Nikon D800, 36 Megapixels and Blur

[ChadS]

I don't. What gives you that impression? At the very worst, the increased pixel count will only result in the *same* resolution as before -- it can never be worse.

At a certain point the noise issue will bite you at too high a resolution and too low light level. True, you can always integrate the signal over several pixels but the net result of that is a higher noise floor vs. having one large pixel capturing the same number of photons. I'm certain that for bright light conditions it should be possible to reach ~100 MP given the sharpness of the best lenses and improvements to sensors that we know how to do. However, I fear for low-light conditions.

[HDNitehawk]

I don't. What gives you that impression? At the very worst, the increased pixel count will only result in the *same* resolution as before -- it can never be worse.

I wouldn't think it would, and I think you would be correct. I would think that at some point in the MP and sensor war that a persons pictures resolution wouldn't improve by more MP, instead what they were able to achieve would be limited by there skills and ability.

Why I thought the article was saying it would is the following quote:

"While its high pixel count of 36 megapixels gives the D800/ D800E resolution unrivalled by previous digital SLR cameras, a side effect is that bokeh and blur are made that much more obvious."

Why would blur be much more obvious, unless the comparison is drawn against the D700 that the D800 is replacing, and the D700 is very bad in this area. I doubt that is the case, not that I would know.

[jrw]

Well, I have learned something today. After reading this thread earlier today I started wondering if I would see much difference due to pixel size when using the 100-400 on a 5Dii and comparing shots to the 70-200 IS ii on a 7D which has slightly smaller pixels than the D800. Everything was tripod mounted for this testing. I maxed out the zoom on the 70-200 and set the 100-400 to closely match the framing. On shorter distance to subject shots the 7D combo had nice clean images. Once I went out to long shots, over a 100 yards, I started to notice a decline in the shots from the 7D when looking at the images with 100% view. There was an increase in noise. The 5D combo just kept going. I can only assume that I hit the limit of what the 200 mm focal length was able to deliver to a smaller pixel in comparison to what the 320 mm focal length delivered to the larger pixel on the 5D. I realize that using two different focal lengths introduced another variable into the comparison, but I must admit that one reason I use the 7D is for the ability to have a high pixel density on a cropped sensor to allow more pixels on a small target for cropping in post processing. The other is for action shots at distances less than where it started to have issues. I also put the 70-200 on the 5D afterwards as a reality check and did not have the same issue. In real world shooting it isn't a problem for me given the required subject distance to encounter this issue.


Just thought I'd pass on the info.

[Raid]

@jrw. Just wondering what ISO, speed and fstop were you using?

Also as the focus distance is increased lots of other things come into play. Always double check your equipment and setup.

Have you swapped lens on the bodies, to see if anything changes?
Have both lenses been calibrated for the respective bodies?

Tony

[Jon Ruyle]

At a certain point the noise issue will bite you at too high a resolution and too low light level. True, you can always integrate the signal over several pixels but the net result of that is a higher noise floor vs. having one large pixel capturing the same number of photons. I'm certain that for bright light conditions it should be possible to reach ~100 MP given the sharpness of the best lenses and improvements to sensors that we know how to do. However, I fear for low-light conditions.

There seems to be a lot of confusion about this topic. As Daniel said, image quality will only be higher or the same, not lower, with more MP. Noise will not be worse just because there are more pixels. The only downside of more MP is that it takes up more data.

[Busted Knuckles]

How would distance from the camera effect IQ - more specifically signal to noise ratios.

There might be something about the angle of the photons hitting the sensor but oof that is way down the list of IQ issues.

Is this not more of a lens issue?

[ChadS]

@JRW The image circle created by the lens on the camera is showing the same image - you're just seeing less on the crop. So the fact that the IQ on the 7D failed before the 5DII just tells you that your lens is not capable of making a 5 micron detail on the sensor size at that focal length - hence the discussion about having the same "resolution" even though you have more megapixels.

@Jon I'm not confused at all. In terms of SNR a single large-area sensor will maintain a higher SNR than the integration of a bunch of itty-bitty pixels. The reason has to do with amplifier noise and dark noise. In the absence of any photons striking the pixels but with the amplifiers turned on, there will be a background signal that appears due simply to shifting electron energy states accumulating on the various capacitors, etc. This is referred to as dark noise. With a single pixel you've got a single amplifier chain and a single source of dark noise. If you have 100 pixels you're going to have 100 separate dark noise sources. The signals from these 100 pixels will not all be the same - it's a random process. Therefore, they're not correlated and when I add them all together there is a greater noise floor from the ensemble than on the single pixel. (it all goes back to the math theorem that says that a^2 + B^2 >= |A+B|^2).

I'm not suggesting that the new Nikon is at this limit yet. I clearly stated that I think 100 MP is fine with current technology. However, the 1Dx is almost certain to have much better low-light performance even after downsampling the Nikon's image to match the total number of pixels.

[jrw]

@ Raid - AV mode, ISO 100, f5.6 on both lenses, CR2 output, variable cloudiness all shots at 1/500s or faster. Same tripod used, mounting plates attached to tripod ring on lens, camera tethered and focus confirmed for every shot. All lenses are focus calibrated to each of my bodies just wanted to have remote release and ability to confirm focus on a bigger screen. To be more specific the OOF regions became grainier. The in-focus area started to show spots of different colours across uniformly coloured objects (cars) as well as smooth lines becoming rough and jagged (license plates, body panel edges, and printed signs throughout the parking lot). Just to be clear, these were parked cars and the 5D combo did not exhibit the same behaviour on the same targets. The 7D only showed these at greater distances. With closer targets there were no issues. The overall amount was not terribly great, but I was specifically looking for such issues at the 100% zoom level to see if it does exist and I found it.

70-200 was put on the 5D afterwards to see if problem was with lens. No problem there, just with the 7D. Does not rule out possibility of 7D having some other issue. Don't have other 7D bodies to confirm or deny. Post processing is able to clean up the shots with no difficulty, just wondering what is going on. A question without an answer is something that seizes my focus until solved.

Perhaps someone who is more familiar with the equations of optics wouldn't mind stepping in?

What I saw isn't making me want to sell my 7D. It will continue to function the same as it always has, and results when using it may improve the more I understand what it is capable of doing.


EDIT: I see someone with more specific knowledge replied while I was pecking away at the keyboard.

[jrw]

@ Raid - AV mode, ISO 100, f5.6 on both lenses, CR2 output, variable cloudiness all shots at 1/500s or faster. Same tripod used, mounting plates attached to tripod ring on lens, camera tethered and focus confirmed for every shot. All lenses are focus calibrated to each of my bodies just wanted to have remote release and ability to confirm focus on a bigger screen. To be more specific the OOF regions became grainier. The in-focus area started to show spots of different colours across uniformly coloured objects (cars) as well as smooth lines becoming rough and jagged (license plates, body panel edges, and printed signs throughout the parking lot). Just to be clear, these were parked cars and the 5D combo did not exhibit the same behaviour on the same targets. The 7D only showed these at greater distances. With closer targets there were no issues. The overall amount was not terribly great, but I was specifically looking for such issues at the 100% zoom level to see if it does exist and I found it.

70-200 was put on the 5D afterwards to see if problem was with lens. No problem there, just with the 7D. Does not rule out possibility of 7D having some other issue. Don't have other 7D bodies to confirm or deny. Post processing is able to clean up the shots with no difficulty, just wondering what is going on. A question without an answer is something that seizes my focus until solved.

Perhaps someone who is more familiar with the equations of optics wouldn't mind stepping in?

What I saw isn't making me want to sell my 7D. It will continue to function the same as it always has, and results when using it may improve the more I understand what it is capable of doing.

[ChadS]

@jrw Gawd I hope you're not referring to me!

By all means, don't sell your 7D - well, unless you want to give it away to me! Think of it this way, the lens is the lens and it knows nothing about the sensor behind it (assuming the image is in-focus on both). There's really nothing the body brings to the image - which is why the camera manufacturers had to work so hard to differentiate pro-level film cameras from entry-level bodies (autodrive, better light sealing, etc.) Fundamentally, a film camera body is an empty box... In the digital world you can think of most of the body as an empty box and the sensor is the film.

The pixels just take whatever light comes through the lens and dices it up into little squares. It just so happens that the 7D has smaller pixels and so chops the light in the center part of the image into smaller pieces. The 5D2's squares are larger but they cover more of the image circle (but not all, of course).

The pixels know nothing about the lens just like the lens knows nothing about the pixels. If I were to put a $10 lens on my camera I'd expect the details to be blurred all over heck. I'm still dicing the blur really, really small on the 7D but all I'm really doing is discretizing the blur from the lens. I could double the megapixels on my sensor and wouldn't record any more information - just a larger file.

The difference between the 7D and the 5D2 using your 70-200 and your 100-400 is right on the brink of the ability of those two lenses to resolve features. A more valid test would be to tripod mount the lens, then swap bodies for the same exact shot lens-to-lens. You'll see that if you zoom in on the 7D you'll get equal or greater detail on the 7D than on the 5D2. At some point the blur of the lens may make the two equal.

This is why Bryan commented that birders may not want to upgrade to the 1Dx. Most of them have really good glass that can resolve really small features on the sensor. So a higher density sensor like the 7D, 1D4, etc. will allow you to "crop zoom" the bird and still get more details. The rather large pixels on the 1Dx will not allow as much "digital magnification."