jonnyapple said:
That is what I was implying, satellites, but only accidentally because I misspoke. What I meant to say was that it may beat 12MP printed even if the 36MP sensor had a lower signal to noise ratio for individual photosites. Sorry about that.

I hope we do see the day when a 36MP sensor beats previous generation 12MP sensors pixel for pixel, but I don't think that's what would happen with the first iteration. Prove me wrong, Nikon. ;-)

now that makes sense! haha thanks for clarifying.

Nikoner said:
I hope this question fits here, may be FX owners can chime in with their experiences. I do have a vague idea of what to expect from a D700.

But say I choose to shoot a D800/85mm f1.8D in DX mode at a low light sporting event, will the camera retain whatever low light ISO it may have in FX mode?

What are the ramifications of using a D800 in DX mode for low light/high speed events?

Thanks.

It basically becomes a D7000- it just clips (turns off) the pixels not used. Computer cropping would probably be a better solution for what you described. The sensor noise would be the same but cropped, it may be a bit more noticeable as each pixel would be larger. Technically it would be the same noise characteristics.

In a few weeks or months reviews of the new Nikon FX senor will be available and then we will be better able to judge it's high ISO noise compared to current FX and DX sensors. Nikon may even include a brief comparison in its advertising. I would expect the D4/D899 to be something alike a one or two stop improvement over the D3/D700 since that seems to be about norm for a major sensor change. And I think that would equate to a 2 to 3 stop improvement over the DX D7000. Gaining a stop or two or three of cleaner higher ISO should be useful to those who shoot in low available light. I would like to see a relatively noiseless ISO 6,400. But we will just have to wait and see.

In a few weeks or months reviews of the new Nikon FX senor will be available and then we will be better able to judge it's high ISO noise compared to current FX and DX sensors. Nikon may even include a brief comparison in its advertising. I would expect the D4/D899 to be something alike a one or two stop improvement over the D3/D700 since that seems to be about norm for a major sensor change. And I think that would equate to a 2 to 3 stop improvement over the DX D7000. Gaining a stop or two or three of cleaner higher ISO should be useful to those who shoot in low available light. I would like to see a relatively noiseless ISO 6,400. But we will just have to wait and see.

I hope this question fits here, may be FX owners can chime in with their experiences. I do have a vague idea of what to expect from a D700.

But say I choose to shoot a D800/85mm f1.8D in DX mode at a low light sporting event, will the camera retain whatever low light ISO it may have in FX mode?

What are the ramifications of using a D800 in DX mode for low light/high speed events?

Thanks.

That is what I was implying, satellites, but only accidentally because I misspoke. What I meant to say was that it may beat 12MP printed even if the 36MP sensor had a lower signal to noise ratio for individual photosites. Sorry about that.

I hope we do see the day when a 36MP sensor beats previous generation 12MP sensors pixel for pixel, but I don't think that's what would happen with the first iteration. Prove me wrong, Nikon. ;-)

TaoTeJared I think he was implying that the technology would be concurrent, not years more advanced so no we are not. If i misunderstood my apologies...

The 16mp sensor of the D7000 beat the older 6mp sensor of the D40 hands down.

Evidently we are living in that parallel dimension. :)

I don't know if the 36mp monster will eclipse the 12, but I'm willing to bet the rumored D4 will.

One thing we can count on, is that technology will drive forward and become better, faster, and eclipse previous generations which are only years away and not the decades us mere mortals are measured in.

"so a 36MP FX sensor could conceivably beat a 12MP FX sensor in printed low-light images (the only really fair comparison), even though each photosite has 1/3 the area and a higher signal to noise ratio than the 12MP sensor"

maybe in some evil parallel dimension

Artifacts in photos abound. Noise levels are often referred to but very little is done to describe what the noise LOOKS LIKE. The most unnatural artifacts I have seen introduced to a photo is from the incorrect scanning of film to convert to digital. As most of the scans I have ever done were from 35mm full frame I would think that equates to FX. I do think our experts could do better in describing their aversion to and remedies for noise.

I might disagree with the last paragraph there, jonnyapple. To quote Drab above: "Sensor size does not matter in the pixel noise game. Individual photosite size matters."

The size of the photosite (pixel size) has a far (FAR!) greater impact on SNR than the number of pixels, all other things being equal (same technology, etc.) That's why FX sensors generally outperform DX sensors with similar number of pixels.

Another reason, as I mentioned in a previous post, is that when we downsample that 36mp image to a 12mp image, the reduction of noise in real-life isn't as much as one might expect.

See for example real-life testing done by Phil Askey on this subject. He showed that a 2:1 reduction in image size only yielded a 4% improvement in SNR. You'd have to quadruple the sensor size to see a 20% SNR improvement when downsampling. These numbers are far less than most forum discussions often suggest (e.g., 1-stop noise improvement cited above for the hypothetical D800 when downsampled by 2.25:1).

I'm hoping the D800 will actually debut some new noise control technology vs. the D7000, and not depend on getting better SNR from mega downsampling.

I'm sorry I misunderstood you, then. I thought you were saying that larger photosites wouldn't have a better signal to noise ratio than smaller ones. I'm still fairly certain I disagree with you, though, as I discussed in this topic:
http://nikonrumors.com/forum/topic.php?id=2235

Maybe we agree without me knowing it, though. More megapixels means you have more information for the noise reduction algorithms you mentioned, so a 36MP FX sensor could conceivably beat a 12MP FX sensor in printed low-light images (the only really fair comparison), even though each photosite has 1/3 the area and a higher signal to noise ratio than the 12MP sensor.

Easy. D3s has better SNR vs. the D300 in large part because it has a larger pixel size, 8.45 µm^2 vs 5.5 µm^2.

The situation Drab is talking about however is different. It assumes two cameras (D7000 vs hypothetical D800) having comparable pixel size & technology, but with the D800 having a larger sensor size.

I don't think you've got the right idea about signal to noise, Ade. Most noise sources don't increase as fast as the signal does when scaling up sensor size, and some sources wouldn't change much at all (shot noise on the ADC, for example). With your explanation, how could you account for the differences in signal to noise between, say, a 12MP D300 an a 12MP D3s? There is no question that signal to noise gets better for photodetectors as sensitive area (volume, actually) increases.

Noise reduction via downsampling sounds good in theory, but doesn't really pan out in real life. Partly because there are many kinds of noise, not all of them are random, and not all of them can be effectively reduced by downsampling.

This notion also assumes the "best case" scenario where noise is "averaged out" by the downsampling algorithm, but in photography we typically don't use averaging when reducing pictures as it doesn't look as good "to the eye" (bicubic is the most common approach used in printers, for example).

While a 2x larger sensor gathers more light, it also gathers more noise! Hence the measurement is a ratio (signal-to-noise) which is independent of sensor size.

Also, you can't really get 1-stop improvement in noise by downsampling a picture from a larger sensor in half. Even in the theoretically perfect case scenario I think the amount of noise would be reduced by the square-root of the size difference.

Worse, in empirical "real-life" tests I've seen, a 2:1 reduction in picture size only reduces noise by about 5%. Hence many of us still buy expensive noise-reduction software.

Thanks for clearing all of that up, Drab!

studio460 said:
Just as a telescope with a 12" mirror, will gather more photons than a telescope with only a 6" mirror, correct?

Right.

Drab said:
Picture noise will be ~1 stop better on the D800, as the sensor is 2.25x as large and will have gathered 2.25x as much light (information).

Right. I'm confusing photosite size with sensor size. Just like a telescope, right? A bigger overall sensor (even if consisting of about the same pixel-density), still represents more light-gathering "wells," in aggregate. Just as a telescope with a 12" mirror, will gather more photons than a telescope with only a 6" mirror, correct?

Ok, I read an article on DX vs FX. I guess I never fully understood the differences, though in principle I did. But thanks for your info, it helped understand the article I was reading.

studio460 said:
Well, that's my question as well, Drab. If the pixel-density is approximately equivalent between the two sensors, DX and FX, aren't the noise levels also going to be approximately equivalent?

This comes right back to the point I failed to make clear: Which noise levels are being asked about?

Noise is measured by the ratio of intentional signal to random signal.
Intentional signal = electrical charges generated by actual captured photons.
Random Signal (generally called noise, but I'm avoiding that to stem confusion) = electrical charges generated by leakage currents and thermal pollution.

Pixel noise will be approximately equivalent (need I disclaim again "assuming the same sensor technology"?) as a D7000 photosite will receive the same number of photons as a D800 photosite.

Picture noise will be ~1 stop better on the D800, as the sensor is 2.25x as large and will have gathered 2.25x as much light (information).

Drab said:
The D800, if it comes as rumored, however will basically have D7000-sized photosites. Meaning pixel-noise should be pretty similar between the two (assuming similar sensor technology). Picture-noise should be lower on the D800, however as the larger sensor (again think of the two photos I described before) allows more light, more signal, per photo than the DX one would.

Well, that's my question as well, Drab. If the pixel-density is approximately equivalent between the two sensors, DX and FX, aren't the noise levels also going to be approximately equivalent? In other words, the presumed 36MP D800 sensor, should have about the same noise floor as the DX sensor of similar pixel-density (e.g., D7000), yes? I mean, won't the D800, as expected, basically be an "FX" D7000, simply yielding larger print sizes?

mirtos said:
If this has been explained somewhere else, please just point me to a link. I'd like to understand a comment I have been seeing on the main rumors section when people talk about the d800. How its ISO in FX would be equivalent to a higher ISO in DX.

Ehh, define noise. ;)
Not to derail the question off the bat, but a clear definition of noise is crucial to understanding the question I believe you are asking. Think of a 24MP D3X and a 12MP D300. Use both cameras equipped with equivalent field of view lenses, let's say a 35mm lens on the D3x and a 24mm lens on the D300. Take the same picture, print the same size.

Now you have two photos, both 8"x10" let's pretend. One is made of 24 MP, one is made of 12 MP.

It is not hard to visualize how the photo with twice as many pixels could have significantly more noise per pixel than the photo with less pixels, yet still have less noise per picture. Because each pixel is that much smaller, and therefore less significant to the overall picture.

So what is noise? ;)

mirtos said:
FX is a larger sensor, so I would think that it would need more ISO to have equivalent noise (or lack thereof) than a smaller sensor. Or am I missing something?

Sensor size does not matter in the pixel noise game. Individual photosite size matters. Historically this has been indistinguishable from sensor size as Nikon has never made a FX sensor with photosites as small as their DX sensors. The D800, if it comes as rumored, however will basically have D7000-sized photosites. Meaning pixel-noise should be pretty similar between the two (assuming similar sensor technology). Picture-noise should be lower on the D800, however as the larger sensor (again think of the two photos I described before) allows more light, more signal, per photo than the DX one would.

If this has been explained somewhere else, please just point me to a link. I'd like to understand a comment I have been seeing on the main rumors section when people talk about the d800. How its ISO in FX would be equivalent to a higher ISO in DX.

FX is a larger sensor, so I would think that it would need more ISO to have equivalent noise (or lack thereof) than a smaller sensor. Or am I missing something?