MIT/LL #3 CCID-20 W67C2 TEST SUMMARY ************************************ Revised 21/10/98 Started 16/9/98 J.R.BARTON MITLL3_TEST_SUMMARY.TXT CCD DESCRIPTION: This is the third 2Kx4K 15 micron pixel MIT/LL CCD received as part of the UofH contract with Lincoln Labs (the first chip being a setup grade, the second being the engineering grade W20C2, now on the telescope as MITLL2). MITLL3 is a deep depletion, high resistivity CCD with a reported thinned thickness of 40 microns. SUMMARY OF CHARACTERISTICS OF INTEREST TO OBSERVERS *************************************************** CAUTIONARY NOTES MITLL3 has been a difficult CCD to set up and several compromises have been made to provide generally acceptable operation. These are discussed in the appendix. Observers need to know that.... * Hot pixels cause saturated columns and weaker vertical trails in bias frames. In long dark frames weaker hot pixels may also saturate pixels locally about the hot pixel site, faint hot pixels become more prominant and some that are undetectable at short integration become prominent. A defect map may be needed to assist observers in the placement of their obejects. * The trailed charge from hot pixels can appear to start up unexpectedly in windowed images when the hot pixel is not contained within the window. The trailed charge is, however, in the expected column containing the hot pixel and is of the expected intensity. * Binning along the row will not work. * Binning along the column will produce a substantial brightening of the background right across the image once the saturated columns of the major defect are shifted into the readout register. This is due to readout register overload. Binning should not be used for science readouts. * The linearity correction factors must be carefully applied. * Cosmic events can be very extended (104 pixels is the record length to date). * The CCD needs a few hours cleanout after powering up before attempting the most sensitive observations. FINAL READOUT NOISE, GAIN AND LINEARITY # ## ### SPEED INT GAIN RON ALPHA SAT'N READ RATE READOUT TIME (sec) (us) e/ADU e-rms Ke- (us/pix) full CCD 10E6pix NONASTRO 1+1 4.7 5.2 -0.63 150 6.5 58 40 FAST 2+2 2.2 2.9 -0.15 148 10.5 108 71 NORMAL 4+4 1.1 2.0 0.24 62 16 143 77 SLOW 12+12 0.36 1.6 0.15 23 32 285 101 XTRASLOW 48+48 0.088 1.3 0.03 6 104 924 208 NOTES: # The unusual variation of ALPHA with readout speed arises from the nature of the linearity error being corrected. At NORMAL, SLOW and XTRASLOW readout rates the linearity correction is done well and results in a linearity of better than 0.1% rms error over the full range of data counts to 65535. At the FAST aand NONASTRO readout speeds the correction factor is a compromise (a better correction would require the solution of a cubic function rather than a quadratic) providing a corrected linearity of about 0.5% rms over the full 65535 counts for FAST and up to 30000 counts for NONASTRO. Improved linearities of about 0.25% rms may be obtained up to data values up to 50000 counts in FAST or 22000 counts in NONASTRO by not doing any linearity correction at all. NOTE: ALPHA values are x10^-6 ## saturation figures apply for data values of 65535 counts except for the NONASTRO speed where 30000 counts is the limit above which CCD well size is exceeded. ### the readout times quoted include a reasonable overscan. The times are those calculated by OBSERVER software. - full (window MITLL_FULL) is a 2120 pixel (incl 72 pixel overscan) by 4096 rows unbinned - 10E6pix (window <MITLL_CENTRESTRIP) is a strip 250 pixels wide down the centre of the CCD plus 50 overscan pixels per row by 4096 rows, all unbinned QUANTUM EFFICIENCY Quantum Efficiencies (%) at spot wavelengths (nm) 300 320 340 360 380 400 420 440 460 480 500 550 600 700 800 900 1000 1050 At 160K, 6/10/98 0 0 0 0 5.0 18 31 39 48 55 61 73 81 87 83 59 16 2.1 COSMIC RAY EVENTS The cosmic ray event rate is about 830 hits/1000secs. As the readout time may be significant, the time should be calculated as the exposure time plus half of the readout time. Compared with normal thin epi CCDs the observer will notice that a significant number of cosmic events are very extended and that some of these form cute curly trails. Cosmics have been seen up to 104 pixels long. All have the usual cloud of electrons surrounding the main path. This cloud is possibly more intense in the thick deep depletion device than in the thinner epi CCDs. DARK CURRENTS Once the CCD, operating at 160K, has cleaned itself out for a couple of days following power-on, the ultimate dark current falls to less than 0.4 e-/pix/2000sec. See below for the power-on cleanout rate. The dark current is featureless over the area of the CCD except that its intensity reduces towards the edges of the chip. POWER-ON CLEANOUT The cleanout rate following power off (for one minute) then power-on, with a cold CCD, is indicated in the following table of dark current vs time after power-on, compiled from dark frames ranging from 10 seconds up to 2000 seconds duration: TIME AFTER INTENSITY POWER-ON (e-/pix/2000sec) 5 mins 50 10 24 20 12 30 8.4 45 6.0 1 hour 7.6 2 4.2 4 2.4 6 1.3 8 1.1 24 0.66 2-3 days 0.3 to 0.4 The CCD requires some hours after power-on to sufficiently clean itself so as to attain a dark current low enough to avoid degradation of the readout noise on longer low background exposures. TRAPPING SITES A quick look for trapping sites has revealed only a small number of obvious traps. The co-ordinates given are with respect to the first pixel out (of output amplifier A) being pixel (1, 1). Those with unknown Y co-ordinates have been detected only by their trailed charge into the Y overscan region. ( 472, 488) consuming about 160 e- ( 726, 1064) 250 ( 766, 240) 2000 (1032, ????) (1272, ????) (1441, ????) (1952, ????) (1956, ????) There are probably many smaller traps that have remained undetected. These may become prominent in charge shuffling readouts. HOT PIXEL DEFECTS In the following defect lists the co-ordinates are with respect to the first pixel out of AMP-A being pixel (1, 1). Charge intensities apply to NORMAL readout mode and a CCD operating temperature of 160K. Intensities will be different at other readout rates and temperatures. 1. BRIGHT DEFECTS LIST See the appendix for a more complete description of defects found earlier in the CCD setup, some of which have since changed but may return. Several CCD columns show trailed charge on bias and dark frames due to the presence of hot pixels. The trails are brightened columns that follow the defect site in the readout. They follow after the defect site because the slower readout allows sufficient time for the defect to deposit charge in the column pixels as these pixels are shifted through the defect site. For all but the major defect there is insufficient time to accumulate a measurable charge in pixels ahead of the defect site as each column pixel is swept rapidly through the defect site during the fast cleanout process. Only the major defect is sufficiently intense to deposit a measurable charge ahead of the defect site. The following CCD columns trail charge on bias and dark frames due to hot pixels at the locations indicated. The intensity measurements (e-/pix) apply only for a NORMAL speed readout and the CCD operated at 160K. Column 168 due to hot pixel at ( 168, 3473) .... NOT SEEN LATELY 239 due to hot pixel at ( 239, 4096/7) trails 12 e-/pix 349 due to hot pixel at ( 349, 3920/1) trails 4 e-/pix 733/4 rejected columns due to a major defect at (733/4, 3821) trails 240 and 280 e-/pix in columns 733/4 1050/1 due to hot pixel at (1050/1,1219) trails 1 and 3 e-/pix 1091 due to hot pixel at (1091,1255/6) trails 3 e-/pix 1138-51 including rejected columns affected by the MAJOR DEFECT centred at (1144,1228) - see below for intensities 1157/8 due to hot pixel at (1157/8,1233) trails 12 and 3 e-/pix 1213/4 due to hot pixel at (1213/4, 437) trails 3 and 2 e-/pix 1298 due to hot pixel at (1298,1505/6) trails 4 e-/pix 1316/7/8 including rejected column due to hot pixel at (1317, 155) trails 1, 80 and 3 e-/pix 1497 rejected column affected by the major defect at (1497, 3477) trails 550 e-/pix 1922/3 due to hot pixel at (1922/3, 1893) trails 6 e-/pix just beyond the defect but trails fade to zero over last columns read out 2048 rejected column due to some sort of intermittent edge effect The MAJOR DEFECT's intensity profile, in e-/pix, was measured in three places, one well ahead of the defect site, another just above the defect site and a third beyond the defect site. The results were: COL NUMBER 1138 9 1140 1 2 1143 1144 1145 6 7 8 9 1150 1 WELL AHEAD 1 460 9300 430 1 JUST ABOVE 1 4 10 580 SATR 540 14 4 1 BEYOND 1 2 4 12 140 SATR SATR SATR 270 13 5 3 1 1 2. TABLE OF HOT PIXEL DEFECTS FOUND IN 2000 SECOND DARK FRAMES Peak search routines were used to detect and list hot pixels and cosmic ray events in 2000 sec dark frames at a CCD temperature of 160K. The routine rejected known defective columns from inclusion in these lists. Columns 733, 734, 1143, 1144, 1145, 1146, 1317, 1497 and 2048 were completely rejected from the listing owing to severe defects at (733/4, 3821), (1144,1228), (1317, 155), (1497, 3477) and an erratic last column 2048. Processing several such repeated dark frames, enabled a table of repeating defects to be compiled. In the following table defect co-ordinates are based on (1, 1) being the first pixel read out of the CCD's AMP-A. The third figure is the defect's intensity, measured in ADUs (gain 1.75 e-/pix) above the bias level. The figure for count (cnt) is the number of dark frames displaying the defect. Thus, cnt = 4 indicates that the defect was found on all four dark frames whereas a cnt = 2 indicates it appeared on only two frames and may therefor be a remnant of a cosmic ray event or perhaps a pixel a little distant from an erratic hot pixel. ( 209, 127) 112 cnt = 4 (1316, 154) 65132 cnt = 2 (1295, 178) 11625 cnt = 4 ( 599, 278) 316 cnt = 4 ( 517, 315) 198 cnt = 4 ( 604, 369) 422 cnt = 4 (1082, 401) 31 cnt = 3 (1213, 437) 23267 cnt = 4 (1956, 517) 6884 cnt = 4 (1952, 519) 986 cnt = 4 (1335, 553) 534 cnt = 4 (1553, 582) 191 cnt = 3 ( 823, 638) 946 cnt = 2 ( 607, 699) 669 cnt = 2 ( 811, 778) 533 cnt = 4 (1580, 807) 143 cnt = 4 ( 191, 821) 29 cnt = 4 ( 835, 938) 199 cnt = 4 (1165,1054) 215 cnt = 4 ( 982,1130) 1152 cnt = 4 (1105,1145) 97 cnt = 4 (1933,1171) 5555 cnt = 4 (1131,1176) 38 cnt = 4 ( 754,1193) 2287 cnt = 4 (1027,1204) 135 cnt = 4 (1100,1205) 3328 cnt = 4 ( 957,1210) 91 cnt = 2 ( 956,1210) 100 cnt = 2 ( 786,1211) 24 cnt = 3 (1051,1218) 31956 cnt = 4 ( 969,1220) 391 cnt = 4 ( 952,1225) 1605 cnt = 4 (1147,1232) 67 cnt = 2 (1157,1232) 65138 cnt = 4 (1086,1234) 544 cnt = 4 (1154,1237) 153 cnt = 4 (1112,1244) 151 cnt = 4 (1043,1249) 49 cnt = 3 (1136,1251) 21 cnt = 4 ( 47,1252) 25 cnt = 2 (1122,1253) 55 cnt = 4 (1013,1255) 95 cnt = 4 (1091,1255) 24382 cnt = 4 (1050,1299) 256 cnt = 4 ( 253,1349) 235 cnt = 4 ( 857,1353) 148 cnt = 2 ( 871,1353) 104 cnt = 4 ( 857,1354) 96 cnt = 2 ( 802,1370) 222 cnt = 4 ( 807,1387) 610 cnt = 4 ( 242,1405) 36 cnt = 4 (1860,1474) 5392 cnt = 2 (1562,1490) 203 cnt = 4 (1298,1503) 65138 cnt = 2 (1298,1504) 65138 cnt = 2 (1298,1507) 65132 cnt = 2 (1390,1521) 26 cnt = 3 (1913,1599) 307 cnt = 4 (1923,1893) 5317 cnt = 4 ( 710,2238) 1157 cnt = 4 ( 391,2289) 998 cnt = 4 (1699,2351) 450 cnt = 3 ( 688,2572) 530 cnt = 4 ( 683,2630) 1559 cnt = 2 (1591,2813) 26 cnt = 4 (1272,2957) 5121 cnt = 4 ( 268,3051) 130 cnt = 4 (1599,3143) 598 cnt = 4 (1309,3239) 40 cnt = 4 ( 965,3293) 2475 cnt = 4 (1842,3345) 102 cnt = 4 (2008,3432) 6581 cnt = 4 ( 168,3472) 2595 cnt = 4 ( 749,3614) 99 cnt = 4 (1441,3624) 2962 cnt = 4 (1441,3628) 96 cnt = 4 (1441,3632) 40 cnt = 4 (1032,3678) 245 cnt = 4 ( 732,3820) 65132 cnt = 4 ( 736,3821) 3063 cnt = 4 ( 349,3920) 11740 cnt = 4 (1123,3951) 680 cnt = 3 ( 239,4096) 65138 cnt = 4 CLOCK INDUCED DARK CURRENTS These have not seen on this CCD as they are too weak to show without binning and since binning is not recommended there is no problem. SATURATION STRIPES Saturation stripes in the row direction across bright saturating pinhole images used to simulate bright point sources have not been detected. No bar patterns or regions of depressed or raised bias level were seen when the local spot was overexposed to 50Me-/pix. RESIDUAL IMAGES There are no signs of residual images in 300 sec dark frames taken immediately after a pinhole exposure 350 times chip saturation, i.e. at 50Me-/pix. VERTICAL AND HORIZONTAL CHARGE TRANSFER No direct measuremts have been made. Cosmic rays appear crisp with no smearing detected either in X or Y. APPENDIX - GENERAL DISCUSSION ***************************** 1. LIGHT EMITTING DEFECTS OR HOT PIXELS: The following was compiled after the second cooling of the CCD. Some of the defects have changed again after the third cooling but are listed here as they may return to their original form sometime in the future. The defect intensities are those seen using a FAST readout speed with the CCD at a temperature of 160K. * At (1144, 1228) the MAJOR DEFECT appears to be a light emitting defect (LED) which severely affects a block of about 7 by 7 pixels about this location. The defect spills saturated charge into pixels above and below this location and into adjacent columns. In fast cleanout a trail of approx 9500 e- /pix extends up to the readout register. Below the defect three columns of saturated pixels extend down into the vertical overscan. Columns either side of these have decreasing charge. About four columns are severely affected and a further six have less than 10 e-/pix induced in them. On longer exposures the saturated charge spillage extends up towards the readout register by several hundred pixels in 2000 second exposures but remains well clear of the readout register. * At (1317, 155) a hot pixel is, currently, fairly dormant and generates a trailed charge of only 80 e-/pix with columns either side being weakly affected. However, during the first cooling of the CCD this defect was very strong. It generated a saturating charge affecting a block of 5 by 5 pixels about this location and extending up and down from this location on longer exposures. Trailed charge was very weak above the defect and about 2500 e-/pix below with about 25 to 50 e-/pix appearing in columns 1316 and 1318. On longer exposures the saturated charge spillage extended up towards the readout register by about 130 pixels in a 2000 second exposures but although it remained well clear of the readout register it caused a severe brightening over about 50 rows at the top of the image. * At (1295, 178) another hot pixel, currently fairly dormant, but during the first cooling affected a block of 5 by 5 pixels about the location of the defect. On longer exposures saturating charge spilled up and down the two columns. Above the defect, trailed charge was too weak to be seen, but below, a trailed column pair of 700 and 1000 e-/pix was seen. * At (1213, 437) and (1213, 438) another hot pixel, less intense but produces about 25 e-/pix in the two columns below the defect. * At (1952, 519) a defect produces 3000 e-/pix at the site in a bias frame and causes a trailed charge of about 25 e-/pix extending down but weakens on the way to disappear. * Colums 733 and 734 appear about 2 or 3 e-/pix brighter than the surrounding pixels in a bias frame for an unknown reason. * At (1149, 1233) a hot pixel of about 100 e-/pix * At (1923, 1893) and (1922, 1893) a pair producing about 4000 and 2000 e-/pix and about 30 and 20 e-/pix trailing charge respectively * At (2008, 3432) and (2007, 3432) a pair producing about 2500 and 400 e-/pix and about 25 and 8 e-/pix trailing charge respectively * At (1497, 3477) a hot pixel generating about 25000e- and producing a trail of intensity 400 e-/pix below it * At (168, 3473) and (168, 3472) a pair producing about 4000 and 2500 e-/pix and about 12 and 6 e-/pix trailing charge respectively * At (733, 3821) and (734, 3821) a pair producing about 70000 and 80000 e-/pix and about 800 and 1100 e-/pix trailing charge respectively * At (349, 3921) a hot pixel with 500e in a bias frame with a trail of about 12 e-/pix Most of the above defects trail charge into the vertical overscan. 2. DEFECT LOCATION The table of hot pixel defects was determined by means of peak search and zap routines and a process peaks facility. Peak search software scans the image for pixel values greater than a preset value (20 ADUs in this case) above the bias level. When one is found adjacent pixels are checked, the peak pixel found and its coordinates and intensity logged (but not logged if the peak pixel lies on one of the rejected columns). A block of 7 by 7 pixels centred on the peak are then zapped, i.e. set to the bias level. The software then resumes the search. Ultimately all pixels greater than the preset level are zapped. Some defects will be counted more than once if they extended beyond the 7 by 7 block. Also within the 7 by 7 block, other pixels that may be catagorised as defective are not listed. The process peaks facility takes the zap tables from several images and looks for repeated events based on recurring pixel coordinates. Those that are found on two or more tables are listed in the processed peaks table. NOTE: some are multiple hits on the same major defect, i.e. the one at about (1441, 3630). The process of finding defects is not perfect and the defect tables may need to be refined in a manner more useful for observing. These defects may need to be entered into a defect map for the device. This map could include the defective columns, which in many cases are very useable over the length read out ahead of the defect centre. 3. PROBLEMS ENCOUNTERED WITH THIS CCD a) INITIAL INSTALLATION: On removing the protective shorting flexlead, pin 19 of the Nanonics connector (serial clock, phase 3) on the CCD partially pulled from the socket. This lifted the Nanonics pin from the AlN package, pulling the gold metallised connecting pad off with it and leaving both hovering a millimetre above the package. Repairs to the connection were made with gold-loaded epoxy and the floating pin was bonded to adjacent pins with vacuum grade epoxy to secure it. This Nanonics connector should never again be unplugged. b) FIRST IMAGES * Initial binned images looked bad. They showed a very broad saturating stripe down the centre of the array. This was accompanied by a strong smearing or brightening in the X direction almost the full length of the chip. In longer exposures this major defect caused a saturating charge to leak up the column toward the readout register. When, after about a couple of hundred seconds exposure time, it reached the readout register extremely strong X direction smearing or brightening wiped out the top one-third of the chip. * Three other "hot pixel" sites produced strong trails in Y and smearing in X in binned images. The character of these defects differed from that of the major defect. In fact, there appeared to be three types of defect with differing responses to changes in operating parameters. In long exposures these defects spread saturated charge up and down their columns, the extent of the spread being roughly proportional to exposure time. * Several other lower intensity hot pixels were seen in bias frames and showed bright trailing charge. Many more were seen in longer dark frames c) PROBLEMS REVEALED DURING TESTS: * The intensity of the hot pixels was a function of exposure time and even after the operating parameters were optimised more than 80 are still seen in 2000 second 160K dark frames. They are also a strong function of clock voltages and unfortunately the voltages required to provide a reasonable well size (see below) produced many more hot pixels than was initially hoped. * At least two of these defects are unstable in that they were very prominent the first time the CCD was cooled and powered up. Then they were so intense that the maximum exposure time would need to be limited to less than 300 seconds otherwise the saturating charge would reach the readout register. However, these defects practically disappeared on warmning, pumping and re-cooling the dewar and were even too weak to cause a trail of charge. Unpredictably, they have since re-appeared after a third re-cooling but not with the intensity seen after the first cooling. * Extreme background brightening results when the readout register is overloaded with excessive saturated charge, and no adjustment of operating parameters was able to correct this. By restricting the image area well size to 150Ke- the saturated columns from the defects were reduced to below the threshold level of readout register overload and the brightening was thus eliminated. * Well size was remarkably affected by the particular combination of parallel shift phases held high during integration. Only two combinations permitted the full 145Ke- well size. The mystery is why the poor well size of the remaining four combinations, all of which are sequenced through during a parallel shift, doesn't degrade the well size in the process of reading out the image. * Unexpectedly, the pixel well size was considerably different between the two halves of the CCD, the difference ocurring abruptly midway down the CCD. Clocks providing a well size of 145Ke-/pix over the upper half produced only about 90Ke-/pix on the lower half. Clocks with a much greater swing were required to raise the well size of the lower half to 145Ke-/pix. As a result the number of hot pixels was increased quite considerably. * QE tests using diode mode at room temperatures showed a very poor QE with results only about 2/3rds those measured cold by Lick. As the CCD was cooled to 160K the QE steadily increased ultimately attaining values close to those of Lick. This is contary to Tektronix CCDs that show a fall in QE as they are cooled. The QE shortward of 950nm continues to imcrease as the CCD temperature is reduced, as measurements at 150K have recently shown. * The output amplifier of the CCD chip was found to have a slow recovery after resetting saturated pixels and this caused subsequent pixels to be measured low by a few 10's of ADUs in the worst cases. To avoid the need to add several (about 6) microseconds to the conversion time of each pixel to enable the amplifier to settle, the order of serial shift and amplifier reset was reversed within each pixel time so that the required settling time was provided with only a 1.5 microsecond increase in pixel time above that required for MITLL2. * Hot top rows in the CCD image area occured when the serial register was left runung during long (dark) exposures. Bright centres of "illumination" appeared across the first few lines to be read out and these peaked 7000e-/pix with many over 1000 e-/pix and a lot above 100 e-/pix. The effect was completely eliminated by stopping the serial register during exposures. * A brightening in the bias level for the first 1 - 200 top rows of images was found only when the CCD was strongly illuminated for bright flat-fields frames. The top few rows have about 20 - 30 e-/pix in them and this reduces quickly, but it needs one to two hundred to completely disappear. This brightening occurs only when the serial register is left in a stopped state during these bright exposures and is completely eliminated if the serial register is instead left running. * Last row defects occured that showed as a heavy band of charge trailing into into the vertical overscan starting at the last row to be read out of the CCD. These caused correspondingly broad bands of faint charge to be smeared over the image area resulting from the chip cleanout. Most of these defects were elimated but some still appear. They seem to be a function of chip illumination and can show up after periods of over-illumination of the CCD, i.e. after changing shutters. d) RESULTING OPERATING RESTRICTIONS AND COMPROMISES * Due to the long recovery time after after resetting a saturated CCD output amplifier, the clocking has been re-organised as described above and as a result binning in the row direction is not possible - it just doesn't work. * To overcome the readout register overload problem, described above, the upper limit of the image area well size needed to be carefully controlled to ensure that the saturating columns from the major defects were below the threshold of severe readout register brightening. Although this adjustment eliminated brightening in unbinned images binning in the column direction will still overload the readout register and generate a brightening of the background of images. This brightening is caused by actual electrons and the shot noise thereof renders column binning useless except for quick looks and perhaps high background imaging. * Since the remedy for the brightening of the bias level over the few few rows causes the hot top row problem, one or the other of these failings had to be tolerated. Bright top rows were considered more tolerable than hot top rows since the former have so far appeared only on bright flat fields whereas the latter appear on all exposures of any reasonable length. * When measured intensity divided by exposure time is plotted against exposure time, a system with no linearity problems produces a straight line of slope zero. On CCDs of low sensitivity (GEC, Thomson, Tek) non-linearities show roughly as a straight line with non-zero slope. This non-linearity is fairly well corrected by the use of the ALPHA correction coefficient applied to the solution of a quadratic equation. With the high sensitivity MITLL chips, the plot made from data covering the full 150Ke- well size appears as an inverted parabola and ideally the linearity should be corrected by the solution of a cubic function using, presumeably, two correction coefficients. The continued use of the ALPHA correction coefficient results in a comparitively poor correction of the non-linearity on NONASTRO and FAST readouts. Perhaps this needs more work..... * The trailed charge from hot pixels which appears to start up unexpectedly in windowed images placed below (after) the hot pixel is explained as follows. It is the result of fast cleanout being applied to get rapidly down to the window making the trailed charge from the hot pixel too small to be seen. However, once the window is reached the slow readout is able to generate trailed charge that can be seen. This causes the trail to start within the lowered window. The start of these trails lack the bright head of the hot pixel and if the number of real parallel shifts at fast readout is small the bright head is seen disconnected from the trailed charge within that window. * The increased parallel clock voltage swing, required to properly correct for the poor well size in the lower half of the imager, caused an increase in the intensity of existing defects and the appearance of new defects. e) CONCERNS * The major concern is that hot pixel intensities appear to be somewhat unstable from one cooling to the next and perhaps even change with strong illumination. Two very strong defects almost disappeared after the first warmup only to reappear after a third cooling. * If, oneday, a severe backward step occured resulting in very intense hot pixels, the CCD may be rendered unusable due to the readout register overload problem. A reduction of image area well size may well be required to get the CCD back to a useable condition. * * * * * TIME WILL TELL * * * * * *