The bottom line is that the data from vircam we have received so far indicate that the detectors are functioning well and that given the right type of calibration good scientific results can be extracted from them. Some highlights are: (1) crosstalk -- there doesn't appear to be any. It is very difficult to say that it is completely absent as we don't have very high quality twilight flats yet in any of the filters. But with the data we have to hand we can say that there is no obvious crosstalk, even from very bright objects. Our current expectation is that correcting for crosstalk will be unnecessary. (2) persistence -- this is quite minimal. Observations in J of a very bright star at 14:06:41 -36:22:11.4 with magnitude J=0.4 were done with 1s and 10s exposures. This would be massive overkill in most situations, but in this case it serves to show how small the persistence problem is. After each exposure a number of dark frames were taken. Because the filter wheel took about 45s to move into position, there about 50s or so elapsed between the star and the first dark exposure. In the case of the 1s exposure, the first dark frame showed some persistence but it was barely discernable. In the case of the 10s exposure there were persistent images on the first 4 dark frames (there were about 25 seconds between the start of each dark exposure). A very rough fit showed an e-folding rate of about 34s. By the 5th dark frame it was pretty much gone. Obviously such a small amount of persistence for such a bright object is very good news, but because there was a delay to let the filter change to dark we couldn't see how much there was straight after the star exposure. Data in the Eta Carina region containing large numbers of bright stars on jittered observation frames, where the delay between exposures is minimal, have no visible persistence at all. These objects aren't generally as bright as the star used in the first test above, but they are much more characteristic of bright stars expected in normal observing. It therefore looks like persistence correction will also be unnecessary. When observing a linearity sequence we interleave the linearity observations with exposures of a constant integration time to monitor the stability of the light source. In past experiments the monitor exposures have shown an increase in the flux over time, which could either indicate that the light source really is brightening or a certain level of persistence in the detectors. In the most recent experiment, the sequence was done twice -- once with increasing exposure time and once with a decreasing exposure time. If there were a real persistence issue, then one would expect the flux in the monitor exposures to show an increase and then a decrease with time. A preliminary analysis of this sequence shows that there is a very slight increase in the monitor flux over time, but that it is less that 0.2%, ruling out any measurable persistence effect. (3) linearity -- we only received a good set of linearity observations a few days ago, so comments here are not very complete. A series was taken on 20080817 in the narrow band filter. A monitor exposure was taken in between each step in the linearity sequence. The sequence started at 2 seconds and went up to 80s and then back down again. We have done a quick analysis using a 4th order polynomial model and the results are: -- the sequence did not quite reach saturation level for any of the detectors, so it is not possible to say much about that -- all of the detectors show a distinct non-linearity of generally between 2-4% at the 10000 ADU level. In the case of detector #13, it is more like 9% at that level. The linearity error estimate is very good with the exception of detector #5. In previous linearity sequences this has shown a kink in the linearity curve at the bright end and probably requires a higher order polynomial than the others to describe it. -- the bad pixel masks generated by the recipe shows an odd horizontal band (i.e. not a channel as these are vertical) about 120 pixels high at the bottom of detector #4 which does not flat field well. We have noticed this before and it needs a bit more investigation. It is visible on both dark and flat frames. -- detector #6 shows a large number of bad pixels, mainly because of the well known dodgy channel (#14) problem, i.e. at certain illumination levels the pixels jump +/- 32k ADU. Also, detector #13 still shows a low-level channel problem in some darks, e.g. VIRCAM_IMG_DARK158_0001.fits. (4) night sky fringing: an assessment of this hasn't really been possible owing to the lack of good twilight flats and to the fact that most of the astronomical observations that have been done have been on extended objects. Observations of 47Tuc (where most of the detectors don't have anything extended on them) don't seem to show any fringing in J, but many more observations in all the broad band filters will be needed to progress this further. (5) twilight flats: a few attempts have been made here. There are sequences in J, H and Ks, but we have had to push very hard to get anything out of them. There are two main reasons for the problems. The first is the lack of dark exposures that match the exposure parameters of the flats. This is serious as you can't just substitute any old dark with similar exposure parameters. You have to have the right one. The second problem has been that most of the twilight flat observations have been done without jittering the telescope. This makes the flat exposures useless. We have been able to construct reasonable flats for J and H, but for both science and publicity purposes we need better observations. (6) astronomical observations: we have spent a reasonable amount of time trying to reduce some of the pawprints and tiles that we have. Some have come out well and the results look very encouraging, however: -- the majority of the observations have been done without accompanying dark exposures. We need a better protocol for observers to follow. Dark exposures (at least 5) must be taken with the correct exposure parameters. This is crucial. -- the jitter offsets on some of the observations are actually smaller than 2-3 sigma radius of the stellar profiles of stars in the field. This makes sky subtraction using the simple algorithm in the summit pipeline impossible. There are better ways of doing this and obviously this won't be such an issue with the science pipeline. It would be best though if jitter offsets could be chosen a bit larger during testing. -- the fields chosen have almost all been of extended objects, for quite obvious reasons. Although these make good publicity pictures, if there aren't observations chronologically near of fields that can be used for an offset sky, then getting a good sky subtraction is not possible. Obviously the summit pipeline does not have the ability to handle true offset sky observations per se, so we have to do this by hand currently. It is notable, though that although we were told that the VLT software couldn't be amended to deal with the idea of offset skies, it seems that it has been in the case of HAWKI and it would be nice if we could revisit the idea for VISTA. -- current observations seem to show a problem with detector #16 in that it doesn't seem to flat field well. Detector #16 has a large quantum efficiency gradient across the detector (almost a factor of 2) and our feeling is that this is probably a linearity issue. So far we have had to reduce the astronomical observations without the aid of a linearity curve (as we didn't have a good linearity sequence available until a few days ago) so our feeling is that the poor flat fielding in detector 16 will improve greatly when this is applied. With a 3rd-order linearity fit, a compromise given the shortcomings mentioned previously, chip #16 image quality looks a lot better implying the problems could well be due to linearity. We will test this further in the next few days. (7) other artefacts: all of the detectors have pit-like -ve stars which nearly flatfield out but on one set of J flats apparently moved half a pixel. We need to investigate this further. (8) geometry: the pixel-arrays on the chips (but not necessarily their physical mountings) are mutually misaligned by a maximum rotation of 0.15 degrees, derived from inspection of WCS solutions. There is no evidence so far of non-coplanarity, but final judgement reserved pending data with full AO tuning. (9) astrometry: the WCS is well described by the ZPN projection and our measured distortion coefficient is within about +/-1 of the predicted value 42 radian/radian**3 optical distortion. Although still early days, after calibrating some dither stacks from the Eta Carina region against 2MASS, it appears that the systematic astrometric residuals are already at the <~50 mas level over the whole detector array. (10)photometry: the same J-band Eta Carina data was photometrically calibrated against 2MASS using, for now, WFCAM proxies for the colour equations. Coupled with provisional gain estimates, these suggest that the overall throughput of VISTA is some 0.5 magnitudes better than that of WFCAM, at least in the J-band (see attached table for details). det no. gaincor gain#1 gain#2 band magzpt rms no.stds airm see ell (e-/ADU) #1 0.8369 3.997 3.700 J 23.653 0.070 580 2.13 0.99 0.13 #2 1.3219 4.383 4.294 J 23.610 0.053 1094 2.13 1.00 0.16 #3 1.2213 4.123 4.156 J 23.405 0.050 491 2.13 0.98 0.14 #4 1.0017 4.741 4.398 J 23.386 0.042 781 2.13 1.02 0.16 #5 0.9634 6.039 4.282 J 23.638 0.059 930 2.13 1.00 0.14 #6 0.8785 4.387 4.162 J 23.570 0.059 841 2.13 0.94 0.18 #7 0.8273 4.356 3.978 J 23.556 0.062 1127 2.13 1.01 0.16 #8 1.0231 4.629 4.329 J 23.529 0.050 829 2.13 0.98 0.16 #9 0.9926 5.049 4.695 J 23.511 0.055 676 2.13 0.94 0.13 #10 0.9214 4.578 3.997 J 23.582 0.071 167 2.13 0.94 0.16 #11 1.0336 5.117 4.760 J 23.569 0.049 378 2.13 0.94 0.17 #12 0.9236 4.357 3.932 J 23.523 0.053 479 2.13 0.98 0.15 #13 1.3584 6.704 5.871 J 23.613 0.053 851 2.13 0.96 0.13 #14 1.0026 5.164 4.763 J 23.581 0.049 622 2.13 0.98 0.13 #15 0.8915 4.296 4.041 J 23.579 0.041 456 2.13 0.98 0.16 #16 1.1480 5.602 5.535 J 23.567 0.107 446 2.13 1.04 0.15 #total_av_zp 2008-07-14 J 23.565 0.088 10748 2.13 0.98 0.15 gaincor - computed from twilight flat gives relative gain in ADUs gain#1,2 - two different attempts at measuring real detector gain magpzpt - that brightness star that gives detector 1 ADU per second rms - scatter of photom solution for detector (2MASS contributes most) no. stds - no. of 2MASS stars used in photom solution see - average FWHM of stellar images on detector ell - average ellipicity of stellar images on detector for comparison WFCAM average magzpt in J = 23.0 average gain is ~4.5 e-/ADU