[*isteve]

Very interesting.

How is this, could you elaborate ? How do they relate ?

The architecture and pixel density of a D3X is roughly equivalent to that of a 12MP APS camera.

[CFWhitman]

resorting to f stop/fov is tempting, i was about to fall into that trap too, but i thought again about what jeffkrol said (i had a vague impression from reading past threads that he is not exactly a troll ), and after some consideration it suddenly made sense. let's try to look at it this way:

Bringing angle of view and the size of the glass is valid, so it is not a trap to fall into. I certainly never thought that jeffkrol was a troll. He has interesting information to bring to the discussion.

you have a photosensitive area, we do not care about the individual pixels just now, let's assume there's an infinite number of them, and we can agregate them as we wish (or the other way around:divide the area into pixels as we desire). think of aps-c vs digital p&s (huge difference), to make it more obvious. now, assume we don't care about field of view for now, all we want is to obtain some image on the sensor, so we take a 50mm/2.8 with an image circle very capable of covering both, we assume it is perfect (no focus plane curvature, no vignetting,etc). it is clear that for both sensors, this same lens at the same aperture will deliver the same amount of light per area unit (meaning, the same exposure at the same "iso" will give the same "density" of latent image on the sensors). now comes the issue of useful signal: it is obvious that there will be a smaller absolute amount of information delivered on the fingernail sized p&s sensor as opposed to the big aps-c sensor -- proportional to the absolute area --, so to get the same size of final print, we are actually "stretching" the signal more in the case of the smaller sensor (remember, we are not talking megapixels yet), or, to put it another way, less photons will contribute with information for each area unit of final print, for the same size of print obtained from both sensors, because the density of information (light, not pixels!) per area unit on both sensors is _the same_, so in theory we get less useful information per square inch of print from the smaller sensor, because there simply isn't as much as in the case of the bigger one, to begin with.

This is all true, but it relates to print resolution (be it optical or digital) and field of view rather than ISO performance.

By the information shown in this thread it seems that better high ISO performance from larger sensors has a twofold cause.

1. Higher pixel density introduces more noise because of the number of pixels per unit of area, not because of the size of pixels. At high ISO values there is not enough signal (light) available to compensate for the added read noise. This will give lower pixel density an advantage in high ISO applications until read noise is reduced beyond the level where it will be a factor because improvements to read noise seem to affect large pixels and small pixels equally, and there are always more small pixels. (Little known is that the reverse effect of better low ISO performance can also come from higher pixel density).

2. Larger area sensors allow you to use larger lenses, which can collect more light, to cover the same field of view. This is an advantage to the extent you have longer lenses of the same f-stop available. With the same lenses you can either get a wider field of view where everything is magnified less, including both the subject and the noise, or crop the photo and get similar results (assuming similar pixel density).

This is further proof that in photography, everything is a trade off.

[nanok]

Bringing angle of view and the size of the glass is valid, so it is not a trap to fall into. I certainly never thought that jeffkrol was a troll. He has interesting information to bring to the discussion.

ahem, sorry about that: i was referring to my own brief misconception, not yours (i wouldn't know how you think anyway, and i would not dare to presume, especially as i don't know you anywhere near well enough). jeffkrol's comment does sound trollish at first glance (even only because it is so short an "cutting" ), and yes, the information and insight he brings is definetly very interesting (i know i have learned a thing or two). sorry again for sounding presumptious, not my intention at all.

This is all true, but it relates to print resolution (be it optical or digital) and field of view rather than ISO performance.

the point was that it does not relate to print resolution, and definetly not to field of view, it seems i haven't made myself clear at all . it is also not solely about high iso performance, but more generally about snr (or noise) performance. let me try to put it right:

- print resolution: the above was just an example, think of it this way: print/digital display or whatever you use to see the final result is the same, same size, same resolution, etc for both sensors in the test case above. hence it cannot be about print resolution, as it does not change during the "gedankenexperiment"
- fov: i tried the "one lens" approach to eliminate any temptation about bringing the difference in lenses in the discussion. fov does not influence the outcome of the test, it's just an image, i honestly don't see how the image's fov matters. the quantity of information sent to the sensor is not dependent on the fov in any way, but only on the fstop, it might be tempting to think a wide angle lens will "cram together" more details compared to a tele, but for most intents and purposes, i guess saying that the amount of detail in a real world scene is infinite, at any given scale, is quite accurate (because of the real limitations of our lenses, sensors, and ultimately visible light wavelengths), so we can imagine the scene being used for the test is a true fractal, and at any fov there will be the same amount of detail; further more, this is not even relevant to the test, but it was just to get it out of the way.

By the information shown in this thread it seems that better high ISO performance from larger sensors has a twofold cause.

1. Higher pixel density introduces more noise because of the number of pixels per unit of area, not because of the size of pixels.

the point jeff is trying to push is that higher pixel density does _not_ (have to) mean lower performance (more noise). when pushing iso values up, as stated by you bellow, read noise (amplif. noise) becomes an issue, and things do tend towards "higher density means more noise" in practice. i guess what we are trying to figure out (or at least what i am) is if this limitation is really technical or just "incidental" (is technically easy to overcome, but, for instance, commercially not (yet) desireable/lucrative enough).

if your intended meaning is that, except "read" noise, the higher pixel density means higher noise "per pixel" because, very simply, the amount of information available to the same actual area of sensor is limited and the same, regardless how many pixels you cram in, i think that is true but, as you noted: sometimes, at low iso, higher density appears to deliver better overall performance: could this be because read noise in that case is not relevant and the overall efficiency of the sensor/amplif/adc setup is better with higher density (to some extent), because it gets closer to actually reading the whole information actually available (closer to exhaustively gathering info from the sensor), as now each individual "photosite" receives proportionally less information, but we have a similar "read chain" behind each photosite as with the lower density sensor, so "the match is better made", so to speak?

however (see bellow)..

At high ISO values there is not enough signal (light) available to compensate for the added read noise. This will give lower pixel density an advantage in high ISO applications until read noise is reduced beyond the level where it will be a factor because improvements to read noise seem to affect large pixels and small pixels equally, and there are always more small pixels. (Little known is that the reverse effect of better low ISO performance can also come from higher pixel density).

at the risk of being really annoying: if we are talking about comparing pixel-per-pixel, we are comparing apples to rocks imho. if we are talking about comparing intended output, for two sensors of the same size but different density, it is actually more likely that, due to advances in technology (and who knows due to what other non-technical reasons governing the release of those advances ) the higher density sensor will perform better. can somebody here who owns both k10d and k100d (or other 6mp pentax) do a direct comparison (i know i would be interested)

same intended output means obtaining the same size, type (dpi etc) of print/file from both sources, using all information available from both (ofcourse, the higher mp source will be downsampled to match the lower one, so some information is lost along the way)

2. Larger area sensors allow you to use larger lenses, which can collect more light, to cover the same field of view. This is an advantage to the extent you have longer lenses of the same f-stop available.

for the sake of avoiding confusion: larger area sensors _require_ you to, by design. it is the way they work. you either use those lenses, or you are not using that sensor, but a "crop" of it (so a smaller sensor). this is to say the lens does not define why the sensor gets more light, it just comes as a consequence. the lens simply comes as a design complement to the sensor, failing to fulfill the lens requirements invalidates having a larger sensor (actually makes the larger sensor inexistent, effectively). because we tend to do similar things with different format cameras, and expect them to be more or less interchangeable for most purposes (which, in some cases, and from some point on, is wrong ), we tend to think of lenses in terms of f stop and field of view (so relative terms, so they can be carried over across formats and compared directly), not absolute terms (how much actual light an 80mm/2.8 lens puts on the 6x6cm photosensitive area compared to a 50mm/2.8 on the 24x36mm). further more, regardless of which lens you would use on the smaller sensor, you will simply not be able to get the same snr, for the reasons i explained previously (at least not with normal current technology, i would expect something near the lines of foveon -- multilayered sensors -- to be able to go around it someday, by replacing sheer area with a third dimension, but i fear that day is not so near )

With the same lenses you can either get a wider field of view where everything is magnified less, including both the subject and the noise, or crop the photo and get similar results (assuming similar pixel density).

the "noise" does not come through the lens, as part of the signal on the sensor. everything the lens projects on the sensor is "useful" information, by definition. the noise is "born" at the first receiver in the system, and that is the sensor in our case, even the concept of snr is meaningless if you remove the receptor (sensor, photosensitive area of any kind) from the equation.

This is further proof that in photography, everything is a trade off.

this is true for all practical applications.

let me state again that many of the arguments i make are not to invalidate something you said, it's just to make sure "we're clear" (and sometimes to make sure i am getting it right).

[Jonson PL]

Looking at the high ISO performance of the D3X is some indication of whats possible in the next APSC sensor,

The architecture and pixel density of a D3X is roughly equivalent to that of a 12MP APS camera.

The Nikon D3X has :
24.5 million effective pixels
2.8 MP/cm² pixel density



Pentax K200D has :
10.2 million effective pixels
2.8 MP/cm² pixel density


Nikon D90 has :
12.3 million effective pixels
3.3 MP/cm² pixel density


Nikon D300 has :
12.3 million effective pixels
3.3 MP/cm² pixel density




So basically you’re saying since having close to the same pixel pitch from the APS-C cams, should yield around the same result as D3X on the high Iso aspect

[ManuH]

So basically you’re saying since having close to the same pixel pitch from the APS-C cams, should yield around the same result as D3X on the high Iso aspect

Only the APS-C crop from the D3x should yield the same result as an APS-C cam. If you look at the DxO mark, the D3x has a much better high ISO performance than any APS-C camera. That's because they test for the SNR to a fixed printed size.

[Male Gorilla]

"Pixel density" and "single sensor size" are almost the same only for CCD censors, where the sensor size is 70 to 92 percent of the pixel size. CMOS sensors can have the sensor size varying from 15 to 70% of the pixel size, and this makes huge difference between different kinds.

[Sean Nelson]

So basically you’re saying since having close to the same pixel pitch from the APS-C cams, should yield around the same result as D3X on the high Iso aspect

...except that the image from the full-frame D3X needs to be enlarged less to get a final image of the same size - that reduces the impact of whatever noise is present in the first place.

The only situation in which that's not true is if you consciously decide not to enlarge images as much from an APS-C camera.

[Jonson PL]

Only the APS-C crop from the D3x should yield the same result as an APS-C cam. If you look at the DxO mark, the D3x has a much better high ISO performance than any APS-C camera. That's because they test for the SNR to a fixed printed size.

I haven’t followed DxO Mark closely, but had a look.

As written earlier on Pixel pitch :

The Nikon D3X, has
24.5 million effective pixels
2.8 MP/cm² pixel density

The Pentax K-m has
10.2 million effective pixels
2.8 MP/cm² pixel density

And the K10 around the same.


When I look at the graph over SNR at DxO, the Nikon D3X, Pentax K10 and K-m are fairly close. There is still a difference, but they’re closer than the next jump up to a model like the Nikon D700.


Everybody should go for what would be right for them. But I feel that a better supplement to my APS-C cam, would be a real high-Iso cam like D700.

"Pixel density" and "single sensor size" are almost the same only for CCD censors, where the sensor size is 70 to 92 percent of the pixel size. CMOS sensors can have the sensor size varying from 15 to 70% of the pixel size, and this makes huge difference between different kinds.

Thanks, good point.