I guess my other comment got lost. What I was saying is that 12 megapixels entails more detail too, so it sort of cancels out with the more noise at high ISO. In other words, you need to downsample to 6 MP to compare with same-ISO images of the f31, or to print at the same size. And in low ISOs, you do gain a detail advantage.
More pixels in the same size sensor is good, as long as dynamic range keeps the same. Especially now that storage space is so cheap. More per-pixel noise doesn't mean more image noise.
andyo, you're meaninglessly digressing. Simple physics shows that the smaller 12mega sensor receives less photons per pixel than a useful 5-6 mega sensor. In order to make the 12mega sensor usable, you then have to amplify readings raising the noise floor and increasing those "hot" pixels.
It's pointless to have more detail when you can't see any of it. It's even more pointless to downsize through resampling in order to "average" out detail that wasn't there to begin with. One of the main strengths of the SuperCCD is for high ISOs, not low; it almost totally blows out at low ISOs.
Just browse dpreview.com and you're see empirically that bigger sensors and lenses are the way to go, not higher megapixels.
This camera has 2x more pixels than the F31. That means each pixel has half the area. And it means the camera needs twice as much light to produce the same result. As a result, unless this camera has a very improved sensor (which I really doubt), it's ISO 200 is equivalent to F31's ISO 100 in terms of sensitivity, but with much more noise.
That not only increases noise, but also lowers dynamic range, contrast, and color accuracy.
And finally, as Yoshi has already said the bottleneck is always the lens construction, so it doesn't matter how many pixels your sensor has, what defines the sharpness of the image is always the lenses.
As you can see increasing resolution is not always the best option.
The 6 megapixels and good optics of the F31 are sufficient for almost everyone except maybe professionals and very demanding amateurs.
How many comparisons of downsampled images to the same pixel dimensions have you seen? All of them are at 100%. Look at it this way. You'll see more noise in the 100% crops, but there's also more detail. The object in the picture is being shown by more pixels. I never said you get more detail at high ISO in the image. I said that more pixels entails more detail, but it sort of cancels out with the more pixel-level noise. Hence, probably no more image detail, but also no less.
And at low ISO, you *do* get more detail. Also, more pixels means better handling of resampling, like when you straighten a picture or correct perspective.
Dynamic range is another question altogether. If there's lower DR, then nothing can be done, but people complain about pixel-level noise, when they don't realize that objects in tehe image themselves are being drawn by more pixels.
You can't possibly believe what 95% of dpreviewers say. Look up John Sheehy's posts on the subject, mostly on the 40D forum. He knows what he's talking about.
To make something else clear (regarding Ricardo's post). Even if the lens doesn't resolve enough detail for such a sensor (which is doubtful, but anyway), you will preserve more detail when doing the resamplings I mentioned in the previous post. But in any case, the lens might be cheaper, but it's also easier to make. Compact camera lenses have more resolution than dSLR lenses. I don't know if this one in particular is good enough or not, but I wouldn't go about making blanket statements like that.
Reader Comments (Page 1 of 1)
andyo @ Oct 26th 2007 9:30AM
I guess my other comment got lost. What I was saying is that 12 megapixels entails more detail too, so it sort of cancels out with the more noise at high ISO. In other words, you need to downsample to 6 MP to compare with same-ISO images of the f31, or to print at the same size. And in low ISOs, you do gain a detail advantage.
More pixels in the same size sensor is good, as long as dynamic range keeps the same. Especially now that storage space is so cheap. More per-pixel noise doesn't mean more image noise.
NuShrike @ Oct 26th 2007 4:48PM
andyo, you're meaninglessly digressing. Simple physics shows that the smaller 12mega sensor receives less photons per pixel than a useful 5-6 mega sensor. In order to make the 12mega sensor usable, you then have to amplify readings raising the noise floor and increasing those "hot" pixels.
It's pointless to have more detail when you can't see any of it. It's even more pointless to downsize through resampling in order to "average" out detail that wasn't there to begin with. One of the main strengths of the SuperCCD is for high ISOs, not low; it almost totally blows out at low ISOs.
Just browse dpreview.com and you're see empirically that bigger sensors and lenses are the way to go, not higher megapixels.
Ricardo @ Oct 26th 2007 6:28PM
It is not the same thing.
This camera has 2x more pixels than the F31. That means each pixel has half the area. And it means the camera needs twice as much light to produce the same result. As a result, unless this camera has a very improved sensor (which I really doubt), it's ISO 200 is equivalent to F31's ISO 100 in terms of sensitivity, but with much more noise.
That not only increases noise, but also lowers dynamic range, contrast, and color accuracy.
And finally, as Yoshi has already said the bottleneck is always the lens construction, so it doesn't matter how many pixels your sensor has, what defines the sharpness of the image is always the lenses.
As you can see increasing resolution is not always the best option.
The 6 megapixels and good optics of the F31 are sufficient for almost everyone except maybe professionals and very demanding amateurs.
andyo @ Oct 26th 2007 7:38PM
How many comparisons of downsampled images to the same pixel dimensions have you seen? All of them are at 100%. Look at it this way. You'll see more noise in the 100% crops, but there's also more detail. The object in the picture is being shown by more pixels. I never said you get more detail at high ISO in the image. I said that more pixels entails more detail, but it sort of cancels out with the more pixel-level noise. Hence, probably no more image detail, but also no less.
And at low ISO, you *do* get more detail. Also, more pixels means better handling of resampling, like when you straighten a picture or correct perspective.
Dynamic range is another question altogether. If there's lower DR, then nothing can be done, but people complain about pixel-level noise, when they don't realize that objects in tehe image themselves are being drawn by more pixels.
You can't possibly believe what 95% of dpreviewers say. Look up John Sheehy's posts on the subject, mostly on the 40D forum. He knows what he's talking about.
andyo @ Oct 26th 2007 7:48PM
To make something else clear (regarding Ricardo's post). Even if the lens doesn't resolve enough detail for such a sensor (which is doubtful, but anyway), you will preserve more detail when doing the resamplings I mentioned in the previous post. But in any case, the lens might be cheaper, but it's also easier to make. Compact camera lenses have more resolution than dSLR lenses. I don't know if this one in particular is good enough or not, but I wouldn't go about making blanket statements like that.