Sorry, I didn't see your post until today.
I suppose in a sense one could describe all ISO settings in digital cameras as fake, or bogus. The sensitivy of the light gathering capacity of the photoreceptors does not change with changing ISO settings.
Not of the photoreceptors, but as far as the product of the camera is concerned, ISOs can be very different in the shadows due to read noise differences. Trying to use the bottom 1/16th of the sensor's range at ISO 100 instead of at ISO 1600 leads to a relatively poor capture with some cameras.
At a fundamental level, there is only one ISO sensitivity in a digital camera and only one best and correct exposure for maximum S/N and highest dynamic range.
All ISO settings higher than base ISO (which is ISO 125 for the Canon 5D and ISO 200 for the D3) merely represent the camera's attempts to compensate for underexposure.
The 5D is more like ISO 84. It actually meters and has enough headroom for the RAW data at ISO 120 when it is set to 100, but it has more headroom available in the sensor for its "ISO 50" (metered for ISO 60, and has only enough headroom to be a minimum of about 84).
When I set my 5D to ISO 3200, I'm really just sending an instruction to the camera's processor that subsequent shots will be underexposed by 5 stops. If I'm shooting RAW I don't have any control over the way the camera's built-in processor will compensate for such underexposure.
No, but I believe it does use a different amplification at the photosites for ISOs 50, 100, 200, 400, 800, and 1600, with a consequent different in absolute noise floor.
Another way of looking at this is that fundamentally the difference between the 5D and the D3 is that the 5D will only accept instructions (from the user) on how to deal with 5 stops of underexposure, where as the D3 will accept instructions on how to deal with as much as 7 stops of underexposure. (It's a more advanced robot ).
True, but in type, they seem to be similar; different ISOs by different gains (at the photosites) for multiple ISOs, unlike all compact cameras and most other Nikons and other brands of DSLR, which only seem to vary gain (if at all) before going into the ADC.
But I've got my own way of compensating for underexposure, so I think it's perfectly legitimate for me to say the following shot was taken at ISO 256,000.
I don't think you can really say that, because you collected photons multiple times. There is no real difference (except in read noise) between shooting one frame at 9 seconds and 9 frames at 1 second. Your ISO is 256000 divided by nine, if you took nine exposures. I have not yet investigated the difference between additive stacks and median stacks (I just have assumptions, such as median possibly being more accurate when you have a lot of wild, stray, extreme values in your system, and that additive potentially increases bit depth while median does not, meaning that median will be more likely to lead to quantization), but I am pretty certain that it is a bit of a stretch to identify the exposure index of a median stack by an individual exposure.
That's a total of 10 stops of underexposure, the first 5 handled by the camera's processing and the additional 5 handled by me and photoshop .
How is this possible considering the 5D only has about 10 stops of DR or less?
ISO 100 has that limitation; the sensor doesn't. 5D ISO 1600 goes about 2.5 stops deeper than ISO 100 for usable signal, in an absolute sense. If you could get the quality of readout at ISO 100 in its very deepest shadows that you get of the same absolute levels at ISO 1600, you would have an amazing ISO 100.
Then, as I implied, your exposure index is really the sum of the ISOs of the individual shots.
And let's not forget, most of these measures of dynamic range assume 100% pixel view, or nearly so. These DR values don't mean that you can't separate half a frame that is black from half a frame that is white, but 20 stops below sensor saturation by binning down to two pixels!
I noticed this defect when I tested my second copy of the 5D. The first copy had unacceptable banding in deep shadows. I returned the camera and found the second one was better regarding the ugly banding, but had this defective corner. I considered returning the second camera, but decided against it on the grounds that possibly all 5D sensors would have one defect or another. This was a budget Full Frame sensor after all.
I'm sure it could be calibrated out, but none of the common converters allow that, AFAIK. It's probably just a blackpoint issue, and with Canon's blackpoint offset in their RAW files, you could probably calibrate it out without losing anything near black.
The fourth image is the 100% crop showing a fair amount of noise. What would you say, film at ISO 800?
I don't remeber exact film qualities. Last time I shot film was 1976-1981.