Signal, Bias and Saturation

Starting to look into well capacity and relinearization a bit.  Curtis collected a series of non-destructive scans staring at the cold blackbody source at a variety of temperatures.  The result for the median detector pixel is show in the plot below:

In each of these curves 3s separates one data point from the next.  Flux here is broadband, with no intervening optical elements between the blackbody aperture and the array.  Haven't done a flux calculation yest, but we could do a QE estimate from this data as well.

Up to about 30K there isn't much coming off the blackbody above whatever the background light leak in the dewar is (~80e/s).  Signal is pretty linear for the 35K curve, and we appear to be clipping at the source-follower for the 40K and 45K curves.  We'd need to drop the detector bias below 2V if we wanted to get a smooth debias fit to the data.

The right-hand axis is scaled to give an idea of equivalent electrons, and the swing appears to be about 300k electrons.  Since it will cost us some responsivity to deepen the well further, we should consider how big a background we want to consider handling.  If we ran at -2V (or even -3V), we could conceivably stay reasonably linear up to 300k electrons, which at a 2Hz scan rate and 50% quantum efficiency would work out to something like 1.2M ph/pix-s.  Is that enough?

Noise v. Detector Bias

Noise measurements using all-destructive scans have been collected over a range of detector bias values.

Values reported are measures of noise obtained though the 'many pixel' method used at UofR.  A set of 200 all-destructive reads is obtained at each bias, with consecutive scans differenced and analyzed to yield an areal standard deviation.  This sdev is divided by root-2 and reported as read noise here.

FWIW, the plot below shows a not-terrible exponential fit to the data.  The blackbody was switched off during this measurement, but there is probably some low-level light leak in the test dewar so it may be worth spot-checking this data when the device is taped off for a dark current measurement.

Bias values are relative to VDDUC/VRST setting of -2.221V.  VDET values are more negative than VDDUC/VRST, but biases are plotted as magnitude.  

Relevant log pages:  EXES Vol.3, pp.94 - 95

Data collected at T = 7K, with a three second scan time.  Values are in raw DN (1DN ~ 35 electrons)