The noise in G5 seems to be a function of the leakage current rather than the temperature. This first figure shows the Front Slow LLD rate (noise) vs. time as the spectrometer is first warmed up by about 6 degrees then cooled down again. The other two traces are the front and rear reset rates. Notice how leakage current leaves the front and goes to the rear as the temperature changes.

This is an ambiguous result: the noise could be dropping because the temperature rises, or because the front segment leakage current drops.

But notice what happened during the rampup of the high voltage (second figure). For a short while after segmentation, the noise counts for G5 were quite low, then as the voltage ramped up from 3500V to 4000V, and the leakage current started shifting from the rear to the front, it got noisy. We can make the detector quiet down at any time by bringing the voltage back closer to segmentation.

Notice that the well-behaved detector (G1) is quiet from segmentation onwards, and that the other bad front segment (G3) is bad from segmentation onwards. G3 might be cold, but it might be noisy for any other reason, too. Is it worth it to measure the temperature of the G3 FETS?