Ah, so your misunderstanding goes deeper.
1) readout noise (it's constant for every given sensor)
2) thermal noise (changes according to environment, continuous work, etc)
3) shot noise
Shot Noise - AKA Poisson distribution variation, AKA random packet count...
I've been mostly talking about shot noise that potentially increases
non-linearly in function of the projection surface and depends on a
snip
to handle shot noise down the line. So: FF = More light = More noise
/ Less density = Worse shot noise handling.
Shot noise will not be worse on FF because more area evens the average. Only light intensity dictates how random the photons are, and it only matters in weak light, say shooting in star light. Shot noise is so small I’m not even sure the current sensor technology can pick it up. All the same, shot noise is an argument for FF/low density; given the same technology, the bigger light well will collect more photons and show less deviation. The smaller the light well due to higher resolution or smaller sensor size, the fewer photons collected and the greater the deviation (noise) proportionally.
Moving to the other two noise types (being the ones somehow
considered more important by most of people around here):
More important because it’s visible...
1) readout noise is claimed to reduce on a "larger pixel sensor". I
have my doubts about that. Readout noise depends on the circuitry,
material and a lot of other parameters. I tend to believe it's going
to be more or less the same on FF and APS-C sensor of the same
generation.
Readout noise also depends on the proximity of the circuits. Just as light falls off with the square of the distance, so does every other EM field. The larger, lower density sensor has the circuitry farther apart (with a stronger signal), and as such the EMF interference is weaker. The noise created off sensor remains, but the sensor will send a stronger signal. You simply cannot ignore the fact that the S/N ratio is what matters. All noise suppression schemes rely on the strength of the signal in contrast to the (hopefully) weaker noise to have any success. Higher density smaller sensors will have more readout noise because of the closer circuits and transistors causing EMF interference.
2) Thermal noise: given a larger chamber, more light, bigger shutter,
higher resolution and other factors, I actually thing it's going to
increase on FF. (Unless Sony comes up with some pretty interesting
solutions to the problem).
Thermal noise is easier to control. A higher resolution sensor will be subject to greater thermal loads than a lower density of the same size do to the high transistor count. However, the larger FF shutter and mirror motions funnel in more cool air, the larger chamber contains a greater air volume which will take longer to heat during use, and the FF sensor gains beneficial surface area to allow heat to dissipate more rapidly compared to a smaller format.
Something about SSS now:
1) If APS-C needs a 2mm gap, FF needs a bigger gap to achieve the
same results. I hope the geometry is quite clear here...
Are you going to shake more because you are holding a FF? The geometry is quite clear, regardless of whether the sensor is FF or APS. The distance between the rear element and the sensor is fixed and so is the distance to the lens' nodal point. The shake both in distance and time period, while varying from person to person, is a fixed average. As I said, 2 degrees of swing around a fixed fulcrum distance will always be the same result. If an APS-C needs to move 2mm, so too will the FF. Due to the design of some lenses, there could be some vignetting or some distortions softening the image in the corners, but this will hardly be universal. Instead it will depend on the lens in question as well as the relative instability of the photographer. Even so, none of this means that there will be much more motion for a FF compared to an APS-C because we are still talking about the same shake.
2) Now - with a bigger gap and a bigger sensor it turns out we need
more precision. More precision means: more computing power to
calculate the compensation and a more precision in sensor movement
coordination. The last part being purely mechanical leaves me
somewhat doubtful about the results...
More precision will certainly come with each generation, but this is independent of the sensor size. Power drain may occur from stronger motors that will have to accelerate a more massive sensor tray. This may or may not affect battery life; I don't know the mass of an APS-C assembly, nor do I know the draw of the current motors so I have no way to calculate it.
All of these theoretical arguments are moot, proof is out there. There are FF sensors and results can be seen all over the net. The D3 is far less dense than the A700 and its high ISO shots are much lower in noise. The greater distances, larger light wells, and greater cooling surfaces are contributing to high signals with low noise with great results. The higher density 1DsIII produces more noise than the D3 because of the tighter proximity of circuits and the somewhat higher thermal loads of the extra transistors, but then it's newer design and technologies help it match the noise performance of the older, lower resolution 5D. The 5D mkII will likely be better than the 1DsIII in noise performance, while not matching it in absolute detail, but all these differences are only viewable at 100% or really huge print sizes. All evidence indicates better FF IQ performance compared to APS-C, the cons are in cost and size of supporting hardware. Will FF become much more main stream over time? Yes. Will APS-C be gone any time soon? No.
--
-Dylan Anderson
