FYI I just tried on a DIGIC4 camera (at wide angle, with the lens covered), the JPEGs are black, the DNG doesn't show any circular shape. ISO80, 1s and 1/125s. It's a bit newer than your S90, they may have reduced this effect on dark images, who knows.
Very kind of you to do the test. In general, lens-covered fast shots at low ISO you can expect JPEGs to be completely black so they're not a good test method for this // that's been my experience. If you have an open CCD like the 101a, then flooding with uniform light does show the donut in a normal exposure with a JPEG. Hence the only consistent way is to do this test on a regular camera is with DNGs and CR2s, but you have push the tone curve very hard left, especially at low ISO. I noticed that this effect, and its behavior under different ISOs too, is quite different between the 101a and 100c, so it's not surprising you don't see anything obvious (did you push the tone curve completely left to get a bright image, and it's uniform ??).
Can you re-do the test on 100c, with a stock CHDK 1.1 build (in order to enable the firmware's lens distortion correction)?
Yes I thought of this since the 100c does have distcorr disabled (you have a good memory!), and there could be some influence ... makes sense to do this.
Does temperature have any influence on this? Where is the CCD temperature sensor of your 101a camera? Is that cooled as well?
I've been doing a lot of different temperature tests recently. In general the answer to your 1st question is no, but with a "but." The thermistor is right on the CCD chip itself and is accurate (however FYI, I've determined that the reading CHDK presents is ~12C too high // I had a discussion about this with reyalp long time ago and now I am 100% convinced the reading is too high by that amount). CCD temperature only makes a big difference for long exposures, and higher ISOs to some extent. Some noise sources are: dark noise, photon noise, CCD-FDA noise, current source noise, power supply noise, CDS S&H capacitor switching noise, electromagnetic interference (from switching power supplies), ADC quantization noise, readout noise, etc. At 1/125 regardless of ISO, the dominant noise is the sum total of "electronics noise," which is pretty much everything above except dark noise. Dark noise starts to have a dominant effect in longer exposures and increased CCD temps. Hence my 1/125th darkness shots on the 101a yield approximately the same noise histograms whether at -20C or +30C. You could try increasing ISO, say to 400, making the electronics noise "brighter." However, I noticed that at 400 on the 100c the effect is not obvious, but it is obvious on the 101a. So yes, putting a stock CHDK 1.1 on the 100c should shed some more light.
