A patent application filed by Canon was recently published by the USPTO (App. No. 12/188,385) that may give us some insight as to how Canon will solve the abysmal autofocus performance of Canon’s Live View mode. If you are unfamiliar about how Canon’s Live View AF system currently works (or doesn’t work), you can read my prior rant in Photography Bay’s Canon Rebel XSi Review. For some insight on how Live View AF should work, see the Sony A350 Review.
Nothing works faster for DSLR autofocus systems than the tried and true phase-difference method. If you don’t know what phase-difference AF means, the you can read about it on Wikipedia; however, it’s not essential to understand all the technical details to get excited about Canon’s new patent.
The fundamental problem with making phase-difference autofocus work in a Live View mode is that the AF sensor receives light from the mirror, which also reflects light up into the pentaprism for the optical viewfinder. When you press the shutter button, the mirror flips up and you get a brief moment of black out in the view finder while the shutter opens and exposes the image sensor to the light from the lens.
When you think about this process, which I have overly simplified for illustration purposes, the problem with Live View mode and fast autofocus is apparent. When you set your Canon DSLR to Live View mode, the mirror flips up and the shutter is opened to directly expose the sensor to light from the lens. The only difference here is that the image is not yet recorded, but rather displayed on an LCD screen for you to frame, focus and then finally capture your image. Without the mirror in place to reflect light to the AF sensor, you cannot obtain fast and accurate phase-difference autofocusing. Accordingly, Canon has been forced to suffer (as have the rest of us) with the sluggish contract detection autofocus in Live View mode.
Canon’s US Patent Application No. 12/188,385 claims to have solved this predicament by putting [Updated: Thanks to Micah for pointing out this error in the comments.] the phase-difference AF sensor below the image sensor rather than above it. Additionally, Canon has plans to put a semi-transparent mirror in place a different configuration to allow simultaneous transfer of light to the image sensor and the AF sensor.
The novel part of the patent above seems to be that they are using a mirror that is partially transmissive in a different configuration for live view. It looks like the mirror flips in the opposite direction for live view and the AF sensor is positioned optimally for this, with the secondary mirror’s placement higher up the main mirror compared to the standard position of center. [Ed. See Micah’s comments below.]
New Shooting Modes
If the camera design in this new patent application ever comes to fruition, we will get 3 shooting modes out of it:
1. Standard Viewfinder Mode: In this mode the main mirror will be in its normal position, but light will pass through it to a sub-mirror, which will in turn reflect down to the phase-difference AF sensor.
2. Fast AF Live View Mode (shown in the top image): The smaller sub-mirror will move out of the way and the main mirror will rotate to a different angle to take advantage of its translucent properties, which will allow light to pass through to the image sensor, while at the same time reflecting a portion of the light to the phase-difference AF sensor. This is what we all want – fast, accurate autofocus during Live View.
3. Full Live View Mode: Both mirrors are rotated/retracted out of the sensor path to allow full exposure of the sensor while observing Live View. This will likely come into play when shooting video or when zooming in during macro or still life shooting for critical manual focus.
When Will We See It?
Who knows if we will even see it. The patent application was filed August 8, 2008, so that’s fairly recent in patent world. Canon files a lot of patents though and many inventions never see the light of day in the form that they are embodied in the original patent application.
This method makes a lot of sense though. Canon needs to make a Live View system that works with fast AF. If the light splitting mirror is really the cake and icing that this patent touts it as, then Canon could reinvigorate its waning EOS Rebel line at PMA 2009 with a feature like this. We will know in a couple of weeks if we’re going to see this feature in a first-half 2009 camera from Canon. Otherwise, we’ll all be playing the waiting game.
Unfortunately, I have no idea how close Canon is to putting this feature into a market-ready camera. Only the guys at Canon know this – and they aren’t talking.
What Do You Think?
Is this the holy grail or just another useless feature?
Do you think it will work from a technical point of view?
What problems do you see?
Did you read the patent?
Looks very much like the Olympus E-330 — which people panned as looking weird, as it didn’t have a pentaprism hump.
For me, the 330’s live view works very well and I’m not sure why Olympus has gone away from that kind of LV, leaving it to Sony and now Canon to carry on. Surely it’s more expensive to build — but the advantages are great.
Live View would be a VERY useful feature for people like me whose aging eyes can benefit from this technology.
I think live view is great for hard to reach places, however the Live view on my D3 isn’t something I use to any regular degree at all.
Personally I think Sony has already hit the nail on the head with their live view Autofocus. In their system they move a Smaller CCD sensor into the viewfinder which is what feeds your live view display while the rest of the camera (including AF module, and Metering system) function the way they are intended. If you really want to use LiveView isn’t that what we are looking for anyway?
There’s no need to redesign the wheel. Just my 2 cents.
@Eric: I think you might misunderstand the standard SLR AF technology. At least that’s the impression I get from reading this.
Current SLR tech is to put the AF sensor in the bottom of the mirror box. The middle of the main mirror allows some light to pass through to a secondary mirror that is positioned at a right angle to the main mirror. This secondary mirror reflects light down onto a sensor at the bottom of the mirror box.
An AF sensor in the bottom of the mirror box is the norm. Nikon, Canon, Pentax, Sony/KM, and even Olympus use this arrangement. The first four definitely never put an AF sensor above the mirror. Olympus I’m not sure about.
There are two detriments to this standard layout: most of your light is reflected up into the mirror box, and only a small amount is picked up by the af sensors. This is why lenses of 5.6 and higher suck for AF in low light.
The novel part of the patent above seems to be that they are using a mirror that is partially transmissive in a different configuation for live view. It looks like the mirror flips in the opposite direction for live view and the AF sensor is positioned optimally for this, with the secondary mirror’s placement higher up the main mirror compared to the standard position of center.
There are some up sides and down sides to this. On the plus side:
a) fast AF in live view!
b) potentially better low light AF
c) the placement of the secondary mirror is closer to the pivot point of the main mirror, which could on it’s own enable higher frame rates.
On the down side:
a) much more complexity, which increases the chances for catastrophic failure
b) the increased complexity could add weight to the mirror assembly which could negate any increase in frame rate and maybe even lower it
c) this is only optimally applied to a crop sensor. This could make even a 1ds series camera even more bulky. This is useless on a 1d series camera.
It’s clear that Canon is developing a faster AF system for LiveView (they already stated that months ago at Photokina) as well as improving Movie mode, because they know that those features will have Huge future and demand
LiveView has enormous potential.
Sometimes is impossible to use the viewfinder, or just much easier to use LiveView.
Also the information displayed in LiveView can be a lot more than in the Viewfinder.
Each way to frame and shoot has its own advantages, so they both MUST be available considering the customers have a very wide range of different needs / tasks / jobs
Thinking that a feature “should not” be included in a DSLR just because “it does not belong to the DSLR world” or one doesn’t use it, it’s a mistake due to having somewhat narrow perspective of other people’s needs, the trends of technology, the demand and the companies’ business.
This improvement indeed will help a lot the Movie/Video mode in DSLRs, wich is also going to be improved and implemented in future DSLRs, despite some people like it or not.
(if you don’t like, just don’t use the feature. It’s that simple. And if you don’t want to pay “more”, don’t worry, it won’t make much difference to today’s DSLRs, so the “extra” price won’t show up)
LOT OF CUSTOMERS, both Amateurs and Professionals (Photographers and Videographers) can get lot of advantages of such implementations, wich will offer a level of quality and features that are not still available under several thousands dollars (and some are even inexistant yet)
So, GREAT NEWS that Canon is really working on it.
Hope that Nikon and other companies do the same.
I never understood why contrast-detect autofocus performs so badly in SLRs. On my compact, the contrast-detect method works in the blink of an eye. The point may be moot though. Canon’s most recent SLRs have a “fast autofocus” mode that drops the mirror for focussing. This interrupts the live view, but its so fast its hardly noticeable.
I ha ve owned a eos RT (real time) that was a camera that use a semi translucent mirror to reflect 1/3 of the light to the visor and 2/3 to the film.
one word : it was fantastic, fast, silent,lightweight, and one of the best camera i ever had….
c’mon Mr Canon, with the increase of sensibillity of the sensor, if it had worked fine with the Velvia 50 it should work with the Cmos of the 5 D MKII :
a EOS RTd = 21mp/ 6fps- no black frame-100% visor-Cfast slot-manual control of speed/aperture/iso/wb/AND PLEASE GIVE ME BACK THAT 1/500s FLASH SYNC
that would be a killer
For FF camera bodies, they should make them bit bigger to allow this system get inside.
Otherwise it would be Terrible (and Ridiculous) to have Faster LiveView AF in xxD than in xD DSLRs !!
But for me the main reason to greatly improve AF is to shoot video mode with fast AF. This is what everyone wants.
However, the mirror in front of the sensor will reduce the image quality, so it is useless for this purpose.
So, to improve AF in video they’ll have to invent something else. Or just implement what they use for pro camcorders – external AF module.
That’s a very good point about video. If the camera could capture video in this mode, it could AF. AF during video would be a killer feature. That is, if it can keep at least 90% of the light going to the sensor and still work. Otherwise video quality might suffer from the loss of light. Also AF might not work so well in low light if the light is all going to the sensor. Catch-22.
It’s good to see that Canon is innovating. Both big camps are.
I’m compiling a history of SLR auto-focus and found your web article helpful in that in pointed me in a new avenue of research. The US patent you cited led me into a labyrinth. If you follow the related application & foreign priorities there is a very interesting field of research to mine.
The patent application to which you refer is preceded by a patent issued Aug. 2008, but applied for in Aug.2004. Clearly Canon seriously considered this solution at least 6 years ago.
If you look carefully at the kinematics of the reflex half-mirror and sub-mirror locii, their swept volume is approximately 30% bigger than a simple reflex mirror articulated at its upper edge. The design can be accommodated in the body of a APS-C DSLR, but either a APS-H or Fx body would need to be enlarged.
The other snag is the beam splitter property of the reflex half-mirror. The optical viewfinder image would be dim. The patent application refers to a technology rarely used in the camera industry, variable reflection. What Yasuo Suda proposes is a coating whose reflectivity varies when a charge is passed across it. That is primarily the difference between the original patent and the later application. The kinematics and software are identical.
There is however a potential snag with his ingenious work around. The half-mirror will carry a charge that will attract dust. The charge will have to be earthed to the chassis, and therefore the sensor carrier isolated. But if you isolate the sensor carrier you are faced with the problem of charge difference during the exposure. This will make the sensor more prone to attracting dust. The sensor carrier will no longer be in static charge equilibrium with the chassis.
I rather suspect this is the principal reason Canon have not put their ides into practice. You can bet your bottom dollar though that they will have built prototypes. Canon would want to know the technology was potentially feasible prior to filing their applications.
There is of course another reason why this PDAF Live View solution has not been built into a commercial DSLR. When you stop to think about it, the DSLER is a hybrid technology. If you had an EVF with a fast enough refresh rate and high enough resolution the reflex mirror and OVF would be superfluous. I have a gut feeling Canon, Nikon & Sony are planning on going down either the Panasonic Lumix G2 route, once EVF’s are technically equivalent to OVF’s, or maybe the Sony NEX route.
Meanwhile Sony have devised a simpler and more elegant solution in their Penta-Mirror Live View PDAF, incorporated in the a350 & a390. Sony’s solution is cheap, and effective, and does not produce a dim OVF image.