Patent: Quad pixel AF sensor

[yeahright]

I'm still trying to understand this: So, you're saying you have a horizontal row of 50 pixels, each pixel with 2 sub pixels. One sub pixel is sensitive to the rays from the left of the main lens, and the other sub pixel sensitive to the right of the main lens. I'm wondering what type of micro lens above each of these 2 sub pixels can be so selective? Wouldn't each of the micro lenses have to reject half (at least) of the light, and how would they make such an optical structure that would be so completely effective in the splitting when they sub pixels are both next to each other in the sensor? I think the physical construction of the two sub pixel micro lenses and how they split the sensation of light from the left & right side of the main lens is the crux of what I need in order to really understand it.

I also assume that you could shift your computational logic by 1 (or more) whole pixel to the left or right to have another 50 pixels (25 to the left, and to the right) to give you a new AF value, correct? That is, the 2 sub pixels of a single pixel could be used by up to 50 different sets of AF logic if they wanted to design that many AF points, correct?

Here is an article including a (simplified) image of a DPAF pixel with microlens:

Canon EOS R: A deep-dive Q&A session with the Canon engineers

Canon's new EOS R full-frame mirrorless camera has stirred a lot of buzz, and I'm sure readers will have as many questions as we've had about it. I had a chance to sit down with a panel of top Canon engineers at the…

www.imaging-resource.com

There is actually one lens on each dual-pixel, and it does not reject light (which would reduce sensor sensitivity) but rather directs the light to the appropriate sub-pixel.

Yes, you could move your 50-pixel AF 'sensor' by only one pixel, however, in this case the result would be almost the same, because 49 pixels are the same. So I assume that there is a sensible amount of pixels between AF positions (maybe at least half the length?), and I assume this is why the number of AF points is significantly lower than the number of horizontal sensor pixels. You could also vary the length of the sensor which would have an impact on the selectivity of the AF point and on the probability of finding a target.

 

[Rzrsharp]

Here is an article including a (simplified) image of a DPAF pixel with microlens:

Canon EOS R: A deep-dive Q&A session with the Canon engineers

Canon's new EOS R full-frame mirrorless camera has stirred a lot of buzz, and I'm sure readers will have as many questions as we've had about it. I had a chance to sit down with a panel of top Canon engineers at the…

www.imaging-resource.com

There is actually one lens on each dual-pixel, and it does not reject light (which would reduce sensor sensitivity) but rather directs the light to the appropriate sub-pixel.

Yes, you could move your 50-pixel AF 'sensor' by only one pixel, however, in this case the result would be almost the same, because 49 pixels are the same. So I assume that there is a sensible amount of pixels between AF positions (maybe at least half the length?), and I assume this is why the number of AF points is significantly lower than the number of horizontal sensor pixels. You could also vary the length of the sensor which would have an impact on the selectivity of the AF point and on the probability of finding a target.

They fix the location and number of pixel for AF point is a choice of faster AF. It will reduce the process timing. There have no extra sensitive pixels.

 

[usern4cr]

Here is an article including a (simplified) image of a DPAF pixel with microlens:

Canon EOS R: A deep-dive Q&A session with the Canon engineers

Canon's new EOS R full-frame mirrorless camera has stirred a lot of buzz, and I'm sure readers will have as many questions as we've had about it. I had a chance to sit down with a panel of top Canon engineers at the…

www.imaging-resource.com

There is actually one lens on each dual-pixel, and it does not reject light (which would reduce sensor sensitivity) but rather directs the light to the appropriate sub-pixel.

Yes, you could move your 50-pixel AF 'sensor' by only one pixel, however, in this case the result would be almost the same, because 49 pixels are the same. So I assume that there is a sensible amount of pixels between AF positions (maybe at least half the length?), and I assume this is why the number of AF points is significantly lower than the number of horizontal sensor pixels. You could also vary the length of the sensor which would have an impact on the selectivity of the AF point and on the probability of finding a target.

Thanks, Yeahright! That was an excellent article on AF, and some other good info on the R mount. Now I do feel that I understand enough of how dual pixels work. It also makes is pretty obvious that a quad pixel would work the same way, but with the 4 pixels in the 4 corners of the pixel (in an "X" pattern) instead of in a cross "+" pattern for maximum packing of pixels on the sensor.

A little thing I noticed about the R mount (over the EF mount) is the shape of the 3 mount ridges that hold the lens in place once properly inserted. The R mount has a much longer ridge at the top of the circle than the EF mount. This helps a *lot* (IMHO) since the top of the mount is where the weight of the lens is trying to pull the lens away from the mount and only this ridge is holding it on. The ridges at the bottom don't hold it on from the downward pull of gravity much at all, and are appropriately smaller. Those big long lenses stress the mount a lot, especially if you don't choose to use the lens' tripod collar when there is one (as I try to do).

 

[canonmike]

If this is implemented in the R1, then it would help to define the "groundbreaking" new AF system, as Maximilian mentioned. Since each "quad pixel" would have to cover more area than a "dual pixel" for all the circuitry then I could see the R1 coming in at the 20+MQP (where a pixel is a quad pixel) which is ideal for 4K video as well as plenty for professional stills use. It wouldn't surprise me if the marketing department then decides to call this a "80MQP" sensor and asks the programmers to come up with a new Bayer decoding (up-res'ing) to get 80MP to display & advertise.

Just keep in mind that the IQ and dynamic range will be affected by the increased circuitry on the sensor if that circuitry blocks more of the total light sensed by the sensor. And more false artifacts will occur if the Bayer decoding shifts from 4 cells (RGBG) to 16 cells(RRRRGGGGBBBBGGGG) - but I don't think the marketing department will mention that!

You know, at some point there's going to be so much circuitry and high resolution that I can see Canon finally coming out with a BSI (back-side illuminated) version of their sensors. That's really the only way to keep getting more complexity and resolution and image quality in a FF sensor as you approach the limits of what a non-BSI sensor can do. We may be seeing Canon's first BSI introduction, which would be quite an announcement in itself! Do I think this will happen in the R1? - No, but it sure wouldn't surprise me. And if they want to get a 45M QP sensor and beyond then I think they may be forced to use BSI in the future.

""""You know, at some point there's going to be so much circuitry and high resolution that I can see Canon finally coming out with a BSI (back-side illuminated) version of their sensors. That's really the only way to keep getting more complexity and resolution and image quality in a FF sensor as you approach the limits of what a non-BSI sensor can do. We may be seeing Canon's first BSI introduction, which would be quite an announcement in itself! Do I think this will happen in the R1? - No, but it sure wouldn't surprise me. And if they want to get a 45M QP sensor and beyond then I think they may be forced to use BSI in the future. """"

Well, my fellow CR member, you sure called it and back in Oct, 2020, no less. BSI on its way in the R3.

 

[usern4cr]

""""You know, at some point there's going to be so much circuitry and high resolution that I can see Canon finally coming out with a BSI (back-side illuminated) version of their sensors. That's really the only way to keep getting more complexity and resolution and image quality in a FF sensor as you approach the limits of what a non-BSI sensor can do. We may be seeing Canon's first BSI introduction, which would be quite an announcement in itself! Do I think this will happen in the R1? - No, but it sure wouldn't surprise me. And if they want to get a 45M QP sensor and beyond then I think they may be forced to use BSI in the future. """"

Well, my fellow CR member, you sure called it and back in Oct, 2020, no less. BSI on its way in the R3.

Thanks, Canonmike! I'm probably just one of the many who have predicted that Canon would have to eventually come out with BSI sensors as you can only have so much circuitry blocking your photo receptors before there is insufficient light for them. But thanks for seeing my post on it and mentioning it.

I also remember "guessing" early on that the price of the RF 100-500L would be the same as the RF 70-200L ($2,700 as I figured rounding was appropriate). So I missed by $1 on that one. Well, that one was just a lucky guess (I've guessed prices on others that have missed as well).

The one I like to remember was when I saw the early patent post for the RF 800mm DO lens. I noticed 2 things: it didn't have any IS elements indicated in it (obviously they did add IS to it in production) and that there were absolutely no lenses in the back of it so that I predicted it would have a collapsing ability to shorten it's length by ~30% or so for storage. That one I nailed!

Now, if I only knew which way the stock market was going ...

 

[canonmike]

Thanks, Canonmike! I'm probably just one of the many who have predicted that Canon would have to eventually come out with BSI sensors as you can only have so much circuitry blocking your photo receptors before there is insufficient light for them. But thanks for seeing my post on it and mentioning it.

I also remember "guessing" early on that the price of the RF 100-500L would be the same as the RF 70-200L ($2,700 as I figured rounding was appropriate). So I missed by $1 on that one. Well, that one was just a lucky guess (I've guessed prices on others that have missed as well).

The one I like to remember was when I saw the early patent post for the RF 800mm DO lens. I noticed 2 things: it didn't have any IS elements indicated in it (obviously they did add IS to it in production) and that there were absolutely no lenses in the back of it so that I predicted it would have a collapsing ability to shorten it's length by ~30% or so for storage. That one I nailed!

Now, if I only knew which way the stock market was going ...

Ha! Ha!. When you figure that out, I hope you share your info with the same insight and enthusiasm as your camera predictions. Now, go get to work on it.