[Pavel]

Thom Hogan seemed to have launched a rumor that Nikon D400 sensor will have obscene number of pixels (24 mpixels) but all is well, because they may be used for binning (combining pixel signal from neighboring pixels on a sensor level to reduce noise and increase sensitivity at the cost of resolution). Presumably you will be able to switch binning on and off, giving you a really high res camera (THEORETICALLY good bye to my current lenses - they could not keep up) at low ISO. When photo needs low noise at high ISO, you go to binned mode. Low noise at low light is my personal holy grail - the ONLY new feature I would like in my new camera. It is what I hoped for for years. Gaining 2 f-stops from binning would be all I want. So - consider me hyped

[matthew]

Pixel binning is an interesting idea, and Phase already has models that do this. Of course, the Sony sensor in the D7000 (and Pentax K-5) is sufficiently better than the one in the previous generations that it's not necessary to have fundamental leaps in technology for some pretty major improvements. They would be nice, of course, and it's still no D700.

[Pavel]

no D700 and thus no need to switch to heavier camera and heavier, more expensive lenses. All I want is 2 or preferably 3 extra f-stops. (Modes hope, isn't it? )

[Kombisaurus]

I think the principle of binning a great idea, and really provides a great argument for the flexibility of high-resolution sensors. But should the functionality be built into the camera? I'm not convinced yet (although I admit I haven't researched it).

I mean, downsampling a photo in post (which is pretty much the same thing as binning) does a good job of noise reduction. Simply by increasing the resolution of a sensor without proportionally increasing per-pixel noise levels will automatically instigate "binning" when resampling the image during post processing to its final size (for web, print, etc).

The obvious downside of having this done in-camera is that resolution is lost in the image. Of course the same thing happens when you do the binning later in post processing, but in the latter case at least you can still go back to the original resolution if you need to (albeit with higher noise levels). You don't have this choice if the binning is done in-camera.

The positives I can see of having the process done in-camera are smaller file sizes (but storage gets cheaper every day), and if the process is performed directly at the pixel level before the signal turns digital then there is potential for some improved noise-reduction (which could be dramatic, or could be nominal). But if the process is done any time after the signal turns digital, then the performance won't be any better than could be done in post-processing (and could be significantly worse, as the camera will have limited processing power and needs to stay very responsive to the user, so the binning algorithm will need to be kept simple).

I'll definately be keeping an eye out for how this might unfold though... I love the fact that cameras are requiring less and less light and producing cleaner images these days. Imagine the possibilities if you could shoot clean images at 102400iso and higher. You could shoot using natural light in almost any situation, hand-hold shots at night with no flash, and go beyond how we see the world with our eyes. Exciting stuff.

[Pavel]

great post Adrian. I understand that binning is very different from downsampling, because as you say, it happens BEFORE conversion to digital form. My understanding is that the effect is much bigger than the effect of downsampling. I am thrilled with the possibilities, as I often photograph at the limits of my camera sensitivities, having to make uncomfortable compromises.

[shuttertalk]

Isn't it just another type of interpolation? Fuji have been doing something similar for ages - way back my S5000 only had a 3 megapixel sensor but they would interpolate to 6mp and call it "SuperCCD HR" or something like that.

[Pavel]

obviously I am out of my depth here, but it is my understanding that this is NOT interpolation. This process takes place as I understand it BEFORE pixels are defined.

[matthew]

Fuji has always danced to the beat of a different harmonica, and you've gotta love them for it. Their digital compacts (and SLRs, for that matter) use two different pixel sizes that lets them mix up their output in interesting ways. The current crop have three different modes that mix binning and interpolation to produce different results – excellent cameras for anyone who's technically inclined and looking for a fight.

Some of the Phase One MF digital backs do pixel binning to improve their low-light performance, and I understand that it makes a significant improvement over what's possible from just downsizing the image post-capture. I have to admit that I don't pay much attention to them and their technology, though, since I'd have to win a significant prize in the lottery before I could go shopping for one.

[Pavel]

Interesting Matthew. Going strictly on memory (and you know how bad I am), I am not aware of superior light sensitivity of Fuji's compacts. I am aware of better dynamic range of the DSLRs and better colors(?). Binning would interest me only if it would give me a camera with significantly better performance at low light and/or significantly better dynamic range with one shutter press in a hand-held situation. Otherwise, I amy enjoy reading an article about a technology, but in itself it will not excite me. Phase one is theoretically interesting and it will be downright exciting when you GIVE me a set.

[shuttertalk]

The current crop have three different modes that mix binning and interpolation to produce different results – excellent cameras for anyone who's technically inclined and looking for a fight.

I haven't looked at their technology recently, but back then they used to have a hexagonal sensor "pixel" arrangement and then use adjacent "pixels" to interpolate to get a larger resolution image. The verdict back then was that it didn't gain any additional detail, just increased the file size - but good to see they've made some headway since then.

[matthew]

Well, I didn't actually say that Fuji's approach was particularly successful…

There are so many pixels out there these days that there really should be something better than the standard bayer array to manage them. Hopefully the next generation or two will have something new.

[matthew]

Thom Hogan seemed to have launched a rumor that Nikon D400 sensor will have obscene number of pixels (24 mpixels) …

So now that Sony has production cameras using a 24MP sensor, does anyone doubt that it will also be in the D300s replacement?

And while the Sony 16MP "magic" sensor that's shared by the Nikon D7000, Sony A55 & NEX-C3, and Pentax K-5 has exceeded a lot of expectations for what a 1.5x sensor can do, there's nothing exotic or edgy about it - no binning, same old GRGB Bayer pixel array. Could Nikon still have something special planned for the Sony 24MP sensor?

(And of the 16MP cameras that use the same sensor, it's the Pentax that scores the highest marks on its DxOMark testing.)