sonyalpharumors

I’m on my cell so I can’t do this, but go on to Sony VideOn technical web site. Search for “Juan Martinez” and “Hugo Gaggioni”.

They are Sony authorities in sensor design. Especially Hugo. He is an industry genius.

Their videos are very technical and they both discuss how important photosite size is to the overall performance of the whole sensor. Especially signal to noise ratio and dynamic range. They will show you very nice diagrams to help you understand it too.

Have you seen Canon’s new FF camera? It does NOT have 36 megapixels on it. Why? Because it was designed to have MAXIMUM dynamic range a signal to noise ratio. (Incredable low light abilities) Something a 36mp can’t do to the extent they are aiming for.

I very much doubt these Martinez and Gaggioni contradict the technical information sheets they produce.
And the simple fact is that a Nikon 1 sensor is just as efficient per unit area as a D3S (AND D4), despite having pixels up to 6 times smaller. And I very much doubt the new Canon 1DX beats the efficiency of the former mentioned cameras.
As for DR, it’s also a fact that the FF camera with the largest maximum DR is still the D3X, which has pixels half the size of those found in a D3/D700 or the later released D3S.
In fact, the APS-C D7000 rules them all regarding maximum DR, despite having a sensor that is 2.3 times smaller and having pixels 4.7 times smaller.
The just anounced D4 and 1DX have lower pixel counts for speed, speed, speed.. and a bit of marketing. There’s also a reason why Nikon tends to artificially limit their max fps for their cheaper FF cameras, such as the upcoming D800. It’s not as if they haven’t proven capable of handling such amounts of data (again see Nikon 1) or mirrors flipping that fast (see even old SLR cameras) at even lower price points.

Try on DXOMark to compare sensors from 10 to 24mp, Sony 24mp on FF and APS-C is the best on DR. More Megapixel do not means more noise and lower DR. New 16mpx is far better then old 10Mp on lowiso, even 14MP CCD is better then 10MP CCD on low ISO. 24mp APS-c is in first incarnation, second will be much better then 16 MP on current cameras. 36Mp FF will be better then 24mp, and “MP war is dead” is heard from Canon and Nikon because they don’t compete with Sony in CMOS sensors MP count, and choose 18MP and 16MP because of speed ( Spoon is right on this ). In two to three years 36MP FF and 24MP APS-C will be low count sensors.

Sensors have improved allot in the past few years but what has REALLY improved in the past 3-4 years is noise reduction and post processing.

I’m not saying that Sony’s new 36mp sensor isn’t good. I’m sure it is. It’s readout system and post processing are probably amazing too.

However, if that same system were applied to a 16mp sensor of the same technology. The 16mp WILL have better low light properties. (Remember,…all other circuits and readout techniqes being equal)

Each pixel will gather more light, more charge and deliver a higher s/n ratio.

Remember, large solar panels collect more light and produce more electricity than small panels. Sensors are the same principle on a pixel level.

The larger photosite=larger photon bucket.

Seriously guys….this is under the concept that ALL OTHER TECHNIQES BEING EQUAL on and after the readout.

On this video below, Sony’s Juan Martinez (Pro Broadcast group)explains why larger photosites are better for low light performance. In fact, he is comparing the small photosites on a Canon 5D II to a Sony FS100 which has MASSIVE,…gigantic photosites. The sensor photosite size contributes greatly to the incredable low light performance of the F100 over the Canon 5D II.

This is straight from Sony itself
http://pro.sony.com/bbsc/video/videoChannelSearchResults.do?refinecat=&srchTerm=juan+martinez&omniturePageName=vid-int%3Ahome

Video name: “NEXFs100 Camcorder”

Take not of position time 11:10. Juan Martinez states:

Quote:

“The pixel on our super 35mm image sensor is FOUR TIMES LAGER than a typical DSLR pixel.”

“So not only are they skipping a bunch of pixels,[Canon 5d II] they have less sensetivity,… because this (Sony super 35mm sensor) has much bigger photosite area.”

“And because more light, more photons fall on this bigger area and we can capture in lower light”

“And it goes beyond that. Because these photosites are bigger I can capture MUCH HIGHER DYNAMIC RANGE because the size of the bucket, the site where the photons are landing is converting into electrical charges….”

“The larger the bucket, the larger the charge you are accumulating.”

“So you are given much more exposure latitude or a engineers like to call it, dynamic range…”

I dunno. you guys gotta argue with sensor engineers on this. It is widely understood that larger photosites collect more light than small ones. And remember, the more you pack into a sensor, the smaller the micorlenses are and the more space you need for grid circuits themselves. Yes, there are more “gaps” in between the vertical and horrizontal track in between photosites that waste precious space.

Today, the limitations of smaller photosites are being offset with higher and cleaner amplification and VERY good noise reductution like corelated double phase sampling techniques. (this is why small photosite sensors perform very well today)

But again, all that same post processing applied to a 18 megapixel FF sensor will yeild INCREDABLE sensitivity and clean high ISO quality.

Again,..that is why Canon is heading down that road with their 1DX which will be a low light monster and something that a low light rock concert photog will LOVE to have.

You completely misunderstood what he really said and the implication of his words. A 5DMkII is subsampling in video mode, meaning it’s not using all the pixels and thus not using the whole surface, thus not gathering all the incoming light. The FS100 uses all its pixels. So in effect its using a larger sensor surface. Ofcourse the larger pixels are a benefit here, because they account for the larger surface (especially if resolution would be equal). Like I said, those larger pixels will be better, but more smaller pixels (which the 5DmkII doesn’t offer in video mode) can equalize that out.
Again it comes down to keeping something constant, in the case of this topic, the *sensor surface*. For a given surface, image noise and pixel size are largely uncorrelated. The practical examples I have given (and I can give many more) are the ultimate proof.
And here’s what Sony Semiconductors said about it in its documentation last year:
http://i43.tinypic.com/m75hf9.jpg

So in reality Martinez doesn’t disagree with me at all. Just after 10 mins (ad quite a few more times) he even specifically adds that larger pixels contribute to a larger surface. In other words, surface counts, which is what I’ve been saying all along. But in this topic the surface is largely constant.

And as for DXO, the latest tiny pixel sensors equal or spank the best large pixel sensors available. Factor 10 difference in size or not. Think about it, a >120MP FF camera with low light properties as good or better than a D3S. And here we are discussing the tiny differences between 12, 16, 24 or 36P at FF?

It’s interesting but is there something missing? Smaller photosites have more horizontal and vertical columns and smaller microlenses.

There are gaps in between rows & columns for circuit control. A 10 megapixel FF sensor will use a higher percentage of it’s given surface area than a 20 or 30 mp and have larger microlenses. The higher the pixel count, the higher the surface lost on non-photosite silicon.

I’m not saying you are wrong. Not at all. I’m simply saying there is a trade off to everything you design into a sensor.

The Canon 1DX has an incredable low light performance. They supposedly even found a way to produce “gapless” rows microlenses and photosites to make those damn things EVEN LARGER! lol I mean Jeez,…those guys really pushed the envelope to make those 16 million suckers huge.

Are you saying that it’s low light performance numbers are “not” because of it’s massive photosites? Is it because of some other reason?

If so, why did they work so hard to keep them as large as they could,…especially if it didnt matter much?

BTW,..this is a fantastic discussion. It’s a nerdy one but I’m glad we are having it.

So, if you had to speculate, why did Canon go the “huge photosite” / “low megapixel” road on their FF new baby?