Michael's minutes included these questions: > We all agreed that further progress on image quality will require >one or more dedicated people to investigate the problem(s). In the >meantime, it would be very useful to have a summary of: > -All known and possible problems affecting image quality > (wind-shake? Primary mirror support? Not enough tension > in the secondary truss? Catastrophic slippage giving rise to > double images? Focus instability? High-speed vibrations caused > by the drives?) > -What is currently known (and not known) about their contribution > to the error budget. > -What we need to do to get more information about these things. > Without this in hand, it is difficult to make decisions of the >relative importance of these different items. This is actually a message I composed a few weeks ago, just before the AAS meeting. It was originally written as an attempt to get my own thoughts in order following the various Shack-Hartmann tests last quarter. I considered sending it out to the user community, but there was something of a deluge of messages at that time. However, I think it addresses most of the questions from the minutes, including those for which we don't yet have firm answers. I'm trying to sum up what we currently know about various things that have an effect on image quality, so that we can figure out which is the tallest pole in the tent and try to identify areas where it's possible to get the most improvement for the least cost. What we know about collimation: 1) Coma is probably the worst aberration we suffer. Astigmatism is close behind, but harder to fix. Coma is caused by tilting or translation of the secondary and primary mirrors relative to one another; either mirror could be moving and cause the same effect, and therefore it's sometimes hard to identify the reason for a change in coma. 2) The secondary mirror tilt required to correct the coma changes by significant amounts on a timescale of a few days, and also shows large changes as the telescope cools over the course of an evening. 3) The secondary encoders indicate no correlation between the SH-measured coma and the secondary mirror tilt relative to the mirror frame. This is the single strongest result we've gotten from the Shack-Hartmann monitoring program of Autumn 2002. 4) The primary LVDTs do show some variation, with a slight tendency toward a bi-valued distribution. However, the motion is less than a micron on each actuator, and shows no clear correlation with the SH results. The worst PMSS oscillations only cause the star to wobble on the SH video by a few tenths of an arcsecond. I believe (but I have not confirmed) this indicates that the primary can't be tilting far enough to produce a significant change in coma. When the secondary is tilted by the smallest amount significant for collimation, the star motion is on the order of 10 arcsec. 5) My information on mirror translation as opposed to mirror tilt is a little less solid since both the primary and the secondary only offer translation measurements in one direction. However, the information that is available matches the two previous conclusions: the secondary translation relative to its frame does not correlate with Shack-Hartmann collimation measurements, and the translation seen on the primary is probably too small to produce the observed changes in collimation. 6) The strain gauges on the secondary support vanes do show occasional jumps in resistance which might indicate that the entire frame of the secondary is being moved around. Because only twice-daily monitoring of the strain gauges was possible, we have not firmly linked any of these jumps to a change in collimation. Nor have we been able to find a clear correlation between the strain gauge resistance (or the collimation) and weather conditions, although there might be a weak link between resistance changes and humidity. We need more data with better time-sampling in order to confirm any such relationships; all we know for now is that it's plausible that the support vanes could be causing the changes in collimation. In the three observed cases where the resistance changed on the strain gauges, all four accessible strain gauges (out of the total of eight) changed at the same time in the same direction by similar amounts. We therefore hope that it is possible to judge the behavior of all eight support vanes by monitoring only one strain gauge. What we know about the other effects of mirror tilts: 1) Tilts of the tertiary mirror cause pointing changes without collimation changes. We have evidence from the echelle that large changes in tertiary tilt (several hundred arcseconds) happen on a few-day timescale, although it's usually consistent during a single night and showed excellent repeatability during recent pointing models. Pointing appears to be more consistent at NA2, and when it does have problems, those are often attributable to motors rather than mirrors. I have no strong evidence that tertiary tilts at NA2 have a significant effect on pointing. Tertiary tilt inconsistencies are worst at TR4, showing egregious non-repeatable behavior within a single night, especially at low altitudes. Pointing accuracy at this port may never been better than 10-15 arcseconds. In November, I implemented tertiary sag coefficients at all three instrument ports. I have limited evidence from pointing models that the sag behavior is consistent over a timescale of weeks or months for NA1 and NA2, and therefore I hope the sag coefficients will improve the accuracy of NA1 pointing relative to NA2. I haven't been able to confirm that result; adjustments to the tertiary tilt constant are still needed every week or two to improve echelle pointing. The echelle "tilt" changes tend to be largest in the Y-tilt direction, which might indicate that the problem is actually in the repeatability of the tertiary rotation rather than tilt, although we have not seen any such problems in multiple engineering tests. It's also possible there could be other sources of pointing error which can be fixed by changes in tertiary tilt, but are not caused by irregularities of tertiary tilt or rotation; however, I do think the tertiary is the most likely culprit. 2) PMSS oscillations frequently cause degradations of image quality. These oscillations are clearly visible on the Shack-Hartmann video screen, at levels of a few tenths of an arcsecond. In science, most of the time the effects are quite subtle, making images only slightly out-of-round or masquerading as poor seeing. However, given the evidence of the dataq and the confirmation from the SH video that PMSS oscillations truly do translate into image motion, I believe this is having a significant impact. The oscillations tend to happen in three kinds of situations. All of these are worse when the PMSS is poorly tuned, but even under the best circumstances we usually have some of this going on. The first situation is high altitude, where oscillations usually start showing up somewhere in the region 80-83 degrees. The second situation is windshake; unavoidable in itself, but sometimes it sets off oscillations which are much too slow in damping out. The third, least common situation is oscillations at low altitudes, often arising after a long slew in azimuth. These tend to produce the most obvious effect on science, although again it sometimes masquerades as seeing because observers often attribute it to high airmass. What we know about focus: 1) The current adjustments for focus as a function of temperature are better than nothing, but not adequate by themselves. I believe we spend a considerable fraction of our observing time away from best focus, mostly because of temperature changes. My experience is that keeping the telescope in focus requires checks at least every hour and sometimes every quarter-hour, especially on nights when the temperature is changing atypically or the seeing is exceptionally good. 2) My sense for the scientific instruments is that changes in focus can usually but not always be predicted by changes in the difference between the air temperature and the structure temperature. When the air temperature warms suddenly, the focus goes up until the telescope has time to re-adjust. During the evening when the telescope is not yet in equilibrium, it usually starts out warmer than the air with a negative focus value, then the value moves higher as the telescope cools. If the telex fans are turned off in the middle of the night, the focus tends to move upward. The above statements are based entirely on intuition and anecdote. I spent 5 hours one engineering night doing focus after focus to quantify this relationship, and couldn't prove anything (the temperature was unusually stable that night, and so was the focus). Nor have I been able to plot this correlation using focus data from our logs; all I got when I tried (in 1999? 2000?) was a scattergram. Perhaps I should try checking that again with more recent focus data. Currently, the autofocus algorithms are based only on the structure temperature; the air temperature is not accessible to that algorithm, so it would have to be re-written to test the helpfulness of my intuition. 3) The Shack-Hartmann's focus appears to behave differently than the other instruments. Notably, it tends to start high in the evening when the telescope is warm and move downward as the telescope cools; this occurred with the instrument mounted at both TR2 and TR4. I cannot think of any explanation for this. The focus should be dominated by telescope rather than instrument effects. The echelle's behavior has generally been similar to the other instruments, as far as I could tell. Since spherical aberration is supposed to indicate the ideal position of the focal plane, I might be able to tell if this phenomenon is caused by some internal temperature-related expansion within the instrument. However, focal plane motions need to be quite large (millimeters) in order to make significant focus changes, and I don't see how temperature effects internal to the instrument could be making such a difference. 4) The focus vs. altitude algorithm appears fairly good; I have never had any strong evidence, either from science observations or from engineering, that it was imperfect. However, I would like to check it more closely. With the software tools currently available, I cannot compare the requested focus sag versus the secondary encoder positions; but that by itself wouldn't be a sufficient check, since sag in the truss and vanes might also affect focus. Therefore a direct empirical check is needed, on a night with good seeing. I would prefer to use spicam than the Shack-Hartmann, partly because of the strange behavior mentioned above, and also because spicam can measure focus at lower altitudes. What we know about motors: 1) For the most part, it seems that the motors either work adequately for tracking, or they don't work at all. The very short period between the appearance of the first image-quality problems and the complete breakdown of both azimuth motors in November 2002 is a case in point. 2) When the motors do produce image quality problems, it's often in conjunction with the PMSS, as mentioned above. Long azimuth slews at low altitude sometimes set up mirror oscillations. The PMSS behavior at high altitude tends to be worse in the north and south (where azimuth is tracking more rapidly than usual) and not so bad at similar altitudes in the east and west (where the tracking direction is primarily in altitude). 3) Tracking irregularities in the enclosure motors can be felt fairly easily by a person standing in the dome during observations. Theoretically, these should not affect the telescope because it's decoupled from the dome; however, we know the decoupling isn't perfect because we've occasionally seen footsteps in the dome produce twitches in an untuned PMSS. Didn't Karen once report seeing a little motion on the SH video when someone was walking around in the dome? So it would be good to think of a way to check whether the slight lurching of the enclosure is feeding over into our images. 4) The rotator is the most erratic of the three axes, although its behavior is worse on slews than on tracking. By itself, this is usually not enough to affect image quality even on long exposures, since the science instruments are on the optical axis. However, secondary problems arise when the guider, which is placed further from the optical axis, corrects rotator eccentricities by making adjustments in azimuth and altitude. The drift produced in this way is generally a few tenths of an arcsecond in a quarter hour, somewhat smaller than tracking inaccuracies without the guider running at all. But it would be nice if it could be better, either through software (which would require having two guide stars in the field at the same time -- not always possible) or through repair of the rotator. 5) Bad magnesensor readings usually cause egregious pointing and tracking errors which are immediately noticed. However, we had some instances lately where the first bad reading didn't ruin the pointing right away, and it was only when the second or third error showed up that the problem became obvious. I have to wonder if there could have been small tracking errors after the first bad measurement which were not glaringly obvious, but might have affected image quality. This is a fairly rare situation, however. Our image quality is reasonable now, but I believe if we can address some of the problems above we would be able to reduce the number of occasions of bad "seeing," since some of those are in fact focus or image motion problems. We would also make it possible to take better advantage of the best genuine seeing conditions that might present themselves. APO APO APO APO APO Apache Point Observatory 3.5m APO APO APO APO APO This is message 651 in the apo35-general archive. You can find APO the archive on http://www.astro.princeton.edu/APO/apo35-general/INDEX.html APO To join/leave the list, send mail to apo35-request@astro.princeton.edu APO To post a message, mail it to apo35-general@astro.princeton.edu APO APO APO APO APO APO APO APO APO APO APO APO APO APO APO APO APO