1、1UVIT detector system Engineering Model test and calibration.J. Postma and J. HutchingsUniversity of Calgary, and NRC Herzberg InstituteThis report describes the test and calibrations carried out at Routes, DFL, and the University of Calgary, in October - December 2008. The scope and aims of the wor
2、k are laid out in the calibration test plan. The data are stored and available for further work, and the test details are recorded in spreadsheets. The results and images in this report were generated during the calibration period and subsequently. The EM was shipped to IIA and tested there in Janua
3、ry 2009. The EM uses an NUV photocathode, and all work was done in the wavelength range 190-270nm.The box below gives the overall summary of the work. This draft March 23, 2009. 2Figures 1 to 4 show the Calgary vacuum facility used, and photos of the hardware in the facility. Figure 5 shows the GSE
4、display, in the calibration facility area. Fig 1. Schematic of the Calgary facility. Only the UV arm was used, with a deuterium lamp. Fig 2. REA box in clean area outside tank3Fig 3. Vacuum tank with lid slightly raised. UV light channel is on the right.Fig 4. EM detector and HVU (below) inside the
5、tank, on the movable mounting platform. Light port is on the right. 4Fig 5. GSE control and display screen.DFL dataWe first show data obtained at DFL during the TVAC testing. Figure 6 shows dark frames (entire CMOS image raw readouts), displayed as plots of all 512 individual pixel rows. The initial
6、 row is offset to a lower value, but the rest lie at approximately the same level. There is a signal ramp along the rows, rising to the right, which is present in all data. The photon events are superposed on this, and proper event centroiding requires subtraction of the local background. This is me
7、asured, as described in Postma et al 2007, as the minimum corner value in the 5 x 5 pixel box centred on each photon event. Figure 6 displays the dark pixel rows at cold and hot temperatures, and gives some statistics of the images. The behaviour does not change significantly with temperature, but a
8、 number of hot pixels are seen in the hot images. The hot pixel values are much lower than the threshold values used for event recognition, and do not affect the operation in any way. The dark images are referred to as background or bias images. Figure 6 shows the darks for 3 different gain settings
9、. These scale well with gain and show no problems for use in photon counting. 5Fig 6. Dark frames at cold and hot temperatures.Fig 7. Dark frame data at 3 different gain settings. Normal operations will be at settings 1 or 2. 8 is the maximum. 6Fig 8. Dark single frame on left. Single frame with une
10、ven light illumination on right.Figure 8 shows individual dark and light frames. The dark frame contains single bright pixels that will not be identified as photon events, which are seen in the light frame. Fig 9. Left: event total signal for 3 algorithms. Right: Event maximum pixel signal. 7There a
11、re 3 templates for event centroiding, as described by Postma et al (2007). They are designated 3cross (3C), 3square (3S), and 5square (5S). Figure 9 shows the total signal for the same events, within the three templates. As most of the event signal lies within the 3C template, little more is added i
12、n the others. There is a spread of event signals, and a threshold value is set by inspection of these distributions, which also contain a spike of low energy events which are noise and not real photons.Events are recognized as maximum pixel values surrounded by 8 lower values, and a separate thresho
13、ld is set for this value, from the distributions in the right panel of Fig 9, again in order to eliminate low signal non-photon events. These events are shown with gain setting 1. Gain setting 2 will double the signals. As the A/D converter saturates at 128000, gain 2 is useable here. Should the det
14、ector response drop during use or ageing, use of higher gain may be helpful. Fig 10. Flat fields with the three different templates L to R: 3C, 3S, 5SFigures 10,10b shows the resulting image accumulating many frames of data with uniform illumination of the detector. The nature of the centroiding alg
15、orithms leads to systematic errors, which underpopulate the CMOS pixel boundaries, as discussed by Postma et al, and demonstrated well in these EM data. While corrections are possible for the 3C and 3S images, the 5S results give a fairly good result without correction. The value of the smaller temp
16、lates lies in being able to detect more photons separately in crowded or bright fields of view, and will be selected beforehand, with the scientific aims in mind. All algorithms use the corner minimum of 5S for background level estimation. 8Illumination with low and high count-rate sources was done
17、to measure these. The low count rate source was fibre optic UV feed into chamber, not shining directly on to detector. The high count rate source was DFLs vacuum chamber mass spectrometer, likely shining directly on to detector.Algorithm Low Rate1 High Rate23x3 Cross 19.72 c/f 1411.58 c/f3x3 Square
18、21.82 c/f 1356.91 c/f5x5 Square 21.63 c/f 1215.40 c/f1. At this low of count rate, the lower rate for the 3x3 Cross is likely due to the shape energy being less for that algorithm, thus less able to satisfy the energy threshold. In the plots of Event Count vs. Frame Number below, you can see the cos
19、mic hits superimposed. Cosmic counts rate in space can be calibrated in orbit, and as these should be relatively constant, can be corrected out of the photometry. Also, all voltages were run at nominal settings: Anode = 5000V, MCP = 2020V, Cathode = -400V, full frame (512x512) at 28.7Hz.2. The MCP v
20、oltage was lowered to 1950V for this bright source (mass-spec). Imaging at 2020V would BOD soon after the voltage ramped. Cosmic hits are not noticeable at this count rate, but in space the measured cosmic count rates from in-orbit calibration viewing a dark (background) sky can be used for photomet
21、ric correction. We describe low MCP voltage operations in more detail later. Fig 10b. 1/32 pixel resolution cuts through 3C-centroid flat fields. X-cut on the left and Y-cut on the right. 9Fig 11. 3C images at low count rate, clockwise: image at 1 pixel resolution; minimum corner values (i.e. the bi
22、as image); number of photon events per frame for 55000 frames; the max-min corner values (high points indicate double-photon events). 10Low MCP voltageReducing the MCP voltage reduces the signal gain, and this allows exposure to brighter light without draining the MCP charge and shortening its lifet
23、ime. We did extensive calibrations of this, reported in the light-spot data later in the report. Figure 12 shows the reduction on whole-field counts under uniform low level illumination, with MCP voltage variations. This offers an order of magnitude extension of dynamic range. Fig 12. Photon events
24、counted as function of MCP voltage, under low level flat-field illumination. As there is little crowding of events, all three algorithms have the same response. Nominal operating voltage is 2020. Figure 13 shows the spreads of event signals for the range of MCP voltages used from the full-field low illumination case. These plots show only the events above the threshold that eliminates non-photon faint events.
