On sensor phase detection is the future tech. The NEX-FF and mirrorless ILCE camera will rely on that. As you know the Canon 70D has likely the most advanced solution called “Dual Pixel CMOS sensor”. In that case On-sensor phase detect AF is available on every pixel and adopted for initial subject acquisition.
DxOmark (Click here) posted an interesting Canon 70D vs Sony A77 comparison to see if the Canon tech is as good as it sounds on paper. The full test video page can be seen here. These are some of their findings:
Autofocus accuracy on fast appearance
– In Bright light conditions the Sony AF system is faster
– In Low Light conditions the Sony SLT-A77 performed nearly as well as the Canon, despite only 30% of the light entering the mirror-box being used for AF in the Sony.
Autofocus accuracy for tracking
– In Bright light conditions the Canon EOS 70D is slightly slower as we’ve seen before, while the Sony SLT A77 reacts quickly but displays lens breathing (a change of angle of view during shifting of focus).
– In Low Light conditions the Canon EOS 70D is clearly the better at focusing of the two on test.
DxO conclusion is:
For video enthusiasts, this camera is one of the best choices available on the market currently and at least is a viable option to the entirely likeable and very capable Sony SLT models.
Overall the A77 proves to be a better performer in normal shooting conditions and shows it’s limits only when it gets…dark. Note that the Canon 70D is a brand new camera that also costs $300 more than the two year old A77. It will be interesting to see how the soon to be announced NEX-FF and ILCE7/9 cameras will perform with their new on sensor phase detection tech.
88 Responses
at-least colors look like colors on the a77
Colors Yes, BUT
FIRST – Thank You DXO for providing
great Independent Comparative Analysis
Side by side is undeniable.
Wow, what a difference
SLT Video autofoucs technology looks terrible
This AutoFocus? Is that the best SONY can do for us on SLR products.
I wonder of the A99 autofocus would show as POORLY
We will wait for the new Sony product shipment
We will wait for the 1.1 Firmware –
We will wait for all the
Quality Third party independent testing and Comparisions
for new SONY product offering
We’ll have to wait and see.
Did you get a chance to look at all the tests/video on DxO?
I bought the Sony a-99 specifically for its “revolutionary” AF for a project that could benefitted from it enormously. It wound up in exactly zero finished shots–it kept searching for the wrong subject and when it did find a subject it would jerk into place to make the shot unusable then go out of focus for any and no reason whatsover–if it ever found focus on a given shot to begin with. If it’s not blindingly beach bright it is absolutely useless. You can forget about filming in a well lighted daylight interior. Oh, and the facial tracking feature? The box around the face shows up in all RAW video feeds so if even if the AF worked (and it doesn’t) you still can’t use it with a Ninja. Now the Canon 70d can’t output raw at all but at least it has semi-decent compression, unlike Sony’s dreaded, virtually uneditable in-camera compression. In short, if you need AF for a video project that isn’t on the surface of the sun never, EVER use a Sony camera.
So does this mean that without the use of SLT in the new upcoming SONY cameras we should find more improvement in low light?
Not necessarily, as it will be a different implementation.
Good point. Well i cant see it getting any worst. There must be improvements with every new implementation.
Agreed.
So, the A77 is an excellent camera. The 70D also. And the successor to the A77 will probably pull ahead again.
Well, this test quite clearly shows that A77 successor without full-size PDAF sensors will have a huge issues with focusing – and even more so: during low-light situations.
The test only clearly shows Canon’s implementation quality. It’s not necessarily indicative of the overall state of this technology.
For Sony to switch to that technology means for me that it is at least as good as what they have now. Otherwise they’ll have a major screw-up on their hands.
It’s true that it’s only an implementation sample, but at the same time – it’s one of the most advanced OSPDAF technologies we’ve seen up to date, seems significantly ahead of what OSPDAF Sony does offer right now.
“For Sony to switch to that technology means for me that it is at least as good as what they have now” – That’s not true, and you should know it. Sony already been doing “newer replaces better” kind of progress in past. So I wouldn’t be surprised. Besides – all of the manufacturers could be found doing the same thing one time or another. Honestly I think that for Sony advantages of OSPDAF are not in speed or accuracy, but rather prices and sensor output image quality (no more SLT mirror on the way – though in this tests it looks like Canon is the one having half-mirror in front of the sensor, lol).
Well, I just voiced an opinion about the future, so I am not sure if it can be “not true” right now.
Yes, it is possible that Sony screws up, but this would be too monumental (AF is an important feature after all), hence my opinion that they’ve thought that through.
Otherwise, yes, Canon seem to have done a good job, which is something I certainly can’t say about DXO 🙂
Well, Sony already screwed big time with SLTs at least twice – with A99 and A77, both times an important aspect – high-ISO image quality – came out to be a fail comparing to the direct competitors.
So failing AF wouldn’t be any surprise IMHO. Especially when right now NOONE makes any reliable AF tests (the one DXO made here is the best one I seen in 4 years or so) – so even if they would fail – there’s hardly any way to prove it or show to the customers.
Hmm, good points. Well, not much left to say here, let’s wait and see what they bring and then we can discuss the issue properly.
In did. And for the sake of A-mount: I hope they won’t screw it.
Sony also lacks lenses with stepper motors for the A-Mount, they won’t launch the bodies until they also have the lenses.
Agreed that they shouldn’t switch if the new system can’t match the old system…
But that doesn’t mean that will be the case.
They may very well be willing to partially sacrifice autofocus performance and take a step backwards in order to gain the other benefits of going completely mirrorless… Including the possibility of reduced production costs (by bringing the A-mount in line with the E-mount), and better low light performance (by not sacrificing 30% of the light).
In the eyes of many, the SLTs made sacrifices compared to dSLRs — including worse high ISO performance.
I can imagine someone 4 years ago saying, “Sony will only switch to translucent mirrors if the high ISO performance can match traditional dSLRs” — Yet we know that Sony sacrificed high ISO to make dSLTs work.
And they might be sacrificing the best possible AF performance to get mirrorless A-mount to work.
I suspect that the new A-mount will have very good autofocus “for a mirrorless camera.” But since we have *never* seen any mirrorless camera AF truly match a dSLR…. I’m not betting that this will be the first.
We should know pretty soon — Sony’s mirrorless AF answer should be implemented in the NEX FF and the A7/A9.
Well, this video clearly shows that holding a white paper infront of any lens gives nothing to focus on. Furthermore it shows, that the 2 years old A77 has a much better white balancing. What the video does not show is that in very low light conditions you won’t see anything with your OVF. With an EVF you’ll see – a little bit noisy – something where you can focus on.
When I compared the viewfinders of SLT A37 and EOS 1100, I came to the opposite conclusion:
First things I’ve learned was that in low light, both viewfinders were much darker than the room actually was. Between “both are bright enough” and “both show nothing useful”, there was a margin where I could see only noise in the A37’s EVF, but still something with the EOS. About as dark, but at least not completely covered by the noise.
But to be honest, I own the A37, know the behaviour of the EVF at low light and was a bit disappointed about the OVF. I expected it to cope much better in these conditions.
Sorry, I forgot to mention:
Both cameras I compared had the 18-55mm kit lenses. A faster lens helps both viewfinders. I checked this with short and long end of the zoom and the 35mm/1,8 on the A37.
The key here is the Canon can AF @ F/1.4 through F/5.6 in video mode, whereas the Sony is limited to F/3.5 or max aperture of the lens.
Nothing like buying an F/2 lens like the 24/2 and the video AF stuck @ F/3.5.
The Canon’s not without issue though; it is smooth for slow moving objects but faster ones it has trouble with.
I hate AF in video mode anyway – would rather watch something slightly out of focus than seeing a camera continually AF all the time. Too distracting.
when the paper is brought nearer to the camera, A77 got no words to focus to, just the white paper.
Doraemon, you are here! Give us a Sony camera from the future!
use your dokodemo door to peek into Sony’s project Lab
They were using the same sheet of paper, a full size sheet for both cameras. The split screen was created from the two seperate test of the the two cameras. So each camera did in fact have the same sheets to focus on. It is only because of where they edited the split screen that it looks like Sony was just doing a blank sheet. I admit the test at first glance looks like the Sony was put at a disadvantage, but that was just the impression from editing, not reality.
This test is near useless, the “Focus Target” if you can call it that, is all over the place.
very sloppy!
“- In Bright light conditions the Canon EOS 70D is slightly slower as we’ve seen before, while the Sony SLT A77 reacts quickly but displays lens breathing (a change of angle of view during shifting of focus).”
Focus breathing is a characteristic/condition of the lens, not the camera. Unless the 70D has some active cropping mode to compensate for breathing, I don’t see any reason to point out the appearance of breathing on the A77
The Canon was using a prime lens and the Sony a zoom, so the difference in breathing is not unexpected. I suspect with a prime lens that Sony would be as good from that standpoint. Obviously both would be better with true cinema lenses designed to eliminated breathing.
I love how Admin picked the ONLY video where A77 somewhat fails. In all of the other it outperforms on-sensor PDAF BY A MILE.
Some sample how it performs with most basic focusing in low-light: https://www.youtube.com/watch?v=aLykgfjzyrI
Striking difference!
And regarding that low light comparison:
“When tested under lowlight conditions (20 lux, roughly the equivalent to shooting at dusk), the Sony SLT-A77 performed nearly as well as the Canon, despite only 30% of the light entering the mirror-box being used for AF in the Sony.”
Are they serious? The A77 smokes the 70D in that clip. Am I missing something?
Yea, comments are completely off comparing to the videos. People complain about that all over the Internet.
“”Are they serious? The A77 smokes the 70D in that clip. Am I missing something?””
I think you are missing that they are talking about AF speed not sensor performance .Though thee is no doubt on a similar gen sensor the A77 is disadvantaged by the light wasting SLT. The result of that is that the high ISO is actually a little lower than the 70d
http://www.dxomark.com/index.php/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/%28appareil1%29/895|0/%28brand%29/Canon/%28appareil2%29/865|0/%28brand2%29/Nikon/%28appareil3%29/734|0/%28brand3%29/Sony
“”Are they serious? The A77 smokes the 70D in that clip. Am I missing something?””
I think you are missing that they are talking about AF speed not sensor performance .Though there is no doubt on a similar gen sensor the A77 is disadvantaged by the light wasting SLT. The result of that is that the high ISO is actually a little lower than the 70d
http://www.dxomark.com/index.php/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/%28appareil1%29/895|0/%28brand%29/Canon/%28appareil2%29/865|0/%28brand2%29/Nikon/%28appareil3%29/734|0/%28brand3%29/Sony
Ok, I just went through the whole “review”. What a joke, seriously. And these guys are considered to be professionals.
If it is a joke, what should we conclude about the guy who sees that it shows Canon “good work”, “implementation quality” … LOL!
Jean, it is very simple. I am trusting the videos (hence Canon’s “good work”), but I am disagreeing with DXOs commentary (hence the “joke”). Is that explanation simple enough for you?
Yeah, I watched all the videos. I actually walked away impressed by the A77 and confounded by the DxO assessment. The Canon 70D is amazing. Really amazing. However, it was really only better (marginally) in the tracking tests. I really think DxO got the two cameras mixed up in the first low-light test. Switch the camera names and their assessment is spot on. The Canon performed *almost* as well as the A77–a big achievement for OSPDAF!
Weird. Oh, well. If the Sony implementation is at least this good I’ll be happy. Also I hope the PDAF points are spread out like the Canon but I suspect we’ll see a cluster like the M1
Also the Canon is significantly under exposed. Is that just because the 40mm has a worse transmission rate? Would have liked that addressed.
Canon got issues with adjusting exposition in video. Transmission rate of the lens doesn’t have anything to deal with that.
Odd that this hasn’t made it to the Canon Rumors site….
Just think how good the A77 would be with the promised but never seen new firmware.
Come on Sony, what are you waiting for, don’t leave us out in the cold, show the world that you will support your existing users. Give us that Jpeg engine we have all been waiting for and show us that you CAN do it.
Firware promised by whom ??
*Firmware
JPG engine AND a new tracking AF firmware – we knew it was bad, and this test only confirms that.
The first one year I was shooting jpeg and could never understand why people were always complaining about the quality of the jpegs. The files were big, sharp, rich in colour and responded well to Lightroom improvements. Then I read and read and read and educated myself and decided to give RAW a try. Now I can never look back.
It really comes down to this – if you are serious about your photography, you must shoot RAW. In any other case you are fooling yourself. And if you own the a77, the word “jpeg” should simply be outside of your vocabulary.
At this point I couldn’t care less about the jpeg engine. It’s just irrelevant and even if the camera had no ability to save jpeg files at all, it would be all the same to me. If I were using a point-and-shoot I’d understand the importance of jpegs, but for the a77 – c’mon, if you want better jpegs you are doing it wrong, or using the wrong equipment.
Exactly. I only use JPeg for the most casual of shots.
+1 even a DXO auto adjust is a night and day improvement (and adds seconds to your workflow).
Amen Ken!
I cannot quite see the point of that comparison, even if it was knowledgably done (what I don’t know as I’m not really interested in going into detail there). The A77 is one of the last cameras of its kind and, to my knowledge, Sony’s phase detection AF, while possibly quite decent, wasn’t the most advanced to begin with. The Olympus E-M1, on the other hand, already implements the first generation of what will probably be the future of Sony’s AF throughout both A-Mount and E-Mount systems. And that’s the one AF we need to know more about, how good is it and how good can it become, and not only in comparison to Canon’s new on-sensor AF technology, but to the best-of-breed conventional DSLR PDAF.
Right. Because we should totally spend our time comparing shipping products with theoretical products that haven’t been announced, much less delivered to customers.
Remember this set of tests compares the A77 in the ‘normal (mirror down) mode’ against the 70D in live view (mirror up) mode. For photos (not video) the more realistic comparison should be both cams in mirror down mode.
TERRY
Not if you want to see how the latest breed of OSPDAF cameras performs compared to Sony’s SLT.
So far, the experience with OSPDAF in APS-C cameras has been so-so. Since Sony have leaked that they will abandon the mirror and instead use OSPDAF in the new A-mount cameras, a lot of us are really interested in seeing how well OSPDAF behaves in new cameras, compared to SLT. And this test gives us exactly this type of information.
This test is flawed! One: the canon side has a lot more text for the auto focus to lock onto. Two: when the tester moves the paper back and forth the most of the time the A77 does not have anything to focus on either it’s just white paper or the paper is not even in the focusing area. Three how are the focusing points setup ?
I really hope you’re joking. Both cameras are shooting the exact same thing. You’re just seeing the left half of the Canon’s shot and the right half of the Sony’s. They didn’t set up a split-screen in the cameras and then record that.
While it is a split screen, the person’s hand and most of the paper is generally left of the center. And the Canon was set up physically to the left of the Sony, and thus probably had more to focus on most of the time.
That being said, I wonder why he picked the one video where the Sony doesn’t do as well. The A77 clearly beat the 70D in 3 of the 4 videos (even though the DxO writeup said the Sony did “almost” as well in the first low light test, when it actually did much better than the Canon).
Watching just the video embedded here, you’d think the A77 got smoked, but in reality it won most of the comparisons.
Despite looking not-so-professional, this just seems like a reason to regret the departure of SLT technology. Despite the light-loss, it looks like it still would have something to say concerning autofocus and I honestly don’t believe Sony can come up with drastically a better on-sensor autofocus system…
As far as I know there is nothing on a dedicated PDAF sensor that cannot be put into a imaging sensor and EVFs are getting better all the time. My view is not should/will mirror systems disapear from most* cameras but when. Image sensor PDAF is inevatble as are electronic shutters …
* there is likely to be a continuing market for traditional SLR as OVFs do have some advantages but this will srink to a specialist area of the market.
Yes it’s not that I don’t agree that everything nowadays point toward the complete disappearance of the mirror (fixed or not). It’s just that, given the CURRENT performances of the on-sensor VS dedicated PDAF, I guess the mirror still have a point.
And this doesn’t mean it won’t be ditched sooner or lather… I simply believe this is premature. I’ll be glad if Sony disproves me with the first mirrorless Alpha (or even the apparentely soon to come Nex-FF but, as of now, I think chances are it won’t happen so soon.
+1
CanonRumors is saying that the next EOS-M will also come with phase-detect.
It will be an interesting race to watch. I’ve complained about NEX’s lens range, but it’s far ahead of EOS-M’s. NEX-FF will be starting at the same line, but with some nice pieces of glass. Fun awaits!
remember, the eos m with phase detect will take almost all Canon lenses. just imagine the benefits of that. all the little primes that they have. I think the upcoming eos m could be a nex killer. which is why Sony is pushing ff I guess.
imho, with more and more PDAF/CDAF on Sensor cameras shipping. The Auto Focus Test and Compare will get better and more useful. Just as LENS test/reviews have become.
This is SONY open door to LEAD. On LOW Light AF and CAF.
The Advantage “Should be” with the new SONY 70-200mm G2 lens (A-mount). SONY should have “Tuned” it to the new Sensor. and will have the advantage on the first to ship. It will be a bit of “Proof of Tech”. also needs to be Sharp at 200mm @ f2.8.
Just hope SONY also updated the 24-70 to Weather Sealed and sharper at f2.8 on High Res 32mp +. So it can be class Leading as well.
So, it will be interesting.
With E-Mount shipping first (size and weight reduction for Full Frame). My guess is the “Need” for Speed , High Res (36mp+), PRO stuff is on the A-Mount. Just a guess.
This tracking clip is a ridiculous test. The hand which is holding the paper is always on the Canon side of the screen, which must be big advantage for the Canon sensor/focusing to be able to follow the movement. How about trying it from the another side also?
no, both cameras must have been seeing all of the scene during the test. they were just edited together like that afterwards. think about it.
Do not forget, Sony autofocus works just only f 3.5 . This is the biggest pain .
This statement is not accurate. Autofocus works at the lowest aperture possible for the lens, down to 3.5. For example, if you use the 18-135 lens and you are zoomed at 135mm, the aperture can not be opened further than 5.6, but despite that autofocus works perfectly. If you use the 16-50 however, autofocus will always work with f/3.5 despite the fact that the lens can go down to f/2.8.
Yes you are right , but mostly lens is not sharp straight from f5.6 ,and if you wish use auto focus you stock with f5.6 , only change to manual mode let you go over f5.6 .
Sometimes you have low-light situation and you wish go with f1.8 and you can not use auto focus if you turn on you jump to f3.5 . Stupid . For me alpha system is the worst for video . Video quality is terrible . Of course for someone who can live without auto focus alpha offer something what others do not have it . I have fs100 and I thought a99 gonna be nice as B camera . I was so disappointed when I saw sample .
The big usability advantage that the 70D has is the touchscreen. Being able to rack focus with a touch to the back is pretty slick.
Guess I know which camera to buy if I want to film a piece of paper moving back and forth!!
Question: (disregarding Sony’s horrible performance in the tracking mode…)
In the first two tests on the DxO webpage, Canon is pretty slow in nailing down the focus, even in bright conditions. Is this the limit of what their on-sensor technology can do, or are they deliberately smoothing out the focus path because it’s video? I.e., would it be the same in the single-shot mode?
If so, the technology is just not ready to take over PDAF for still photography.
The 70d shows it’s superior video AF abilities by not completely blurring before catching focus while tracking. This may in part be due to the small number of AF points on the a77.
As a A-77 and NEX 5N owner I cannot really tell any differences in noise level, however the NEX 5N seems to auto focus better in very low light.
Fail — No reliable conclusion available! The test lacked necessary controls.
The left frame focused a full human body holding a piece of paper as it walked, which took up nearly the entire frame. The right frame focused only the top sliver of a piece of paper, which took up substantially less of the frame. I’d expect the left and right frames to focus more or less the same as in this video even if the cameras were swapped onto the opposite tripods.
A more accurate way to test AF of these cameras would be to tape the paper to the guy’s shirt and film him walking back and forth twice – both times from the same unmoved tripod. The first time would be with camera #1 on that fixed tripod, and the second time would be with camera #2 switched out for camera #1. Also, all lights should be the same (e.g. Film both reels within a minute of one another or else use blackout curtains on any windows).
The same test would also need to be repeated in various colors, or temperatures, of light (sun, florescent, halogen, etc.). The sensors’ color perception may well affect AF performance, and this should be accounted for.
There may be other issues, but those pointed out here are alone sufficient to suggest that nothing can be drawn from this test other than the fact that this particular video recorded as it played.
Poor testing is a common issue with DxO Mark. I don’t know why they tend to be so unscientific while also lacking transparency in most of their data. A much more reliable and simpler thing to do would be just go to the a quality camera store and play with both cameras so that you can negotiate a good deal on your favorite pick.
to me the a77 is the clear winner. (and I don’t own it, so I’m not defending my gear here.) under most conditions it’s way ahead of the Canon. the not so smooth tracking isn’t a surprise considering the fewer focus points and the fact that the sheet of paper that they are using is partly all white.
This test doesn’t prove almost anything (except that the A77 is faster focusing).
If you check the setup they have the two cameras side by side on a tripod, 70D on the left A77 on the right. When doing the test the person moving the paper is at the left of the cameras (at the side of the 70D) this give the 70D more “moving subject” than to the A77… just saying…
no, both cameras must have been seeing all of the scene during the test. they were just edited together like that afterwards. think about it.
The point here is that canon made a good job developing this sensor. If their on-sensor solution is about as quick as the off-sensor af of the slt now, how long will it be before it’s on par with canons off-sensor af?
No, I didn’t watch the clips on my phone.
With this sensor technology canon seems well prepared for when the time comes to go mirrorless+evf.
What a crappy unprofessional test! The power of OSPDAF is much better illustrated in the following Canon 70D video review by Tony Northrup: http://www.youtube.com/watch?v=NjR2OztuXtA
What a biased and unscientifical test. The lens aren’t comparable in any possible way…
Well, Sony fans should feel better because the latest tech from Canon is just slightly better(if any) than 2 years old Sony. But to be fair, SLT is the main and the only focusing module for A77 while dual pixel focusing is just secondary for Canon 70D. Progress is good, whether it comes from Canon or Sony.
@just me
+1
I agree, corporate competition is good for everyone. All photographers benefit from it.
The only thing this poor DxO test shows is that they (DxO) have no methodology ti provide such test. ROTFL of the year!
It reminds me their raw software that cannot properly resolve colours.
The test is only for the video autofocus. Which is really different from the one in photo mode.
Because of the following:
– video autofocus must have about 0.5 – 1 sec delay before reaction. This is required in order not to always immideately activate the focus shift to the random pedestrian, who crossed the line.
– video autofocus must work slow enough, in order to avoid rapid focus switch, which looks unpleasant in video.
Last, the test was done simply in the inaccurate way. So difficult to compare.
Yeah but now imagine how the EOS M2 with the 70D sensor will destroy every other mirrorless system in AF performance.
Hmmm – I saw only that EOS was fast in close ups – and was incredible slow at middle. So my opinion test is bullshit – just ONE moment – STM40 MDF – 0,3 – SAL1655 2.8 – 1,0
This are different lenses with different parameteres – and Sony loose in 0,5 – 1 m distance. Is it funny – or not? If looks to lens spec.
For me – dxolabs failed at all. They specially made not two videos side by side, but made montaged version to hide Canon AF errors.