FIG. 5. Read noise vs temperature in SRP mode, R-G mode, and Fowler mode with one sample pair and Fowler mode with 32 sample pairs <strong>for</strong> Lawrence 2m, Lawrence 3 m, standard cryo-CMOS, and TRW 2 m muxes.is about 4 e or 27 V. Below 18 K, there was some instabilityin the Fowler modes. Lawrence 3 m material was thenext best, although it has somewhat higher noise and moreinstability in the SRP mode. Lawrence 2 m and Lawrence 3m materials originate from the same manufacturer, differingonly in epitaxial thickness. Standard cryo-CMOS materialwas well behaved in the SRP mode, but shows substantialinstability in the Fowler mode. We compared itsper<strong>for</strong>mance to CRC-590 and 644 muxes we have testedearlier also standard cryo-CMOS and they are roughlycomparable. TRW 2 m material per<strong>for</strong>med well in the SRPmode. It was reasonably good above 15 K in Fowler modes,but showed an annoying noise peak at 25 K. Below 15 K, theinstability in the Fowler mode was a little larger than <strong>for</strong> theLawrence materials.IV. FPA TESTSSBRC indium bump-bonded low doped (1.7– 2.010 14 cm 3 ) and undoped (1.010 14 cm 3 ) FirebirdInSb to CRC744 muxes <strong>for</strong>med from Lawrence 2 m andLawrence 3 m materials. The best six FPAs using CRC-744 muxes and InSb detectors have been tested at the University<strong>of</strong> Rochester. Tests <strong>of</strong> noise, mux stability, and QEwere made as the Dewar warmed up from 5 to 50 K, afterexhaustion <strong>of</strong> the LHe supply exactly as described <strong>for</strong> thebare mux tests. For noise and mux stability tests, the detector3570 Rev. Sci. Instrum., Vol. 68, No. 9, September 1997 IR <strong>focal</strong> <strong>plane</strong> <strong>arrays</strong>Downloaded 21 Jun 2012 to 128.151.144.191. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
FIG. 7. The reset voltage measured at the output amplifiers vs the resetvoltage applied to the integrating node at 14.5 K <strong>for</strong> FPA CRC 744-40716.FIG. 6. Instability in Fowler mode with 32 sample pairs.viewed our cold dark slide. For the QE tests, data were takenthrough a 3.3 m filter viewing the laboratory. It took about30 min <strong>for</strong> the <strong>arrays</strong> to warm up from 10 to 20 K duringwhich time 12 data sets were acquired. More detailed andaccurate tests <strong>of</strong> read noise, QE, and dark current were madewith the temperature stabilized around 14.5 and 28.5 K. Theoperating voltages and currents <strong>for</strong> the <strong>arrays</strong> are presentedin Tables II and III.The FPA per<strong>for</strong>mance was more stable than that <strong>of</strong> thebare muxes. They showed smaller differences between twosuccessive frames in the same mode and smaller Fowler‘‘bias’’ frame values than the bare muxes. Most <strong>of</strong> the FPAsshowed low noise, small dark current, and high QE. TheFPAs made from Lawrence 2 m muxes had hundreds <strong>of</strong> e <strong>of</strong> dark charge in the dark images, which caused unexplainednonlinearity in dark current versus integration time curves. Inaddition, some FPAs made from Lawrence 2 m muxesshowed a quite wide distribution <strong>of</strong> zero bias points, whichrequired operation at very high applied detector bias 800and 1250 mV. There<strong>for</strong>e, the FPAs made from Lawrence 3m muxes were judged to give better overall per<strong>for</strong>mance.Among all the FPAs, CRC744-40716, which has aLawrence 3 m mux bonded to a Firebird low doped InSbarray, exhibits the best overall per<strong>for</strong>mance. Its best readnoise <strong>of</strong> 5 e was obtained using the Fowler sampling modewith 64 sample pairs at 15 K. The dark current images <strong>for</strong>this array were reasonably uni<strong>for</strong>m and there were only a fewhot pixels. At both 14.5 and 28.5 K, the 3 upper limit to thedark current was 0.2 e /s/pixel. The average quantum efficiencyin the whole array is 87% at both 14.5 and 28.5 K.The central part <strong>of</strong> the array has somewhat lower quantumefficiency 78% at 14.5 K and 82% at 28.5 K and it decreaseswhen the temperature is reduced. This behavior isanomalous, and is under current investigation. The powerdissipation and load driving capability at 14.5 K was alsomeasured <strong>for</strong> this array.The following discussions will concentrate on CRC744-40716.A. CalibrationThe dc gain <strong>of</strong> the multiplexer was determined from theslope <strong>of</strong> the reset level measured at the output amplifiersversus the reset voltage applied to the integrating node seeFig. 7. At both 14.5 and 28.5 K, it was 0.87 throughout thewhole array.To calibrate our voltages to the charge collected on theintegrating node, the noise 2 versus signal method was used.The read noise in Fowler mode with one sample pair and thesignal were measured in three 2525 pixel boxes. The slope<strong>of</strong> the noise 2 versus signal gives the system gain in ADU/e .This in<strong>for</strong>mation, in conjunction with the dc gain measurementwhich gives the conversion ADU/V, determines thetotal capacitance <strong>of</strong> the integrating node.The noise 2 versus signal curve was measured at both14.5 and 28.5 K. Figure 8 shows the data obtained at 14.5 K.The capacitances calculated within three 2525 boxes centeredat column,row(x,y)(50,50), 128,128, and225,225 are summarized in Table IV. Since the capacitanceRev. Sci. Instrum., Vol. 68, No. 9, September 1997IR <strong>focal</strong> <strong>plane</strong> <strong>arrays</strong>3571Downloaded 21 Jun 2012 to 128.151.144.191. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions