We all know Sony’s solution to archive phase AF by using a semitransparent mirror. And looking at the latest Panasonic patent it looks like there are even smarter solutions! The sensor itself can reflect some of the light to the phase AF sensor (N.13 on the picture). There is a sort of reflective material between the pixels. That solution is called “Front side illuminated sensor”. There is also a second solution where some of the RGB pixels have been replaced with Phase-AF-Pixels like it works on the [shoplink 25297 ebay]Fuji’s F300EXR (it is a compact camera)[/shoplink].
Now, I doubt that Panasonic will ever introduce such a phase detection system. The current [shoplink 23568]Olympus E-P3[/shoplink] and [shoplink 16254]Panasonic GH2[/shoplink] have already an exceptional fast contrast autofocus. My bet is that in a few years every camera will use contrast AF only. Phase AF is only a waste of space and resources!
More about the patent at Egami (Click here to read the translation).
73 Responses
I love Panasonic’s camera engineers!
I think I’ve seen a similar solution patent somewhere, I just don’t remember which brand it was.
The major problem with contrast detect AF is that it becomes considerably slower in low-light situations. And when you are recording a video and you end up filming the contrast AF in action it’s pretty depressing and ugly.
It was a Oly patent!
http://photorumors.com/2011/11/16/fuji-organic-sensor-casio-circle-panorama-and-olympus-phase-detection-af-patents/
http://egami.blog.so-net.ne.jp/2011-11-09
So I guess it will be a m4/3 feature, no? I think can expect this to become reality sooner.
I think that Nikon1 camera uses that phase-AF-pixels type. If it works, I think that reflection is better than lose some pixels. If reflection doesn’t have a downside as well.
I think it was Olympus…
It’s funny then why Nikon 1 system uses hybrid AF, Phase Detect for good light and Contrast Detect for low light?
I am having the J1 and when light goes down, the camera switches to Contrast Detect.
I think Contrast Detect although slower, but it is still able to lock focus in low light condition, while Phase Detect, like my A77 completely refuses to AF in low light sometimes.
I think you got it backwards. Phase detection is worse in poor light than contrast detection. That is why dslrs have such poor focus accuracy in low light. Sure the focus time is fast, but that is because it is focusing at an inaccurate distance.
Contrast detection seems slow in low light because it is constantly trying to get accurate focus. If the programmers just let it guess like PD, then the focusing time would be better.
It would be nice if we could auto focus for still shots, with our SHG lenses on future mirror-less cameras…
Exactly..I would love to use my 50-200swd on a m4/3 camera..rather than buying an e-5
this would be a very good move if Oly could back support all their 43 glass!
So the big advantage with phase-detect is *direction*. CDAF is great, but fundamentally it revolves around resolving a blurry image to a sharp one, and there’s no way to know which direction will sharpen things up until you pick one. (I assume there are some heuristics but they are just guesses.) The current bodies manage by scanning the full focus range VERY quickly. So if you can get juuust enough phase detect focus to know which direction you’re headed, contrast detect can get you to the right point with unbelievable speed on the new lenses.
Well said!
Exactly!!!
> So the big advantage with phase-detect is *direction*.
PDAF doesn’t provide direction either – it provides only the distance. Or you forgot the “lens hunting” already?
It is that PDAF algorithms at the moment are better developed than those of the CDAF and are able to guess direction better.
I’m no specialist in AF, but it could also be that PDAF provides some extra info CDAF isn’t capable of delivering to AF, thus they might be complementing each other.
PDAF’s distance for sure can be useful to AF: it would allow to make one big jump to the target and then fine tune focus with the CDAF.
“PDAF doesn’t provide direction either – it provides only the distance. Or you forgot the “lens hunting” already?”
quite wrong
PDAF most certainly provides direction to move a lens focussing element by analysing the strength of contrast from multiple detectors. The detectors neither know nor care how close of far away an object is, theyre merely configured to seek the strongest contrast with a given algorithm.
Lens hunting, is more a feature of an inability to lock onto sufficient contrast.
PDAF knows both to which direction and how much focus is misaligned.
http://graphics.stanford.edu/courses/cs178/applets/autofocusPD.html
http://www.dpreview.com/news/2010/8/5/fujifilmpd
And that’s why it’s also much better in tracking moving target because it knows instantly to which direction target is moving out of focus.
Uhm… “a small embedded processor” right there – which calculates the direction. It is not a feature of PDAF itself – you need the algorithms to do it. And the algorithms are all imprecise and probabilistic and thus sometimes AF does “hunt”.
The mentioned algorithm – Cross-correlation – if you care to read the wikipedia’s page, you would have immediately understood that it can’t work reliably in the real world. If it did, we’d had the perfect OCR and voice recognition software decades ago. And we still do not.
P.S. OK, I understand that photographers are not obliged to understand how their gear works in depth. But still. 🙂
The PD routines are pretty well known at this point. They do not “hunt” in and of themselves. Indeed, that’s one of the problems with PD: once the lens is assumed (an important word) to be at the place where the PD routine told it to go, most systems just do a “within tolerance” check and stop if they are. If the subject moved, obviously, that detection triggers another PD sequence. But hunting, no. Canon seems to have looser tolerances than Nikon, but both have tolerances that can and do miss precise focus. Moreover, PD is really designed to a specific aperture (f/5.6 for Nikon), and fast lenses will tend to have spherical aberration that typically results in “back focus” results. That’s usually “fixed” by in-camera algorithms, and again the word assumption comes into play (assumes that the spherical aberration is exactly the same between all samples of a lens, which we know isn’t true).
You’re a great debater: your claim to wisdom now solely depends on the definition of what constitutes PDAF (and whether PDAF implementations commonly depend on algorithms that may or may not exhibit hunting behavior). Mmmmh. 🙂
Exactly. And this is why PDAF is not a waste of space and resource yet. It will be when a m43 camera does AF tracking as well as my 7D and when this time comes I’ll give up my DSLR to have only m43. Until then… I need both
Is that “front-illumination” the same thing that noise disadvantages make the development of back-illuminated sensor
I was worrying the same thing too……….
Maybe they should use a backlit sensor to put those AF pixels so that it’s a “front and back lit illumination sensor”. High tech, eh? What was wrong with film and manual focus again??
My understanding is that back-side illumination essentially moves the non-sensing part of the chip (connecting wires etc) to the back of the chip, allowing you to get larger photo-sites on the same size die. I also believe that the size of these wires is constant (or at least linear with respect to MP) and not affected by chip size. So while BSI is a huge win on phone and compact sensors, not so helpful on large sensor cameras.
Size of wiring and other components around actual light gathering photo-site should be able to shrink with more advanced manufacturing process having smaller so called half-pitch.
And in system camera’s big sensors those have apparently taken so little sensor area per pixel to make effect small in overall and that effect should get smaller.
But tiny sensor compacts have so little area per pixel that in there area loss was such that when CMOS sensors were starting to appear in compacts they had worser noise than CCD using cameras. Good example is Canon SX1 vs SX10.
Here’s good simplified explanation of differences between front and back side illuminated sensor:
http://www.cameratechnica.com/2011/06/23/technology-demystified-backside-illuminated-sensors/
Thanks for the explanation 🙂
Actually, I thought it’s interesting if they took advantage of the disadvantages in non-BSI sensor
I see two problems with the idea: (1) fill area is likely compromised; and (2) the need for telecentricity is likely increased (exactly the opposite of why Olympus said they designed 4/3 ;~).
If you lose enough fill area, yes, noise could increase. If you lose light due to lack of telecentricity, yes, noise could increase.
The other strange element to the design is that the focus sensors aren’t at the focal plane distance, which is typically where they would be in a PD design. So there’s something else going on here than a standard PD design. I’m plane-tripping today, so I don’t have time to look at the patent closely to see what’s really going on.
I think the Contrast Detect AF doesn’t do a good job with moving subjects, does it?!
@renato: It’s not exactly true that CD is “slower” than PD in low light. Both systems have light sensing that is area dependent and frequency dependent. You can have CD systems that collect more light and have more sensing in both directions than a PD system, or you can have PD systems that have more light collection and more discrete frequency sensing than CD. It depends upon what you design (and sensor size and megapixel count plays a role in CD).
@promit roy: It’s also not true that PD has a direction advantage over CD. While that was true some time ago, there is now at least one algorithm (and I think multiple) that can compute direction from contrast info.
Both PD and CD have pluses and minuses. I suspect we’ll see more and more hybrid solutions ala the Nikon 1, which uses both.
I cannot imagine an algorithm that can computes direction from contrast info without requiring needing multiple samples. Any more information would be appreciated!
There is no such algoritm, he’s talking out of his buttocks. Its theoretically impossible to infer the direction without taking multiple samples in both of them.
Looking forward to Fuji’s organic sensoramachacallit-thing
That is a most interesting development then, which tends to suggest that there’s not a particularly strong case for PDAF simply from theoretical possibilities. Still advantages with tracking AF maybe?
The advantage of CDAF is accuracy. None of the front and back focus problems that plague PDAF cameras.
If the PDAF sensors is built onto the main sensor (what Nikon and Fuji has did), then there won’t be F/B focusing issue.
Alignment problems will cause F/B focusing issues, but the main problem is that the mechanical parts aren’t perfect (and probably wont ever be).
Sorry, but that’s not true. PD is designed for specific apertures (typically f/5.6). When you have lenses that are far from that (e.g. f/1.4) you usually get back focusing due to spherical aberration. You can put a “correction” for that into lens table in the camera, but you’re putting a fixed correction that might not apply exactly to all samples of the lens. That’s just one of the reasons why we have a Focus Adjust function in the higher end cameras.
@Thom – I think you are mixing the phenomenon of “focus shift” as observed in lenses like Canon 50mm 1.2L with the more general problem of front and back focus. Sure in some lenses focus shift due to spherical aberrations can be a significant source of front or back focus, but in general the loose tolerances are the main reason for front and back focus.
While there is a Optics designed for the use of PDAF, it must be present in a cameras.
All I know us that even the top mft cameras (GH2, EP3) is very poor at tracking focus. This is one of the main remaining weaknesses of the format. Please do whatever you need (PD, black magic, alien technology etc) to make it better!!!!!
CD requires very fast refresh (typically 120 fps or higher) to do reasonable tracking. That’s partly because it needs a quick data stream in order to do reversal and speed calculations. 60 fps is better than 30 fps, and 120 fps is better than 60 fps, but we need faster. Thing is, the faster you make the video stream off the sensor, the more data bandwidth you need internally. The Nikon 1 blows away anything we’ve seen previously (even a D3) in terms of internal bandwidth, so coupled with the PD cheat in reasonable light, it does actually focus quickly and track reasonably.
“Phase AF is only a waste of space and resources!”
If you call every single PDAF lens ever produced a waste of space and resources and if you are happy with those lenses being thrown away to the garbage bin because they literally take several seconds to focus on CDAF cameras, then yes, it’s a waste of space and resources.
So let’s throw our expensive lenses away and replace them with the ninth m4/3 14-42 kit zoom!
Phase AF (and CD AF) is in the body, not the lenses. However, the way the body control focus on a lense influence how suitable a lense will be for CD AF. That might change though with better CD algorithms, particulary if CD-based systems can guess in which direction they should focus as Thom Hogan writes earlier in this thread.
Actually, no. There’s a huge mechanical difference between PDAF and CDAF lenses. They just work differently – different motors, different gears, different electronics. You can get a PDAF lens to focus in CDAF mode, but it’s going to be slow, and it’s going to have horrible consequences on the mechanism, wearing it out and breaking it tens of times faster than if used in PDAF mode. Some lenses, like the Little Tuna (Oly 150mm) won’t even focus on m4/3 for some reason. A friend of mine has been told by an Oly repairman not to use SWD lenses on m4/3 bodies because the mechanism will die in a year of regular usage.
Furthermore, it’s easy and quick to focus a lightweight CDAF lens, but if you venture into the realm of fast and serious (meaning heavy) glass, you run into faster mechanical wear, as well as faster battery wear. And you don’t have a lot of battery power in m4/3 bodies.
I don’t see PDAF going away any time soon, if ever.
> A friend of mine has been told by an Oly repairman not to use SWD lenses on m4/3 bodies because the mechanism will die in a year of regular usage.
In GH2 Zuiko 12-60mm actually focuses as fast (or faster) as contrast AF compatible Leica D 14-150mm but instead of smooth constant focusing it happens in fast jerks which literally sound like they could tear focusing mechanism to pieces if used lot.
But SWD refers to just type of focusing motor, beside normal cheap micromotor and with less mechanical parts like gearing it should be able to do focus hunting of contrast AF better if control electronics of the lens would be designed to do that in addition to PDAF’s ability to directly jump to right focusing distance in one move. (that 14-150mm has Panasonic’s equivalent to SWD)
Well, the strain on the mechanism is supposedly 10-100 times as much in CDAF as it is in PDAF due to constant jerking and tiny movements. That just can’t be good on the lens. I have no idea if the remark about the focusing mechanism breaking in a year is true, but it very well might be.
PDAF could be used to nail the initial focus lock, and CDAF could then be used to get fully accurate focus extremely quickly… But I don’t know if it could be done with the current lenses.
BTW, when I tried the 12-60 on the E-PL1, it took about 2 seconds on average to focus in reasonably decent light. I’ve only used the new Oly primes on the E-P3, so I can’t tell how it behaves there.
Again, if you design a lens solely thinking that it will be forever used with PD, you might use a motor that would have issues with CD. However, even when the 4/3 cameras were designed, surely Olympus understood that Live View was CD. Thus, if they continued to design lenses with motors that couldn’t handle CD, they were essentially committing a major design error. Half of good SYSTEM design is a strong understanding and anticipation of the likely future. It’s why you have to have forward thinkers–Steve Jobs comes to mind–in a leadership role in decision making.
If true, that would have been a design error by Olympus, as it means that they did not anticipate moving from PD to CD (e.g. did not anticipate m4/3 when they designed 4/3 lenses). But I’m not sure it’s true.
m4/3 is Panasonic’s lovechild. Also, most of the lenses were designed before LV and CDAF, early in the previous decade.
The newer Oly lenses are CDAF-compatible, like basically all the kit and low-end lenses. The only mid-level lens that was designed with CDAF in mind was the 14-54 MkII. While that one does focus faster in CDAF than the MkI, it’s still a rather slow focuser (due to heavier glass elements), nowhere near the speed of m4/3 lenses.
Not a design error at all, since some the Oly PD lenses are renowned: the 12-60mm seems to have almost legendary qualities. Its performance would not have been feasible using CDAF-compatible technology at the time.
Olympus pioneered the transition phase to mirrorless. Managing this transition was the challenge then, designing “error-free-in-hindsight” lenses was not.
Personally I think that PDAF will always have an advantage of fast focusing without hunting. So I dont think it will go away completely. But it will be used mainly in “pro” cameras for film, sports and reportage.
The massmarket will be 100% CDAF.
But I dont understand the patent. Even if some light is reflected from the main sensor to the PDAF sensor, the reflected light is not in the focus plane any longer. Some additional tricks must be added to the patent to make it work.
It should be possible to compensate as long as you know how much off from the focus plane the sensors are.
I think the main point of Sony’s semitransparent mirror (or so called translucent) is not to have phase detection but an optical view finder, something this solution does not provide. They are not really comparable.
AFAIK none of Sony’s translucent models has an optical viewfinder; they all have an EVF.
The semi-transparent mirror don’t open when exposing like the usual SLR. It still have to let through as much light as possible to the sensor, so if they implement OVF, it will surely be really really dark.
Um, sony’s SLT cameras don’t have optical viewfinders – they have an EVF.
“I doubt that Panasonic will ever introduce such a phase detection system.”
I doubt that Panasonic or Olympus will ever introduce a phase detection system that works with 4/3 lenses. Which is a great shame since:
1) There are several patents and working solutions which show it can be done with a little R&D devoted to it.
2) There are people who have invested a lot of money in 4/3 lenses and are now being left in the cold. That’s not the way to treat customers. Is our investment in m4/3 lenses and accessories going to be a waste of money as well? Even Sony, which is notorious for its poor support, did make a PDAF adapter.
3) Those 4/3 lenses have excellent optical quality which is not mathced by their m4/3 counterparts. Why not give people who want highest image quality possible with a 4/3 sensor a chance to have it?
4) PDAF is still somewhat quicker than CDAF in certain applications. Why not cobine them the way Nikon did in order to get maximum focus speed available?
Therefore I don’t think “Phase AF is only a waste of space and resources”. Backward compatibility is a way to ensure people stay in your system. And sometimes old technology can be reintroduced in such a way to greatly improve your product – see semi transparent mirror as a good example.
I agree completely with Miroslav. PD AF on MFT would make all the excellent FT lenses work well on MFT, doubling (at least) the number of lenses and adding fast zooms and fast tele lenses to the system.
It would also make it much more likely that current FT-users would switch to MFT when they upgrade. Seems like mister MFT rumors have missed some important points on this one.
Three years after the introduction of the first CDAF-only camera system, there’s still no proof that CDAF will ever become able to focus bigger lenses with larger and heavier moving optical groups as quickly as any supersonic-drive PDAF can. By bigger lenses I mean stuff like a 300mm f/2.8 or even just a 50-200mm f/2.8-3.5.
If mirrorless systems want to be the future of photography, they need such lenses, and they need to focus them quickly. As of now, only Sony NEX has the ability – but only through the new SLT-mirror-fitted, PDAF capable A-mount adapter.
In the long run, we’ll have to wait and see whether PDAF will go or stay. I cannot see it disappear anytime soon, though. And, from a personal point of view, because much of what I do is a kind of telephotography for which the existing mirrorless lenses simply are neither good enough nor fast enough, I cannot see a CDAF-only system replace my DSLR for the foreseeable future.
Because of the way how it works contrast AF will always be handicapped by it being unable to directly know how much and to which direction to move focus.
Especially with moving target there’s notable advantage in PDAF because it doesn’t need to do hunting to know where target is moving in relation to current focusing distance.
Because of that it’s hard to see plain contrast AF ever replacing PDAF completely.
Speed at which lens focusing elements can be moved is entirely secondary and if those can move faster PDAF can always tell faster than contrast AF to where they should move.
You’re confusing two problems. The issue you’re referring to is mass and force, which has nothing to do with CD or PD speed and everything to do with the motor you use. Moreover, most modern designs now use focus elements that are small and don’t move much, even in the big lenses. A whole bunch of intersecting issues start to encroach as you go “exotic” (stabilization elements, fast apertures, and more), but you’re still stuck with a simple truism: CD works faster if the video stream is faster. As I note above, making the video stream faster means that you need way more bandwidth internally. The Nikon 1 is current state-of-the-art in that respect, though I suspect we’re going to get other Nikon cameras shortly that push things even further.
Note that you can (and I do sometimes) use Live View (which is CD focus) on a D3s or D3x and a 400mm f/2.8. It works just fine, and it works at the speed of the video refresh.
Neither phase nor contrast detection works without the proper software behind that. And from the algorithmic point of view, phase detection is a root search function, contrast detection is a minimum search function which is always much more CPU-intensive and therefore slower. Thus, phase detection AF is always the better solution.
Never bet against increased CPU ability. Never. It improves tangibly every 18 months.
This brings us back to the question raised above: did Olympus design lenses that will have issues when we start throwing 240 fps using CD at it? If so, then someone made some very bad decisions at Olympus, because this is an area that should have easily been predictable.
You are hinting at “design issues” caused by “bad decisions” that simply aren’t there. Again.
Having 240fps processing will never be an issue at all. If a lens is designed for 120fps, it will be controlled at 120fps even if the processor of a newer camera is faster; it will not have “issues”. A lens-design is a balanced concept, and it would be ridiculous to assume all parts should be able to keep up with all future enhancements.
Such speculation reeks slightly of FUD-spreading, I think.
Interesting discussion but the patent link does not seem to be working. What I find interesting in the sketch provided is maybe a clue of where Pany may be going with their next GH3 body? Note it is a squared off body. While the hump in the middle for EVF is maybe still there or a large rangefinder body but the projection in the back that many object to has been removed so similar to some Canon and Nikon DSLR viewfinder designs?
It is my hope and desire that Panasonic, if not Olympus as well, both figure out a way to get PDAF to work on their micro 4/3 cameras, and quickly.
While I do NOT think it is necessary for all camera models to have PDAF, having it, at least, on a mythical “pro spec” micro 4/3 camera would be very welcome indeed, as it would allow those very nice Olympus high grade and super high grade lenses to focus properly on micro 4/3, especially the lenses with the SWD motors.
Honestly, I think CDAF only is stupid. A hybrid solution, which is fairly easy to implement (witness the new Nikons) is the best way to handle “legacy” lenses as well as the newer CDAF only glass, and makes your camera bodies far more versatile, with no weight penalty, and only a minor increase in cost.
I would LOVE to see PDAF on a future Lumix “pro spec” weather and dust sealed body. If the rumors about the next “X” lenses being weather and dust sealed is true, a weather and dust sealed micro 4/3 body may not be long in coming.
– Olympius
+1
And I’m sure anyone currently using FT systems would agree. For Olympus to have developed a system (mFT) that is not compatible with their better lenses (FT’s HG and SHG), and then telling those who use them they should migrate to this incompatible system has always been nonsense.
CDAF is also poor for sports and birding, the first a necessity for anyone with kids, the latter popular with nature lovers and many retirees, two large markets that will stick with their PDAF DSLR’s until this is remedied. For the Admin to not have noted this is a miss on his part.
Please, Panasonic, develop this technology and also share it with Olympus to grow mFT.
Olympus did not create Micro Four Thirds; it was clearly a Panasonic development that Olympus adopted. Also, Olympus worked much earlier and harder on the issue of backwards compatibility with adapted 4/3 lenses than did Panasonic – see the lens compatibility charts. (I have an Oly 50-200 that cannot AF on my M4/3 Panasonic body, but would if I bought a Pen body). So, criticism of their treatment of 4/3 users is somewhat misdirected. It wasn’t completely under Olympus’ control.
re: Phase AF is only a waste of space and resources!
As has been well stated by several here: A hybrid solution is critical to marry 43 and m43 lens systems. Important to me for use of my 43 lenses.
Not yet stated: Such a hybrid solution could provide the error information that would allow dynamic correction of PDAF front or back focus issues with the correction information stored by lens. That is: if PDAF system indicates focus, but CDAF system shows out of focus, CDAF system can be used to find true focus and the difference for that focal length and lens can be stored as a corrective factor for the PDAF system. If enabled, you might need a lens specific reset option for when a different copy of a specific lens gets used, so correction data collected from the prior lens is not used for the new lens.
Let’s hope they are close to a good implementation!
Peter
You don’t really correct for focal length, you correct for aperture. Though I suppose some zooms do have different aberrations at different focal lengths, so maybe you have to correct for both. But the first thing you correct for is aperture.
You guys talk like mirrorless cameras are the end-all for all cameras. There is not one mirrorless camera out there that can AF a 400mm 2.8 while shooting at 9fps on continuous focus while maintaining that burst rate for several seconds. Parlor tricks like using the electronic shutter still can’t hack it. I’m an M43 owner, but some of you guys need to be realistic. Mirrored camera flat out work. Not to mention they use less battery power since they’re not draining the battery with an LVF or EVF.
Sorry, but you’re incorrect. There is certainly a camera that can do it, and it’s called a Nikon 1. It’ll actually do 60 fps.
The system isn’t perfect, as I’ll report when my review comes out. But it’s darned spooky at times. At my typical settings I’m running 25+ frames at 5 fps and often holding focus on fast moving objects. If Nikon can do that on a US$900 consumer camera, one wonders what they can do with a US$8000 pro camera.
When I first heard about Nikon’s hybrid system (almost three years ago, by the way), I immediately recognized the advantages to the approach. Done right, you should be able to cancel out the disadvantages of each system while retaining the advantages. But it takes a lot of sensor/CPU horsepower to pull off. As I wrote above, never bet against semiconductors (CPU, memory, bandwidth, efficiency, etc.). They’ve been on a constant improvement line for decades, with little sign of slowing. Thus, if the thing that’s keeping you from doing Algorithm X is horsepower, you need to design that today in anticipation of tomorrow’s chips. Designing to what’s available today is wrong, and has always been wrong in high tech.
Are you using a 142mm f/1 lens on that Nikon? 🙂 It’s not hard to keep focus if you more or less have the entire scene in focus.
Very interesting contribution Thom, thanks. It is very easy to get frusfrated at the “slowness” of development, but really: things are changing very fast. Not all engineers will see the whole forward picture, so it is quite possible that there are PDAF lenses that won’t cope with the rigors of CDAF. Olly won’t be alone there.
I’m really looking forward to trying out a Nikon 1. I doubt that I’ll buy one, because for me the main reason for moving up from a P&S was to gain DOF control. But it almost feels to me that Nikon is using the camera as a test bed before they introduce similar features on a larger sensor camera. As you point out, the capacity now to handle the info stream from a smaller sensor will grow, and in 18 months or so they can put the same abilities into place on an APSC or FF sensor.
Proof the E5 is history.
And phase detect is still good for video.