Hey guys, been reading about the 60D, and just wanted to make a post about higher pixel density sensors, so, this is my 1st post. :-D
I thought about the whole down sampling thing, and applied my further maths knowledge too it. I have come to the conclusion that a 40mp APS-C sensor image down-sampled too 10mp would create twice as much noise as a 10mp APS-C sensor when both are based on the same technology.
The amount of noise is basically the standard deviation of the pixel value from the actual value (if the pixel had no noise). This allows me to apply my statistics knowledge to the problem. I'm going to use some abbreviations for this maths as otherwise I'll be writing loads out.
SqRt= Square root
Now, imagine one pixel. Lets give it a SD of 1 (it's level of noise). Now imagine we quadruple the pixel density, there are now four pixels, because we have quartered the area of each one, it's noise level or SD increases by 4 times. So each of the four pixels have a SD of 4.
Now, say I want to use those four smaller pixels to create one big pixel using an algorithm to try and get one lower noise pixel. Basically I just find the M of the 4 pixels.
Here comes the issue, the SD of a sample mean (basically, the SD of the one pixel created from the 4 smaller ones) equals: SD/SqRt(number of pixels). If we apply this we can find the standard deviation of the one virtual pixel created from the 4 smaller ones:
4/SqRt(4) = 2
So as we can see the one physical pixel should, theoretically, have half as much noise as the one "virtual" pixel created from the four smaller pixels. I therefore propose that there is a very very sound reason for creating a newer sensor with a lower pixel count. This also implies that Canon has been doing great work at getting higher pixel density sensors that create lower or similar noise in print than older cameras with lower pixel counts.
I am fairly sure that this is correct, however, if I'm wrong, I would be happy to be corrected. :-)