One of the things I've been pondering over the past few weeks is, which is preferable: high ISOs or IS? This issue seems to divide enthusiasts, and I myself have firmly taken sides—I love IS, as I've written ad nauseam. But why do I? Am I really justified? Of that, I'm not so sure. Does it suit my own particular way of working? Does it address shortcomings in my technique? (Or my liking for coffee?) Or do I just like the idea of it, somehow?
I want to caution you against putting too much stock in tonight's little trial, which I'm about to relate. There are far too many uncontrolled variables here to claim that this proves anything. In fact, I suspect it's pretty much impossible to level the playing field and make all these variables objectively fair. But I haven't tried very hard: I'm not even using lenses with roughly equivalent angles of view (although they're both the same focal length).
Here's the shot that started it all, which I admit I thought was enormously impressive when I took it. This is the Nikon D700 with the AF-Nikkor 35mm ƒ/2 at ISO 6400. All of the following are screen shots from ACR, and in each case I shot several shots and picked the best one.
If it looks like it was taken at night, that's because it was. This was a good three hours past sunset. The exposure ended up at 1/60th at ƒ/4.
Then I took the Pentax K20D and 35mm ƒ/2.8 DA Macro and set the ISO at two stops slower, ISO 1600. I used the same aperture, but slowed down the shutter speed by two stops, to 1/15th, to compensate. But I turned SR on. (That's "shake reduction," what Pentax calls IS.)
You can see the angle of view is narrower due to the smaller format. Both shots were handheld.
Here are a few 100% comparisons: in all three cases, the D700 shot is on top and the Pentax detail is on the bottom. (Sorry if I didn't manage to grab exactly the same areas. In all three cases I tried to grab roughly the same pictorial area, not the same size sections.)
To me, it looks like the K20D at 1600 compares favorably to the D700 at 6400—it has better sharpness and more detail. This looks like a win for SR to me, for this kind of shot—a static scenic in extremely low light.
I have to admit I'm a bit disappointed that the Nikon didn't do better here. Here are a few things that could be going on: Maybe 6400 is past the sweet spot for the D700 (that is, maybe the D700 would beat the K20D by more than two stops if there were more light and I were using both at lower ISOs). Maybe the newer (c. 2007) Pentax lens is better than the older (c. 1987) Nikkor (I did give the Pentax DA Macro a rave review when I wrote about it). Maybe I'm better at hand-holding 1/15th with IS than I am at handholding even 1/60th without it (I think it's pretty obvious that this is true). Maybe noise reduction and sharpening would change the results or at least ameliorate the differences. Maybe I messed up the test in some other way I'm not aware of.
Here's what I'd do to make the trial a little more carefully: I'd use a 50mm Nikkor to make the angles of view more similar; I'd use a vintage 35mm on the Pentax to make the lens comparison more fair; I'd make duplicate shots on a tripod, focus bracketed, as controls; I'd photograph in a little better light to make the shutter speed of the Nikon 1/125th instead of 1/60th; and I'd use 3200 and 800 instead of 6400 and 1600.
More to follow (well, it's a blog—there's always more to follow). Stay tuned.
P.S. There's one criticism of this test you shouldn't make, and that would be to say that I could just use VR on the Nikon. Nope, because Nikon does not now and never will put VR in any of the lenses I want it in—not a chance. That's why I like IS built into the camera body, because it's usable with all lenses—including the ones I like. So the Nikon's only hope of competing is with higher ISOs alone.
Okay, here's another test, in which I tried to be more rigorous. For this one I used vintage, manual focus lenses on both cameras—a Nikkor 50mm ƒ/2 AI and an SMC Pentax-M 35mm ƒ2.8. The Nikkor was used one stop down and the Pentax lens wide open, which would give the optical advantage to the Nikon.
And in this case I compared the Nikon D700 at ISO 3200 and 1/125th with the Pentax K20D at ISO 800 and 1/30th, SR on. Both shots were corrected (well, sort of!) in raw but what you see here are screen shots from ACR. In each case I focus bracketed, and took seven shots and picked the best one.
Again the angles of view do not quite match, but they're closer.
100% comparisons, Nikon on top, Pentax below:
A much closer call this time. Having pored over both shots in various areas at various percentages, I'd again have to give the nod to the Pentax, but the Nikon is clearly holding its own. In this case, having shot the Nikon at 1/125th, there is no sharpness advantage to SR; noise and detail are ever-so-slightly worse in the Nikon image, and sharpness is better—the latter probably a consequence of the somewhat better lens stopped down one stop vs. the Pentax moderate WA used wide open.
From these trials I'd say that the Nikon almost but not quite maintains a two-stop advantage over the Pentax in terms of high-ISO performance—call it a stop and two-thirds to be on the safe side—as long as the ISO 6400 setting is not used. This accords just about perfectly with my sense yesterday that ISO 2500 would be a sensible setting on the D700, 2500 being a stop and two-thirds more than 800, which in turn is a very usable speed on the K20D.
Featured Comment by Ry H.: "Well, either one of those photos viewed at that close crop appear more like oil paintings than photographs to me, which is probably a big reason why I try not to pixel peep too much. That said, the Pentax K20D and that old M 35mm ƒ/2.8 lens seem to acquit themselves pretty well, especially considering that that particular lens is pretty pedestrian for a Pentax prime.
"On a different note, while I have seen a few examples of excellent shots which were only possible with a high ISO, I tend to believe that there's a reason photography is known as 'painting with light'— it doesn't work so well when the light is lacking. If people are using ISO 12500 and are getting good pictures of their children playing in a dark room, all the more power to them. Myself, I'll wait until morning."