I enjoyed reading Mark Segal's piece on noise, comparing the 1DsIII and D3. It would appear that he may have put to rest the claims that the D3 sensor is so much better at handling noise than Canon sensors. He says:

"Much as expected, once roughly equivalent exposure techniques are used, differences of noise performance between the Canon 1Ds3 and the Nikon D3 are not remarkable, save for the very noisy results obtained from the Nikon for ETTL beyond ISO 80016. Comparing the 1Ds3 and the D3 for noise using the ETFR strategy with the Canon, the differences are insignificant."

The total number of stops between the lowest ISO and the highest in the normal (non-boost) range in the two cameras is only one stop, in the normal range. For the 1DsIII, it starts at 100 and goes to 1600- 5 stops. The D3 starts at 200 and goes to 6400- 6 stops.

Given the 1DsIII is a 21 megapixel camera and the D3 is a 12 megapixel camera, it is actually remarkable that the Canon performs so well. Seems to me the initial hype on this site and others claiming the D3's huge advantage in noise handling may well have been exaggerated. Mark's article points out that both cameras are equally as good when proper exposure is taken into account.

I was a bit disappointed that Mark did not venture into the 'push processing' area. Those of us who used film are aware that the ISO of the film was not an absolute. The type of chemical development of the film and the time duration of development could be varied according to the exposure.

If you happened to have an ISO 200 film in your camera but needed an ISO 800 exposure, you could shoot the whole roll as though it were an ISO 800 film and then tell the lab technician to process it accordingly.

I never actually got around to comparing (for example) a dedicated ISO 800 film with an ISO 200 film push processed to ISO 800, to see which was better, although the thought did occur to me.

I therefore find it curious that reviewers, when comparing cameras with different maximum ISO settings, usually refuse to display any image from a camera which doesn't boast a particular high ISO setting.

It seems we've gone from a situation of limited processing potential (with regard to film) to an even more limited processing potential with regard to digital, in this respect of ISO.

Ray,

I'm not sure that I understand your point. Digital is not film. The processes that take place chemically when one "pushes" film have no direct analog (no pun intended) to digital.

All sensors have a "native" sensitivity, usually the lowest regularly set ISO setting. Anything above that is based on the camera's circuitry cranking the gain on the analogue signal before the A/D converter, along with various black magic incantation which the camera maker includes to suppress noise and retain detail lost due to the decreased S/N ratio.

After that there's no "push processing" left for the user to do to increase "sensitivity". It's all just image manipulation to suck detail out of the shadows if needed. Not something that's meaningful in a comparison test situation.

Michael

If you look at Emil Martinec's analysis of noise in digital cameras (to which Michael has referred), it is not surprising that the D3 and 1DsIII have similar noise in tests. Assuming rough parity in sensor technology, both sensors have the same total area and would collect the same number of photons, and photon sampling noise (shot noise) is the major determinant of noise in the image except for the deepest shadows. With smaller pixels, the 1DsIII has more noise per pixel but a lot more pixels. If the 1DsIII image were downsized to match the resolution of the D3, pixel binning would reduce the noise.

Michael,
My analogy is that the RAW digital image is like an exposed film which hasn't been developed. It has the advantage that it can be developed and redeveloped as often and in as many different ways as one likes, whereas film can only be developed once.

Whilst the presence of a particular ISO setting may be an indication that the camera has been designed to 'push process' the data before the RAW image is written, there is sometimes not a lot of difference between an underexposed image at a lower ISO and the same exposure at a higher ISO, after appropriate post processing, noise reduction and EC adjustment etc.

If one were in a situation with a 1Ds3 where a fairly fast shutter speed was required in poor light, would one not take the shot on the basis that the camera didn't have an ISO 6400 or ISO 12,800 setting?

Are you just looking for the number of available stops below neutral gray?

I get the impression that the noise and dynamic range differences between two cameras is not a constant at every ISO setting. My own tests comparing the 5D and D3 were all done at ISO 3200 and higher. I got the impression the D3 produced images about 1/3 to 1/2 a stop cleaner in these dark conditions.

My guess is that the 1Ds3, underexposed by 2 stops at say ISO 3200, would produce results no noisier than the D3 at ISO 12,800, after downsampling the 1Ds3 image to the same size as the D3 image, a process which results in the throwing away of some degree of noise.

I agree and to simplify the practical application of what I think you are saying would that a 1Ds3 owner wanting to get ISO 12,800 should expose at 1600 and then use the exposure control in ACR or Lightroom to add two stops of exposure.

The issue I am not sure about is this. A D3 exposes at 12,800 by boosting the S/N ratio at the sensor. A 1Ds3 can achieve an equivalent 12,800 ISO only by increasing the exposure in the raw file by two stops from ISO 3200. Can these two results be truly compared on an equal basis?

Not really. ISO 6400 is the top true ISO.


I have not analysed any ISO 3200 image from an 1DsMkIII, but I read somewhere, that 1600 is the top true ISO. If that is so, then the two stop difference remains.


This situation is screaming for a detailed, thorough comparison. The presence of higher ISO gain itself does not guarantee corresponding quality.

We had a small discussion about this over at photo.net. Not exactly the same thing, but the OP took a "correct" exposure at iso 800 and also one at 100 and 1600 pushed and pulled to iso 800, and compared them for noise.

I think if one is trying to simulate an ISO 6400 or 12,800 setting with a 1Ds3 by underexposing, there's probably no advantage in using ISO 1600 rather than ISO 3200 because there's little chance of blowing any highlights.

I've never actually compared identical exposures at ISO 1600 and ISO 3200 (the ISO 1600 shot being underexposed 1 stop). I've just accepted the opinion of the more technically knowledgeable people on this forum that it is the case, that ISO 3200 is basically ISO 1600 underexposed one stop, when shooting RAW.

But I recall John Sheehy mentioning that some folks who had taken the trouble to compare shots had reported that there was a very minor and subtle improvement in choosing ISO 3200 instead of underexposing one stop at ISO 1600.

Since I'm not into extreme pixel peeping, as you know, I've never considered it a useful allocation of my time to check this out .

OMG - my head was spinning by the time I finished reading that article on Noise About Noise.

I know gas prices have gone up - but keyboard letters are still cheap -lol. All the abbreviations made my head spin. Expose-Right, Expose Left, Expose-Right+ is so much more readable than the way it is.

And ETTL is given a new definition on top of all that. Gosh, make it easy on the readers, some will say thanks Thanks . . .

As for the Nikon D3 noise levels, I liked what was said in this article comparing it to the 1Ds3. But in my wedding work I find the top ISO with the 1D3 and the 1Ds3 is 1600 while I find the same comfort with the D3 at ISO 3200. I would say at specifically ISO 1600 and ISO 3200 the D3 is about 70 percent of a stop better. Not quite a stop, but enough for face details to stay intact in printed large albums. Something that does not happen with the 1D3 - but DOES with the 1Ds3 simply because of it's file size even though ISO 3200 is not a profiled ISO setting that complies with the normal range of the camera.

I remember reading the Michael R said he was going to do his own report on the D3. I was looking forward to that - is it still coming?

Peter

That's not clear. Are you saying that the D3 is one stop (or close to one stop) better than both the 1D3 and 1Ds3? Have your comparisons involved same exposures at ISO 1600 and 3200 (with all cameras), same image sizes and appropriate processing of RAW images rather than in-camera processing of jpegs?

The main reasons for using a high ISO are because the subject is not completely stationary or (if the subject is stationary) because one doesn't have a tripod for the long exposure that the poor lighting conditions require.

In such circumstances, the choices are, when flash is not available or appropriate; risk getting a blurred shot using a shutter speed that is too slow; risk getting unacceptable noise and a tonally degraded result by using a high ISO setting; don't take the shot.

My concern would be that some photographers using a D3 might happily take a shot at ISO 6400 and get acceptable results for their purposes, whereas other photographers in the same situtaion and with the same purpose in mind, but using a 1Ds3, might be deluded into thinking that the shot is not worth taking because their camera's maximum 'real' ISO setting is only ISO 1600.

It's a big assumption to make that a 12mp camera at ISO 6400 is likely to produce a better picture than a 21mp camera at ISO 1600 underexposed 2 stops (or at ISO 3200 underexposed 1 stop).

This is somewhat strange, as the 1DMkIII goes up to 3200 with real ISO gain, while the 1DsMkIII stops at 1600 (I have not analyzed the latter myself, but I do accept it). Thus this categorical statement is rather dubious.



I wonder if you can support your statements with directly comparable raw images.

(Note, that I don't own either of these cameras, and I don't care for their reputation the very least.)


Do you care to explain to me, what a "profiled ISO setting" means, and what is the "normal range of the camera" to comply with?

Ray, think of the sensel dimension. The D3 is performing a "hardware binning", which is inherently better than downresing the image.

The truth will become obvious, when Nikon comes out with a 21MP FF camera, which will be far inferior to the D3 in term of ISO performance.

All else being equal. which it rarely is.