Nikon 70-200mm f/2.8E FL ED AF-S VR MTF Tests
Well, we all want to call it the 70-200 VR III, since it replaces the Nikon 70-200 f/2.8G AF-S VR II. But the new lens is officially the 70-200mm f/2.8E FL ED AF-S VR. It is most definitely a different lens; it’s 22 elements include one fluorite, one high-refractive, and six ED elements, along with new coatings. There’s also fluorine finishes on the front and rear elements to prevent smudging. I’m not a fan of this in general (they seem far more prone to coating scratches), but some people are. Not to mention it has electronic aperture control (of which I am a fan). And, of course, an improved 4-stop VR system.
So most of these improvements are things you’ll need to find out about in lens reviews, which this isn’t, if you’re considering upgrading to this lens with it’s Sonyesque $2,800 price tag. I’m just here to tell you how it performs on an optical bench, what it’s resolution is, and stuff like that. Maybe in a few weeks, we’ll take one apart and look into that ‘improved weather sealing’ claim. But for today, we’re just going to look at the optics.
MTF Results
To make things simple from the start, I’ll compare the MTF curves for the new FL lens on the left to the MTF curves of the Nikon 70-200mm AF-S f/2.8 G VR II lens on the right.
At 70mm
The 70-200mm VR II was at its best at 70mm, and the new lens is even better on the MTF bench, although just a bit. I doubt that anyone is going to notice much difference in their photographs, but they might see improved corner sharpness.
At 135mm
The 70-200mm VR II was at its worst at 135mm, and the new lens just mops the floor with it in the middle of the zoom range. It’s not just hugely sharper in the center; it’s sharper 2/3 of the way to the edge of the image than the old one was in the center. Now I know a LOT of people use their 70-200mm lens just as a 200mm lens, but honestly, if you ever visit the middle of the zoom range, you will notice this difference. It’s night and day.
At 200mm
The VR II was quite good at 200mm, particularly at the center. The new FL lens is perhaps a bit better in the center, but it maintains its sharpness much better as you go away from the center. Maintaining sharpness this well away from center is pretty amazing for a zoom lens.
I’m going to have to talk about this MTF curve a bit, because you may have noticed that the sagittal and tangential curves (dotted and solid lines) aren’t quite together in the center of the FL lens, which is unusual. The reason, in this case, is a bit of ‘off-centering.’ I’m not going to use the word decenter, because that implies optical decentering, which this isn’t, exactly. Rather the sharpest point at 200mm is, for most copies, not quite at the center of the image, but just a few mm (in sensor distance) from the center.
Remember, the graphs I show you are averages of 10 copies, each of which is averaged at four rotations; so the average of 40 MTF curves. If you look at a typical single rotation MTF reading they mostly look like this:
So technically, all of the 10 copies we tested were very slightly decentered at 200mm. But we usually think of decentering as causing a soft image, and in none of these lenses could the image be considered soft. It’s rather that the area of maximal sharpness was just a bit away from the center.
Because the sharpness is maintained so well from side-to-side, I very seriously doubt anyone could find that the center was a bit off-axis in a photograph in any of these ten copies. That doesn’t mean there aren’t going to be some that are significantly decentered at 200mm, though, because lots of lenses made. But again, decentered or not, they were all just damn sharp at 200mm.
Field of Focus Curves
I wanted to post the field-of-focus curves, for those of you who are interested in the geeky stuff, to show just how flat they are. Again, this would be good in a prime lens, but for a zoom they’re pretty spectacular.
This is one copy, but a very typical copy. And yes, most had a very slight field tilt at 200mm, but little or no tilt at shorter focal lengths. Again, you can harp on the negative, but for a zoom, this less field tilt is less than we expect to see.
OK, Fanboys, Here You Go
Yes, I hate myself for doing this but if I don’t someone is going to dig up old results and do it anyway. What follows is a comparison of the Nikon 70-200 f/2.8 FL VR to the standard bearer of 70-200 f/2.8 lenses, the Canon 70-200 f/2.8 IS II.
At 70mm
The new Nikon is a tiny bit better in the center but definitely is a bit better off-axis in the middle frame at 70mm.
At 135mm
This demonstrates just how much of a difference Nikon has made in the middle focal range. The Canon was clearly better than the Nikon VRII, while the Nikon FL is clearly better than the Canon.
At 200mm
Despite the oddity in the Nikon 200mm graphs, I think the Nikon is clearly a bit better at 200mm. As always, let me emphasize this is ‘lab better.’ Actual photographs with all of the variables that introduce, I doubt the difference at 70mm or 200mm would be of any visible significance. At 135mm, though, I think they might be.
And as always, this has been an MTF test; it was only an MTF test. Had this been an actual photograph you would have been instructed that all differences were simply because of poor technique.
Roger Cicala and Aaron Closz
Lensrentals.com
November, 2016
130 Comments
Omesh Singh ·
Hi Roger, thanks for always being right at the cutting edge of information 🙂
Have you guys gotten the 19mm PC-E lens yet? I’m interested to see how good their first ultrawide Tilt-shift is. (I’m assuming 19mm counts as ultra-wide)
Roger Cicala ·
Omesh, we’ve gotten a few in, but not enough to test yet. Maybe next week. I expect good things – they would have made sure it was at the least competitive with the Canon 17.
Lee ·
Given the price it’s a pretty catastrophic misstep if they didn’t.
Omesh Singh ·
Hi Roger, thanks for always being right at the cutting edge of information :)
Have you guys gotten the 19mm PC-E lens yet? I'm interested to see how good their first ultrawide Tilt-shift is. (I'm assuming 19mm counts as ultra-wide)
Roger Cicala ·
Omesh, we've gotten a few in, but not enough to test yet. Maybe next week. I expect good things - they would have made sure it was at the least competitive with the Canon 17.
Lee ·
Given the price it's a pretty catastrophic misstep if they didn't.
obican ·
Ugh, this is a hard one.
Roger Cicala (from LensRentals) says that the new Nikon 70-200/2.8 (Buy link here) has a Sonyesque pri.. not good enough.
Let’s try it with more links.
Roger Cicala (from LensRentals) tests the new Nikon 70-200/2.8 (Buy link here) that you can mount on a Sony via the new AF adapter (Buy link here) and found it even better than Canon (Buy link here).
Needs to be more sensational but you’re not really giving me much.
Oh, Author. Roger Cicala part!
Roger Cicala (from LensRentals) says Sony RX1R (Buy link here) is as good as Medium Format (B&H link to the whole medium format section) when shooting real pictures.
Any good?
Roger Cicala ·
Obican, you know they’re just going to start cut and pasting your comments as their headlines 🙂
Piotr Krochmal ·
Roger, there one more link missing: for tests we are using (link) 🙂
Will Humber ·
If I know a certain editor, the headline you’ll see will be ‘Roger Cicala says, “The 70-200 G Master is…really…great.” ‘
obican ·
Ugh, this is a hard one.
Roger Cicala (from LensRentals) says that the new Nikon 70-200/2.8 (Buy link here) has a Sonyesque pri.. not good enough.
Let's try it with more links.
Roger Cicala (from LensRentals) tests the new Nikon 70-200/2.8 (Buy link here) that you can mount on a Sony via the new AF adapter (Buy link here) and found it even better than Canon (Buy link here).
Needs to be more sensational but you're not really giving me much.
Oh, Author. Roger Cicala part!
Roger Cicala (from LensRentals) says Sony RX1R (Buy link here) is as good as Medium Format (B&H link to the whole medium format section) when shooting real pictures.
Any good?
Roger Cicala ·
Obican, you know they're just going to start cut and pasting your comments as their headlines :-)
Piotr Krochmal ·
Roger, there one more link missing: for tests we are using (link) :)
Will Humber ·
If I know a certain editor, the headline you'll see will be 'Roger Cicala says, "The 70-200 G Master is...really...great." '
dyswim ·
Thanks for the work on this Roger. Extremely useful and interesting information for the community that complements field usage based reviews. While you don’t like doing the C vs N stuff (because your damned if you do, damned if you don’t), comparing the lens to its predecessor and most direct competitor gives appropriate context.
A spectacular lens – along with the lack of focus breathing, reduced flare, better VR, faster AF etc. Nikon has given its users a great new tool.
dyswim ·
Thanks for the work on this Roger. Extremely useful and interesting information for the community that complements field usage based reviews. While you don't like doing the C vs N stuff (because your damned if you do, damned if you don't), comparing the lens to its predecessor and most direct competitor gives appropriate context.
A spectacular lens - along with the lack of focus breathing, reduced flare, better VR, faster AF etc. Nikon has given its users a great new tool.
Said AZIZI ·
A comparison with the G Master will be very welcomed.
Thank you Roger !
Junz Inc ·
Yes, I also hope to see that Roger.
Thank you.
Roger Cicala ·
Well, it’s not pretty. The 70-200mm https://uploads.disquscdn.com/images/dee115f056905e74d9a581d0b741d44de6050bad87a328cb5d1eb6bd5ce68623.png https://uploads.disquscdn.com/images/e4e23fcba52b0bee246992d26445571be981d765e76dccb2be4f4e98ac4c262b.png https://uploads.disquscdn.com/images/42d95e5a8c871c7fe04782a64172a0fd369cb7cccd15d5335dd2e1d1dd80e88f.png G Master isn’t really all that great.
Marlon Richardson ·
Probably worth putting this in review simply because of all the Sony interest.
Roger Cicala ·
Let’s kind of keep this here for right now – I haven’t published it because I’m still a little uncertain about the results. Sony has suggested that a change in cover glass thickness might improve the results some. Not dramatically, but some, off-axis. This lens also has two focusing motors and we have to focus it electronically via a camera to test. I’m not absolutely certain that ‘setting it at infinity focus on the camera’ and ‘manually focusing on an object at infinity’ are absolutely the same. So take these lab results with a grain of salt. On the other hand, they do seem to agree with what we see in real world results.
I try hard not to put out anything until I’m just absolutely certain our results are correct. We’re doing some stuff here that’s pretty cutting edge, honestly. No one does 4 rotation MTFs, for example. I’m pretty certain these are good results, but not absolutely certain. So I’ll post them in this discussion on my own site but I’d rather not see them reproduced all over the internet yet.
Roger
Marlon Richardson ·
Wow, this is more complicated than you’ve let on. I agree we should wait till you can do a test that you feel good about standing behind.
Roger Cicala ·
I think there’s another thing worth considering here that I didn’t mention. An early batch of 70-200 GMs came out and then the supply seemed to dry up. I honestly will want to test another batch of 10 later before I make any conclusions. And just so I’m clear, I have absolutely no knowledge of why availability seems to have slowed to a trickle. But just because it’s the way I’m wired, I’ll want to test another 10 copies after we get resupplied before I make any definitive statements.
Brian Smith ·
Would it possible (time wise) to take all three lenses in question (Canon/Nikon/Sony) and focus on a detailed scene and post the images for comparison. This would be an apples to apples comparison.
Roger Cicala ·
Wouldn’t that involve 3 different cameras? Hardly apples to apples. It would be a good system to system comparison, and I guess there are some people deciding which system they want.
But I have suggested some of the phototechs shoot all 3 on a Sony camera with adapter. The adapter adds another variable, of course, but it has some good practical implications.
Brian Smith ·
You are correct. I meant to say Canon (with adapter) vs Sony. Yes, the adapter “may” play a role and work against the Canon. That being said, if the Canon with an adapter is much sharper than the native Sony that says a lot about the performance of the Sony.
Of course you could throw in the Nikon (with an adapter) or just compare it on say a D810 just for fun.
For what it’s worth I think your results will remain roughly the same based on the images posted thus far on other sites. So far they have been mediocre, at best. Some actual owners are reporting hit or miss behavior with focusing as well.
Regardless, I’m pretty sure I speak for many when I say we would love to see a real world comparison with adapters. Perhaps a portrait shot and then a fixed landscape shot?
I’m super disappointed I waited 3 months for most likely what will end up being an overpriced, underperforming lens. 🙁
Thanks Roger!
Brian Smith ·
Do you have a rough ETA on when the comparison might be posted? When reaching out to Sony what have they said about the long delay?
Roger Cicala ·
Wouldn't that involve 3 different cameras? Hardly apples to apples. It would be a good system to system comparison, and I guess there are some people deciding which system they want.
But I have suggested some of the phototechs shoot all 3 on a Sony camera with adapter. The adapter adds another variable, of course, but it has some good practical implications.
Brian Smith ·
You are correct. I meant to say Canon (with adapter) vs Sony. Yes, the adapter "may" play a role and work against the Canon. That being said, if the Canon with an adapter is much sharper than the native Sony that says a lot about the performance of the Sony.
Of course you could throw in the Nikon (with an adapter) or just compare it on say a D810 just for fun.
For what it's worth I think your results will remain roughly the same based on the images posted thus far on other sites. So far they have been mediocre, at best. Some actual owners are reporting hit or miss behavior with focusing as well.
Regardless, I'm pretty sure I speak for many when I say we would love to see a real world comparison with adapters. Perhaps a portrait shot and then a fixed landscape shot?
I'm super disappointed I waited 3 months for most likely what will end up being an overpriced, underperforming lens. :-(
Thanks Roger!
Peter Kelly ·
“I’d rather not see them reproduced all over the internet yet”?
I think your hope is about as likely as emptying a bucket of mercury with a sieve!
Peter Weir ·
The Sony 70-200 GM might be a quite difficult lens to properly test on an optical bench.
As you mention, the GM contains two focusing motors:
1) a ring ultrasonic motor, similar e.g. to Canon’s USM, driving lens elements located near the front of the lens
2) a linear motor driving lens elements located closer to the bayonet mount.
In a phase-detect AF system, the body’s AF unit determines the lateral optical distance — a.k.a. “phase” — between the current lens focus position and the desired focus position, and then orders the lens to move that distance. The lens’s built-in AF controller translates that lateral distance information into the concrete, focal distance-dependent mechanical longitudinal movement of the lens elements that would, in theory, cancel the lateral phase difference.
From a control theory point of view, a SLR’s PDAF system is an “open loop” system, where the SLR doesn’t check if the distance moved by the lens elements actually achieves the best focus. The camera and lens system just assume that the initial phase distance information is always translated into a perfectly matching lens movement. That assumption was maybe sufficient during the film era, but the increasing resolution of digital cameras has made more and more apparent that there may, in fact, exist a residual focus error after the lens has executed the focus instruction sent to it by the body. SLR manufacturers have tried to compensate for that by introducing e.g. AF micro adjustment facilities in their bodies that allow the user to somewhat adjust for the focus veriability occuring with their specific camera and lens combinations.
Mirrorless cameras typically have a contrast-detect AF unit. In a CDAF system, the camera instructs the lens to “wobble” its position, and determines the direction and position of best focus by examining the actual picture sharpness variation at the chosen focus point. A CDAF system is thus, inherently, a “closed loop” system that repeats the “measure defocus and then move the lens” process until the best focus is actually achieved. Thus, CDAF doesn’t require gimmicks like AF microadjustments. A CDAF system also has a few disadvantages:
1) the iterative focus approach can make CDAF slow compared to PDAF systems
2) a lens designed for CDAF should have a focusing mechanism that efficiently supports the tiny to-and-fro movements required for the wobbling AF convergence process. Conventional, unit-focusing lenses where the entire optical system must be repeatedly accelerated from standstill, then accurately stopped, are diasdvantageous for CDAF, as their high inertia would require very powerful, energy-hungry motors to implement high-speed wobbling. CDAF lenses thus often have an internal focus lens design, where only a lightweight, and therefore light inertia lens group is driven longitudinally by a linear AF motor.
At the Photokina a few weeks ago, I had an interesting chat with someone at the Sony stand about that lens.
I asked him whether the ring USM was used only with PDAF-capable bodies like the Sony A7RII, A7II, A6500, A6300 etc., and that the linear motor near the rear of the lens was intended for CDAF-only bodies like the A7R and A7S. He explained that it wasn’t the case, and that the Sony 70-200 GM actually uses the two motors even with CDAF bodies. The linear motor is used, he said, to implement a novel, electromagnetically- instead of helicoid-driven “floating element” aberration correction system. Floating elements typically implement a tiny variation of the distance between select lens elements in what is, otherwise, a conventional unit-focusing lens. This variation is designed e.g. to compensate the subtle aberration pattern changes that occur e.g. between infinity and minimal focus distances. As the minute lens movements require more complex, and very accurate focusing helicoids, floating elements were typically implemented only on high-end, expensive lenses — e.g. 28/1.4, 85/1.2 etc.
The 70-200 GM takes advantage of the fact that all Sony bodies, even if they have PDAF sensors, can use CDAF to optimalize the focus point. The ring USM, linear motor and camera CDAF thus work in concert to achieve the best focus position and minimize aberrations by driving the “floating element” using a linear motor, in a “closed loop” fashion, using feedback from the camera body.
I think one consequence might be that an optical bench would have to accurately emulate the CDAF-like “closed loop” mechanism to tell the 70-200 GM’s linear motor-driven floating elements to “wobble” and slide to the proper position required to minimize the residual aberrations at the selected focus distance — e.g. infinity…
HF ·
I don’t see the purpose of writing the many obvious things down here. In addition regarding open-loop look here:
https://www.dpreview.com/articles/5402438893/busted-the-myth-of-open-loop-phase-detection-autofocus
Peter Weir ·
That article would have a bit more credibility if it addressed /why/ an AF microadjustment-like calibration is necessary if an SLR AF system is, as the article author claims without any real substantiation, closed loop.
Unfortunately, the only way the article author addresses the issue is thusly:
“[..] I’m not trying to explain why a closed-loop system requires lens calibration for accuracy. For the purpose of this article, I accept the fact that lens calibration is essential for accurate focus, whether the system is open or closed-loop.”
Nice way to elude the ESSENTIAL issue. The author apparently doesn’t even realize that the “closed loop” system he describes can’t even detect that actual, accurate focus is achieved, hence the need for “solutions” like “calibration”, “offset”, “microadjustments” to compensate e.g. for lens variations.
What the article author says, in essence, is that “if we ignore the accuracy errors (which require e.g. calibration to be eliminated), then there are no accuracy errors in a SLR AF system. Ergo, as we decree that there are no accuracy errors, the system must be closed loop”.
#FAIL
HF ·
Doesn’t matter. Based on the input data it achieves the goal by using information in a feedback loop to achieve focus. Whether this is the true focal plane or not is irrelevant for the method being in use.
The exact focus plane can’t be found MILCs, too. Focus is where our numerical algorithms think maximum contrast is located. Depending on the size of your focus point, this changes. Depending on light, this changes (esp. if noise creeps in). Lens motors are not perfect, too. People in contact with camera developers confirmed the closed loop approach.
Marianne Oelund ·
PDAF servo is closed-loop. Period.
Without that, the camera would be unable to detect subject movement during focus, which modern PDAF systems are designed to do, even in single-shot or non-servo mode.
Further, anyone who has carefully observed the focus-drive behavior that occurs during repeated testing, knows that additional movements can and often do occur to further refine focus after the initial movement. No controls engineer (which I have been for several decades) in their right mind would intentionally forego valuable information which can be so easily obtained while the focus group is moving; one does not place a powerful processor into a long wait loop when there is useful activity it can perform.
Finally, in the case of large initial mis-focus, it is implausible that any PDAF system could initially arrive at a sufficiently accurate value for the lens command, due to blurring of the images at the AF sensor.
The existence or non-existence of an AF micro-adjustment feature is an entirely separate consideration, with no bearing on the control law design other than to add a bias to the error term.
Peter Weir ·
Marianne, I have enormous respect for you, based on the remarkable competence and intelligence evidenced in your posts on DPreview over the years.
When you write something, I thus sit down and take notice, as it’s quite clear to me that you’ve got what it takes to back up what you write.
In this case, I humbly submit that we might *both* be correct, in the sense that a modern DSLR might be performing both closed-loop and open-loop AF.
The Abstract section from US patent #6,603,929 dated Aug 5, 2003 hints at how such a scheme could be implemented:
“The camera has two focus detection units: the first is a low-precision detector, and the second a high-precision one. The control circuit performs closed loop control for driving the stepping motor [..] when using the information from the first (low-precision) focus detection unit. and performs open loop control for driving the stepping motor when using the information from the second (high-precision) focus detection unit.”
https://patents.google.com/patent/US6603929B2/en
This Lensrentals blog article
https://www.lensrentals.com/blog/2012/08/autofocus-reality-part-3b-canon-cameras/
shows that the AF variance of many (most ?) Canon DSLRs is higher than what you can achieve using contrast detect of careful, live-view manual focusing.
Closed loop control is essentially a process designed to measure and /eliminate/ errors.
The fact that, when pixel-peeping hi-res images, there’s a significant residual, uncontrollable variance — i.e. random residual error — in the focusing results of many Canon DSLRs would seem to indicate that the “measure, minimize / eliminate, repeat” looping part required for high-precision focusing doesn’t seem to exist on many Canon cameras, and that therefore, the system should rather be called “open loop”.
Nikon-brand cameras, of course, might be behaving differently.
Marianne Oelund ·
That’s correct that Canon offer several different levels of precision in their AF sensors, and I know from experience with the 1D Mk III that I owned back in 2007, that their AF system routinely uses information from more than one level of detector.
Reading into the description you found, I would take it to mean that the final focus movement based on the high-precision detector – once near-focus has been achieved with the low-precision detector – is done open-loop. This is reasonable, as the movement needed at that point would be very small, high detail is available to the detector, and the calculation can be done with sufficient precision. There must be a final point which the system arrives at, which it deems “good enough” that no further refinement is of value. Relying on the high-precision detector to determine that point makes sense.
Nikon have not yet produced an AF system with more than one level of detector precision – other than a small modification which “stretches” some of the f/5.6 detectors to see into the f/8 circle as well.
On a slightly different topic, the reason why AF fine tuning is needed, has nothing to do with the servo loop operation. It is down to second-order optical effects such as aberrations and distortion. The AF optics have no “zoom” capability to accommodate variations between lens designs such as exit pupil distance, and distortions inherent in wide-angle lenses cause difficult design issues for the AF system. The AF system uses a parallel optical path which must be artificially aligned to “sync” to the behavior of the imaging path, and due to compromises in its design, there will be discrepancies which are sometimes significant. Exacerbating those difficulties, is the fact that optical elements used in the AF module are low quality molded plastic exhibiting high levels of CA and spherical aberrations. All things considered, it is amazing that PDAF can work as well as it does.
Roger Cicala ·
Thank you Peter. That’s exactly the kind of thing I considered when I decided to hold those results. Our process, I THINK, should be accurate (it is for other Sony EM focus lenses): We set the lens to infinity focus via the camera, then fine-tune focus with the optical bench. It works fine for the others. But the others aren’t twin motor so there is that question.
Ed Bambrick ·
So twice the likelihood of something going wrong and errors. Well done Sony!
Ross Hamamura ·
Roger , did you Share the GM vs FL ED … MTF
at 200mm f2.8 somewhere ?
Roger Cicala ·
Let's kind of keep this here for right now - I haven't published it because I'm still a little uncertain about the results. Sony has suggested that a change in cover glass thickness might improve the results some. Not dramatically, but some, off-axis. This lens also has two focusing motors and we have to focus it electronically via a camera to test. I'm not absolutely certain that 'setting it at infinity focus on the camera' and 'manually focusing on an object at infinity' are absolutely the same. So take these lab results with a grain of salt. On the other hand, they do seem to agree with what we see in real world results.
I try hard not to put out anything until I'm just absolutely certain our results are correct. We're doing some stuff here that's pretty cutting edge, honestly. No one does 4 rotation MTFs, for example. I'm pretty certain these are good results, but not absolutely certain. So I'll post them in this discussion on my own site but I'd rather not see them reproduced all over the internet yet.
Roger
Marlon Richardson ·
Wow, this is more complicated than you've let on. I agree we should wait till you can do a test that you feel good about standing behind.
Roger Cicala ·
I think there's another thing worth considering here that I didn't mention. An early batch of 70-200 GMs came out and then the supply seemed to dry up. I honestly will want to test another batch of 10 later before I make any conclusions. And just so I'm clear, I have absolutely no knowledge of why availability seems to have slowed to a trickle. But just because it's the way I'm wired, I'll want to test another 10 copies after we get resupplied before I make any definitive statements.
Peter Kelly ·
"I'd rather not see them reproduced all over the internet yet"?
I think your hope is about as likely as emptying a bucket of mercury with a sieve!
Peter Weir ·
The Sony 70-200 GM might be a quite difficult lens to properly test on an optical bench.
As you mention, the GM contains two focusing motors:
1) a ring ultrasonic motor, similar e.g. to Canon's USM, driving lens elements located near the front of the lens
2) a linear motor driving lens elements located closer to the bayonet mount.
In a phase-detect AF system, the body's AF unit determines the lateral optical distance — a.k.a. "phase" — between the current lens focus position and the desired focus position, and then orders the lens to move that distance. The lens's built-in AF controller translates that lateral distance information into the concrete, focal distance-dependent mechanical longitudinal movement of the lens elements that would, in theory, cancel the lateral phase difference.
From a control theory point of view, a SLR's PDAF system is an "open loop" system, where the SLR doesn't check if the distance moved by the lens elements actually achieves the best focus. The camera and lens system just assume that the initial phase distance information is always translated into a perfectly matching lens movement. That assumption was maybe sufficient during the film era, but the increasing resolution of digital cameras has made more and more apparent that there may, in fact, exist a residual focus error after the lens has executed the focus instruction sent to it by the body. SLR manufacturers have tried to compensate for that by introducing e.g. AF micro adjustment facilities in their bodies that allow the user to somewhat adjust for the focus veriability occuring with their specific camera and lens combinations.
Mirrorless cameras typically have a contrast-detect AF unit. In a CDAF system, the camera instructs the lens to "wobble" its position, and determines the direction and position of best focus by examining the actual picture sharpness variation at the chosen focus point. A CDAF system is thus, inherently, a "closed loop" system that repeats the "measure defocus and then move the lens" process until the best focus is actually achieved. Thus, CDAF doesn't require gimmicks like AF microadjustments. A CDAF system also has a few disadvantages:
1) the iterative focus approach can make CDAF slow compared to PDAF systems
2) a lens designed for CDAF should have a focusing mechanism that efficiently supports the tiny to-and-fro movements required for the wobbling AF convergence process. Conventional, unit-focusing lenses where the entire optical system must be repeatedly accelerated from standstill, then accurately stopped, are diasdvantageous for CDAF, as their high inertia would require very powerful, energy-hungry motors to implement high-speed wobbling. CDAF lenses thus often have an internal focus lens design, where only a lightweight, and therefore light inertia lens group is driven longitudinally by a linear AF motor.
At the Photokina a few weeks ago, I had an interesting chat with someone at the Sony stand about that lens.
I asked him whether the ring USM was used only with PDAF-capable bodies like the Sony A7RII, A7II, A6500, A6300 etc., and that the linear motor near the rear of the lens was intended for CDAF-only bodies like the A7R and A7S. He explained that it wasn't the case, and that the Sony 70-200 GM actually uses the two motors even with CDAF bodies. The linear motor is used, he said, to implement a novel, electromagnetically- instead of helicoid-driven "floating element" aberration correction system. Floating elements typically implement a tiny variation of the distance between select lens elements in what is, otherwise, a conventional unit-focusing lens. This variation is designed e.g. to compensate the subtle aberration pattern changes that occur e.g. between infinity and minimal focus distances. As the minute lens movements require more complex, and very accurate focusing helicoids, floating elements were typically implemented only on high-end, expensive lenses — e.g. 28/1.4, 85/1.2 etc.
The 70-200 GM takes advantage of the fact that all Sony bodies, even if they have PDAF sensors, can use CDAF to optimalize the focus point. The ring USM, linear motor and camera CDAF thus work in concert to achieve the best focus position and minimize aberrations by driving the "floating element" using a linear motor, in a "closed loop" fashion, using feedback from the camera body.
I think one consequence might be that an optical bench would have to accurately emulate the CDAF-like "closed loop" mechanism to tell the 70-200 GM's linear motor-driven floating elements to "wobble" and slide to the proper position required to minimize the residual aberrations at the selected focus distance — e.g. infinity...
HF ·
I don't see the purpose of writing the many obvious things down here. In addition regarding open-loop look here:
https://www.dpreview.com/ar...
Peter Weir ·
That article would have a bit more credibility if it addressed /why/ an AF microadjustment-like calibration is necessary if an SLR AF system is, as the article author claims without any real substantiation, closed loop.
Unfortunately, the only way the article author addresses the issue is thusly:
"[..] I'm not trying to explain why a closed-loop system requires lens calibration for accuracy. For the purpose of this article, I accept the fact that lens calibration is essential for accurate focus, whether the system is open or closed-loop."
Nice way to elude the ESSENTIAL issue. The author apparently doesn't even realize that the "closed loop" system he describes can't even detect that actual, accurate focus is achieved, hence the need for "solutions" like "calibration", "offset", "microadjustments" to compensate e.g. for lens variations.
What the article author says, in essence, is that "if we ignore the accuracy errors (which require e.g. calibration to be eliminated), then there are no accuracy errors in a SLR AF system. Ergo, as we decree that there are no accuracy errors, the system must be closed loop".
#FAIL
HF ·
Doesn't matter. Based on the input data it achieves the goal by using information in a feedback loop to achieve focus. Whether this is the true focal plane or not is irrelevant for the method being in use.
The exact focus plane can't be found MILCs, too. Focus is where our numerical algorithms think maximum contrast is located. Depending on the size of your focus point, this changes. Depending on light, this changes (esp. if noise creeps in). Lens motors are not perfect, too. People in contact with camera developers confirmed the closed loop approach.
Marianne Oelund ·
PDAF servo is closed-loop. Period.
Without that, the camera would be unable to detect subject movement during focus, which modern PDAF systems are designed to do, even in single-shot or non-servo mode.
Further, anyone who has carefully observed the focus-drive behavior that occurs during repeated testing, knows that additional movements can and often do occur to further refine focus after the initial movement. No controls engineer (which I have been for several decades) in their right mind would intentionally forego valuable information which can be so easily obtained while the focus group is moving; one does not place a powerful processor into a long wait loop when there is useful activity it can perform.
Finally, in the case of large initial mis-focus, it is implausible that any PDAF system could initially arrive at a sufficiently accurate value for the lens command, due to blurring of the images at the AF sensor.
The existence or non-existence of an AF micro-adjustment feature is an entirely separate consideration, with no bearing on the control law design other than to add a bias to the error term.
Peter Weir ·
Marianne, I have enormous respect for you, based on the remarkable competence and intelligence evidenced in your posts on DPreview over the years.
When you write something, I thus sit down and take notice, as it's quite clear to me that you've got what it takes to back up what you write.
In this case, I humbly submit that we might *both* be correct, in the sense that a modern DSLR might be performing both closed-loop and open-loop AF.
The Abstract section from US patent #6,603,929 dated Aug 5, 2003 hints at how such a scheme could be implemented:
"The camera has two focus detection units: the first is a low-precision detector, and the second a high-precision one. The control circuit performs closed loop control for driving the stepping motor [..] when using the information from the first (low-precision) focus detection unit. and performs open loop control for driving the stepping motor when using the information from the second (high-precision) focus detection unit."
https://patents.google.com/...
This Lensrentals blog article
https://www.lensrentals.com...
shows that the AF variance of many (most ?) Canon DSLRs is higher than what you can achieve using contrast detect of careful, live-view manual focusing.
Closed loop control is essentially a process designed to measure and /eliminate/ errors.
The fact that, when pixel-peeping hi-res images, there's a significant residual, uncontrollable variance — i.e. random residual error — in the focusing results of many Canon DSLRs would seem to indicate that the "measure, minimize / eliminate, repeat" looping part required for high-precision focusing doesn't seem to exist on many Canon cameras, and that therefore, the system should rather be called "open loop".
Nikon-brand cameras, of course, might be behaving differently.
Marianne Oelund ·
That's correct that Canon offer several different levels of precision in their AF sensors, and I know from experience with the 1D Mk III that I owned back in 2007, that their AF system routinely uses information from more than one level of detector.
Reading into the description you found, I would take it to mean that the final focus movement based on the high-precision detector - once near-focus has been achieved with the low-precision detector - is done open-loop. This is reasonable, as the movement needed at that point would be very small, high detail is available to the detector, and the calculation can be done with sufficient precision. There must be a final point which the system arrives at, which it deems "good enough" that no further refinement is of value. Relying on the high-precision detector to determine that point makes sense.
Nikon have not yet produced an AF system with more than one level of detector precision - other than a small modification which "stretches" some of the f/5.6 detectors to see into the f/8 circle as well.
On a slightly different topic, the reason why AF fine tuning is needed, has nothing to do with the servo loop operation. It is down to second-order optical effects such as aberrations and distortion. The AF optics have no "zoom" capability to accommodate variations between lens designs such as exit pupil distance, and distortions inherent in wide-angle lenses cause difficult design issues for the AF system. The AF system uses a parallel optical path which must be artificially aligned to "sync" to the behavior of the imaging path, and due to compromises in its design, there will be discrepancies which are sometimes significant. Exacerbating those difficulties, is the fact that optical elements used in the AF module are low quality molded plastic exhibiting high levels of CA and spherical aberrations. All things considered, it is amazing that PDAF can work as well as it does.
Roger Cicala ·
Thank you Peter. That's exactly the kind of thing I considered when I decided to hold those results. Our process, I THINK, should be accurate (it is for other Sony EM focus lenses): We set the lens to infinity focus via the camera, then fine-tune focus with the optical bench. It works fine for the others. But the others aren't twin motor so there is that question.
Arthur Meursault ·
So twice the likelihood of something going wrong and errors. Well done Sony!
Ross Hamamura ·
Roger , did you Share the GM vs FL ED ... MTF
at 200mm f2.8 somewhere ?
Brian Smith ·
This is the first I am hearing about the 70-200 GM not being very good. We are talking about the brand new Sony 70-200 2.8 GM, correct? The lens that is impossible to find at this point.
Overall the GM does not hold a candle to the Canon and Nikon? That’s disappointing as I have been waiting for a preorder to arrive for 3 months now!
I think a lot of us would be interested in a blog post about the GM specifically, with your thoughts.
Will Humber ·
There have been a few threads over on DPReview that showed fairly disappointing results but I had chalked that up to the DPReview compression and/or poor technique on the part of the reviewers. Guess it might actually be from the lens, =(.
Brian Smith ·
I just started a thread basically stating the exact same thing. I chalked it up to user error or internet compression when posting. Turns out the subpar images may very well be the lens itself.
I really hope Roger does a blog or comments in more detail. I’d like his take on how it compares to the Canon and Nikon.
Brian Smith ·
I just started a thread basically stating the exact same thing. I chalked it up to user error or internet compression when posting. Turns out the subpar images may very well be the lens itself.
I really hope Roger does a blog or comments in more detail. I'd like his take on how it compares to the Canon and Nikon.
Brian Smith ·
Removed.
Ed Bambrick ·
Roger, why not appease the prime fans and compare the new Nikon with the usual suspects?
Otus 85mm, the new Nikon 105, the Apo 135 and the Nikon 200mm. Maybe some of us are willing to carry 20 lbs of lenses.
Ross Hamamura ·
Roger Cicala, because of the a99II … I am curious about the 70-200mm f2.8 A-Mount SSM2 at … 200mm f2.8 because I am one of those that live a lot at that setting. 🙂
Have a LINK to that MTF ?
Ed Bambrick ·
The f4 version of the Sony lens was even worse.
Said AZIZI ·
A comparison with the G Master will be very welcomed.
Thank you Roger !
Junz Inc ·
Yes, I also hope to see that Roger.
Thank you.
Roger Cicala ·
Well, it's not pretty. The 70-200mm https://uploads.disquscdn.c... https://uploads.disquscdn.c... https://uploads.disquscdn.c... G Master isn't really all that great.
Brian Smith ·
This is the first I am hearing about the 70-200 GM not being very good. We are talking about the brand new Sony 70-200 2.8 GM, correct? The lens that is impossible to find at this point.
Overall the GM does not hold a candle to the Canon and Nikon? That's disappointing as I have been waiting for a preorder to arrive for 3 months now!
I think a lot of us would be interested in a blog post about the GM specifically, with your thoughts.
Arthur Meursault ·
Roger, why not appease the prime fans and compare the new Nikon with the usual suspects?
Otus 85mm, the new Nikon 105, the Apo 135 and the Nikon 200mm. Maybe some of us are willing to carry 20 lbs of lenses.
Ross Hamamura ·
Roger Cicala, because of the a99II ... I am curious about the 70-200mm f2.8 A-Mount SSM2 at ... 200mm f2.8 because I am one of those that live a lot at that setting. :)
Have a LINK to that MTF ?
Arthur Meursault ·
The f4 version of the Sony lens was even worse.
Piotr Krochmal ·
Thanks!. I have question as Pentax user. 15-30 and 24-70 are pentamron lenses. There are some myth that they are better since lack of floating VR element and better pentax coating. It will be nice to check if its right.
But most interesting are 70-200 and 150-450 lenses as they are fully pentax optic.
What about that tests? I’m very keen on your scientific approach, so I’m waiting 🙂
FantasticMrFox ·
I second this – ePhotozine tests showed quite a difference in optical performance between the Tamron and Pentax/Tamron version of the lenses, would be interesting to see whether it’s just sample variation or whether there’s something to it.
Roger Cicala ·
Unfortunately we don’t have a Pentax mount for our optical bench, so I’m unable to do that comparison.
Piotr Krochmal ·
Roger nooo… Please make one. KAF is very easy. All lenses have mechanical aperture coupling. KAF 4 was shown but only one lens have electronical aperture. So fulfill that quest enter in world of pentax lenses.
Roger Cicala ·
Well, each mount has to be manufactured optically flat, etc. They cost about $1,000 each. And OLAF is, well non-profit would be a charitable description.
Will Humber ·
Wait, are you suggesting that a young company with a niche market and massive startup costs wouldn’t be an instant cash cow? =)
Roger Cicala ·
Yeah, except way more than that. It’s not even a cash cricket. But it’s a fine line, too. We’re getting more and more work but a lot of that work comes with nondisclosures.
Will Humber ·
Wait, are you suggesting that a young company with a niche market and massive startup costs wouldn't be an instant cash cow? =)
Roger Cicala ·
Yeah, except way more than that. It's not even a cash cricket. But it's a fine line, too. We're getting more and more work but a lot of that work comes with nondisclosures.
Piotr Krochmal ·
Roger, Pentax have one of best forum society. I will write there and I think we will find supporters 🙂
Piotr Krochmal ·
Roger, Pentax have one of best forum society. I will write there and I think we will find supporters :)
Piotr Krochmal ·
Thanks!. I have question as Pentax user. 15-30 and 24-70 are pentamron lenses. There are some myth that they are better since lack of floating VR element and better pentax coating. It will be nice to check if its right.
But most interesting are 70-200 and 150-450 lenses as they are fully pentax optic.
What about that tests? I'm very keen on your scientific approach, so I'm waiting :)
FantasticMrFox ·
I second this - ePhotozine tests showed quite a difference in optical performance between the Tamron and Pentax/Tamron version of the lenses, would be interesting to see whether it's just sample variation or whether there's something to it.
Roger Cicala ·
Unfortunately we don't have a Pentax mount for our optical bench, so I'm unable to do that comparison.
Piotr Krochmal ·
Roger nooo... Please make one. KAF is very easy. All lenses have mechanical aperture coupling. KAF 4 was shown but only one lens have electronical aperture. So fulfill that quest enter in world of pentax lenses.
Gandalf ·
it better be good for the price…
Gandalf ·
it better be good for the price...
Ireen APPLY it Yourself ·
They both seem to be very nice lenses. Shame they are WAY out of my league, from a price perspective.
Brad ·
Are there any other good products (3rd party or otherwise) in the 70-200 range for Sony? I was considering a Sony mirrorless but this report gives me pause.
bala ·
The beauty of Sony mirrorless cameras is that they, as some people put it, have a ‘universal mount’ – there are adapters available for pretty much any mount out there. Here’s one that’s been recently announced for Nikon mount
https://www.dpreview.com/news/4063594549/fotodiox-introduces-full-automatic-nikon-to-sony-e-adapter
So, now you can adapt this beautiful Nikon lens to your Sony camera.
Recently I bought a Canon EF 24-70mm f/2.8L II USM lens instead of Sony counterpart for my A7R2 because it’s cheaper and also I have a Canon 80D, which I can share this lens with now.
bala ·
The beauty of Sony mirrorless cameras is that they, as some people put it, have a 'universal mount' - there are adapters available for pretty much any mount out there. Here's one that's been recently announced for Nikon mount
https://www.dpreview.com/ne...
So, now you can adapt this beautiful Nikon lens to your Sony camera.
Recently I bought a Canon EF 24-70mm f/2.8L II USM lens instead of Sony counterpart for my A7R2 because it's cheaper and also I have a Canon 80D, which I can share this lens with now.
ohm image ·
Does anyone need another lens blog? Brilliant.
Oleg ·
Please test the new Sigmas!!!
Roger Cicala ·
I’m trying. Can’t keep the 85 in stock long enough to test it. 🙂
Andrej Belic ·
What’s up Roger, did You become a Nikon Fanboy? You are going to upset the Canonistas 😉
Just joking, I’m a huge fan of Your website and please go on with direct comprisons
Roger Cicala ·
I’m a ‘good lens’ Fanboy!
Andrej Belic ·
What's up Roger, did You become a Nikon Fanboy? You are going to upset the Canonistas ;-)
Just joking, I'm a huge fan of Your website and please go on with direct comprisons
Roger Cicala ·
I'm a 'good lens' Fanboy!
Phil ·
Hi Roger. Very useful information as usual. But what about the focus breathing? That was in my opinion the biggest flaw in the second version. Also, can you comment on the light transmission?
Thanks!
Phil
Roger Cicala ·
Phil, you know I don’t do like actual useful tests 🙂 But I bet 6,234 lens reviewers are gong to measure that to the nanometer.
Phil ·
Hi Roger. Very useful information as usual. But what about the focus breathing? That was in my opinion the biggest flaw in the second version. Also, can you comment on the light transmission?
Thanks!
Phil
Roger Cicala ·
Phil, you know I don't do like actual useful tests :-) But I bet 6,234 lens reviewers are gong to measure that to the nanometer.
bdbender4 ·
No offense, but I just wonder why anyone who has been here before asks: Hey Roger, can you test this that and the other thing? Sony fans, I feel your pain, but this is a Nikon test.
Søren Stærke ·
Nice comparisons Roger, some really stunning results come to think of it.
Roger Cicala ·
I'm trying. Can't keep the 85 in stock long enough to test it. :-)
bdbender4 ·
Sony. Sony Sony. Sony Sony Sony.
I’ll say it a little less politely than Roger did, below. This is a Nikon test. If you have been here before, you know that jumping on the bandwagon to ask Roger to test this that and the other thing is just not helpful, and can preclude a useful discussion of the test itself. So please don’t.
“Hey Roger, can you test the Porsche Carrera G? How about Swarovski binoculars, in comparison to Zeiss?”
Roger Cicala ·
Let the record show I am completely willing to test a Porsche Carrera G. Porsche? Are you listening Porsche?
bdbender4 ·
Hmm, well a Porsche test would have to be in secret.
Hypothetical: I set up a test with Porsche for you, and of course I would come along (we’re talking REALLY hypothetical here, I am not in any position to do that).
We get back to the showroom, and try to have a conversation about what we just found out. If it wasn’t in secret, a whole crowd of BMW fans (and a few Alfa die-hards) would come flooding into the Porsche showroom, and start having a loud unending ongoing conversation about BMWs, including really geeky technical BMW cylinder head details. The Alfa folks would try to be heard saying “hey, Alfa makes the best cylinder heads and has for years and years”.
This would make it darn tough to have a useful Porsche conversation. Just sayin’. 😉
bdbender4 ·
Hmm, well a Porsche test would have to be in secret.
Hypothetical: I set up a test with Porsche for you, and of course I would come along (we're talking REALLY hypothetical here, I am not in any position to do that).
We get back to the showroom, and try to have a conversation about what we just found out. If it wasn't in secret, a whole crowd of BMW fans (and a few Alfa die-hards) would come flooding into the Porsche showroom, and start having a loud unending ongoing conversation about BMWs, including really geeky technical BMW cylinder head details. The Alfa folks would try to be heard saying "hey, Alfa makes the best cylinder heads and has for years and years".
This would make it darn tough to have a useful Porsche conversation. Just sayin'. ;)
Jonas ·
Is your optical bench equipped with a Porsche mount?
bdbender4 ·
Sony. Sony Sony. Sony Sony Sony.
I'll say it a little less politely than Roger did, below. This is a Nikon test. If you have been here before, you know that jumping on the bandwagon to ask Roger to test this that and the other thing is just not helpful, and can preclude a useful discussion of the test itself. So please don't.
"Hey Roger, can you test the Porsche Carrera G? How about Swarovski binoculars, in comparison to Zeiss?"
Carleton Foxx ·
One of the characteristics of a great lens is the clean, gentle way that it handles the falloff between in-focus and out-of-focus; especially when it comes to faces. Is there a way to tell from these charts which lenses might possess that quality?
Roger Cicala ·
You can get a bit of an idea from the field of focus graphs – the color fall off is a reasonable representation of how sharpness falls off.
Carleton Foxx ·
You are like the Galen of the optical world—you heal the worried souls of photographers as you deftly slice away the tumors of internet folklore and lance the boils of outdated belief. When are you going to open your lens lamasery so that we can all come and sit around in togas and drink wine out of rams’ horns while you sing Zernike polynomials set to music.
Roger Cicala ·
TOGA!! TOGA!!! TOGA!!!!!
Roger Cicala ·
You can get a bit of an idea from the field of focus graphs - the color fall off is a reasonable representation of how sharpness falls off.
Carleton Foxx ·
You are like the Galen of the optical world—you heal the worried souls of photographers as you deftly slice away the tumors of internet folklore and lance the boils of outdated belief. When are you going to open your lens lamasery so that we can all come and sit around in togas and drink wine out of rams' horns while you sing Zernike polynomials set to music.
Lee ·
Thanks for the update, Roger! Quite the dramatic difference, that’s always exciting to see. I was expecting progress more in the “small but definite” range (if not the “hmm I guess it’s better in some ways? *shrug* ” territory of the 24-70E).
Nikon does seem to be on a good run at the moment. Maybe the Art series and the 35L II scared them straight.
Lee ·
Thanks for the update, Roger! Quite the dramatic difference, that's always exciting to see. I was expecting progress more in the "small but definite" range (if not the "hmm I guess it's better in some ways? *shrug* " territory of the 24-70E).
Nikon does seem to be on a good run at the moment. Maybe the Art series and the 35L II scared them straight.
donpnz ·
Thanks Roger! I’m saving the $$. Happy days..Nikon getting something right
donpnz ·
Thanks Roger! I'm saving the $$. Happy days..Nikon getting something right
Ryan Mack ·
Would love to see how the new 70-200 stacks with various TCs. I’m most curious to see how the 70-200E + TC-20E III compares to the 80-400 or 200-500.
Ryan Mack ·
Would love to see how the new 70-200 stacks with various TCs. I'm most curious to see how the 70-200E + TC-20E III compares to the 80-400 or 200-500.
Yangyang Zhang ·
Are you planning to review the Tamron 70-200 G2 and make the comparison between these two lenses?
Yangyang Zhang ·
Are you planning to review the Tamron 70-200 G2 and make the comparison between these two lenses?
Ben Cushwa ·
Hey Roger, have you had a chance to get one apart yet? I’m really curious to see if it lives up to Nikon’s claimed level of weather sealing.
Thanks!
Ben Cushwa ·
Hey Roger, have you had a chance to get one apart yet? I'm really curious to see if it lives up to Nikon's claimed level of weather sealing.
Thanks!
Username_Here ·
I don’t read MTF charts enough to know off hand if they’re always done at a specific f/stop or opening diameter, or specific DoF? Maybe I’m blind, but I can’t find the f/stop used mentioned anywhere… Also, is this at infinity? How do they vary with focal distance since I assume they’re all internally focused?
Username_Here ·
I don't read MTF charts enough to know off hand if they're always done at a specific f/stop or opening diameter, or specific DoF? Maybe I'm blind, but I can't find the f/stop used mentioned anywhere... Also, is this at infinity? How do they vary with focal distance since I assume they're all internally focused?
Hans J ·
wow cool
Hans ·
wow cool