the output end of the fiber as well and convert the telescope beam fromn f/36 to f/5 then back to f/36 for an (existing) long-slit spectrometer (CGS4). Sam felt that 67% efficiency was reasonable if you used a fast (f/5) spectrometer customized for an IFU so only one focal conversion was necessary. The French developed a cylindrical lens scheme which also has higher optical quality than resin casts. Up to 10th-order aspherics can be applied to these lenses.which have been made up to 0.4mm in diameter, w/ 1-2 micron deadspace between each strip. Two arrays are crossed at 90 degrees to form a square grid. One question to consider is which array is better: square or hexagonal? One might suspect that square micro-lens arrays would have worse scattered light because of their sharper corners. There are currently 5% variations in the transmission of different lenses, which can of course be removed w/ flat fields.

The MPI-Garching  instrument SYMPHONY is having their fibers tapered: the fiber diameters flare out by a factor of 15x at the input end, and the fibers are polished and formed into hexagons to fully sample the image, all without breaking the fibers! This is a J, H, K high resolution imaging spectrometer. They did this because their alignment tolerances between fiber and micro lenses was extremely high.

Finally, Sam reported on the IoA instrument COSI, which suppresses 100 OH-lines in each of J & H, which will be fed with an IFU. There are at least 3 groups working up IRAF scripts to reduce IFU data. There was also discussion about whether it was acceptable to allow cross-talk at the output. [There is no cross-talk between fibers or between microlenses because the lenslets image the telescope pupil down onto the fiber cores.] Clearly IFU's can be arranged so that adjacent fibers on the sky also appear close to each other on the slit.  Thus there is already cross-talk at the input from seeing especially if you are over-sampling your seeing profile, so cross-talk is then also permissible at the output. Sam was not aware that anyone was designing their fibers systems with anything more sophisticated than ray tracing. He was not aware of any work being done to model scattered light for example.

Cecil then summarized Sebring's activities: a meeting with M3, hiring Jeff Barr (Larry Barr's son who works for NOAO as an achitect) to layout the dome & support building in conjunction with M3; studies underway at Equatorial (Brazil Industry Initiative), Contraves & ROSI (formerly HDOS) for the active optics subsystem, Comsat RSI & L&F for the mount subsystem. Candidates are being interviewed to replace Kitty. Gilberto has designed an f/16->f/9 corrector which currently has good images and excellent UV transmission with 4 lenses and a field flattener/dewar window. It is compact, feeds only the imager, and incorporates an ADC for the imager only in the converging beam. This will be refined.  The spectrometer would have its own ADC, probably before any ``slit" to maximize throughput. Thus, on the 6-sided non-Gemini instrument side we'd have the input beam, optical reimager, IR imager, wavefront sensor, and direct path to an optical MOS.  This leaves one face for the facility calibration unit (a la Gemini) or an AOB-type adaptive-optics unit. Beam redirection is done with an M4 fold mirror. This has to be gimballed, because the facility calibration unit will use this mirror to feed all the other instruments on that side of the telescope.

Cecil reported on a conversation with Boronson (Gemini) about optical detectors.  2x4K 3-side buttable thinned CCD's are available from both SITE & EEV. SITE is delivering the chips and EEV is imminent. The SITE chips have 15 micron pixels, EEV has 13.5 microns but the chip is actually 2x4.5K. The chips are flat to 10 microns. There is fringing in the red; the chips are thinned by the manufacturers. Each chip costs $50K if purchased in a consortium (which seem to form roughly every 18 months or so.) Lincoln Labs is also making similar chips for a consortium led by IfA in Honolulu, which violates LL's mandate from Congress that they not compete with industry. As a result they are being sued by the other CCD companies! So, our optical imager could be a 2x1 or 3x1 mosaic (4x4.5K or 6x4.5K, the latter spanning 7.9x5.9' with 13.5-micron pixels at f/9.)

Simkin then reviewed her progress on revising the Science Requirements Doc. She listed the instruments that had been discussed in Tucson. Baldwin agreed with Elston that the instruments should be ranked by the science that would be lost if each instrument was not available. Cecil felt that an IR imager would be the logical commissioning instrument because the atmosphere would be most stable in that waveband, allowing us to most accurately assess the telescope performance. It was agreed that this instrument could be borrowed for commissioning, i.e. use COB. COB as 0."1/pixel, which might be a little coarse to sample the telescope images. Elston felt that COB's optical degradation will need to be quantified because SOAR's image spec is so tight. This will presumably be done by Gemini because they are planning to commission Gemini-S with it; Cecil will check w/ Doug Simons about this. It was felt that the commissioning camera should be as simple as possible to minimize optical ambiguities. It would be translated to different points of the science field to check telescope optical performance. The science imager could then reimage etc. as required to optimize ITS mission.

Simkin felt that cutting the commissioning phase by the project down to 6 months would simply transfer the burden to CTIO. She reminded the group that the Operations Working Group had expected a 3-year shakedown period, with more and more science being done as time went on.

Baldwin felt that the 2d optical spectrometer (MOS) should be higher on the list than the high-efficiency stellar spectrometer. Diaz agreed, feeling that SOAR would be more competitive with an MOS. The goal should be to keep the latter as cheap as possible, so that we can easily afford it. Cecil noted that one could play the high efficiency of the latter against the spatial multiplex of the former. To decide, we needed to understand target densities on the sky. He is working to quantify this in a variety of environments. The group clearly was favoring an MOS over a high-throughput bore-sight spectrometer. The latter would float to the top of our construction list if funds are limited. Simkin felt that the MOS was an experimental instrument. Do we want first-light instruments to be experimental? Cecil felt that the instrument could be modular, with different modes coming on-line in the future.

Baldwin next discussed the desirability of an natural-guidestar (NGS) system that could be upgraded to a laser-guidestar. It seems that the laser upgrade is comparatively inexpensive. Bladwin felt that we should have the AO system under construction before the telescope is completed. The group felt that we should adopt the formal goal of having the AO system operational by the end of the 3-year commissioning phase. Cecil felt that the AO would remain separate from the other instruments, if only because it is an obvious ``package" for NSF funding as a ``telescope upgrade". Baldwin felt that to prevent this ``decoupling" we should state strongly that we will be beaten into the ground by other 4m telescopes in the south if we don't have AO.

There was further discussion about the Science Requirements doc. It was decided to edit it by circulating around the group. Various time goals were discussed, [none of which were ultimately met because of the holidays.] The document is being subdivided into several sections, which will eventually be recombined.