Does pixel size matter?

[TrishBJ]

I'm looking for advice on the importance, or otherwise, of pixel size. Does a sensor with bigger pixels perform better in low light situations than a sensor with smaller pixels for example, while the smaller pixels do better when capturing fine detail in good light? What would be the performance differences between, for example, a 24 megapixel full-frame camera versus a 36 megapixel full-frame camera. I imagine that given the full-frame sensor size is the same in each case, the higher pixel number in the same space means that the pixels are smaller.

Thanks,
Trish

[Manfred M]

If you had two virtually identical sensors (other than the size of the individual sensor elements), then yes, the larger pixel size would be more sensitive. This is simple physics at play; think of two buckets sitting beside each other, outside in a rainstorm. The bucket with the largest diameter would capture more rainwater than one with a smaller diameter.

There is a price paid for this increased light sensitivity. The resolution of the larger pixel sensor elements (fewer pixels) would be lower than the one with smaller pixel elements (more pixels).

The issue is that different sensors across different cameras and generations of cameras are not directly comparable. As an example the full-frame Nikon D700 with its 12MP sensor had slightly worse high ISO performance than the full-frame D810 with its 36MP sensor (two different generations of technology). So each camera has to be judged individually.

As I am primarily a landscape, architecture, portrait and street photographer, high ISO is less important to me than it would be to a sports photographer or a wildlife photographer. That doesn't mean I don't shoot high ISO shots; I rarely shoot above ISO6400 without a tripod, so I don't need what I would refer to as "stupid high" ISO values. As an ex-film shooter, I managed to get pretty decent shots with much lower maximum ISO films.

[dje]

I imagine that given the full-frame sensor size is the same in each case, the higher pixel number in the same space means that the pixels are smaller.

Yes Trish that is correct. For example, the 36MP D810 has a pixel size of 4.9 um and the 24MP D610 or D750 has a pixel size of 5.9 um.

For a given sensor technology, the larger pixel size gives a higher Signal to Noise ratio (S/N). The lower the S/N, the more likely it is for the noise to be noticeable in a given image. This only becomes an issue for high ISO settings (eg 3200 and above for a FF camera). It's the ISO setting that is important rather than the amount of light. Satisfactory exposure can be achieved in low light with low ISO, wide aperture and long shutter speed which is fine for stationary subjects and a tripod mounted camera. But for sports and wildlife, the use of high ISO is often the only way to get good exposure with a satisfactorily fast shutter speed.

When considering S/N, the viewing conditions also enter into it. eg noise is more noticeable viewing at 100% on a computer monitor than on a 7x5 print. Noise can (and is) smoothed out in processing (either in camera or in a raw processor) but excessive noise reduction can lead to a drop in sharpness.

The difference in S/N between the D610/D750 and the D810 is about half a stop max, not a huge difference.

On the other hand, smaller pixels do give finer sampling of the image produced by the lens on the sensor surface. Nikon's approach with the finer sampling associated with the smaller pixels (D810) is to leave out an anti-aliasing filter which in turn leads to more sharpness. The AA filter is used to prevent artifacts such as moiré occurring when fine regular patterns are found in an image. The finer the sampling, the less likely it is for artifacts to occur. The possibility of moire is not terribly important for landscapes but is more important for say fashion photography.

So it's a bit of a trade-off, and depends what is important with the style of photography for which the camera is to be used.

Dave

[lukaswerth]

This is funny. Not having posted here for some weeks, I just opened CIC in order to open a thread about pixel size. Now I attach my thoughts to this one.
I have a Nikon d800e, my very first dlsr, as I had remained analog for my expressive photography longer than most people (still thinking about analog work from time to time, might somewhat, partially return to it).
I had bought the d800e because of the possibility of printing large. Whow, like medium format, I thought. I like large prints which still remain sharp when you inspect them up close, and I like to think I have this possibility, even if most of my prints remain smaller.
Recently, however, I found myself thinking about having a Fuji x pro 1, a camera concept I also liked. But 16mp? I googled what other people had to say about their Fuji sensors, I found to my amazement that many don't care much about the 24mp sensor a rumored x pro 2 is about to get. Their petty 16 is quite enough, they think.
So I printed some larger details from a d800e imagefile. I printed the same detail at 100%, 200%, 300% (the entire print would have been about 120x180cm/48x72") not interpolated and interpolated all at once (bicubic smoother), and in 3 steps with some sharpening and bicubic automatic - nothing fancy, just roughly, and not genuine fractals or such like.
I went quite compulsive inspecting these test prints, putting two of my near glasses over each other so I could look at them from a distance of 5 inch or so: the difference in all the small prints was, for me, surprisingly little. Even without interpolation a 250% enlargement without interpolation the detail looked quite decent wen inspected from indecently close, hardly any pixelation. And honest to those Olympic gods or whoever divinity else, the 300% 3-step interpolation looked just as a 100% one to me; would I not have done it myself, I would not have known the difference. This means, a Fujifilm 16MP file can be enlarged up to about 80x120cm/32x48" with the same quality. Quite an exhibition size by any standard!
For what the hell then do we need 36MP? Well, there is one thing: I often like to squeeze anything out of my files in development/PP, torture them some might say: steep curves, stark separations by masks, contrast, colour, everything squeezed and stretched to the utmost. When I get lines on the edges I try to get rid of them, but sometimes there is only so much I can do. Then there is the noise of 3200 ISO for such files it is probably not a bad thing to have more megapixels so they can still be blown up to a decent (or indecent) size (the file from which I did the enlargement test was a clean 100 ISO one). And, of course, there might be the occasional cropping.
But then the Fujifilm sensor is also supposed to be quite good at high ISOs.
Another point I am thinking about in this regard is the lenses. My test file was done with a Nikon 50mm 1.8 G at f/5.6 or 8 - sure as hell I am not going to get one of those Sigma Art or even an Otus. Why should I? In the next days, I am rather going to look more closely at some old lenses in my drawer, and I will look for the qualities of haloes, blurs, not-quite-so-technical-sharpness, and I am going to see what difference that makes in relation to what I want, what I expect from my pictures. I might even finally realize a plan I am harbouring since some time, to make some technically really bad lens myself with a fancy diaphragm to boot (I have an old Kodak instamatic which longs to be cannibalized by me). For a long time I made, and enjoyed, and sold, pinhole prints - that's not that effective with digital, but what I enjoyed was the personal prints I got.
Perhaps my overkill of 36MP will just be helpful of rendering very smooth blurs!

Lukas

[IzzieK]

When I was shooting less and everyone seems to come up with better images than I did, I thought more pixels matters. Now since my D800e, now D810 and using my D300s and D90 more, it does not matter to me much anymore, only sometimes...

[Manfred M]

Does pixel size matter?

Maybe, but certainly not if all you plan to do is display the images on a computer screen. A 10MP camera is overkill there.

Now, if you print to large format (which I do), then it does start to matter. My minimum print size for display is A2 / 17" x 22" and when I use my Epson 3880 (native resolution is 360 ppi), I actual start off with a slight amount of enlargement, but the output is darn close to contact print size with images out of my D800. Here you can start seeing the image quality.

Go larger, it becomes even more apparent when close up. The high MP image cameras will allow you to create huge prints with great IQ (comparable to medium format, in some cases). On the other hand. Very few people print and when they do, it is smaller sizes.

Again, honestly, a 12MP - 16MP camera is all most people probably need.

[DanK]

I agree with most of what has been written, but the thread discusses two different things: pixel size and pixel count. The two are obviously related, but your question was about pixel size--really, photosite size

I think the short answer is that for most people, it doesn't matter often, if at all. I say that as someone who shoots with both crop sensor and FF format cameras, in my case, a Canon 5D3 and a 7D (first gen). Yes, the 5D3 handles low light better, and for some purposes, that is very nice. I try not to use noise reduction, and I like having the extra stop or two that I can shoot without it. It will handle large prints somewhat better, but I don't print as large as Manfred, and so far, I have printed only one image for which I think it would have mattered.

I also prefer a large-photosite camera for one other purpose: night photography. I do exposures up to 10 minutes, and in hot weather, the sensor will tend to overheat. That problem is worse with cameras that have small, densely packed photosites.

However, for many purposes, it just doesn't matter, or it matters very little. For some purposes, I deliberately choose the crop-sensor camera because for some uses, the greater pixel density and lighter weight are both plusses.

I am a member of a very old camera club that has some truly excellent photographers. One of the people who routinely wins awards and who does many types of photography far better than I do shoots with a Canon 40D--a crop sensor camera that is now 3 generations out of date. Recently, I inspected some fairly large prints someone had done with images from a small Fuji, and they were beautiful.

So, my way of thinking about it is that larger photosites offer some advantages, but not enough to make all of the other considerations in choosing a camera moot. The people who think they have to get a FF camera to do well are fooling themselves.

[xpatUSA]

Yes Trish that is correct. For example, the 36MP D810 has a pixel size of 4.9 um and the 24MP D610 or D750 has a pixel size of 5.9 um.

For a given sensor technology, the larger pixel size gives a higher Signal to Noise ratio (S/N). The lower the S/N, the more likely it is for the noise to be noticeable in a given image. This only becomes an issue for high ISO settings (eg 3200 and above for a FF camera). It's the ISO setting that is important rather than the amount of light. Satisfactory exposure can be achieved in low light with low ISO, wide aperture and long shutter speed which is fine for stationary subjects and a tripod mounted camera. But for sports and wildlife, the use of high ISO is often the only way to get good exposure with a satisfactorily fast shutter speed.
Dave

I can vouch for the note about noise reduction in the coversion from raw to RGB. I own the noisiest camera on the planet - a Sigma SD9. It's raw-only X3F files are horribly noisy when converted by DCraw, barely tolerable in Sigma Photo Pro and not too bad in Adobe Camera Raw. In addition the "red" channel is sharpened in-camera, whether you like it or not, so that doesn't help.

Foveon technology breaks the "rule". The Sigma SD9 has a 9.12um pixel pitch - bigger even than the much-vaunted Nikon D3/D700! (only 54% fill factor though). However, the extreme processing needed to convert from camera space to RGB totally negates any size advantage

[PhotomanJohn]

Does a sensor with bigger pixels perform better in low light situations than a sensor with smaller pixels for example, while the smaller pixels do better when capturing fine detail in good light?

Trish - The answer to this question used to be yes but that is changing. The sensor surface is composed of light sensitive areas (photo sites) and areas between the photo sites for the signal interconnections to the photo sites. Originally the light falling on the areas between the photo sites was lost. The more pixels the sensor had, the more area was needed to carry the interconnections reducing the percentage of light sensitive surface area of the sensor.

Today the designs are much more efficient. The area needed for the signal interconnections has been greatly reduced so the percentage of active area on the sensor doesn't suffer as much as the number of pixels increases. In addition, modern sensors have a micro lens structures over the surface that funnel the light into the photo sites so the light falling on the inactive areas is not lost. Even though the smaller photo sites collect less light each, there are more of them. So when the larger files is down-sampled for viewing or printing, the smaller pixels are added together to be effectively the same as larger pixels.

So today the low light capability of the sensor has little to do with the number of pixels. The larger number of pixels is able to capture more detail in the image which is a plus. I shoot with both 36 and 24 mega-pixel cameras. The 36 mega-pixel camera gives me a somewhat greater ability to crop my images. But, as has been said by others, most of the time I would do just fine with a camera with less resolution.

I hope this helps and doesn't add to the confusion.

John

[TrishBJ]

Absolutely invaluable information from all of the replies so far!
I currently have a D7000 which is hugely better than my old D70, but still find that noise can be at an annoying level for images taken in lowish light. Most of my photography is landscapes, sunset/sunrises, wildlife especially birds, and people. I'd like to have a go at some wide-field astronomical images such as the Magellanic Clouds or things like planetary conjunctions for which the D7000 doesn't really work well (though it would do better with faster lenses). For images of birds and other wildlife I often have to crop tightly as they may not be conveniently nearby. If I've understood everyone correctly, with the most modern sensors there is little difference between a FF 24 MP and FF 36 MP (all other things being equal) which means that one can't really go so very far wrong in decision making. This is most comforting!

[lukaswerth]

Does pixel size matter?

Maybe, but certainly not if all you plan to do is display the images on a computer screen. A 10MP camera is overkill there.

Now, if you print to large format (which I do), then it does start to matter. My minimum print size for display is A2 / 17" x 22" and when I use my Epson 3880 (native resolution is 360 ppi), I actual start off with a slight amount of enlargement, but the output is darn close to contact print size with images out of my D800. Here you can start seeing the image quality.

Go larger, it becomes even more apparent when close up. The high MP image cameras will allow you to create huge prints with great IQ (comparable to medium format, in some cases). On the other hand. Very few people print and when they do, it is smaller sizes.

Again, honestly, a 12MP - 16MP camera is all most people probably need.

Manfred,

I have the same camera and the same printer, and I cannot help noting that your post ends up hugely exaggerating the difference 36MP make at the enlargement size you mention. Let me put it like this: A 16 MP file would perhaps benefit from a little interpolation at 17x22", but that would probably not create any issue for this very modest enlargement, and given a good camera like the Fujis, I would very much doubt that you would be able to notice any difference however close you get.
I don't know at which enlargement a noticeable difference would start creeping up, but my rough guess would be not before 1x1.5m or so, and then only if you inspect the print very closely.

We all like our big and wonderful cameras, and the d800/e has this marvellous dynamic range which is, for me, when all is said and done, even more important than the pixel rate, but I don't think any more the pixel rate makes the difference in practice we like to think it does.

Lukas

[Manfred M]

Manfred,

I have the same camera and the same printer, and I cannot help noting that your post ends up hugely exaggerating the difference 36MP make at the enlargement size you mention. Let me put it like this: A 16 MP file would perhaps benefit from a little interpolation at 17x22", but that would probably not create any issue for this very modest enlargement, and given a good camera like the Fujis, I would very much doubt that you would be able to notice any difference however close you get.
I don't know at which enlargement a noticeable difference would start creeping up, but my rough guess would be not before 1x1.5m or so, and then only if you inspect the print very closely.

We all like our big and wonderful cameras, and the d800/e has this marvellous dynamic range which is, for me, when all is said and done, even more important than the pixel rate, but I don't think any more the pixel rate makes the difference in practice we like to think it does.

Lukas

Lukas, you are misreading my answer. I don't feel that I print to a large size at all; but others I have spoken to feel that A4 / 8-1/2" x 11" is a large print and the A2 / 17" x 22" size from an Epson 3880 is huge, when it is not. I was merely pointing out that size wise those sizes are effectively almost a contact prints size and more or less the starting point where up sampling takes place. I've seen d800e prints that were 1 m / 40" in size that show incredible detail; to me those are large prints.

I've also printed to that size using 2x and 1.5x crop frame images which look great, but not quite as great as an image done from the FF. This could be due to a number of reasons, but I suspect up sampling softening to be at least part of what I am seeing.

[lukaswerth]

Manfred,

okay, I get you. Also, I must admit I find your remarks somewhat reassuring concerning my camera choice. Let me put it like this: I actually also hope you are right. And, we have the similar ideas about large prints.

When I think back to analog days, I think the d800 series equal 4x5" in detail. God Almighty, when I think about those days of yore and which quality I accepted as a 20x30" print from 35mm...

Lukas