Trusted sources told us that the NEX-5R has a new 16 Megapixel sensor with phase detection AF pixels on sensor! This is pretty much the same technology used by the new Canon EOS-M mirrorless camera (click here to see that cam) that also uses those kind of pixels. It allows a super fast autofocusing and more reliable tracking of moving objects while the contrast autofocus is more accurate. Combine that and you have an almost perfect autofocus!
It’s not confirmed yet but the NEX-6 could use exactly the same sensor!
P.S.: Last year we spotted a Sony patent describing that technology (Click here to read it).
115 Responses
Awesome
1st
Hello eBay…here is my dated NEX-7 for sale…
Let’s not get hasty, if it’s anything like the current on-sensor PDAF sensors, they don’t work in low light–i.e. it reverts back to CDAF.
That said, this is HUGE news for daylight shooting. I’m highly looking forward to this new development.
http://matthewdurrphotography.com
Even that might be a tad early to conclude. 😉
Because as I see it, for the Canon M series it offers no speed benefits over some of the existing CDAF systems (focus is even slower than fulltime CDAF from the latest NEX bodies).
The Nikon 1 series show there is quite a bit of potential to the technology, at least in good light though. So if true, we’ll have to wait and see which of those examples Sony will follow…
Robert Cicala at Lens Rentals speaks about the new sensor technology as well as the steps needed to make phase detection as accurate as CDAF and recalls a Canon patent that described that for phase detection to be accurate, it must have a lens that has a stepping motor in place, so the body knows how far ahead or behind the focus area the lens is at and instructs the lens to accurately point focus.
I think this example was in an article regarding the new 40mm STM lenses. Worth noting as well is that he says that the new STM technology on the bodies and lenses is not backwards compatible with previous bodies (so, a 5D mark II cannot benefit from the STM advantage)
That being said, could this mean the new lenses launched could have STM to improve on PDAF combi with CDAF?
From what I recall, NEX lenses already have stepper motors (as most CDAF dedicated system lenses do).
I strongly suspect Nikon 1 AF speed is due to large motors and light lens elements (due to small sensor and slow lenses), not PDAF
It’s always a combination if factors and I’m sure processing power and read out speed play a role too, but all of that isn’t going to give any CDAF system good follow focus capabilities, so I’m confident the PDAF helps there.
Just look at tyre E-M5. Crazy fast focus but still horrible follow focus.
Correct! Not only phase detect reverts to contrast method on Nikon 1 series, but there is no way in menu to switch it or change it or control it in any way. What would be perfect if a user could selected whether he/she wanted to use contrast of Phase Detect method of focusing at any particular time. Not possible at present time
Olympus, the very first company to offer Live View on a DSLR, introduced the concept of switchable phase and contrast detection years and years ago. I don’t know the implementation on the Sony A’s or Nikon 1’s, but the idea is hardly new.
Maybe this new sensor has been the hold up on releasing high quality lenses for the NEX system?
now you can stack the la-ea2 and get even better autofocus!!!
PDAF²???
yes, this would be great news.. i’m glad i still only have been using my good old nex-5 and only bought the zeiss lens as an upgrade so far. next would be a good telezoom lens and a new body like this one 🙂
Cool… if it is much better than canon implementation… It could be a problem for EOS-M…
I will be surprise if NEX-6 don’t have it since it is a higher model : )
Can’t wait…
Now we need an A-Mount camera with this… Got so much A-Mount Lenses…
So far no on-sensor PDAF worked good enough. It’s either hybrid or completely CDAF if only light goes down a little bit from “optimal” conditions.
I wouldn’t be too excited for on-sensor PDAF Alpha. Unless you really need crippled AF for some odd reason.
Regarding on-sensor PDAF, the Nikon 1 hybrid AF system is what Sony should target on!
hopefully Sony is able to match Nikon’s AF speed.
and not Canon’s ultra slow – slower than MF sytem : )
Not really. Nikon 1 speed was already outmatched by some good setups of Panasonic m4/3.
Besides – Nikon 1 with it lenses got DOF of a Point & Shoot camera, so it’s not relevant for any camera requiring spot-on focus with actual narrow depth of field on the pictures. If NEX would have as tiny sensor as Nikon 1 it surely would have blazing-fast AF as well 🙂 RX100 is a decent proof of that.
There’s more to it than just speed. Follow focus is where the 1 series blows any CDAF system to date out of the water in good light, for the simple reason that PDAF has inherent benefits here. The smaller sensor (DOF effects) only helps but that doesn’t fully explain the difference based on what I’ve seen.
This is also where the potential is for a suggested hybrid system used by future NEX cameras. In good light it could greatly help follow focus without the need for the SLT adapter plus A-mount lenses. No doubt it will be greatly compromised, just like all the existing hybrid systems. But here’s hoping it will atleast *add* capabilities in certain conditions.
What you say doesn’t match with other reviews. Nikon 1 is mostly a CDAF system that isn’t as fast as new-firmware NEX5n or NEX7. The PDAF is HIGHLY specific to locations in the image where several sensor-sites are replaced with (large) PDAF sites. [No, you cannot have a site does both PDAF and Sensor-feedback, yet]
Strangely, although Nikon documents the number of PDAF sites for most cameras, there are only widely scattered claims about the number and size of PDAF sites on Nikon 1. The sites need to be large to work-quickly for PDAF, but that displaces more sensor-cells.
Actually, what you’re saying doesn’t reflect the 1 series reviews at all. It’s a very fast PDAF system in good light with rather good accuracy. In less than ideal light it switches to CDAF, a not so seamless switch in itself and not so hot CDAF results either.
But it proves than in set conditions the system has potential, especially with follow focus.
Dpreview wrote:
“The J1 and V1 feature an all-new adaptive AF system that Nikon is calling ‘Hybrid’ AF. Hybrid AF incorporates both contrast-detection and phase-detection focus technology – the latter being used when the cameras detect favorable lighting conditions. Assuming that the light level is high enough, then the J1 and V1 switch to a 73-point focal-plane phase-detection AF system that offers fast and responsive ‘single-shot’ AF and impressive continuous AF tracking performance. Phase-detection AF is a particular gift for the ‘soccer mom’ and indeed anyone who finds themselves regularly needing to capture moving subjects outdoors, albeit one of the many aspects of the cameras’ feature set which cannot be manually controlled. This sequence of images shows a 10-frame burst at 10fps, shot on the V1 in AF-C mode using the 1 NIKKOR VR PD-Zoom 10-100mm f/4.5-5.6. As you can see, the camera’s hit rate is impressive.”
And the sequence showing a racing bike coming towards the camera is indeed impressive, something every CDAF system would seriously struggle with.
And they continue:
*The intended audience of the J1 and V1 are unlikely to take them to any major sporting events, but this doesn’t mean that their innovative hybrid AF system is pointless.–> **The J1 and V1 are the first mirrorless cameras that we’d be truly be confident about using to capture moving subjects, and potentially this means anything from boisterous children and pets to school sports days and weekend soccer games**<–. The naturally deeper depth of field provided by their 1-inch sensor format and correspondingly short focal length lenses helps of course, but even taking this into account, we're genuinely impressed, especially by tracking accuracy with phase-detection AF."
Pretty clear description.
“Strangely, although Nikon documents the number of PDAF sites for most cameras, there are only widely scattered claims about the number and size of PDAF sites on Nikon 1.”
– that’s because there are no sites on it. Nikon uses lines of PDAF pixels – not “points”. You may have a line of PDAF pixels in the middle of the screen from it’s right border to center – that’s roughly how it’s implemented.
“Actually, what you’re saying doesn’t reflect the 1 series reviews at all. It’s a very fast PDAF system in good light with rather good accuracy.”
– Spoon, at the date of release Nikon 1 had the fastest AF out there. It’s not the case anymore since new CDAF mirrorless where released. And what Beachrider told exactly matches recent reviews I read (none of them in English though, sorry).
As you said yourself – PDAF in mirrorless basically helps with objects tracking, though it hardly have anything to speed or whatever else. And it still doesn’t work in dim light, still doesn’t work well with very narrow depth of field like it happens to be in DSLRs, and still is way behind regular PDAF or even recent iteration of CDAF.
Look at EOS-M or C650D – they got one of the worst movie AF out there despite of using on-sensor PDAF and DEDICATED lenses made especially for this system – magic? Nope, it’s just that on-sensor PDAF currently, and very likely for quite some time isn’t anything to be excited for.
Oh, and as for dPreview review – half of it’s bit dedicated to tracking AF is a joke. Look at these samples – guy riding a bicycle on a shots where depth of field is all the way from 3 meters in front of photographer back to the sea 50 meters behind him – what the heck is it suppose to prove?! You don’t need AF at all to make sequence like that – everything would be in depth of field anyway! (fun fact here – the closer the guy gets to a camera, the less sharp he becomes, lol, I’m not sure if it wasn’t shot with MF in a first place)
Even if Olympus surpassed it in speed (I’ll believe Sony did with the update when someone points me to a test proving it), there it’s no denying the focus on the 1 seriesis crazy fast and helped Nikon a lot with speed, since they haven’t figured out a fast CDAF system.
Imaging Resource measurements confirm the Nikon claim that it’s at the speed level of a top end DSLR, which sofar no mirrorless had been able to replicate in their test, not even tyre Olympus.
No idea what tests Imaging Resource exactly made but I tested it myself and Panasonic GF5 with decent lens got better AF then Nikon 1 – both: indoors and in bright sunlight (although the difference isn’t big). Plus, as I said – other tests confirm the observations.
And You’ll hardly find any English tests now, as everyone who were interested in Nikon 1 already made their tests soon after the release – while all the other people never got much interest in mirrorless with P&S sensor anyway. Guess only press from ”2nd grade world” actually still didn’t got Nikon 1 for testing.
And finally, again, half of these tests are worth nothing due to testers forgetting about inability of Nikon 1 to take pictures with shallow depth of field which would actually prove AF tracking ability in any reasonable way. So instead we got samples showing nice tracking where object moves inside of 50 meters depth of field area. lol
Did you take metering out of the equation in your test? Because the 1 series is notoriously slow at evaluative metering, which in the real world means that many hockey moms won’t fully utilize the speed.
And which lenses were used? Did you measure times? And how?
The whole DOF argument only takes you this far. Just look at the very fast focusing Panasonic compacts these days. Pinhead sensors, much deeper DOF still, yet can’t follow focus worth mentioning.
I stand by what I said. Until someone proves otherwise, I’ll stick with the tests available, stating and measuring the 1 series PDAF to be crazy fast in good light, even compared to current higher end DSLR standards.
it cud be better than my beloved nex 7 than……. 😐
Wow )
Time to sell 5n ))
Wondering if now the LA-EA1 will work faster with this new sensor… Hopefully the NEX-6 uses also the same sensor…and it will be a NEX-7 killer for sure =)
LA-EA1 definitely got faster with newest NEX5n firmware, so this has strong potential. Everything but lens-correction (CA/Vign/Distort) seems OK with Alpha lenses on LA-EA1.
Good news. It is only a matter of time until we see this for the A-mount. Sony will not develop sensors for NEX/Alpha separately but share them like they have done in the past. If the tech is available, there should be some use for it on the A-mount.
There is no use of this on the A-mount line up. Unless they’re ready to get rid of the mirror, I don’t see this implemented on A-mount models.
Still won’t be as fast as a proper dedicated AF sensor, but might make AF usable with LA-EA1 which is considerably smaller than LA-EA2 as well as cheaper. Not to mention no light loss This would probably be enough to convince me to get a NEX again.
The only problem is that lenses with screw drive will not AF.
That looks like a Stargate Portal to another dimension. Or at least some wicked voodoo. Be careful where you point that thing.
If some of the pixels are used for phase detection, then there is light loss, and loss of resolution. If the phase detection pixels somehow double as imaging pixels, then maybe not…
The lost pixels can be interpolated, so you wouldn’t even see the difference.
So it IS voodoo… I knew it.
Interpolation is NOT the same as resolution. No vendor is selling its ability to interpolate for displaced sensor-cells.
He didn’t claim it’s the same, he said you probably won’t notice it. Just like how you won’t notice the hot pixels that were cloned out once a month on your Sony camera.
Yes, true, demosaic algorithms should account phase-sensitive pixels too.
The days of the mirror are counted …
Actually it’s a huge mistake. When the memo came down to kill the 500/8 Mirror lens, it accidentally went to the camera department. “No More Mirrors” was taken too literally.
lol 😀 😀 😀
In the end, slapping mirrors will die sooner than mirror lenses.
Including DSLTs?
This will resolve the 1 major fault with the NEX System, cant wait!
Only other secondary thing would be to slow down the fps so it doesnt hit the memory buffer so fast, otherwise NEX is perfect.
Yes, agree, an option to set to 6 fps would be nice. That shutter is old before I even took one awesome picture. Besides, deleting pictures is slower than to be expected at this framerate.
Yep exactly, its actually too fast, you get 2 of the same shot quite often and fill up your memory card, 6fps +/- would be about right, and a memory buffer that can last much longer.
NEX hack on the way… hopefully.
Hope its better than Canon EOS-M T4i implementation other wise just another useless spec.
From what I understand of the rumored A99 hybrid AF system it sounds much better.
So….will that support phase detected AF on a-mount lens on LA-EA1?
If it were so, it should be a star compare with LA-EA2!! (LA-EA2 is only for those 85, 135, 1680Za without motor)
Not exactly. Full-sized PDAF sensor still have a huge advantage over tiny one – especially in low light. It should make LA-EA1 noticeably better but still nowhere near LA-EA2.
Also it should be a key step for our A/E hybrid camera…
Good if the upcoming alpha camera can remove the translucent mirror
Yeah baby Yeah!
Now I can forget the alpha system and concentrate to the NEX – if this is true!
It’s ok if the sensor keeps its microlens technology. We really need to have it to use our M-mount lenses !
Hope this will be better for M-mount lenses than current 16mp used in 5n. Don’t care about AF anyway
Hahaha…no more translucent mirror! Thank God!
Wait…..what about the A95/99? Will that camera be stuck with a mirror in its sensor path for years to come? Will the A95/99 be Sony’s first full frame and last full frame SLT?
This is hilarious!!
Why is it hilarious? There is a possibility that Sony will migrate all A mount cameras to mirrorless designs faster than its competitors (assuming you can achieve high AF accuracy and AF speed in sensor). That will be a big reason to buy the A mount cameras – after all, there is a need for bigger bodies. Looks like Sony is on track to becomming a significant player in many aspects of the imaging field (pro/amateur picture/video, etc.). I guess when they finally put their mind to it and started adapting a strategy of some sort, things are beginning to mesh as well. Let’s hope that bigger and better things come out of Sony in the future. Competition and choice is good.
.
I suspect that Sony is basically working off the plan they developed six years ago, or at least five years ago. For two, maybe three, years I have thought that the A100, A700, & A900 were more in the nature of shots across the Canikon bow, not the course Sony’s cameras would take. (The 2009 cameras were treading water in economic bad times.) Now we are beginning to see the full extent of Sony;s strategy. On all fronts.
“Hahaha…no more translucent mirror! Thank God!” – nope, not yet.
“This is hilarious!!” – 1. no idea why would it 2. it’s not.
I strongly suspect that when we all look back at Sony’s SLT concept, we will all laugh at how short lived it was. Thats all im saying.
I doubt it. On sensor PDAF system will never match the rumored 102 points AF system of SLT A99. Not this year. Not in the next 10 years.
On sensor PDAF system is a compromise too. Every AF point on the sensor would mean thousands of pixels that would have collected light are removed to make space for one on PDAF point.
Well?…when you take into account that these sensors have so many pixels (photosites, really) on them that are “barely being used, it’s not much of a loss in the end.
Example: “Real world” end result in resolution between Sony’s 16mp and 24mp sensors is really nothing. We have test after test that show very little difference between the two in the final image.
If Sony dedicates several thousand photosites on a 24mp sensor,….I guarantee non of us will ever see the loss.Not even the deepest pixel peepers will see it.
If you think of how much resolution is lost on the lens, how much is lost on Bayer pattern and how much is lost on simple Nyquist frequency cut-off….on sensor PDAF is only a small fraction of all of that.
If you put 102 points on the main sensor, each taking precious space away from light gathering ability of the sensor, then you are losing the light anyway, just like you are losing the 1/2 stop light with SLT mirror. Plus we don’t even know if it will work as well as SLT AF
Whatever happened to the E24Z lens report that claimed the prototype lens had a built-in leaf shutter? Could this be related?
If you put 102 points on the main sensor, each taking precious space away from light gathering ability of the sensor, then you are losing the light anyway, just like you are losing the 1/2 stop light with SLT mirror. Plus we don’t even know if it will work as well as SLT AF
Wow.. Really?
Let’s see what they’ll make.
BTW, what if the sensor would be awesomely great?
We don’t need SLT any more, is it?
This spells the end of semitransparent mirrors covering the sensors in Sony cameras. Finally. The A99 will probably be the last one.
Doubtful
very doubtful indeed.
This might improve the usability of alpha lenses on nex with much simpler adaptors.
Apart from on sensor PDAF it means that new generation (third? Exmor) sensor will be implemented in new NEX cameras. New sensors should be noiseless and with even more improved DR and color…
“Noiseless”? lol
If Sony use same PDAF on sensor technology as Canon M it would be much worser than PDAF solution in Nikon 1.
Nikon 1 has the advantage of smaller 2.7 crop sensor with deeper DOF than APS-c
It’s not DOF what counts.
What counts is:
1. The weight of focusing elements
2. Distance which focusing elements need to move
DOF does matter because the focus can be off, but you won’t be able to tell because more (the subject) is still in focus. But when you start cranking DOF down to really tight levels, then it is more important for your AF to be spot on.
Didn’t Fuji do this first in one of their EXR compacts a couple of years ago ? Nikon also do it…. and Canon too… but Sony has a patent dated after Fuji already brought it to market ?
From what I’ve read, Fuji’s didn’t seem to work too well, Nikon’s is good for tracking, and Canon’s seems to be poor.
So how will Sony do ?
Fujifilm announced their camera with hybrid focusing in july 2010. Sony submitted the patent in 2009. So while it may be true that Fujifilm was earlier, without seeing their patent (assuming they have one), there is no way of knowing.
That being said, I expect them to have (slightly?) differing patents (technologies) anyway.
*fingers crossed*
Will this require lens (or lens firmware) updates to fully support it?
I guess new lenses with a different motor.
As far as I know, NEX lenses already have stepper motors and for as long as CDAF is atleast part of the focusing equation (the next years, I expect), that’s the best to use.
I expect the 7 iterations to come in 2013.
[Meant for Geir below.]
While some might feel such will render their NEX-7 “obsolete”, it is Sonys first venture into the technology and even though nikon, canon and fuji is already using it – an early venture. There are no indications that the implementation is perfect. I would go so far to say that while it probably will be a slight increase in speed, it won’t be enough to warrant an upgrade unless other factors put the scale in that favor.
It however bodes well for the next revision of the 7 series, in say 2014-16. Then the technology will have had a couple of iterations in the direction of perfection. That is when I’ll buy one. As I knew when i bought the nex-7, there will soon be pd-af on sensor. 🙂
I guess it will take untill the third generation or so of this technologi with phase detection AF pixels integrated on sensor, untill we can completely get rid of the dedicated phase detection AF module in the SLT-range of cameras!
This technique is probably not fast and light-sensitive enough until then, to replace the separate phase detection AF module.
But it is a good, interesting and valuable step forward in the evolution of digital cameras.
This must be the future of digital cameras (and by that the dead of optical viewfinders)!
Sony rocks!
Will 5R have a touch screen?
I’d be surprised if it didn’t, seeing as the 5N had one.
Holy shi
So can I assume the LA-EA2 will NOT be compatible with the new NEX sensor? Will it cause problems with overlapping one system over another when both has phase detection? Also, if that is the only improvement over the previous sensor, then I’m out of the running to upgrade considering I have the LA-EA2 on my 5N. In fact, I use it in that form 99.99% of the time. The problem with the NEX line is the lack of lenses. I love the technology but it’s like having a Ferrari with bicycle wheels. I hope the integration of phase detection means more lenses. For me personally, if they were to come out with a zoom lens that equates to 24-27mm or 70-200mm with the 1.5x crop sensor and with a constant aperture of f/2.8, I would have enough confidence to finally release the LA-EA2 from my 5N.
nice – but will this work all the existing e-mount lenses? or will it require new e-mount designs?
That would be horrible especially for the folks that shelled out $1000 for the Zeiss 24mm
Even if existing lenses didn’t work with the PDAF sensors (which hasn’t been confirmed either way), they will work fine with contrast AF. The Zeiss will work fine.
I doubt (like manny here) that the PDAF on sensor will be as fast as a dedicated PDAF sensor.
I see no reason why the PAF on sensor will give problems with the LA-EA2 adapter, the on sensor focussing system will be shut off when the adapter is used.
IBIS on the NEX-6 would be nice…
First Fuji, then Nikon, followed by Canon, and now Sony.
Hybrid AF definitely seems to be the way of the future.
Lets hope Sony’s implementation is as good as Nikon’s.
If so, two words: SPORTS & ACTION!!
We’ll have to see what effect the on-sensor PDAF has on the sensitivity.
e.g. The canon 650d with PDAF has significantly lower performance at high iso compared to its predecessor; the 600d despite canon claiming the opposite.
Based on the always reliable RAW comparisons at Dpreview? 😉
No
Would love to see the source then.
Chipworks has published some photos from teardown of 650d sensor. The number of PDAF pixels is very small…
Prepare for bitter disappointment.
The on sensor PDAF will only marginally help reduce hunting when focus is way off target and with very wide apertures as it is notoriously inaccurate. So may be the 50/1.8 experience will be somewhat better say 50% better on average or alike. Hopefully will help more with 85/1.8 like 2x or 3x improvement there.
On sensor PDAF, CDAF, plus the best manual focus (peaking) ~ looks like a big win for NEX cameras.
Does anyone else here wonder if PDAF on sensor AF speed is correlated to sensor size (a bit like CDAF is)? Reason I raise this is the Nikon 1 seems to focus fast, but the Canon EOS-M and 650D are slower than most current generation CDAF cameras, even in good light (in all fairness, a lot slower – both NEX and M43s are faster).
Well the Canon was very slow.
I agree. I sure hope it is nothing like the EOS-M because the Canon autofocus looks like it might be the slowest mirrorless camera…even slower than the Fuji X-Pro1.
Of course it will be faster than the EOS M. Even a NEX with CDAF only is already faster.
A friend of a friend who’s got some inside information told my friend that the sony sensor-based PDAF was engineered to be superior to the competition in orders of magnitude, and that there’ll be no significant difference between the focus speed of the DSLT cameras and those using the new sensors!! The source my friend that it was an engineering miracle at Sony.
The DSLT era may be short lived after all.
If the same sensor is deployed in A99, I wish it had mirror-up & main sensor LV function – Mirror-less A-mount camera?!
I shoot mainly portraits & landscapes and I can live with slow focusing. SLT is needed for video and quick AF only.
“If the same sensor is deployed in A99…”? That doesn’t even make sense, mate. lol And of course the A99 will have main sensor LV, jeez it has only 1 sensor.
I’m not really excited about this news. It’s a good to have thing, but nothing much to excited about. On sensor PDAF replaces SLT? I rather have the mirror. Actually I would rather have a mirror that only reflect infrared for PDAF focusing that theoretically has no visible light loss.
If you use IR to AF:
1. Invitation for front/backfocus
2. Inability to focus under luminiscent and photodiode light…
though…
Well, I am aware of the drawbacks of IR AF, that’s probably the reason why nobody is interested in making such system. But there may be solutions…
1. PDAF IR + on-sensor CDAF should give more accurate results.
2. IR assist lamp. As a bonus feature, it’s invisible and unobtrusive.
The bad things:
1. PDAF is sensitive to chief ray angle. Expect Sony to introduce largeish retrofocus lenses to make them work with this PDAF.
2. CDAF can be made much better without compromising sensor… It’s pity that Sony programmers are not good enough. Most people understand CDAF as of 198x, with 1 MHz processors…
The good things:
In theory, one can extract slighty more color (the PDAF pixels being panchromatic) and DR from that sensor…
I would prefer RGBW sensor (with 1/4 W pixels) and excellent AF software.
That’s exactly what’s in my mind.. on-sensor PDAF may compromise sensor performance. The effect might be negligible though.