SD15 AFE Exposure versus ISO setting

I tested my SD15, inspired by this test by new member @calebistaken on a Quattro:

https://www.dpreview.com/forums/post/62420091

Shot a target on the wall, constantly lit by warm LEDs at 8-1/3 EV lighting value. Used Aperture Priority, f/4 with no EC. Shot between 50 and 3200 ISO, doubling each time for total of 7 shots. Note that 50 and 3200 ISO are termed 'extended' settings. I'll call the others 'native'.

Examined the raw histogram of each shot's green layer in RawDigger - with the levels shown in EV relative to a reference value of 2400. (the SD15 sensor saturates at just below 2400 - a surprise to anybody who thinks that value should be 4095).

The SD15 turned out to be just like the Quattro:

Native ISO and Level EV:

100: -1.4

200: -1.4

400: -1.4

Here we see that the programmable gain amplifier (PGA) changes as expected to keep the ADC output constant in conformance with the lighting value and aperture. This does NOT mean that the sensor exposure remained constant - because the shutter speed did change proportionally.

Continuing up to 1600 ISO:

400: -1.4

800: -2.3

1600: -3.2

Here we see that the PGA gain does not change, because the raw levels are approximately halved for each increase in ISO. This behavior is identical to that of Merrill and other non-AFE models, I believe.

Extended ISO.

Turns out that, for 50 ISO, the sensor exposure going from 100 to 50 does NOT double! Instead it goes from -1.4 to -0.9 EV, about half a stop. No idea why but at least 50 ISO doesn't throw away a whole stop of so-called headroom by so doing.

For 3200 ISO, the sensor under-exposure continues downward to -4.1 EV. At that setting, the raw and the embedded image noise is horrible, with a SNR between 1 and 2 in the shadows. SPP 5.5.3 however does a reasonable job with that:

Initial image (embedded) versus loaded image (converted) ... ignore Capture Date

Knowledge is power ...