 |
brainiac
Offline
[X]
|
jorkata wrote:
If that's the case indeed, larger pixels still rule when it comes to noise, no 
Larger pixels only rule if there are just as many of them. At the sort of sizes we are dealing with here, larger pixels don't necessarily rule, because the sensor size is constant. 10,000 smaller pixels might be less noisy, overall, than 1000 larger pixels. In a sensor of a given generation and total size, changing pixel size has no effect on noise (because the greater number of smaller pixels perfectly compensates for the fact that they are noisier). So, no, larger pixels DON'T RULE, for any given sensor size, they just give you less maximum resolution. Larger sensors rule, larger pixels don't.
Noise is created by the individual pixels, not by the sensor as a whole.
It's the SNR of each pixel that determines how much noise there will be in an image.
But what you ignore is that the more pixels there are, the less apparent the noise of each will be. Given a particular sensel SNR, a hundred of them will give you a much noisier picture than a thousand. For any given sensor size, the number of sensels changes in direct proportion to their area, so the adverse effect on pixel SNR of shrinking pixels is perfectly compensated by the larger number of sensels on the sensor. That's why it's not worth considering what sensel size does to noise, only sensor size and age matter.
All things beeing equal, larger pixels have better SNR - and the end result is less noise in the image.
No - the end result is the same noise in the image, because although your sensels are larger, you have fewer of them, which means they are bigger in your print, and therefore they do more damage PER PIXEL to the image itself.
So, if you reduce the resolution of a sensor (of any size) in order to make the pixels larger, you will have less noise.
No - you will have less noise PER PIXEL, but you will have the same noise PER SQUARE INCH OF IMAGE.
Downsizing an image does not imrove noise dramatically (and you lose detail too).
Downsizing dramatically improves NOISE PER PIXEL which is the measure that you keep obsessing over. It doesn't improve noise PER SQUARE INCH OF PRINT OF A GIVEN SIZE.
>Don't understand - why are guys perpetuating the myth that sensor size is what matters, not pixel size.
Because it's a fact, not a myth.
This is simply not true. Ultimately, the pixel SNR determines the noise.
The noise in an image is affected not only by the pixel SNR, but also by HOW MANY PIXELS THERE ARE. If there are a lot of pixels then for any given print they will be printed small, and therefore each pixel's SNR contributes to the whole IN PROPORTION TO THAT PIXEL'S SIGNIFICANCE IN THE IMAGE. Really noisy pixels can make great images if there are an insane number of them. Really clean pixels can make horrible images if there are very few of them. If you keep sensor size constant while varying pixel pitch, you WILL NOT SEE A SIGNIFICANT DIFFERENCE IN THE AMOUNT OF NOISE IN THE IMAGE, for the reasons spelled out in triplicate above.
As technology advances, small pixels could have the SNR of larger pixels from a previus tech generation. So, it's totally possible to create a 1.6x sensor that has the same noise as the 5D classic from four years ago.
Amen
For that to happen, though, megapixel increases really need to stop.
You fell at the last hurdle. Once more, this has nothing at all to do with sensel sizes and everything to do with sensor sizes, because for any given sensor size, changes in sensel sizes are automatically compensated by the different number of sensels. The 1.6 format noise performance could potentially catch up with an older full frame sensor after many years due to better sensor technology and data-processing REGARDLESS of how many pixels either sensor has. The better technology compensates for the smaller sensor size. The number of pixels is not relevant to that question.
Edited on Oct 16, 2009 at 12:49 PM · View previous versions
|
| Oct 16, 2009 at 12:48 PM |
| |
|
 |
