Digital Sensitometry for Black and White ~ High(ish) ISO dynamic range

[Updated 7 October, 2024 ~ Important correction: Zone 5 at 0EV is 76hex/118decimal, not 7F/127 as I previously wrote.]

Scrolling through various cameras tested on Photons to Photos I couldn't help be notice that the dynamic range decreased as ISO increased.  To see for myself how this worked in relation to B&W, EV steps, and the Zone System I grabbed an old Sony NEX-7 and a newer(ish) Sony A6300 and set them both to ISO 1600.

Here's what I see.

The Sony NEX-7 looks pretty good down to -1EV.  Then the curve takes a sharp hook to the left.  I take this to mean the noise floor of this sensor at this ISO has been reached.

From the Sony A6300 I can see the application of advances in sensor performance.  There is a difference between the 1600 and 100 ISO X positions, but the slopes are fairly equivalent.

Interestingly, I see more dynamic range in my A6300 than reported on Photons to Photos.  There are 12EV from pure white to pure black.

So... why is that?

Well, it has to do with how I measured luminance at each EV step.  Photons to Photos is measuring  pixel to pixel luminance differences and making a different calculation than I.  In my case I use a 20pixel brush to measure luminosity, which averages out noise.

How can I justify this?  Easily, it turns out.  

When we view a photograph, even with a lot of grain (film) or noise (digital), our eyes average out luminance differences and we can see an overall tone.  Sure, if we stick our noses up to the photo we'll see grain or noise, but step back to a "normal" viewing distance and we see things a little differently.

Why is this important to me?

I'm often in dark museums where my little cameras really crank the ISO up.  I've noticed in practice that the dynamic range "felt" broader than reported.  Now I understand a little better why.  Which is good.  I really don't want to spend more money on a camera that might suit just a single purpose (ie: one with reportedly excellent high ISO performance).

Lastly for this blog entry, each high ISO input curve can be tailored to map directly into the linear 1EV step per division Zone System.  The curve shapes are a little different than those at low ISO's, but not terribly much.