(commentary)
Someone asked me why I haven’t picked up a Sony A7s yet. My answer startled them: I see really high ISO values as mostly useless.
Say what? I mean we can now take pictures in the dark, right?
Sort of.
Here’s my problem with ISO values of 12,800 and above: we lose dynamic range with each ISO boost plus we’re in light where the random nature of photons will certainly come into play. We also lose color fidelity, with colors blocking up and going darker. But most of the types of pictures I want to take in “low light” actually have light sources in them. So I'm in a Catch-22 situation: if I set exposure to keep light sources from blowing out, I push the poorly lit portions of the scene I’m trying to capture into the low values where there isn’t a lot of definition due to the few number of bits being used, and thus the dynamic range and shot noise coupled with rounding errors then makes very muddy images as a result. If I set exposure for the mid-range in the scene, I’ll have a lot of blowouts. Remember, our eyes will always go to the brightest areas in a image. Always. So I’m basically forcing my viewer to look at paper (or screen backlighting).
I suppose if I were a private investigator shooting at night I wouldn’t care. Let the streetlights and other light sources blow out, I just want to see the face of the person I’m photographing. But I’m not a PI, I’m more of an art photographer who values very high technical optimization.
The problem is that the higher we push current sensor technology into low light, the more we see the problem of too much dynamic range in the scene for the sensor to capture. At ISO 12,800 we’re typically worse off than the earliest sensors were in terms of dynamic range at base ISO, so we’re years behind state-of-the-art in our image’s technical aspects, just at a higher ISO value. (Update: since someone took me to task on the way I originally wrote this statement, I’ve reworded it; also, let me point out the word “typically” as well as the fact that I’m using my measurement practices for consistency, not someone else’s, such as DxO's.)
That’s not to say that the problem can’t be solved. It can, by a new and different sensor technology. Basically, you need to break the other end of the electron container to fix the problem: you need to lift the electron well size (increase saturation capability), not lower the noise floor. That solution would also apply to my normal landscape shooting, too: HDR would be a thing of the past (though HDR-like processing techniques would still be required to move the captured data appropriately in the more limited range we have available with current output technologies).
I should note that the A7s does increase the electron well size somewhat. That’s one of the things that usually comes from larger photosite sizes compared to smaller ones (all else equal). But what I’m talking about in the previous paragraph is a radical change, probably one that would involve emptying the electron well and setting a rollover bit when it gets full. That’s the only thing that is going to give us a huge dynamic range increase from the sensor at this point.
I’m sure there are shooters who’ll love the Sony A7s. But it’s not for me. Indeed, the D4s and Df are rarely for me, though there are times shooting sports when they are (mostly to keep shutter speeds high in poor lighting).
One other thing I was reminded of when shooting sports in low light a couple of weeks ago is this: with my Retina display and 100% view I can certainly pixel peep and see noise. Yet when I printed those same images at 13x19”, I didn’t see that noise. You certainly don’t see it in the down sampled images I posted on the Web, and you wouldn’t see it in newsprint or even magazine reproductions. So if I don’t see it in output, why am I overly worried about it? ;~)
Don’t get me wrong. I want some high ISO flexibility. But in evaluating what we have now versus where we’ve been, I’d say that the D3s probably had enough of everything for what I’d need a “low-light camera” for. The D4s, 1Dx, A7s, and whatever comes next may push further, but they push further in a way that doesn’t really benefit me all that much. It might benefit you, but more and more I’m doubting even that. A new sensor technology would help us more than pushing the current technology further, I think.
This is just another of those “good enough” problems that the camera makers are fighting. Very similar to the way people stopped upgrading their personal computers when they had enough CPU speed and RAM to do the basic tasks they were performing—mostly Web browsing, email, word processing, and smallish spreadsheets—there wasn’t a need to buy a new computer every year.
We’re now past that same point with cameras. Even for some pros. As I noted, a D3s was probably all I needed for the sports and high ISO work I do. I just don’t have jobs where ISO 12,800 is needed.
