Anonymous, 12 Jun 2021"a sensor with a 4um pixel and a 250mm/f10 lens will always perform better than a sensor ... moreTherotical joker returns..
You have zero idea about image processing Algorithms, judging a camera based on hardware alone is like judging a car based on engine alone.
Anonymous, 13 Jun 2021Personal attack is a no no. I hope you don't fail too much in society and at work with th... moreIt was not a personal attack. Disregarding image quality in a thread about image quality is a plain lack of common sense. The whole story started with image quality in focus.
Anonymous, 13 Jun 2021This Kuro guy is still confusing pixel size with ISO.I suggest you should first understand what is ISO, and how a cmos image sensor works. Then rhe relationship to the pixel size will become obvious.
Kurosava, 12 Jun 2021No, I thought you have common sense. Personal attack is a no no. I hope you don't fail too much in society and at work with this kind of attitude.
Kurosawa, 11 Jun 2021Which means, you will need much longer exposure time to get the comparable image. Actually you... moreWhile you're not wrong, I don't know why you're telling me that since that's not what we have been discussing about. I end our argument here, have a great day.
Kurosava, 12 Jun 2021You have really no idea how pixels work.
Smaller pixels have larger noise because of other f... more"I mentioned it a few times already, but you chose to ignore it."
No, I didn't ignore it, but you were not able to provide a source. As I said the sources of noise are photon noise, read noise and dark current noise. Read noise and fixed pattern noise is the noise you capture when you use a lens cap and a short exposure time, dark current noise is the additional noise you capture when the exposure time gets longer. There isn't another major noise source, unless you can provide a reliable source that says that there is another major noise source apart from read noise and dark current noise. The noise that is caused by the analog to digital conversion is read noise. It also doesn't matter what the reason for read noise, etc is. Small pixels don't really have more read noise than large pixels as I mentioned. That's why equivalence theories work. And this is also supported by measurements from photonstophotos.net . Smaller pixels can even have less read noise.
"Smaller pixel = lees photons = less shot noise"
Less photon noise per pixel, but not per object. The photon noise per object stays constant, no matter which pixel size.
Anonymous, 12 Jun 2021I didn't say that one can ignore read noise. And no, read noise isn't always the dom... moreYou have really no idea how pixels work.
Smaller pixels have larger noise because of other factors - like smaller conversion capacitance. Which is different from readout noise. I mentioned it a few times already, but you chose to ignore it.
Then what you describe is called photon shot noise and it has a lot to do with the pixel size. It is actually better for smaller pixels, since they capture less photons. If number of photons hitting the pixel is S, then photon noise is sqrt(S) and is in direct relationship with the active area of the pixel. With abundance of light it is the dominant noise source. Smaller pixel = lees photons = less shot noise. But then a larger pixel achieves the same by reducing exposure time.
Kurosava, 12 Jun 2021I though that striving for image quality is self-explanatory and does not need to be mentioned... moreYou think we can read your mind? You said small pixel size makes exposure time longer. That statement is wrong. Period. I rest my case.
Anonymous, 12 Jun 2021I didn't say that one can ignore read noise. And no, read noise isn't always the dom... moreIf you paint noise on your wallpaper and then you want to take a photo of your wallpaper (perfect light conditions), you wouldn't expect to see a better signal to noise ratio in your wallpaper photo, if you use a camera with a different pixel size. Though, this example isn't perfect because it only works, if the effective diameter of the lens stays constant.
Kurosava, 12 Jun 2021First, you can't ignore readout noise, since it is one of the major noise contributors. S... moreI didn't say that one can ignore read noise. And no, read noise isn't always the dominant noise source. In good light conditions without shadows, read noise isn't important. Also, read noise isn't necessarily that important, if there is a lot of light pollution. Furthermore, when you take a photo of an object with a 4 micrometer camera + 250mm lens, the object will have a similar amount of read noise as a 0.8 micrometer camera + 50mm lens because in both cases the object consists of the same number of pixels and the read noise per pixel will be similar (between 2 and 1 electrons per pixel). For example, an older iPhone or Pixel phone has a similar amount of read noise per pixel as a Canon M50 according to photonstophotos.net . Apart from photon noise and read noise, there isn't another major noise source. There is dark current noise, but this is only relevant for long exposures and I have seen no source that says that smaller pixels have more dark current per pixel. Indeed, there are many specialized astrophotography cameras with smaller pixels than Aps-c cameras. And not even every camera with large pixels has back side illumination.
"I am really curious how the noise of the object does not depend on the noise of the pixel that captures it."
The effective diameter determines how much light per time can be captured from an object. This doesn't depend on the pixel size as a sensor can only capture the light that goes through the lens. Photon noise is the variation of the number of photons that are emitted/reflected per time, so it has nothing to do with the pixel size. Of course the pixel size doesn't affect the photon noise of an object (unless the pixels are less effective and lose light), otherwise it wouldn't be called photon noise, but read noise or something else. Photon noise is part of the reality and affects our eyes as well.
Anonymous, 12 Jun 2021"a sensor with a 4um pixel and a 250mm/f10 lens will always perform better than a sensor ... moreFirst, you can't ignore readout noise, since it is one of the major noise contributors. Second, even if you ignore the readout noise, a smaller pixel will still have larger noise from other sources like kTc. This is simple semiconductor physics. Then as I already said you will have bloom due to optical crosstalk. Especially when pixels become so small that they are comparable to the wavelength. Plus all the other drawbacks like dynamic range.
The only advantage is the increase sensor MTF, which will allow you to see smaller objects. I am really curious how the noise of the object does not depend on the noise of the pixel that captures it.
Kurosava, 12 Jun 2021Yes, you can capture only as much light as the lens permits. But pixel size makes difference h... more"a sensor with a 4um pixel and a 250mm/f10 lens will always perform better than a sensor with 0.8u pixels and a 50mm/f2 lens"
No, small pixels don't have more read noise per pixel than large pixels. Read noise is the important factor. I have never heard that the fill factor is worse for small pixels. If the fill factor was significantly different for small pixels, then simple camera equivalence theories wouldn't work at all, but they do work for cameras of the same generation.
When you take a photo of a distant object, one doesn't care about the noise per pixel, but the noise of the object. These are totally different things. So, when I ignore read noise and demosaicing and judge the noise of a distant object, it doesn't really matter whether I use a 125mm f/5 lens, a 250mm f/10 lens or a 500mm f/20 lens and it doesn't really matter which pixel size I use. Only when you take read noise into account, a high number of pixels per object will lead to significantly worse results in low-light conditions.
Anonymous, 11 Jun 2021Note that pixel size or sensor size does not affect how much light per time you capture from a... moreYes, you can capture only as much light as the lens permits. But pixel size makes difference how good you can utilize the light that enters the lens. Smaller pixels capture less photons than large pixels. Then smaller pixels have larger noise. So total signal to noise ratio is worse. So no, a sensor with a 4um pixel and a 250mm/f10 lens will always perform better than a sensor with 0.8u pixels and a 50mm/f2 lens, as long as all the rest is the same. The signal level will be the same, but the noise for the smaller pixel will be higher. So overall SNR will be worse. And this is on individual pixel level. If you consider optical and electrical crosstalk gets worse with pixel size going down, then the picture gets even worse.
Anonymous, 12 Jun 2021That wasn't what you said. I quote "With this pixel size all exposures will ve long&... moreI though that striving for image quality is self-explanatory and does not need to be mentioned explicitly. In the end, the purpose is to capture a nice image, not just some noisy gibberish. If the purpose is to have bad images with poor color accuracy, narrow dynamic range, lots of noise and so on, then small pixels are much better than large ones.
Kurosawa, 11 Jun 2021Which means, you will need much longer exposure time to get the comparable image. Actually you... moreThat wasn't what you said. I quote "With this pixel size all exposures will ve long". You didn't mention anything about image quality.
Anonymous, 11 Jun 2021Note that pixel size or sensor size does not affect how much light per time you capture from a... moreNote that even 0.8 micrometer pixels + 250mm f/10 lens or a 2500mm f/100 lens would still capture the same amount of light per time from a small galaxy, the difference would be that the galaxy consists of many more pixels and this leads to more read noise.
Note that pixel size or sensor size does not affect how much light per time you capture from an object, if you use exactly the same lens. A sensor can only capture the light that goes through the lens. So it doesn't really matter whether you photograph a galaxy with small pixels or large pixels as long as the effective diameter of the lens is the same and as long as the number of pixels per angle is the same. When you take a photo of a galaxy, a 54 megapixel full frame camera with 4.0 micrometer pixels and 250mm f/10 lens will perform similar to a tiny sensor with 0.8 micrometer pixels and 50mm f/2 lens.