The Digital Picture review of the 1D X

marekknowak wrote in post #14812337
Shadow recovery isn't high ISO noise level.

The D800 dynamic range is higher than the canon only at low ISO. At high ISO, canon has higher dynamic range, and better shadow recovery.

The tests you see on the net showing how great shadow recovery is on a D800 are done at low ISO.

Which is kinda useless unless you are the kind of photographer who
miss the exposure by 3 stops when there is plenty of light.

I never said shadow recovery is at high ISO. Why would I shoot a landscape at those settings? Surprisingly enough the D800 sensor doesn't produce heat even at long exposure times of 30 minutes. And this is with NR on!

EDIT: I have reference images if you don't beleive me.

What sort of DR are we talking about with the 1Dx and D800? It would be nice to know real world figures, not the total nonsense spouted by the sales departments.

^I have no idea Richard. I can tell you that there are some Nikon converts out there.

While I really would like Canon to fix the banding issue, I'm a bit puzzled why some "landscapers" want ever more pixels. It seems that we'd end up with a camera capable of something like the 800E and then need to replace all our glass to actually resolve close to the camera's performance.

A study published hereshows quite convincingly that focal curvature and accuracy limit the value of small-pixel, high resolution sensors. I'm not saying it can't be done but there comes a time when higher resolution requires simply larger pixels, a longer lens and either a MF back or emulating this by constructing a composite...

On the dynamic range, I suspect that this is limited by the well depth achievable in pixels of a given size - hence the stall in performance improvements. On the other hand, I suspect that pattern noise is deterministic. If so, it can be characterised and then eliminated in post processing. I'd be interested to hear whether any astrophotographers have played with dark / bias / defect frame corrections as a cure for pattern noise.

(... Of course, if pattern noise is deterministic, you might ask yourself why Canon doesn't do it in the internal image processing.)

[QUOTE=noisejammer;148​13230].....I'm a bit puzzled why some "landscapers" want ever more pixels. It seems that we'd end up with a camera capable of something like the 800E and then need to replace all our glass to actually resolve close to the camera's performance....
quote]

It's simple: More pixels gives smoother tonal changes, the whole image becomes less "gritty" (and I'm not talking about noise here). The more pixels the better it gets. Thats why we landscapists lust after medium format or very high pixel counts in a dSLR.

Regarding the glass issue, thats why Canon, Nikon and others are busy upgrading all their lenses, because we now (in Nikons case) or soon (hopefully in Canons case if they can be persuaded to rethink) have sensors that are outperforming the older lens designs.

Sorry - I didn't intend to hijack the thread.

Lowner wrote in post #14813749
It's simple: More pixels gives smoother tonal changes..., the whole image becomes less "gritty"...

Of course, but over sampling the image space is an exercise in rapidly diminishing returns.

I think I may have been less clear than I might. Achieving sharp focus is fine but it's of limited value when focus cannot be delivered over the entire subject. Typically, the focal surface curves around the photographer. Small pixels and diffraction limit the usable f-stop so that the relatively shallow DoF limits the achievable resolution. This seems to be a limitation imposed by physics and affordable lens design.

Large pixels and longer lenses mitigate this limitation (and as a side effect you synthesize a large number of pixels which preserve the tonality gradations you spoke of.)

noisejammer wrote in post #14813853
Of course, but over sampling the image space is an exercise in rapidly diminishing returns.

I don't know what you mean by over sampling. The more pixels the better, its as simple as that as long as the glass can do the job (we are assured it can or soon will). I have read that optical resolution can be raised to whatever level is required so I disagree with "diminishing returns". Certainly none of this technology is going to come cheap, we can agree on that!

Lowner wrote in post #14814051
I don't know what you mean by over sampling. The more pixels the better, its as simple as that as long as the glass can do the job (we are assured it can or soon will). I have read that optical resolution can be raised to whatever level is required so I disagree with "diminishing returns". Certainly none of this technology is going to come cheap, we can agree on that!

I've printed images with Nikon's 50 1.8 G and it's one of the cheapest lenses they have. So far no dimishing returns at all. I even printed 40 x 60 to amuse myself and the IQ is world's apart compared to my 1Ds 3.

noisejammer wrote in post #14813853
Sorry - I didn't intend to hijack the thread.

Of course, but over sampling the image space is an exercise in rapidly diminishing returns.

Oversampling is relative to the actual resolving power of the lens. And the resolving power of the lens is a moving target.

Yes, but remember that landscapers shoot distant objects, which are well beyond the hyperfocal distance even when the aperture is wide open. As a result, field curvature becomes a non-issue.

Even the pixel density of the 18 megapixel crop sensors has a diffraction limited aperture of f/6.9. On a full frame sensor, that would get you 46 megapixels worth of resolution.

Lenses are nowhere near the limits of physics, as is evidenced by the incredible resolving power of lenses on point'n'shoot cameras. The main limitation of DSLR lenses isn't one of physics, it's one of construction. DSLR lenses are larger than point'n'shoot lenses (by a lot), and the surfaces of the elements have to crafted to an extreme degree of precision over a much larger area than for point'n'shoot lenses. That's a technological issue, and therefore is something that is subject to improvement over time.

And the evidence is quite clear that the improvements are making their way towards the less expensive lenses. The 18-55 IS kit lens is a massive improvement over its predecessor in terms of resolving power. It has to be, because the cameras it's being used on are 18 megapixel crop cameras.

Yes, but a higher resolution sensor on an interchangeable lens camera is one that, even if it oversamples the image circle generated by current lenses, may not oversample next generation lenses. The drive towards higher resolution sensors drives lens development as well, and pushes the technology forward. That's a good thing: it makes the resulting camera systems more flexible.