Can you tell how effective a camera's high ISO images are from the specs?

Question

My current camera, a Canon 50D has OK image quality at 1600 ISO and its not really very usable above that.

I know that all the manufacturers publish an ISO range for their bodies. But just because the dial goes to 6400 or even 12800, doesn't mean that it works well at that setting.

Some of the newest camera bodies (at both the pro and serious enthusiast levels) provide amazingly great images at high ISO settings. Are there rules of thumb one can use, like two stops below the max published?

Or do you just have to rely upon reviews to validate the specs?

Answer 1

Generally speaking, I think the 1-2 stops below maximum "native" setting is a good rule of thumb. I use the Canon 7D, and its maximum ISO is 6400. The maximum reasonably usable ISO (except in the case of rather good light, which outside of the single case were you MUST have a certain minimum shutter speed, kind of eliminates the need for a very high ISO in the first place) is ISO 1600, maybe ISO 2500 (although ISO 2000 seems to be fairly bad). On my Canon 450D, which had a max ISO of 1600, the max usable ISO was 800...BARELY. For best IQ, I would rarely use more than ISO 400 on the 450D.

Today, with the likes of the 1D X, the maximum usable ISO seems to roll in around ISO 16000 based on many of the bird photographs works and reviews I've seen, which is a bit less than 2 stops down from ISO 51200. On the Canon 5D III, it seems that ISO 10000 seems pretty excellent, however above that things do start to fall apart, and 25600 is pretty noisy. Again, thats a bit less than two stops.

One of the better reviews of the 1D X that convinced me the 1-2 stops rule still seems to be in effect was by Andy Rouse, a renown wildlife and bird photographer. He is a Nikon user, but is a pretty objective guy. His 1D X review included a lot of owl photos. I think birds are superb subjects for testing cameras as they have lots of contrast and fine detail, and are usually photographed against smooth backdrops (which are generally the worst kind of regions for noise in any DSLR). The performance of the 1D X up through ISO 16000 is stunning....it looks as good as my 7D's ISO 1600, and when it comes to subject detail, even a bit better.

This all assumes the use of RAW. When it comes to JPEG, the story is very different, particularly for the 1D X and 5D III. Canon put a hell of a lot of work into their in-camera JPEG algorithm. I've seen ISO 51200 JPEG photos strait out of the 1D X That looked better than ISO 1600 on just about any previous generation camera. Hardly any noise at all...although many of the shots were of decently lit subjects (I think EVERY example of this that I've seen involved soccer games in one of the EuroZone countries...so bright artificial lighting, lots of bright colors, and in general some mind-blowing ISO 51200 JPEG photos....I'll see if I can dig up some of the links, although I'm not sure if I kept any of them around.) I don't think that Nikon fares as well on the in-camera JPEG front...JPEGs at high ISO on the D800 certainly did not seem to fare as well, although at the same time Nikon cameras do not have as high of native ISO settings, and require digital boost to achieve the likes of ISO 25600 or higher. In the end, I would say all bets are off with JPEG, and unlike RAW, you might be able to get usable "keeper" results right up to the maximum native ISO setting.

In general, I consider expanded or boost ISO settings completely unusable as they amplify noise right along with exposure...however if you really need to achieve a specific minimum shutter speed, sometimes they are the only option (although you could always just use the max native ISO and use a shorter shutter speed anyway, and manage the "boost" in post yourself, extracting every ounce of IQ you can). Just don't expect the kind of IQ that would allow you to blow a photo up to large A1 size and hang it on your wall. ;)

Answer 2

No, you cannot tell performance by specifications alone. You need to see review samples, not graphs and still should be cautious.

It is my observation that a manufacturer tends to be consistent where the stop their ISO range. So, if the standard range of a camera is ISO 100-3200 and another of the same manufacturer is 100-12800, you can expect the latter to be two stop better. If you find ISO 800 acceptable on the former, then you should find ISO 3200 on the latter.

Now reviews tend to show graphs, crops or both. Graphs are the most misleading because they show noise levels without regards to softness introduced by noise-reduction. Some makers have aggressive noise-reduction defaults which make for nice graphs but do not help for making images more usable.

Image crops are better because you can see noise and details at the same time and judge the effect of one on the other. The only problem is that most reviews only show crops made from the default camera settings which vary from camera to camera. Some makers have rather soft defaults and different levels of noise-reduction.

Even with crops taken from RAW files, the converter may behave differently on different cameras, so its hard to tell exactly but in general you can get something that has been processed more consistently. Of course, this only matters if you plan on shooting RAW. If you plan on shooting JPEGs, the best are crops using optimized camera settings as defaults are not rarely the best.

Answer 3

It is not so easy to find consistent test, on different models of different years. The best way is to use some estensive inforamtion repository. For example this one is pretty good.

You can choose brands, camera type (compact, hibrid, etc.) and sort by price.

You can also compare your camera with some other models. The value "Sports (Low-Light ISO)" tells what is the maximum ISO you can use to stay below a certain noise level. So your camera is about 700 ISO, a camera with 1400 ISO can more or less give the same quality of your camera at a doubled ISO.

You can also have detailed graphics clicking on Measurement -> SNR 18%