Cableaddict wrote:
Brainaic, That is absolutely not true, and someone with your extensive experience & general knowledge should know better. Photosite size has a significant effect on noise levels and dynamic range. -almost by definition. This information is easy to find online. Photosite technology is no doubt better now than 3 years ago, but you specifically wrote "of the same technology."
Photosite size has a significant effect on PER PIXEL NOISE. You're right: it is easy to find this information online. Sites like DPReview and Luminous Landscape assess cameras by zooming to 100% and making pronouncements about PER PIXEL NOISE, ignorantly, and wallet-toting photographers avidly read those reviews which inform their opinions. The whole thing is a mass delusion - the blind leading the blind. PER PIXEL noise should only be compared between cameras in the ratio of the number of megapixels. A D3 needs to have much better PER PIXEL noise characteristics than a 1Ds3, JUST IN ORDER TO KEEP UP WITH IT, because in any given print the 1Ds3 pixels will be much smaller, and therefore the noise is far less destructive to the image, since noise averages out.
Never have an opinion based on comparing 100% crops unless the two files have first been resized to the same numbers of pixels. Otherwise a myth develops, and that myth is: "per pixel noise gives a consistent insight into per picture noise, and larger pixels are therefore better".
> I must admit that all the tech stuff I read did not relate to gapless micro lens sensors.
Perhaps Brainiac is correct after all, though only for the very newest sensors, and if so I apologize!
-But I want to read more about this before I fully commit to the idea. I get the basic idea though- If there are no gaps (are these photosites square?) then there's no accumulated area loss.
-----------------------
-note, though, that it is still wrong to call the other poster's info "a myth," since it is correct regarding all older sensors..
The myth is even wronger for the newest sensors, but it is also a myth for older sensors. We can see this by trying to pick two cameras with sensors of the same size from the same manufacturer in which the higher resolving camera has worse PER PICTURE noise. There are remarkably few examples, if any, and many counter-examples. The reason that pixel size has never had a huge effect is that for a long time now the great majority of the sensor area has been harnessed, and the gaps have been very small. For that reason one would have to make a dramatic difference to density in order to see a significantly detrimental effect.
How did the myth grow up? People looked at 100% crops from different cameras with different numbers of megapixels, and failed to notice that they were comparing sections from images at different magnifications. This always handicaps the higher resolving camera. As a result, the benefits of increasing the number of megapixels are underestimated. So I say it again, because it's important: never compare 100% crops if the two cameras have different numbers of pixels. In order to compare 100% crops from two cameras with differing numbers of pixels, it is absolutely crucial to resize the two images to the same number of pixels. The fairest way to do that is to uprez the lower resolution file.
In camera ST-E2 would be useful for me. But then we're back to the issue of pop-up flash and weather sealing. Which, if Canon can put a pop flash in a camera and still seal it, not compromise the VF and also include ST-E2 capability, then they have my permission to proceed. Where's the flippin' camera already!?
d_chiesa wrote:
I understand the 'per surface area' noise argument. But what about dynamic range? That should be reduced at the 'per pixel' level with smaller pixels, and not made up by having more of them...
I don't care much about noise, but better DR would be nice.
Unfortunately, from the standpoint of RAW data... noise & dynamic range are one and the same. DR is the range from the brighest detail you can resolve, to the darkest detail you can resolve. However, noise will disrupt the darkest details you can resolve --- effectively reducing DR. Hence, more noise = less DR.
I'm not a scientist. But I play one on TV. So here's my unscientific opinion:
Imagine 2 full frame sensors: 24 MP and 6 MP. The big photosites on that 6 MP sensor will have a much stronger signal from incoming light (during exposure), just because it has a large surface area to receive lots of photons. Most of the noise will come from the electronics within the camera. If the technology level is equal, then the noise generated by electronics in a camera would be the same for a 24 MP sensor and a 6 MP sensor.
But on that 24 MP sensor, the individual pixels are 4x smaller, therefore its signal is 4x weaker, and will need to have a gain applied that's 4x stronger. Even though the noise generated by electronics in the 24 MP and 6 MP sensors are the same, the output read noise in that 24 MP sensor will be 4x bigger --- simply because a gain 4x stronger would've been inadvertently applied to it.
d_chiesa wrote:
Still, what about DR? Image DR is PER PIXEL DR.
I don't know if DR reduces with pixel size. A smaller well is, well, smaller, so it's no more likely to fill up than a larger one. The maximum number of photons it can hold in voltage will be less, because it's smaller, but it also services a smaller area, so that may compensate at the white end since fewer photons are arriving. Extending the micro-lens width while using the same silicon will increase the odds that a pixel will blow out, but it also captures more photons at the black end of the scale, so to some extent it's swings and roundabouts. My point is that like high iso performance, DR doesn't seem to have suffered as pixel densities have increased. The 1Ds3 perhaps has slightly less DR than some older cameras, but more than most. There seems to be no guarantee that an 13 Mpixel camera will have more dynamic range than a 21 Mpixel camera. Why infer a dependency which doesn't seem to hold in real life examples?
Etadam wrote:
Given a technology, the smaller the pixel, the more likely noise will bias the color accuracy - this is physics.
That is common sense.
But a BIG assumption has been made. You are assuming there is NO improvement in technology. Changing the electronics read noise, improving on micro-lens coverage: all these are improvements.
I may be an optical physicist by profession, but when it comes to photography, I don't give a damn what goes into the chip. The only thing that matters to a photographer is the end result. So, as long as a good change has been made to the technology, it's good for me.
sifpandor wrote:
In camera ST-E2 would be useful for me. But then we're back to the issue of pop-up flash and weather sealing. Which, if Canon can put a pop flash in a camera and still seal it, not compromise the VF and also include ST-E2 capability, then they have my permission to proceed. Where's the flippin' camera already!?
-- Mark
Hey matey, thanks for the info. You don't need a pop up flash for the ST-E2 though, it doesn't have one itself, its just an IR transmitter and a focus assist beam. So if they are hiding a new grip in the 5D mkII moon pic, could that have the transmitter in it for the ST-E2 I wonder....? Yum!
brainiac wrote:
Why infer a dependency which doesn't seem to hold in real life examples?
Not that fast: it does indeed seem to hold. Technology got better, noise, even per pixel, got better, megapixels are more and STILL DR is basically the same. Why it did not increase? Maybe because we do have more MP?
Why do MF sensor have more DR?
Not that it matters a lot to me, but i just don't believe that more MP are necessarily better. Then again, i'll be extremely glad to upgrade to a camera with more of them, 'cause the 6 of my 10D are too few for a good 13x19 print
Etadam wrote:
Given a technology, the smaller the pixel, the more likely noise will bias the color accuracy - this is physics.
thw2 wrote:
That is common sense.
But a BIG assumption has been made. You are assuming there is NO improvement in technology. Changing the electronics read noise, improving on micro-lens coverage: all these are improvements.
I love how this thread takes a different turn .. just before it dies on Sept. 17...
Of course progress in technology plays a high role in noise etc...
Long story short: pls read again my post...
thw2 wrote:
Etadam wrote:
Given a technology, the smaller the pixel, the more likely noise will bias the color accuracy - this is physics.
That is common sense.
But a BIG assumption has been made. You are assuming there is NO improvement in technology.
...forget the assumptions: one BIG error has been made. Two 10% inaccurate pixels are significantly more accurate than one 10% inaccurate pixel. So smaller pixels can afford to be less accurate by some margin, in order to produce the same result. "This is physics". These days this means that added resolution has very very little downside. The "keep the pixels few" crowd are not just ignoring the technological advances, they are also not doing their arithmetic correctly.
If you're going to say things like "this is physics", it would be a good idea to actually get your statistical analysis right, otherwise it's not physics, it's fantasy. I'm afraid that Etadam is still offering a per pixel analysis which fails to take into account the macroscopic effect. Colour accuracy is not a distinct problem.
I do not know much about physics. I can see.
I just compared some 50D ISO 3200 images to those from other cameras. Although they have a 25% higher resolution the ISO noise in a 100% crop is imo as good/maybe better than D3/D700 supports. I do not know how good or bad DR is. The pics looked excellent to me. Maybe my Eyes do not see any difference, my monitor does not support that DR or I just do not care.
But expecting that tech in FF and another stop in same IQ with 5D II makes me feel a gooseskin in my neck!
Me too. But the DR of my 12 bit 30D images was good enough for my customers and me. So if its the same (what it surely will not be) I will be fine, if High Iso makes less noise.
At this point, I just want to see a camera. I don't just mean announced, I want one released. My hat goes off to Nikon for the rapid release after announcement of their cameras. The D90 was announced on the 4th of September. On the 9th I walked into the local camera store and they were sitting on the shelf. The D700 was less than a week as well.
Why is Canon dragging their heels? Do they not think they've built the suspense enough? Maybe they thought they should hold off on production until after they've lost enough market share to make the fight interesting?
I have a broken camera and need a replacement NOW! If I didn't have so much Canon gear (flashes/glass/cords/whatever) I'd have a D700 already and be done.
Incidentally... Some people may not have noticed that the image is updated today... but the amount is so small, it's not really of great value. There's a little more shadow on the left (our right) of the bulge, and the Canon is a bit more clear... Otherwise welcome to yesterday. ^_~
Me too. But the DR of my 12 bit 30D images was good enough for my customers and me. So if its the same (what it surely will not be) I will be fine, if High Iso makes less noise.