Previous news story    Next news story

Video tests of the Canon EOS 650D / Rebel T4i's Hybrid autofocus system

By dpreview staff on Aug 15, 2012 at 19:39 GMT

What improvements has Canon's Hybrid AF system brought to the EOS 650D's usability in live view, and what might this mean for the forthcoming EOS-M mirrorless camera? As a precursor to our imminent 650D/Rebel T4i review, we've published two videos showing how Hybrid AF works, compared both to conventional phase-detection AF and to a contemporary mirrorless rival (in this case the Panasonic DMC-G5). It's a chance see how the 650D performs but also gives an idea of what we can expect from the EOS-M, which uses the same technologies.

For these videos we've used a Canon EOS Rebel T4i with the EF-S 18-135mm f/3.5-5.6 IS STM lens designed specifically for use with the camera's Hybrid AF system. A further explanation of the system can be found by clicking here, but the basic idea is that phase-detection elements on the main sensor help determine the distance of the subject and contrast detection then performs a focus fine-tune, at that distance. The 18-135mm STM's lens' design features a light, internal focus group that can be quickly accelerated and decelerated to suit this autofocus behavior, and uses a stepping motor to allow fast, quiet and precise movement. The EOS-M uses the same technologies and similar designs to offer the same functions.

So, how do they perform?

Canon EOS 650D - Quick AF (phase detection) and Hybrid AF compared

Canon EOS 650D and Panasonic DMC-G5 compared

424
I own it
60
I want it
31
I had it
Discuss in the forums

Comments

Total comments: 243
12
Pritzl
By Pritzl (Aug 18, 2012)

I'm surprised people think STM lenses focus faster. Designed for movie autofocus they are slower and quieter for smoother focus transition while recording.

i.e. you need to test a regular EF lens if you are comparing single shot autofocus AND you probably need to make sure the focus target is inside the limited "+" shaped zone in the center of the sensor if you really want to check the hybrid AF's performance.

3 upvotes
rhlpetrus
By rhlpetrus (Aug 17, 2012)

If that's how the M's AF performs, not good. It looks like CDAF from years ago, even my G9 focuses faster. I don't get it. What does hybrid here means? Is the mirror up and AF is from sensor?

Comment edited 52 seconds after posting
0 upvotes
padoods
By padoods (Aug 17, 2012)

Correct me if I'm wrong, but doesn't the Nikon V1 use hybrid AF also?
Based on what I've read, AF on the V1 is blazing fast. Is this because of the smaller sensor on the V1?

Maybe dpreview can also do a comparison video between the V1 and 650D?

1 upvote
rhlpetrus
By rhlpetrus (Aug 17, 2012)

Nikon 1 has both PD and CD AF off-sensor and the PDAF is blazing fast, the CDAF is not that bad either. It changes to CDAF when light gets too low or contrast is lacking, even though in my exp it works incredibly well.

0 upvotes
matt25
By matt25 (Aug 17, 2012)

one more point: i do think that hybrid AF performance with an ultrasonic EF lens like the short panny zoom used here will be slightly better for still photos AF speed. The STM lenes are not for speed. It may be louder but I think it will be faster, although again I do think it will be slight. Try a 17-40 or something similar to the 14-45 panny you tested.

1 upvote
matt25
By matt25 (Aug 17, 2012)

There seems to be a lot of confusion in these comments about what these videos show.
A couple of conclusions and observations to put this in perspective:
1. Rebel's main and preferred AF mode is thru the viewfinder, which is phase AF, and is very fast. G5 has one AF mode, contrast AF via live view (or an electronic viewfinder, same thing) and it is also very fast on a still subject. Tests show the G3, GH2, G5 are as fast to faster focusing on still subjects than the rebel, 60d, on still subjects.
2. Rebel AF is generally preferred and superior to G5 AF on moving subjects.
3. As a secondary mode of AF on the rebel, live view, still objects AF is v.slow with hybrid AF compared to G5.
4. As to video AF, which is primary benefit to rebel, Hybrid AF may or may not work, though early videos of similar scenes show it favorable to G5 video AF IMHO.
5. Has DP review talked to Canon? Can firmware improve performance?
6. Hybrid is more critical for EOS-M as sole AF method.

1 upvote
Camediadude
By Camediadude (Aug 17, 2012)

I see again you censor my benign and apt observation that your videos are excellent examples of the wadsworth constant run a muck. There is no need for such a long period at the beginning of the video in which nothing happens whatsoever. It bores the viewer to tears and it is a waste of time. It is akin to flooding a gallery with lots of bad shots. You show that you can't take criticism of any kind very well at all here by doing so, by deleting my comment repeatedly. My critique here is not meant to savage, but to help, as a very long term reader here. But, if you do not wish to improve, then I will leave you be to stagnate all you like. I cannot and will not force you to improve, you have to want it yourself, although that is what I thought these here comment sections were partly for. Good luck.

3 upvotes
ThePhilips
By ThePhilips (Aug 16, 2012)

The use of the G5 in comparison suggests that we ought to see its review sooner than later?

1 upvote
Ulfric M Douglas
By Ulfric M Douglas (Aug 16, 2012)

Is it actually using Hybrid PDAF+CDAF, or is it just using three-year-old CDAF tech?

0 upvotes
Marcelobtp
By Marcelobtp (Aug 16, 2012)

Anybody noticed that even in the Hybrid mode the mirror moves?
So probably in the EOS-M it will be much faster, as the nikon 1, and probably the new NEX 5R.
I guess that sony, panasonic and olympus, have much better contrast detection than canon or nikon, so i hope when one of them put this hybrid to work, they will succeed.
But one thing that these videos don't show is the accuracy difference, i think working togheter the accuracy will be always better, with hybrid AF.

1 upvote
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

In "hybrid" mode (right on the top video, left on the bottom), the mirror only moves /after/ focus and exposure. The Rebel series does not have "silent mode" and so the mirror must go down/up after the shot as part of the reset mechanism.

Comment edited 2 minutes after posting
1 upvote
Bob Meyer
By Bob Meyer (Aug 16, 2012)

Yes, even ignoring the delay caused by the mirror, the AF acquisition time on the Canon is dramatically slower than on the Panasonic. Given how many years Canon has been working on CDAF, and how many competing cameras do it better, it's depressing how poor Canon's live view AF is.

3 upvotes
OttoVonChriek
By OttoVonChriek (Aug 16, 2012)

CAF should always be able to achieve perfect focus unless there is no detail whatsoever. Phase focus has several possible sources of error.

The advantage of phase focus is speed. In hybrid AF the idea is reducing the speed CAF takes.

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

Here is a test that would have a decent chance of showing if there is/isn't any improvement due to Hybrid AF. Start with three targets with vertical detail placed at different distances. Make sure two are clearly within the hybrid AF zone and one is not -- perhaps place one in the exact center one a little left and one a lot right. Focus on one, then the other, and then back to the first. Repeat the same test with the target outside of the Hybrid zone so we know it's pure CDAF. If the focus speed is the same inside/outside, then Hybrid is not really doing anything. If the focus speed is different, then you have the likely best case for improvement due to the hybrid PD pixels. Compare to the G5 or Quick AF speed for the same target size/distance (quick AF will need different outer target positioning to hit the AF sensor.)

Comment edited 48 seconds after posting
1 upvote
Donnie G
By Donnie G (Aug 16, 2012)

Why do an apples vs oranges comparison between a DSLR and a mirrorless camera when you could just as easily have compared Canon's new hybrid sensor tech against the tech available in its competitor's DSLRs? When you do that side by side autofocus speed test for video capture, then I'll sit up and take notice.

1 upvote
Amadou Diallo
By Amadou Diallo (Aug 16, 2012)

The mirrorless comparison is an interesting one because Canon's upcoming mirrorless EOS M will share the same Hybrid AF system as the 650D.

2 upvotes
howardroark
By howardroark (Aug 16, 2012)

I just wanted to see the 600D traditional CD AF next to the 650D Hybrid AF to determine if there was an improvement. That might be more useful for those looking at the EOS M since it is likely a Canon body they're looking at rather than the "mirrorless" ecosystem (I think ILC is more accurate representation of what a "mirrorless" camera's selling feature is....just like the SLR is indicative of what makes it special and desirable).

With lenses designed for an EF-M mount I'd imagine any improvement in the 650D AF speed would mean even more in a camera with a shorter flange focal distance.

Comment edited 6 minutes after posting
1 upvote
Bob Meyer
By Bob Meyer (Aug 16, 2012)

There shouldn't be any relationship between flange distance and AF speed. Electrons move at (pretty much) the speed of light.

0 upvotes
howardroark
By howardroark (Aug 16, 2012)

Yes, but focus motors and the glass elements attached to them move through physical space not along wires. Having a shorter distance between the imaging plane and the focus group could make a difference.

0 upvotes
OttoVonChriek
By OttoVonChriek (Aug 16, 2012)

Just a note, electrons do not move a the speed of light, or anything like it. Signals do ;-)

(Think about a long line of newtons balls, the time between a hit at one end and the ball flying off at the other end can indicate a signal speed much greater than the speed of the ball itself).

But this has NOTHING to do with the argument, Bob Meyer is quite right!

Comment edited 60 seconds after posting
0 upvotes
Hansplast26
By Hansplast26 (Aug 17, 2012)

99% of the people do not know the difference between the AF systems. for them it is just AF, and thats how actually anyone else should look at it too. AF is AF and it should be fast, accurate and ideally allow focusing on moving objects

0 upvotes
bcalkins
By bcalkins (Aug 16, 2012)

This is an especially tricky type of test to do and get definitive results. There are so many variations in subjects - some of which favor PDAF and others that will favor CDAF. One hard part to rule out is that CDAF can get 'lucky'. What makes CDAF slow is not knowing which way to move first - but that only makes it slow if it chooses incorrectly to start with. Panasonic seems to be pretty intelligent about the guessing part - which adds to the perceived (and actual) speed. My experience with the Panasonic GH2, Olympus OM-D and Canon 7D is that for static subjects in different lights they can all be fast, and depending on lens can end up hunting for focus. There are certainly cases where the GH2 beat my 7D for focus acquisition in low light, but also cases where the opposite was true. Primes seem to be slower than smaller aperture zooms

In short - I'm glad both technologies are improving and getting combined. All around performance is getting better and better in both camps...

0 upvotes
Bart Hickman
By Bart Hickman (Aug 16, 2012)

I think the reason the Panasonic appears to get lucky is it makes guesses extremely fast. You have to be sure to use a lens that enables this fast guessing.

I don't understand why Canon's hybrid system is so slow. The Nikon 1 cameras are incredibly fast. Perhaps it's a patent issue.

1 upvote
ntsan
By ntsan (Aug 16, 2012)

Well GH2 still focus more accurate than E-M5 even if they use same lens, I guess the algorithm from Panasonic is more fine tuned than other brand's CDAF

0 upvotes
bcalkins
By bcalkins (Aug 16, 2012)

@ntsan: I'm not sure that the GH2 is any more accurate. I haven't noticed my photos being any less sharp with the OM-D...

0 upvotes
Low Budget Dave
By Low Budget Dave (Aug 16, 2012)

Could you do us a favor and let us know your opinion of the test results? I have been reading through the comments, and different people seem to interpret the results differently.

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

Summary:
1. The Hybrid AF is not as fast as Quick AF for this target.
2. the T4i+18-135 STM is slower than the G5+14-42mm for static scene live view AF for these targets (although DPR says this represents their experience with other situations.)

What this means for EOS-M and the usability of the T4i vs. other designs is when the interpretations vary because it then becomes highly speculative or subjective.

0 upvotes
flipmac
By flipmac (Aug 16, 2012)

Basically, Canon's liveview still sucks when compared to the competition, even with the help of on sensor PDAF.

Some possible causes:
- Canon's CDAF algorithm is behind
- the stepper motor in the STM lenses aren't as quick as others in mirrorless lenses
- connections between sensor, processing and lens is slow

That said, there's potential here. When (and not if) Canon improves their CDAF to the same level as others, this hybrid would be their trump card. It can solve the remaining issue with CDAF: continuous AF.

Comment edited 52 seconds after posting
0 upvotes
Bob Meyer
By Bob Meyer (Aug 16, 2012)

I think your "When (and not if)" comment may be optimistic. Canon's had CDAF for what, at least 3 generations of DSLRs? And it was dog slow then, and it's still dog slow. I don't understand why, either. Every other manufacturer has made great strides in CDAF technology in the same, or shorter, period of time.

I've shot EOS since the first EOS film camera, but I'm growing increasingly jaded.

0 upvotes
Area256
By Area256 (Aug 17, 2012)

Yeah I also don't get what's taking Canon so long to get CDAF working. I love my Canon DSLR, but seriously they should have solved the CDAF problem by now - live view is basically not usable except for manual focusing or static subjects when you aren't on a time budget.

I do really hope it's a "when" not an "if" issue, but I'm losing hope that it'll come anytime soon...

Unless they pull something off for the EOS M in the next two months, I suspect they are going to get a lot of negative reviews (although that may finally make them wake up and realize it's important).

0 upvotes
zos xavius
By zos xavius (Aug 16, 2012)

I find this development intriguing. I've used both types of focus extensively over the years and cdaf, while certainly more accurate has some significant failings still. Especially when it comes to tracking as well and low light situations where pdaf rules the day. Having switched to pure dslr shooting, I ultimately prefer pdaf and find it more than adequate and quite fast. I'm not too sold on this hybrid system. Here's the thing: how much shutter lag is induced? Obviously the sensor must be exposed to use the cdaf sites, so clearly the shutter must close and reopen right? Or does it use an electronic shutter with the curtain merely closing at the end of exposure? Also I don't think keeping the shutter curtain open all the time is good for dust. Mirror boxes get dusty fast, and the shutter curtain helps keep a good bit of that away from the sensor. Just some thoughts.

0 upvotes
Bob Meyer
By Bob Meyer (Aug 16, 2012)

What recent cameras have you used CDAF in low light? I find the GH2 and OM-D to be at least as good as my EOS gear, with or without focus assist lights. I'd agree that some earlier cameras, especially Olympus Pens, struggled in low light, but that problem has been solved.

0 upvotes
Sean Nelson
By Sean Nelson (Aug 16, 2012)

CDAF autofocus has always been more accurate, but for a long time phase detection AF had the advantage of speed. That's no longer the case for static subjects - a well-implemented CDAF is now just as fast if not faster than good PDAF systems.

The one advantage remaining for well-implemented PDAF is its ability to continuously track subjects moving towards or away from the camera. If Canon's hybrid system can do that well and still maintain the high accuracy of a CDAF system then it's a worthwhile achievement. I wouldn't rush to judge it without knowing whether they've been able to achieve that goal or not.

But as the processing speeds and algorithms for CDAF continue to improve I think that it won't be all that long before it's able to match PDAF for continuous tracking too.

0 upvotes
Jon Ragnarsson
By Jon Ragnarsson (Aug 16, 2012)

I would have assumed that Canon of all companies would have the knowledge and manpower to create a fast CDAF. Olympus struggled in the beginning with the E-P1, it wasn't until they re-designed the AF mechanism in the lenses that AF speed/accuracy was 'good enough'.
I wouldn't be surprised if we will see CDAF overtake PDAF in the future with better software algorithms, lens motors and processors...

0 upvotes
weddingsbyira
By weddingsbyira (Aug 16, 2012)

As far as focus goes. The T4i blows away my Canon Mark II in autofocus. I recently used the 135mm 2.0 at my kids football practice and focus was dead on at 2.0. I find myself using this camera instead of my 5D Mark II. I can't believe it, but it is true. The added software makes the 5DII interface seem outdated.

1 upvote
Lucas_
By Lucas_ (Aug 16, 2012)

It's a start. Anyway, Panny is quite ahead and Sony ( not shown on this demo ) is easily the master with the SLT design.

2 upvotes
howardroark
By howardroark (Aug 16, 2012)

The "master with the SLT design" is not a compliment. They've mastered Auto Focus in Live View by removing the option to get a partially hazed mirror out of the sensor's way and use all the light coming through the lens. Sony came up with a compromise that was reasonable to some people at the time, but if PD AF sensors can be integrated into the sensor itself for people who insist on using Live View instead of the viewfinder then the SLT design will quickly become a fossil...and rightly so. The SLT design is a compromise that should be made very, very carefully by consumers considering what else is available The master of all the one-legged Ninjais not the same as the Ninja master.

Comment edited 14 minutes after posting
2 upvotes
Dave Oddie
By Dave Oddie (Aug 16, 2012)

The compromise as you out it simply works and works very well. And since it exists and PD AF sensors integrated into to sensor do not I don't think you really have a point.

You are basically suggesting something that doesn't exist is better than something that does.

2 upvotes
howardroark
By howardroark (Aug 16, 2012)

"The EOS 650D has a new 'Hybrid CMOS' sensor that now includes pixels dedicated to phase detection autofocus (in a similar fashion to Nikon's 1 J1 and 1 V1 mirrorless cameras). The Hybrid AF system uses these to set the lens quickly to roughly the correct distance, then uses contrast detection AF to fine-tune focus." -- DPReview

Get with the program.

Comment edited 41 seconds after posting
1 upvote
howardroark
By howardroark (Aug 16, 2012)

And yes, the SLT design does work, but by doing something with light other cameras have wisely avoided: stealing light away from the sensor.

1 upvote
ET2
By ET2 (Aug 16, 2012)

Each AF point on the main sensor takes light away from several thousand pixels. Now imagine putting 20 points on the main sensor. That's taking a lot of light away from the sensor too. Now image putting 102 points on the main sensor (the rumored A99 AF system).

Besides, as you can see, Canon version doesn't perform as well as SLT cameras. Sony has main sensor PDAF patents dating back to 2009. It's rumored new Nex models will incde that, but obviously Sony thinks SLT is superior than on sensor PDAF for A-mount.

1 upvote
Gianluca101
By Gianluca101 (Aug 16, 2012)

Sony slt design it's simply the best option, today, because not only work superb well, but it work with all alpha lens, even minolta.

You have to build an entire system lens with stepper motors for good result with pdaf on sensor, otherwise..... :(

1 upvote
Bob Meyer
By Bob Meyer (Aug 16, 2012)

The SLT designs, unfortunately, share a [potential] problem with all other non-sensor focusing systems: front and back focus. On chip sensors eliminate this. Frankly, if I want good PDAF (and I do), I think I'll stick with standard DSLRs with OVFs. For my needs, at least, I don't see any real advantage to the SLTs. (Yes, I do realize that my needs aren't the same as everyone else's, and the SLT design may be a good compromise for some.

0 upvotes
howardroark
By howardroark (Aug 17, 2012)

Brutal. Sacrificing light, detail, noise....for liveview.

0 upvotes
ET2
By ET2 (Aug 17, 2012)

Funny you say sacrificing detail, when A55 scored higher resolution numbers than both D7000 and K-5 on imaging-resource review.

Sacrificing light, yes, but so is the main sensor PDAF. Each AF point on the main sensor takes away light...

SLT also gains something aside from liveview: 12 fps with AF, EVF, no mirror blur, shorter shutter lag due to electronic first shutter curtain ..

1 upvote
howardroark
By howardroark (Aug 17, 2012)

Okay, I'm not going to quote all of the goofy, quirky garbage you have to do in order to get that 12 fps. The link to the A77 says it all under the heading "Continuous Shooting and Buffering" http://www.dpreview.com/reviews/sonyslta77/10
The A77 was announced in Aug. 2011. The 7D was announced in Sept. 2009. The 7D does 8 fps without an aperture lockout or glitchy display issues. Shutter lag is .061 through the viewfinder and .202 in liveview. The A77 is .054. In two years they really made some leaps and bounds in AF technology. Mirror blur is only a concern in some very specific situations and there is a tried and true method for taking care of that. Sorry, but unless you absolutely require liveview all the time it's a really lousy compromise.
The extra megapixels over the 7D are nice.....up until about ISO 400 when the A77 noise starts taking control and pulverizing detail.

Comment edited 4 minutes after posting
0 upvotes
dz1
By dz1 (Aug 17, 2012)

Sony shouldn't have gone with the 24MP sensor for the A77. That's why there's too much noise because Sony made some poor choices there. Noise on the older Sony A55 is better.

Incidentally, I have seen reviews that say low light performance on the 650D is worse than on the 600D due to all the new AF pixels on the sensor taking away light from the sensor. Now that is a real concern for any buyers. Looks like it is Canon making the lousy compromises.

Comment edited 1 minute after posting
1 upvote
howardroark
By howardroark (Aug 17, 2012)

Even if they use 10,000 dots for the AF that would be 0.01% of the total. Go ahead and use 50,000 if you must and make it 0.27%! Now how much light are those dots losing, around half? How much light must Sony siphon off to get LiveView and AF? And how much sensitivity or accuracy do the AF sensors lose by getting a tiny fraction of the available light? And how much extra noise is there to compensate for the loss of light across the entire sensor for those compromises? Too much in my opinion, but obviously others have different priorities than I do. More power to them. I hope they enjoy their cameras and get some great shots. I think this technology sacrifices too much.

0 upvotes
Gothmoth
By Gothmoth (Aug 16, 2012)

canon has already a few new patents up it´s sleeve.
it´s new technology.. it needs time to mature.

i remember the horrible AF on my Olympus E-PL1... what a useless crap that AF was.
2 models later the E-PL3 autofocus is nice.

but one thing panasonic and olympus can not change in their m43 cameras is the smaller sensor... FAIL.

1 upvote
Winston Loo
By Winston Loo (Aug 16, 2012)

Obviously you've not seen the IQ from the E-M5. In every way equal or better than the 650D. AT LEAST, the AF is fast and it doesnt cause you allergy like the 650D !! Lol.. I think the 650D is a joke and EPIC FAIL!

8 upvotes
FrankS009
By FrankS009 (Aug 16, 2012)

The 4/3rds IQ is fine. The advantage of the smaller sensor is smaller lenses and a smaller, lighter package. Canon and the other camera makers who put larger sensors in their cameras can't have it both ways. It is just the physics of the situation.

4 upvotes
VJVIS
By VJVIS (Aug 16, 2012)

@Gothmoth, I have been using my pan GH2, G2 (don't have that anymore) and recently GF3, for a while, especially GH2. I have printed 30" x 20" photos shot at F1.4 using the 25mm lens and the photos look gorgeous. As a pixel peeper I find it hard pressed to find any difference between my GH2 and the Canon 7D upto ISO 1600 after which 7D gains about half a stop to a stop advantage. But if you are printing large sizes you don't shoot at high ISOs anyway. It is because of Canon fanboys like you that Canon is getting away by putting out crap like this. My friend owns the 7D and he said once the continuous AF improves with CDAF he will be switching. At the moment the continuous AF is definitely better on the PDAF, but static objects there is no way you can tell a difference in real life shooting, and the CDAF on the mirrorless cameras is light years ahead of the CDAF crap that Canon and Nikon are putting out.

Comment edited 3 minutes after posting
1 upvote
Bart Hickman
By Bart Hickman (Aug 16, 2012)

Actually, Canon can have it both ways. If they just match the pixel density of the m4/3 sensor, then they don't need a longer lens for a given FOV because they can crop. If lens sharpness is an issue, only the center of the lens must be sharp enough to support this (and they generally are.). The Canon lens would still be slightly larger to handle coverage at the wide end, but since the wide end will be wider for a given F/L, you'd need fewer lenses in your Canon kit, so it works out about the same.

0 upvotes
random78
By random78 (Aug 16, 2012)

Canon APS-C has only a marginally bigger sensor compared to m43. The difference is fairly small. Based on sensor size you would expect canon aps- to have half a stop of advantage over m43. So given equally good technology canon aps-c at ISO1200 should match m43 at ISO800. And thats just about what you get in G5 vs canon aps-c (the older 12MP m43 sensor was pretty bad though).

However canon is no longer the top of the line in sensor tech so whatever little advantage should be there is lost due to the superior sony tech used in E-M5. E-M5 m43 sensor matches and in some aspects exceeds the canon aps-c sensors.

0 upvotes
Bob Meyer
By Bob Meyer (Aug 16, 2012)

Canon may have a few patents up their sleeve, but their CDAF technology is hardly new. It's been on several generations of their DSLRs, and it's as bad now as is was in the beginning.

0 upvotes
Bart Hickman
By Bart Hickman (Aug 17, 2012)

On the one hand people say m43 has an advantage because it has a much smaller sensor which enables much smaller lenses. On the other hand people say m43 has no real disadvantage because m43 is only a little smaller than 1.6x or 1.5x. So which is it?

The Canon 650 CDAF is slow, but it looks a heck of a lot faster than the previous generation (60D) to me. The 60D is so slow I can hardly see it moving. At least now the 650 looks about the same speed as the Nikon D7000 CDAF. Seems like Canon might be holding back which is a mistake IMO.

Comment edited 4 minutes after posting
0 upvotes
rhys1974
By rhys1974 (Aug 16, 2012)

Phase detection isn't better than contrast detection. It's the new feedback mechanism in the newer Canon lenses and cameras that improves focus. Lensrentals have a fantastic review which exposes the Canon hype for what it is:

http://www.lensrentals.com/blog/2012/08/autofocus-reality-part-3b-canon-cameras

1 upvote
Gothmoth
By Gothmoth (Aug 16, 2012)

i would not call it a hype... because it is actually working.

"The two newest Canon cameras have more accurate phase-detection sensors than their previous cameras. The newest lenses have more accurate focus movement (or provide more accurate focus movement feedback, or both) that takes advantage of those sensors."

of course for the best AF performance you need the latest technology in lenses.. what a suprise!!!

who would have thought that a 24 year old 50mm f1.4 does not work as good as a new lens...... LOL

Comment edited 3 minutes after posting
2 upvotes
peevee1
By peevee1 (Aug 16, 2012)

So far the hybrid AF is a fail.

3 upvotes
ThePhilips
By ThePhilips (Aug 16, 2012)

My reading was that intended purpose was to use PDAF for a larger move and CDAF to fine-tune the focus. It does twice more work, thus slower.

Yes, it is generally fail, but so are the PDAF optimized lenses on the CDAF-only bodies. And that's quite a lot of the lenses, pretty much all 1st and 3rd party lenses.

As a taster of the tech, IMO it is promising. It is promising that both Canon and Nikon try to bridge the PDAF and the CDAF.

1 upvote
Bart Hickman
By Bart Hickman (Aug 17, 2012)

Canon's hybrid AF may be slow, but Nikon's V1 AF is incredible--much faster than even Panasonic when I've compared them. Although Nikon uses *only* PDAF when the light allows, so their definition of hybrid AF is different than Canon's.

0 upvotes
Edoveral
By Edoveral (Aug 16, 2012)

This Hybrid-AF should be able to be a step up when it comes to video AF. But from what I have seen on the net in several tests it's still not there yet. It's nice Canon is finally giving it (and the mirrorless-concept) a try, but to little to late. I'm jumping ship. Just sold my Canon DSLR and switching to this lovely Panasonic G5.
The future is in the mirrorless concept, so I'm betting on the horse that is currently implementing things best. We'll see what the future brings.

2 upvotes
Greynerd
By Greynerd (Aug 16, 2012)

When contrast detect cameras struggle to focus I often move the focus square to overlap the edge of the object which usually solves the problem. In fact I would move the square from where it is on the Canon example to where it is on the Panasonic example to get focus. It does not seem to be a very convincing demonstration. In fact having the focus square in the middle of the body in a same sized image would be a tougher and more equivalent test.

Edit: Also looking at the examples the Panasonic focus square is also larger which helps especially when it overlaps the object.

Comment edited 8 minutes after posting
1 upvote
Amadou Diallo
By Amadou Diallo (Aug 16, 2012)

The Canon uses PDAF to determine initial subject location before calling on CDAF to fine-tune. In the many tests we've run under different lighting conditions and with different subjects, the results you see in the videos we posted do not materially change. We're not talking about small performance differences here.

Comment edited 15 seconds after posting
0 upvotes
Valentinian
By Valentinian (Aug 16, 2012)

Thanks. I understand the first part of your statement: "The Canon uses PDAF to determine initial subject location before calling on CDAF to fine-tune". However, could you say if it is faster and more accurate than Phase AF only and than Contrast AF only?
(itis not, right?)

Comment edited 1 minute after posting
0 upvotes
Richard Butler
By Richard Butler (Aug 16, 2012)

Even for single shot AF, in theory a hybrid system could be faster than just CDAF (since it has a greater understanding of how far away the subject is), while retaining CDAF's greater accuracy.

At present, this potential isn't being achieved.

1 upvote
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

You can see the overshoot and then correction of the "Hybrid AF" in the video. This seems inconsistent with how PDAF should work --which suggests (to me) that the PD sensors did not contribute to focus in this case. You see similar jitter when the subjects outside the Hybrid AF zone are chosen.

1 upvote
Amadou Diallo
By Amadou Diallo (Aug 16, 2012)

Erik, PDAF will not necessarily prevent 'overshoot'. What it will do in this hybrid system is prevent the lens from initially heading in the wrong direction. Once it is 'near' the subject, CDAF kicks in and does its front/back search for focus.

Comment edited 35 seconds after posting
1 upvote
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

These tests look like they start at minimum focus distance. If so there is only one way to move and PD direction would not be of any advantage -- unlike PD distance estimation. Understanding when/if/how much the Hybrid AF does it's thing would be valuable.

0 upvotes
viking79
By viking79 (Aug 16, 2012)

The only real benefit I see to the canon focus right now is for video, where going the wrong way will be very noticeable in the scene. How does it perform during video focus?

0 upvotes
Amadou Diallo
By Amadou Diallo (Aug 16, 2012)

We haven't seen it begin focus in the wrong direction for subjects that are in the central area of the scene where the PDAF sensors are located. We're looking at some options for doing video AF comparisons for a future article about the current state of DSLR and mirrorless AF.

0 upvotes
Greynerd
By Greynerd (Aug 17, 2012)

Thanks for the reply Amadou. Obviously you have found marked differences in varied testing so that is fair enough.

0 upvotes
AmateurSnaps
By AmateurSnaps (Aug 16, 2012)

Lumix fz200 looks great on paper (ignoring the poor initial test photos) but it would benefit from a bigger sensor. Waiting to see canons own sx50.

650D is a good camera but the hybrid sensor seems a half way house approach.

0 upvotes
Blaufeld
By Blaufeld (Aug 16, 2012)

It's simply the new Canon marketing gimmick that has totally failed.
Interesting how this stirred the usual fanboy negationism, tough.

4 upvotes
Pete Mc
By Pete Mc (Aug 16, 2012)

http://www.videometry.net/650D_vs_550D/650D_vs_550D.jpg

Having bought and tested the 650D, I returned it and kept my 550D.
While the features are fantastic, particularly the touchscreen, focus was a real problem, especially because in automode, aperture is wide open giving a small to non-existent depth of field. Video displayed the same characteristics. It may track, but never achieves sharp focus. The sensor appears to require far more light to operate resulting in higher ISO's, longer shutter speeds, and wider apertures, compared to earlier models.

I should say I only tested a single camera, and didn't dare invest time or money in a replacement. I think I'll grab a Lumix FZ200 instead. The lens alone (a 25-600mm F2.8!) would cost 4 times as much for a Canon DSLR. Not to mention 12 fps burst, upto 240fps video...
http://www.dpreview.com/previews/panasonic-lumix-dmc-fz200

Comment edited 1 minute after posting
0 upvotes
zos xavius
By zos xavius (Aug 16, 2012)

Yeah...12fps...in JPEG! Also a 24-600 lens comes with a lot of compromises. I should know. I own a fz-35. I also have a pentax k-7. There is no comparison. Even at screen resolution. If your vision is poor and you do nothing but snapshots than the fz might be adequate. It wasn't for me, which is why I moved on to a real camera.

0 upvotes
EDWARD ARTISTE
By EDWARD ARTISTE (Aug 16, 2012)

yeah something about this post didnt sound right25-60-0 2.8 would have many photogs buzzing, but they arent.

Devil is in the details, eh?

0 upvotes
mosc
By mosc (Aug 17, 2012)

Why do people give equivalent lens ranges without equivalent f-stops? If you want to call it 25-600, then you're multiplying by the crop factor. That's great for comparison with the more traditional film standard and all but why don't you correct the f-stop by the crop factor too? f2.8 doesn't look so impressive when you have to multiply it by 5.2x crop. Your FZ200's f15 equivalent throughout lens speed is pretty terrible for anything beyond broad daylight.

I like bridge cameras. You don't need extra lenses and they work great as a camcorder from the back of the gym. Paired with a flash, they work well enough indoors too (where flash is legal). Still, I would never be bragging about the widest f-stop on one. That's like bragging about the size of your motor in a geo metro. Bridge cameras do lots of things well. The thing they are the WORST at is aperture.

0 upvotes
Hansplast26
By Hansplast26 (Aug 16, 2012)

It shows that Panasonic contrast detection AF is leaps ahead of Canon AF (despite that they use a hybrid AF system).

7 upvotes
Amadou Diallo
By Amadou Diallo (Aug 16, 2012)

That was exactly our point. On paper, the hybrid AF should offer clear advantages but we're not seeing them in this first iteration of the technology.

0 upvotes
Noveenia
By Noveenia (Aug 16, 2012)

Contrast detection does not aim at the speed, but the accuracy, does it?

0 upvotes
Andy Crowe
By Andy Crowe (Aug 16, 2012)

All types of autofocus /aim/ for both, with PDAF it's easier to get speed, with CDAF it's easier to get accuracy.

4 upvotes
Valentinian
By Valentinian (Aug 16, 2012)

"with PDAF it's easier to get speed, with CDAF it's easier to get accuracy"
So, if the hybrid AF doesn't get faster accuracy, what is it good for? (maybe to use the existing canon lenses?)

0 upvotes
Bob Meyer
By Bob Meyer (Aug 16, 2012)

"All types of autofocus /aim/ for both, with PDAF it's easier to get speed, with CDAF it's easier to get accuracy."

Except that today's best CDAF systems, like those on the G5, are every bit as fast as anyone's PDAF when shooting static subjects, while still being more accurate. Only when shooting moving subjects does PDAF really have any advantage over the best CDAF systems any more.

0 upvotes
Bart Hickman
By Bart Hickman (Aug 17, 2012)

Isn't the whole point of high performance AF to be able to focus on moving subjects? Making AF go from 0.3s to 0.2s on static subjects is nice, but it's really a solution to a non-existent problem.

0 upvotes
sproketholes
By sproketholes (Aug 16, 2012)

These videos made no sense to me at all.
one at a time please.

1 upvote
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

Second major flaw in this test. on p159 of the manual Canon warns that the system may have trouble when there is only horizontal detail, I.e. the Hybrid system only sees vertical detail. What did DPR use as a test subject?

1 upvote
MarkInSF
By MarkInSF (Aug 16, 2012)

I'd call that a Canon flaw not a test flaw.

8 upvotes
xvip
By xvip (Aug 16, 2012)

1) in case of PDAF Canon handles with heavier lens elements very well. So I think the weight of moving elements are not an issue.
2) you are right about hybrid AF area
3) CDAF in Panasonic doesn't have a problem with horizontal details...

0 upvotes
jgardia
By jgardia (Aug 16, 2012)

The weight of the elements you move is more important in CDAF than in PDAF, because when you do PDAF, you estimate how much you have to move, and then you do it, but in CDAF you move a bit, check, move another bit, check, etc. So you have to deal with the inertia problems of moving and stopping during the whole autofocus process. in PDAF you accelerate and decelerate only once (each) in the whole autofocus process.

2 upvotes
Amadou Diallo
By Amadou Diallo (Aug 16, 2012)

Erik, we focused on three different objects in the video. And these comparative results match up with what we've experienced in using the 650D while producing the review (coming soon) in a range of real-world situations.

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

So you agree that that 2 of the 3 targets are outside of the hybrid AF zone? That when you are comparing AF speed to the G5 for these 2 targets you are only comparing conventional CDAF speed that has nothing to do with Hybrid AF? That should be noted in the narration. And then there are the issues with the center point (on the boundary of the zone and detail-type). This may or may not represent how fast/useful in Hybrid AF is "real world" situations but it's a bad example for direct comparison. If you want to compare to "Quick AF" you need to choose the /same/ focus point (center for both) and include vertical detail because Hybrid AF is not cross-sensitive.

Comment edited 7 minutes after posting
1 upvote
VJVIS
By VJVIS (Aug 16, 2012)

Why on earth does that matter? There are clearly situations where canon is slower. You just want to find the perfect situation where the hybrid AF is faster? How many times are you going to come across a situation (object like that) in real life. Canon is slower and they are putting out technology that is old and outdated. Stop supporting canon so they start putting out real advances in photography instead of crap like this.

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

xvip: the restricted area and directional sensitivity of the T4i Hybrid AF zone could be a significant limitation in real world use. But as a demo for the /potential/ of Hybrid AF in the T4i, this one is poorly choosen. It's possible it's just as limited with an optimal target - but how do we know?

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

VJVIS, how does this matter? The title of the article is "Video tests of the Canon EOS 650D / Rebel T4i's Hybrid autofocus system". So when 2 of the 3 comparisons don't include the Hybrid zone (but dpreview forgets to mention this), it's a tad misleading. The T4i + 18-135 STM is likely slower in some/many/all situations but it deserves a better test.

0 upvotes
Amadou Diallo
By Amadou Diallo (Aug 16, 2012)

Erik,
As you know, the Hybrid AF system encompasses two different technologies, used in conjunction over a small area of the total scene.
What we think is most useful for a short video (and keep in mind these were pulled from a forthcoming full review with more context) is to see how this 'system' performs against one that is mature. It's not a PDAF vs CDAF shootout, although that's where the comments quickly headed.

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

I thought the point of the shootout was to evaluate/compare the "Hybrid" part of the T4i live view AF system. Because of the targets and focus points chosen, it's not clear that you've done that (i.e. used the Hybrid feature benefits at all.)

0 upvotes
dark goob
By dark goob (Aug 16, 2012)

Canon still sells cameras with mirrors in them?

May I ask... why?

Comment edited 19 seconds after posting
5 upvotes
njkdo
By njkdo (Aug 16, 2012)

Because if not I will not buy Canon anymore.

15 upvotes
Andy Crowe
By Andy Crowe (Aug 16, 2012)

Maybe because the market for DSLRs is still bigger than CSCs?

1 upvote
io_bg
By io_bg (Aug 16, 2012)

Maybe because having an optical viewfinder and a longer-lasting battery is cooler?
(p.s.: I'm a Nikon shooter)

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

Missing the Hybrid AF zone -- look the the diagram of the Hybrid AF area here:
http://2.static.img-dpreview.com/previews/canon-eos-650d-rebel-t4i/images/AFcoverage.jpg (page 3 of the preview) . The outer two targets are definitely not in the active hybrid zone and would not use this focus mode! Even the "center" press is near the top of the zone and may have missed using hybrid mode to any good effect.

1 upvote
Amadou Diallo
By Amadou Diallo (Aug 16, 2012)

Erik, look at the first video again and note that the AF point in hybrid mode is well within the phase detection boundary of the chip. And don't forget that the main advantage of PDAF in Canon's new hybrid approach is that it prevents the lens from heading in the wrong direction. Indeed, we've only seen the long CDAF hunt - focusing all the way out and all the way back in - occur well outside the central area of the sensor.

Comment edited 13 minutes after posting
0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

You have a strange definition of "well within". I overlaid your graphic of the Hybrid area with the image of the focus point from your video here: dpr://galleries/8632621561/photos/2158037. The center of the focus point box is almost exactly on the notched corner of the Hybrid zone. Canon's Hybrid AF may or may not work, but the demonstration here is dubious.

0 upvotes
Amadou Diallo
By Amadou Diallo (Aug 16, 2012)

If we had seen a different result by shifting the AF point by that small a degree we would have said so.

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

That's your excuse for a poor test? That it confirms what you already believe to be true? That's called "confirmation bias."

1 upvote
Amadou Diallo
By Amadou Diallo (Aug 16, 2012)

The point of my previous comment was not that we didn't 'think' it would make a difference. It's that we've tried it and it didn't make a difference.

Comment edited 2 times, last edit 1 minute after posting
0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

Then simply show us and eliminate all doubt. What's so hard about that if you've already done the testing?

0 upvotes
Bob Meyer
By Bob Meyer (Aug 16, 2012)

Erik, don't cling to brand loyalty. It's pretty obvious to everyone else than Canon's hybrid tech simply doesn't work well. Unless you think DPR is simply lying to us for some reason. Even if you were right, it's still a dismal performance for CDAF, compared to the G5 and other recent MILC cameras. And it sure doesn't bode well for the EOS-M.

0 upvotes
chrisnfolsom
By chrisnfolsom (Aug 16, 2012)

This is bizarre - who takes pictures with an slr by pressing on the screen on the back of the camera?
1. The experience would be better looking through the viewfinder.
2. With low light and exposure the mirror will be down longer creating the illusion of speed - this exposure time is very slow - please test this in the sunlight.
3. We need to see if the actual focusing is better for the phase detection. Some final prints from each with a very shallow lens would let us know much more.
4. Please test this with moving objects in movie mode - moving at different speeds in different directions - pausing at different speeds to test overshoot - that would be interesting - as well as their vibration control.

Interesting though - DPR should include some movies of the camera's in actions to put some actual real world association to all those numbers that are shown - I love the numbers personally, bit it is hard to actually know what they all mean - thanks for all the hard work!

1 upvote
grumpyolderman
By grumpyolderman (Aug 16, 2012)

obviously they did press on the LCD to show you (visualise) the differences.....all those other tests are of course interesting but i would say about a 100 times more difficult.....what I find interesting is that they did not compare the speed of the phase on the EOS with the G5....would it stir up too much trouble here?

JL

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

For stills, most people care about total shutter lag rather than AF speed. In live view Quick AF mode, the Canon would lose every time due to mirror flap. There is a reason almost no-one does eye-level AF+lag testing - it's darn hard and the results are lens and target specific.

1 upvote
joharis
By joharis (Aug 16, 2012)

The test is about what improvements has Canon's Hybrid AF system brought to the EOS 650D's usability in live view, and what might this mean for the forthcoming EOS-M mirrorless camera?
In live view you don't look through the viewfinder and the EOS-M doesn't even have a viewfinder. So, what's bizarre about the test?

7 upvotes
waitformee
By waitformee (Aug 16, 2012)

If I have a G5 and I want to know if 650D is better then clearly -- G5 is better.

I am not a G5 owner or supporter but clearly I can say that G5 can take more shots in the same given period of time compared to 650D

3 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

Ideally you'd like to use more similar lenses if you are just comparing the AF system. The 7x Canon lens has larger, heavier elements to move for focus than the little 3x Panasonic. So as a test of systems you can buy today, it's sort of valid although perhaps one of the 10x u4/3 lenses would be more similar in capability. For judging if hybrid AF can improve focus vs optimized CDAF, the lenses are really too different to make any firm conclusions.

1 upvote
Richard Butler
By Richard Butler (Aug 16, 2012)

The STM version of the 18-135mm only uses a single focusing element, so far as we know.

Ultimately, we can only perform this test with lenses that exist, and Canon currently only produces two STM lenses (and it really wouldn't be fair to use the unit-focus 40mm, which is trying to move lots of glass around). The 14-140mm Panasonic zoom is similarly fast, but that's a much more expensive lens, so we didn't think it was a fair comparison.

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

Neither choice is exactly "fair" but it would be interesting to see how much difference the lens size/capability might make. How similar is "similarly fast"?

0 upvotes
Nappe1
By Nappe1 (Aug 16, 2012)

wait a minute, this new Hybrid AF is exactly the same hybrid AF as in my _over 4 years_ old Olympus E-420? It seems to do exactly the same things (first contrast af with mirror up and then pdaf with mirror down.) as E-420 with liveview when the lens is not supported by CDAF only function. Canon seems to be slightly faster, but the difference ain't big enough considering Oly's version has been available over four years already.

So what's the new thing here? Or is it just Canon inveting something again and by having more name trying to get the credit?

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

No, dropping the mirror to focus is what Canon calls "Quick AF". Hybrid AF is the one where the mirror only goes up /after/ the shot. You'll notice the lens overshoots and then iterates a bit to lock focus. That suggests that either the lens focus mass is too great or the phase-detect is not getting a good solution on the target. I wonder if there is any vertical vs. horizontal sensitivity to the Hybrid focus like with non-cross sensors.

Comment edited 3 minutes after posting
1 upvote
jpr2
By jpr2 (Aug 16, 2012)

Nopa, read tech. descriptions of T4i carefully again = the "hybrid" AF of E420, and the hybrid on-sensor PDAF of T4i are different animals :)

0 upvotes
Richard Butler
By Richard Butler (Aug 16, 2012)

Erik, the lens overshoots and iterates because it is a hybrid system - it is conducting a CDAF focus check. It does this in all the tests we've tried.

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

Then what is the Hybrid part contributing? Does the overshoot behavior differ for subjects inside vs outside the center area? It may very well be that Hybrid AF has little or no real effect but you need to do a better job of showing it.

0 upvotes
VideoSam
By VideoSam (Aug 16, 2012)

Panasonic's CDAF has come a long way since the G1. From what I gather, and after taking over 80,000 photos with my GH2, the G5 looks even faster than the flagship GH2. Makes me excited for the GH3 around the corner!

Comment edited 45 seconds after posting
2 upvotes
pannumon
By pannumon (Aug 16, 2012)

The bottom line is that G1 AF is much faster than the hybrid AF on 650D (based on these videos and my experience with G1).

2 upvotes
waitformee
By waitformee (Aug 16, 2012)

G3 AF is super fast.
Some of the compact can AF faster due to a smaller lens. Smaller lens is due to smaller sensor. But this smaller sensor in G3 as compared to 650D does not lose out very much on image quality.

So what is this about Bigger DSLR?

3 upvotes
paqman
By paqman (Aug 16, 2012)

have watched that first video 4 times and have no idea what it's showing. what's with the whole video going blank/black a couple of times?

0 upvotes
Richard Shih
By Richard Shih (Aug 16, 2012)

When the screen is pressed on both cameras, the cameras are each taking a picture.

1 upvote
qwertyasdf
By qwertyasdf (Aug 16, 2012)

I'll give a better explanation:
The screen on the left is using a so called "quick AF", when the user half-presses the shutter button, the mirror flips down so that the main AF sensor can be used, and then the mirror is flipped back up with AF is achieved, thus the screen blacks out.

0 upvotes
qwertyasdf
By qwertyasdf (Aug 16, 2012)

Does a hybrid mean putting 50% effort into PDAF and another 50% into CDAF?

And end up with a system that is 2 times slower than the competition either way?

I'm a Canon user, and has no intention to upgrade my xti.

0 upvotes
Nectar D Or
By Nectar D Or (Aug 16, 2012)

I'm also a Canon user, and have no intention of upgrading my t3i, yet I do understand what Canon had done.

This is a DSLR, not a mirrorless. It has an excellent auto focus system (much better than yours or mine). IN ADDITION it has reasonable focus system for liveview.
So for shooting sports, you'd use your OVF. Yet liveview can be used in many other conditions with the advantage of touchscreen interface.

For me, as a t3i user, this makes a lot of sense.

1 upvote
Ross the Fidller
By Ross the Fidller (Aug 16, 2012)

That's about right. This video shows how slow the Hybrid AF is & since that is the only way the new EOS-M will also focus (using the same hybrid system as the EOS 650D), it's something that needs to be considered before buying the EOS-M. The Panasonic DMC-5 shows how much faster the opposition (to Canon) can focus.

6 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

Since focus speed can be quite lens dependent, it's only directly relevant to the EOS-M if using the same 18-135mm STM on an adapter.

0 upvotes
qwertyasdf
By qwertyasdf (Aug 16, 2012)

Erik,
I think that the STM lenses are specifically made for the hybrid AF, we are probably seeing the best AF performance from canon here.

Nectar,
Yea...actually I was having the EOS-M in mind when commenting, sorry for the confusion

Ross,
Have to give big credit to Panasonic, their CDAF is exceptional, one-shot AF speed is probably in the leagues of 1-series body.

2 upvotes
photonius
By photonius (Aug 16, 2012)

"Does a hybrid mean putting 50% effort into PDAF and another 50% into CDAF?" No.
Standard PHAF works be redirecting part of the light from the mirror in a dSLR to a special array of phase detect sensors. PHAF detects correct distance, tells lens move this far, done (sort of).
With CDAF, you measure contrast on the sensor directly, you move lens a bit, measure, move lens a bit, measure until maximal contrast. Slow, but accurate (and no mirror of course).
With the new hybrid sensor, you make a few pixels on the sensor so that they can detect phase directly on the sensor. So, it should be an improvement over CDAF, since now the camera can tell the lens how much and in what direction to focus, this seems to be followed up with standard CDAF to fine tune focus (at least in the current 650D).
Here are some pictures of the hybrid sensor
http://tinyurl.com/8f2tm9k
http://tinyurl.com/cvxypnd

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

qwertyasdf, simple question: do all lenses focus at the same CDAF speed for u4/3 designs? If not why not? The 18-135mm STM is itself a hybrid lens designed to work with any EOS camera as well as the new systems. While it's faster/smoother than the non-STM 18-135 it's not as highly optimized for fast CDAF as the small Panny.

0 upvotes
VJVIS
By VJVIS (Aug 16, 2012)

I have tested the 14-140 (10X), the regular kit lens 14-42 and the pancake kit lens 14-42, 25mm F 1.4 and the 100-300mm and they are all phenomenally fast and unless you do some specific tests you won't notice a difference in real life situations. I have owned all these lenses at some point. Currently I have the pancake 14-42, 25mm 1.4 and the 100-300mm. I have owned several canon lenses when I had the Canon camera (70-200 F 4), 50 1.8, 100-400 and pentax lenses when I owned the pentax system (to which I moved from Canon) and then to Panasonic. The resolution on the kit lens is better than the kit lenses (except pentax) of other companies, the 25 1.4 is amazing, the 100-300 definitely leaves something to be desired in still photography, but excels in video, but GH2 is better for video resolution than even Mark III at lower ISOs, so there is no comparison there.

0 upvotes
Erik Magnuson
By Erik Magnuson (Aug 16, 2012)

This /is/ a specific test to compare focus speed.

0 upvotes
VJVIS
By VJVIS (Aug 16, 2012)

Doesn't look very specific where they are calculating the actual milli seconds it takes to focus. They are just touching the screen/shutter to see how fast it focusses. I am talking about measuring differences of .1 sec vs .2 sec. Maybe they are doing that as well.

0 upvotes
Sergey Borachev
By Sergey Borachev (Aug 16, 2012)

Hybrid sounds nice and can potentially be great, but not with this T4i and therefore also not for the EOS-M too. It is disappointing to see Canon coming out with one model after another that is not really exciting lately in its affordable camera lines, when others gave us RX100, E-M5, FZ200, ....

Comment edited 14 minutes after posting
3 upvotes
Mattoid
By Mattoid (Aug 16, 2012)

Wait, conventional phase detect doesn't work in live view so you couldn't even record it! Unless you filmed through the viewfinder, which they didn't because they activated focus by touching the screen. Thinking about it, it is phase detect but live view has to be disengaged temporarily. the mirror comes down and focus is acquired. This takes more time than conventional phase detect.

0 upvotes
howardroark
By howardroark (Aug 16, 2012)

This is the new hybrid AF system that has PD AF sensors on the imaging sensor to assist during live view focus.

0 upvotes
Richard Shih
By Richard Shih (Aug 16, 2012)

What you described is exactly what happened. If you look at the first video, when the phase detect camera is pressed, there's a split-second blackout before the image reappears where focus is being achieved through phase detection (and then subsequently the image is taken regularly).

3 upvotes
ageha
By ageha (Aug 16, 2012)

The K-30's LV is much faster than the EOS 650!

Comment edited 18 seconds after posting
2 upvotes
Alizarine
By Alizarine (Aug 16, 2012)

Yes, but the K-30 is not entry-level like the 650D... so the latter will have a hard time against the former, except AF I guess :)

0 upvotes
raimaster
By raimaster (Aug 16, 2012)

nope, entry level is 1100D, both 650 and k30 are in similar level price and target. ONLY K30 has soo many plus (wr body, ovf pentaprism 100% coverage, dual dial, focus peaking, micro focus adjustment, better burst rate both jpeg and raw and ofcoure better high iso and dynamic range.) And 650D swivel LCD and supported by masiv / clever marketing :)

4 upvotes
Richard Butler
By Richard Butler (Aug 16, 2012)

I think we'd consider the K-30 and 650D to be peers. However, that's not relevant to what we're showing here.

Nikon DSLRs also tend to be able to perform CDAF faster than the current generation of Canons (in general - I'm sure there are lens/body combination that suitably advantage one camera and disadvantage the other that you can reverse that if you really try).

However, no conventional DSLR gets close to offering the CDAF focus speed that the best of the current generation of mirrorless cameras, in part because they tend to have lenses that weren't designed for it.

0 upvotes
micahmedia
By micahmedia (Aug 16, 2012)

Thanks for this! Any chance of comparison of continuous tracking performance?

2 upvotes
fmian
By fmian (Aug 16, 2012)

Most that are serious about photography will be better off using the OVF anyway, and most that are serious about video recording will manual focus.
Does not mean this comparison is invalid, but the features tested here are mainly aimed to the general point and shoot consumer.
Just thought I'd get that out at anyone claiming one system is better than the other just by looking at the test above.

0 upvotes
nofumble
By nofumble (Aug 16, 2012)

I use OVF 95% of time, maybe more.

The only times I need focus in Live view are 1. Movie, 2. use a manual lens.

For movie, slow focus is actually a good thing.

For manual lens ... needless to say. Beside Sony NEX, all camera sucks. Canon is less suck because I can have Magic Lantern option.

So fast focus in Live view with the T4i is really a non-issue.

0 upvotes
peevee1
By peevee1 (Aug 16, 2012)

Slow _hunting_ focus is not a good thing in movies. You spend too much time OOF then, especially if your subject moves.

0 upvotes
AngryCorgi
By AngryCorgi (Aug 16, 2012)

I'm much more interested in seeing the EOS-M performing in similar tests to see if they've tuned it more for still shooting (faster, not as smooth) in that application.

1 upvote
tkbslc
By tkbslc (Aug 16, 2012)

If it was made for video, slower is sometimes better. As long as it is smooth and doesn't distract the viewer. But for stills, it's pretty awful performance. I don't get how Canon could release such a product when they knew what they had to compete against. It isn't even in the same class. It's almost like it is just starting when the Panasonic is done.

4 upvotes
peevee1
By peevee1 (Aug 16, 2012)

"I don't get how Canon could release such a product when they knew what they had to compete against. It isn't even in the same class."

Yet most buyers are driven by brand recognition, not by rational testing, so 650D will be sold more than G5 (not to mention that it is so much more expensive that the absolute margins are also higher).

1 upvote
thisisjh
By thisisjh (Aug 16, 2012)

So... the lesson here is... conventional contrast AF including DMC-G5 is better than hybrid AF?

So what is the hybrid AF for?

0 upvotes
Richard Butler
By Richard Butler (Aug 16, 2012)

It shows that a well-designed contrast AF system is faster at single AF acquisition than Canon's current implementation of Hybrid AF.

In principle, it should be possible for Hybrid AF to offer the same speed - but that isn't the case yet. Its strength should be improved performance in movie shooting and continuous AF (not tested here).

8 upvotes
thisisjh
By thisisjh (Aug 16, 2012)

But it looks like it is not even close to the well-designed contrast AF system. Therefore, it may be still worse than non-well-designed contrast AF. Also, why Canon even bother to implement this slow system? Can you turn this hybrid thing off? If you can, is contrast AF alone going to be faster? If it is, why have hybrid in the first place?

3 upvotes
thisisjh
By thisisjh (Aug 16, 2012)

Ok.. I am watching the video again. So, phase detection alone is better. What would be the benefit of using Hybrid system.... For video?

0 upvotes
Richard Butler
By Richard Butler (Aug 16, 2012)

The Hybrid AF system an improvement over Canon's contrast-only system on the 600D, even if you fit the STM lens which should better cope with such focusing, so there wouldn't be any benefit to switching it off, even if you wanted.

Its theoretical advantages would be in continuous AF (for either stills or movies), since it can establish the distance to the subject. But that's something we'll cover in greater depth, soon.

2 upvotes
photo nuts
By photo nuts (Aug 16, 2012)

@thisisjh: Contrast AF alone in Canon DSLRs is definitely slower. You can find that in Butler's reply further down below.

Comment edited 20 seconds after posting
0 upvotes
johnparas11zenfoliodotcom
By johnparas11zenfoliodotcom (Aug 16, 2012)

NOw if only canon/nikon... Just remove the mirrors of their DSLRs, use same mounts, and flange distance something :- ) and use EVF instead.. And have hybrid AF.. Now that would be ideal.. I think the entry level bodies of canon and nikon are small already..

Although i hope nikon will put the AF screw, back on entry level bodies.. So i can use the smaller non afs lenses :-)

Comment edited 1 minute after posting
0 upvotes
Zvonimir Tosic
By Zvonimir Tosic (Aug 16, 2012)

You mean Pentax K-01?
Which is only a first attempt in that direction, thus without an EVF.

1 upvote
johnparas11zenfoliodotcom
By johnparas11zenfoliodotcom (Aug 16, 2012)

Zvonimir.. yeah kinda like that.. but not the body design.. I have seen that Pentax in person.. not looking good nor does it look ergonomic :-)

0 upvotes
snooked123
By snooked123 (Aug 15, 2012)

EOS M's AF will be significantly slower than Panasonic and OMD EM5 which makes me really sad because I was so excited to see a 35mm f2 lens in APS-C mirrorless cameras.

Comment edited 36 seconds after posting
1 upvote
Sabatia
By Sabatia (Aug 16, 2012)

Except at the tele end, the array of lenses for the m4/3s system makes 35 f2 seem far less than exciting or even interesting. Amazing number of quality primes available for m4/3 and now with Pana coming out with a very sharp 12-35(24-70 equiv.) 2.8 IS the fun expands.

1 upvote
Essai
By Essai (Aug 15, 2012)

shame on you Canon. This is unacceptable IMO.

14 upvotes
photo nuts
By photo nuts (Aug 16, 2012)

Yup, you got that right. It is a shame. They just don't seem to know how to handle contrast AF and low ISO sensor dynamic range.

1 upvote
Total comments: 243
12