Dynamic Range (Bill Claff): d3s-d4-d800e-5d3

Started Mar 15, 2012 | Discussions thread
Pierre Sottas Contributing Member • Posts: 787
Re: That's not "more realistic"

Marius D wrote:

There is no "file cooking", it could be explained by a change in sensor design. They added a voltage reference control unit for ADC. Under 800 ISO, lower ADC voltage reference is probably used to obtain intermediate ISOs, instead of using a second stage amplifier. With 35% voltage drop they get 2/3EV ISO gain. Above 1000 ISO, when first stage amplifier reach his max. gain, they progressively reduce the ADC voltage reference. A lower voltage reference can improve read noise characteristics and could explain the "convex" DR curve. Under 800 ISO, all samples are at "full" ISO steps, where voltage reference is at nominal value (no 2nd stage gain). This things could probably be seen on a graph with all intermediate ISO values (1/3 steps).

Adjusting voltage reference to provide gain would certainly help at low ISO and explain the shape of the curves. However I did not follow the last advances in CMOS technology these last years and I do not know if voltage reference can be implemented on the Exmor on-chip ADCs. Also I'm not sure that data obtained at intermediate ISOs will really help here because both amplifiers can (in theory) work in parallel at each 1/3 step.

I also played a bit with some NEF files in Matlab and found similar results, with the D800 having about 0.8-0.9 stops more "engineering DR" than D7000 at low ISOs. This is a bit less than what Bill found (especially at high ISOs) but this remains nevertheless significantly higher than my initial guess. And if we add the nice color separation, the D800 is an impressive camera.

Pierre

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow