High ISO shooting Options

[douglasf13]

Interesting article on D3s high ISO image quality. Rob sees between one and 2 stops improvement over the D3 which is nothing short of amazing considering the level already reached by the D3.

http://www.robgalbraith.com/bins/content_p...d=7-10045-10329

Cheers,
Bernard

Yeah, the D3s looks like a lowlight monster. Awesome work, Nikon.

[Ray]

It does seem that the D3s has about a one stop high-ISO advantage over the D3. However, I think there's something else going on which may not be confined to high-ISO performance, and that's the higher resolution of the D3s. It's particularly noticeable in the Rob Galbrath's shots from the D3s at, for example, ISO 3200 which is sharper (or slightly more detailed) than the D3 shot at ISO 3200. In fact, the D3s shot at ISO 3200 is even slightly sharper than the D3 shot at ISO 1600, but one wonders if the slower shutter speed at ISO 1600 has a bearing on this, considering the lens used does not have VR.

I get the impression from other comments that the D3s has an improved AA filter, and maybe better micro-lenses. If this is so, then one would expect the D3s to be slightly sharper and more detailed than the D3 at all ISOs.

[dwdallam]

I've thought of another reason for high ISO: flash throw. The more sensitive the camera is to light, the less light you need, and this equates to a flash that can throw perhaps hundreds of feet further than when using ISO 100.

I use the Canon 1DS MKIII and it's good at ISO 1600 (it's limit w/o artificial gain at 3200) if you get the exposure right. If you underexpose even one stop, it's crappy. Probably better to use ISO 800 and underexpose. I was in Vegas last week doing the SEMA car show and after I got to wander around Vegas at night. Well, to make a long story short, when shooting w/o a tripod (not allowed on the strip mall areas) and trying to shoot stop action, you really need more light. This is especially true while holding a 1DS MKII with a 70-200 f2.8 IS lens attached.

I can tell everyone here and now, if Canon's next 1DS model has a leap in high ISO performance, I'm in all the way--given that I can afford it. I'll post a couple of pictures later to show the difference in ISO 800/1600.

[Ray]

I use the Canon 1DS MKIII and it's good at ISO 1600 (it's limit w/o artificial gain at 3200) if you get the exposure right. If you underexpose even one stop, it's crappy. Probably better to use ISO 800 and underexpose.
I can tell everyone here and now, if Canon's next 1DS model has a leap in high ISO performance, I'm in all the way--given that I can afford it. I'll post a couple of pictures later to show the difference in ISO 800/1600.

It's always better to use ISO 800 than ISO 1600 if the shutter speed is sufficient at the lower ISO. One stop underexposure at ISO 800 should also be better than one stop underexposure at ISO 1600. However, a correct ETTR at ISO 1600 should be at least marginally better than the same exposure at ISO 800 (which would be one stop underexposure in relation to ISO 800).

The improved high-ISO performance of the D3s for me would result in less reluctance to use the moderately high ISOs of 1600 and 3200 where image degradation with the Canon models tends to become too obvious, especially in wildlife shots where the fine detail of feathers and texture tends to get smeared.

For this reason I'm considering the purchase of a 7D. If the 7D can produce an ETTR shot at ISO 3200 which is as good as an ETTR shot at ISO 1600 using the 50D (discounting the effects of different shutter speeds with regard to freezing of motion), then such improvement is a great bonus when added to the other improvements, such as video capability. I'm not sure I'm going to be able to resist getting a 7D, although it might be good for my character if I were to resist .

[dwdallam]

It's always better to use ISO 800 than ISO 1600 if the shutter speed is sufficient at the lower ISO. One stop underexposure at ISO 800 should also be better than one stop underexposure at ISO 1600. However, a correct ETTR at ISO 1600 should be at least marginally better than the same exposure at ISO 800 (which would be one stop underexposure in relation to ISO 800).

The improved high-ISO performance of the D3s for me would result in less reluctance to use the moderately high ISOs of 1600 and 3200 where image degradation with the Canon models tends to become too obvious, especially in wildlife shots where the fine detail of feathers and texture tends to get smeared.

For this reason I'm considering the purchase of a 7D. If the 7D can produce an ETTR shot at ISO 3200 which is as good as an ETTR shot at ISO 1600 using the 50D (discounting the effects of different shutter speeds with regard to freezing of motion), then such improvement is a great bonus when added to the other improvements, such as video capability. I'm not sure I'm going to be able to resist getting a 7D, although it might be good for my character if I were to resist .

Ray I've had a discussion with people on this site that say that underexposing one stop at ISO 800 is better than a correct exposure at ISO 1600. Also, if noise can be cleaned up, I have no problem with that, but like you said, with the Canon you get a smeared look if you push your luck too far. And there is no way to get rid of that. You just have a crappy image. I think I could even be happy is there was an improvement in ISO 1600. In other words, if you could use ISO 1600 like you do ISO 800 on the 1DS3, I'd be happy with that.

[Ray]

Ray I've had a discussion with people on this site that say that underexposing one stop at ISO 800 is better than a correct exposure at ISO 1600. Also, if noise can be cleaned up, I have no problem with that, but like you said, with the Canon you get a smeared look if you push your luck too far. And there is no way to get rid of that. You just have a crappy image. I think I could even be happy is there was an improvement in ISO 1600. In other words, if you could use ISO 1600 like you do ISO 800 on the 1DS3, I'd be happy with that.

Interesting! I wonder what's causing that. I could understand that the difference between ISO 800 being underexposed one stop and ISO 1600 being fully exposed might be so insignificant that it's not an issue. In such circumstances, it would generally be preferable to underexpose at ISO 800 for the sake of reducing the risk of miscalculated blown highlights.

However, that ISO 800 underexposed one stop should actually be better than the same exposure at ISO 1600, seems contradictory to DXOMark's test results which indicate the DR of the 1Ds3 at ISO800 is 10.07EV whereas the DR at ISO 100 is 9.24EV.

My reasoning would be that the DR at ISO800 underexposed one stop should be 9.07EV, which is slightly worse than the 9.24EV of a correctly exposed ISO1600.

The difference in SNR at 18% gray is equivalent to exactly one stop, ie 27.5dB at ISO1600 and 30.5dB at ISO800, a difference of 3dB being equivalent to one stop.

[dwdallam]

Interesting! I wonder what's causing that. I could understand that the difference between ISO 800 being underexposed one stop and ISO 1600 being fully exposed might be so insignificant that it's not an issue. In such circumstances, it would generally be preferable to underexpose at ISO 800 for the sake of reducing the risk of miscalculated blown highlights.

However, that ISO 800 underexposed one stop should actually be better than the same exposure at ISO 1600, seems contradictory to DXOMark's test results which indicate the DR of the 1Ds3 at ISO800 is 10.07EV whereas the DR at ISO 100 is 9.24EV.

My reasoning would be that the DR at ISO800 underexposed one stop should be 9.07EV, which is slightly worse than the 9.24EV of a correctly exposed ISO1600.

The difference in SNR at 18% gray is equivalent to exactly one stop, ie 27.5dB at ISO1600 and 30.5dB at ISO800, a difference of 3dB being equivalent to one stop.

I see what you mean. That's most likely correct. I was talking about shadow noise only. You get a less noisy image -1ev at ISO 800 than you do at ISO 1600 correctly exposed. Which means you are actually shooting at ISO 1600 with less noise.

[BernardLanguilli...]

I see what you mean. That's most likely correct. I was talking about shadow noise only. You get a less noisy image -1ev at ISO 800 than you do at ISO 1600 correctly exposed. Which means you are actually shooting at ISO 1600 with less noise.

You realize that if your claim is true, the Canon engineers are total idiots?

Cheers,
Bernard

[Jeremy Payne]

You realize that if your claim is true, the Canon engineers are total idiots?

Cheers,
Bernard

The last "true" ISO will often be close to a wash in this regard, no? Isn't that to be expected?

I know that on the D700, 3200 off by 1 stop and 6400 are hard to tell apart, but that's not the case at lower ISOs.

[dwdallam]

You realize that if your claim is true, the Canon engineers are total idiots?

Cheers,
Bernard

It's not my claim. I'm just passing information from a thread here at Luminous.

[Ray]

I see what you mean. That's most likely correct. I was talking about shadow noise only. You get a less noisy image -1ev at ISO 800 than you do at ISO 1600 correctly exposed. Which means you are actually shooting at ISO 1600 with less noise.

I'm not sure you do see what I mean. According to DXOMark, the noise at -1EV ISO800 and 18% gray is exactly the same as at 0EV ISO1600, with the 1Ds3. What differs is the dynamic range which is very marginally greater at 0EV ISO1600. Greater DR usually equates to less noise and more detail in the shadows at darker shades than 18% gray. However, since the difference in DR is less than 1/4 of a stop in favour of ISO1600, it's no big deal. The protection of highlights at -1EV ISO800 might be more worthwhile.

[LKaven]

This has been covered by Emil Martinec at U Chicago here

http://theory.uchicago.edu/~ejm/pix/20d/te.../noise-p3a.html

It is a strange and subtle outcome that read noise should be best around ISO 1600 in many cases, and that an improved noise reduction strategy combines bi-amplified signals at ISO100 and ISO1600 in parallel. Martinec's Teddy Bear experiment shows the benefits of this.

Some has wondered whether Nikon is doing something like this in the D3x, but I've never seen a conclusive answer to that question.

[Jeremy Payne]

This has been covered by Emil Martinec at U Chicago here

http://theory.uchicago.edu/~ejm/pix/20d/te.../noise-p3a.html

It is a strange and subtle outcome that read noise should be best around ISO 1600 in many cases, and that an improved noise reduction strategy combines bi-amplified signals at ISO100 and ISO1600 in parallel. Martinec's Teddy Bear experiment shows the benefits of this.

Some has wondered whether Nikon is doing something like this in the D3x, but I've never seen a conclusive answer to that question.

Can you point out where exactly in the paper it discusses that it is better to push than use "real" ISOs for a single capture? It is an interesting paper, but I don't think it addresses that point exactly.

This post has been edited by Jeremy Payne: Nov 13 2009, 08:24 AM

[Ray]

Can you point out where exactly in the paper it discusses that it is better to push than use "real" ISOs for a single capture? It is an interesting paper, but I don't think it addresses that point exactly.

Quite true. The difference in shadow noise between ISO 100 underexposed 4 stops, and ISO 1600 at 0EV (the same exposure) is enormous. Emil concentrates on this in his image examples.

However, when comparing the noise & DR differences between ISOs that differ by only one stop, it's a different picture. The differences are far more subtle, but still in favour of the higher ISO.

I also get the impression that such differences between one ISO and another a stop higher, diminish as one goes up the ISO scale.