Web Analytics RiceHigh's Pentax Blog: New Technology Breakthrough - Phase Matching AF by Sensor!

Thursday, July 22, 2010

New Technology Breakthrough - Phase Matching AF by Sensor!


(Above: The AF Phase Detection Pixels; Image Source: Fujifilm)

By Fujifilm! See:-

http://www.fujifilm.com/products/digital_cameras/f/finepix_f300exr/features/


Well, if this can be applied to larger sensors/imagers like that used in the IDLCs and above later on (currently the above new technology is used in a DC only), then DSLRs will no need to have all those split-mirrors and prisms as well as sensors in a separate AF detection module. As such, there will be no more mis-alignment issues that cause the AF errors. LiveView mode will be much better to use as there will be faster and more accurate AF. And continuous tracking and predictive AF will also be made possible. What a better world it will be!

Thanks, Fuji, for bring us that ground-breaking technology that I think will open up a new era of new echnology that will come! (Thumb Up!)

Update: There is more technical information on how putting additional "AF" pixels on sensors could do the phase detection and job. Here is an explanation on the Sony patent and design (Yes, it's a Sony, not Fuji. Btw, I don't know if Fuji works in the same or a similar way, but I guess that it should be):-

http://olddo.com/bbs/viewthread.php?tid=13800
(Text in Traditional Chinese, for Google translated English page, click here)

So, I think without micro-prisms for the light splitting, focusing accuracy and low-light sensitivity would be affected. Moreover, diffraction would be a real problem by that design (see this diagram). Let's see what the technology will become and how it would evolve! Maybe later on they put micro-prisms on sensors, just like what the sensor makers did that years ago by adding micro lenses on sensors. Whilst theoretically all these are possible, but then it just boils down to the manufacturing cost and money/marketing concerns again..

Comments (10)

Loading... Logging you in...
  • Logged in as
The big question is this:Would putting AF sensor on the main sensor waste precious sensor space? Would IQ be effected negatively?

Even though Fuji is the first with this camera, Nikon, Canon, Sony all have such patents where phase detect AF is performed by the main sensor. For example, here is an old Sony patent with the same concept.
http://sonyalpharumors.com/sr5-new-sony-patent-sh...
1 reply · active 764 weeks ago
I don't worry about this question. If they managed to do it for a small sensor, which the light receiving area per pixel is already small, it would be a non-issue if they do it for larger sensor, in which there are more room and area for them to play with. :-)
But if you don't use Live View, then the light won't hit the sensor, and it won't work. But fast AF while using Live View or when filming would be nice, specially the latter :)
2 replies · active 764 weeks ago
So, it would be the best technical solution and technology supplement to those ILDCs. Even for DSLRs, the Live View experience will surely be greatly improved.
True, if the image quality is not getting killed by adding this, I hope it's in all future DSLRs.
Well, I in another life developped a fast contrast AF algorithm for microscopes. The current contrast AF algorithms found in consumer cameras may be poor. But they aren't by principle. Contrast AF is the way to go for *fast* and accurate focussing. It is much faster than phase AF because it can use all available light. Yes, it requires some iterative element but as I found in my work, if done correctly, is blazingly fast. The focussing time of a contrast AF module, if done correctly, is less than 10ms (if you can move the lens fast enough so it would depend on size of the lens and how far off prior to focussing you are).

Only because consumer contrast AF is so poor doesn't mean phase AF has any future. It hasn't.
2 replies · active 764 weeks ago
The limitation with CDAF is that the "distance" to the correct focus is unknown, which on the other hand is easily measured with Phase Matching AF. As such, continuous tracking and prediction are rather difficult with CDAF, unless more advanced and intelligent algorithms could be developed.
I know about this why I mentioned the "iterative element" above. Industrial grade CAF measure on the go while they move the lens towards the correct end position, without overshoot and little wrong-guessing the starting direction. And the lens has to move anyway. It depends on the available light and required lens shift how long this takes. All factors combined, CAF should clearly outperform phase AF with a sensor allowing fast readout (more than 100 fps on a region of interest), a decent CPU and an algorithm like the one I invented for my work.
I'm thinking the return of real time exposure and TTL flash control in LX and F3, which had gone since the implementation of phase detection autofocusing
1 reply · active less than 1 minute ago
The *ist D and *ist DS actually have real-time TTL flash exposure control and I believe that they are the last SLRs on Earth that could do that. And, you're right that real-time ambient light exposure does not appear anymore on any SLR since the LX/F3 era in early 80s, btw.

Post a new comment

Comments by