There will be no mirrorless camera body announced ahead of Photokina

[fullstop]

So any notion that adapting for mirrorless will crush your L lens AF speed simply isn't substantiated in LiveView on (say) a 5D4 or 6D2 today.

it is substantiated by the loss of AF speed and performance in Live View [vs. Phase-AF] in both 6D2 and 5D4. and all other Canon DSLRS, with DP-AF and without.

 

[scyrene]

Really, they need to get away from their thinking that releasing one product cannibalizes the other. Does any other consumer electronics company think this way? Obviously not.

We don't know they actually think this at all.

especially if you have a client that looks at the models you're using and insists that you have "up-to-date" equipment

It depresses me that some people are like this - they should be judging a photographer based on output, not equipment.

 

[3kramd5]

it is substantiated by the loss of AF speed and performance in Live View [vs. Phase-AF] in both 6D2 and 5D4. and all other Canon DSLRS, with DP-AF and without.

How are you concluding that’s due to the lens, and not the entirely different AF system?

You have two constants (camera interface to lens, lens) and one variable (AF system used by camera), but blame one of the constants for a performance hit when changing the variable.

I don’t see why the camera would drive the lens slower because it derives its forward or backward signal from a different data source.

I would guess, but not claim substantiation, that there’s some of both going on. The motors in some existing lenses are not optimized to go back-and-forth as frequently as a scene based AF system is capable of analyzing (since the AF sensor is never blacked out). Additionally, and perhaps more importantly, on sensor AF is constrained by the size, geometry, construction, and color filtering of the imaging pixels, whereas off sensor is optimized for AF (big pixels which aren’t blind to most of the spectrum), so it stands to reason OSPDAF will take more refinement. This may be the reason for the common practice of starting with phase and finishing with contrast, a step not done in off-sensor PFAF methodology.

 

[neuroanatomist]

M

How are you concluding that’s due to the lens, and not the entirely different AF system?

You have two constants (camera interface to lens, lens) and one variable (AF system used by camera), but blame one of the constants for a performance hit when changing the variable.

I don’t see why the camera would drive the lens slower because it derives its forward or backward signal from a different data source.

I would guess, but not claim substantiation, that there’s some of both going on. The motors in some existing lenses are not optimized to go back-and-forth as frequently as a scene based AF system is capable of analyzing (since the AF sensor is never blacked out). Additionally, and perhaps more importantly, on sensor AF is constrained by the size, geometry, construction, and color filtering of the imaging pixels, whereas off sensor is optimized for AF (big pixels which aren’t blind to most of the spectrum), so it stands to reason OSPDAF will take more refinement. This may be the reason for the common practice of starting with phase and finishing with contrast, a step not done in off-sensor PFAF methodology.

Might as well try to pull a Möbius strip into a straight line.

 

[fullstop]

@ekramd5
(almost all) EF lenses entire AF drive system - hardware and firmware/control logic is not designed/built for OS PDAF, including DP-AF (live view/mirrorfree) operation.

exceptions: a few more recent EF lenses, notably
a. STM drive (40/2.8, 50/1.8)
b. Nano-USM (70-300 II).
c. latest EF (L) lens releases ("Mk. III") will - hopefully - be prepared in hardware and firmware (upgradable).

There is undeniably a difference in AF performance for all/most EF lenses when used in live view mode on a DSLR (including DP-AF) vs. viewfinder operation. if it was not (also) lens related but solely "a camera thing" there should be no difference. But there is.

It could be tested by checking in a standardized capture situation involving acquiring and tracking a moving subject in the frame on a Canon DPAF-DSLR (eg 5D4) "if and how big" an AF performance difference exists between
1) viewfinder operation (detached Phase AF sensor) and
2) DPAF live view operation
for various EF lenses with different types of AF drives (micro motor, older/newer design Ring USM, Nano USM, ). multiple body and lens copies. Could be a nice "geeky" project for Roger Cicala.

until proven otherwise, i go with the assumption and expectation, that new lenses (presumably in new mount) will have superior AF performance on future Canon mirrorfree FF cameras compared to "legacy" EF lenses. not because of a hollow tube adapter, but because of limitations of (most?) EF lenses AF drive system with regards to mirrorfree operation.

 

[fullstop]

Might as well try to pull a Möbius strip into a straight line.

stylistically, i like your really elegant parable.

but ... don't agree at all with its ironic content in regard to my thoughts on "EF lenses AF performance limitations in live view / mirrorfree operation".

 

[ahsanford]

There is undeniably a difference in AF performance for all/most EF lenses when used in live view mode on a DSLR (including DP-AF) vs. viewfinder operation. if it was not (also) lens related but solely "a camera thing" there should be no difference. But there is.

Ah. I was referring to:

Existing EF lens on recent camera (w/DPAF) in Liveview vs. Existing EF lens on (presumed to have DPAF) FF mirrorless camera (effectively: in Liveview)

FF mirrorless should be at least as quick/responsive/accurate as the SLR in LiveView.

Yes, AF time could surely be impacted by walking away from the OVF / AF array setup. I don't think anyone's contesting that.

Now -- getting to your 'legacy' assertion -- could Canon make lenses that are optimized for DPAF? I guess it's plausible. But it remains to be seen that Canon new mirrorless only lenses will outperform existing EF glass with respect to AF, optimized or not. Canon might just put out STM glass that, however optimized, is no threat to surpassing L ring USM glass. Or Canon might not optimize a thing and just stick with tried-and-true DPAF focusing instructions and lens focusing mechanisms for cost or time-to-market reasons. We won't know until they launch this thing.

My point: the whole 'legacy' position implies that the new glass we'll get with this mirrorless system will surpass existing lenses for performance. That might be a bad assumption. Canon may opt against going super premium with glass when those lenses already exist. They may keep the native mirrorless lens portfolio nice but affordable and point us to the adaptor for the good stuff.

- A

 

[fullstop]

well, maybe i am too optimistic re. Canon. but i do expect future "mirrorfree" Canon lenses ("EF-X") to have AF performance superior to (most) legacy EF lenses.

if not, i'd be the first one to not buy new lenses but use my existing EF glass via adapter.

but, Canon presumably wants to sell lots and lots of new lenses. so they gotta offer some advantages over existing glass. ideally in all categories: IQ, size/weight, AF performance, other functionality (eg communication protocol, wireless ETTL, subject focus distance information, etc.), handling (eg see Nikon Z lenses customer re-assignable focus ring!), wheathersealing etc.

speaking of Nikon Z: first indications are, that AF performance of Z-lenses on Z6/Z7 cameras is not as good as F-lenses on Nikon DSLRs (in viewfinder operation). it appears, this is due to "general limitations/inferiority" of Nikon's (non DPAF) entire liveview implementation. if confirmed in upcoming reviews and real life, it will not help Nikon's sales of expensive! Z lenses and Z system.

one would hope Canon to be "more innovative" and being able to offer new mirrorfree camera system and native lenses that surpass EF lenses' AF performance ... not only compared to "EF in DSLR DP-AF live view mode," but also "EF in DSLR PD-AF viewfinder mode".

if mirrorfree cameras without mechanical mirrorslapping limitations allow for extremely high frame rates then AF system needs "an upgrade" over DSLRs as well, no?

 

[Kit.]

(almost all) EF lenses entire AF drive system - hardware and firmware/control logic is not designed/built for OS PDAF, including DP-AF (live view/mirrorfree) operation.

exceptions: a few more recent EF lenses, notably
a. STM drive (40/2.8, 50/1.8)
b. Nano-USM (70-300 II).
c. latest EF (L) lens releases ("Mk. III") will - hopefully - be prepared in hardware and firmware (upgradable).

There is undeniably a difference in AF performance for all/most EF lenses when used in live view mode on a DSLR (including DP-AF) vs. viewfinder operation. if it was not (also) lens related but solely "a camera thing" there should be no difference. But there is.

Why do you think there should be no difference if DPAF were a worse focusing system?

Have you actually compared focusing speeds of 70-300 II and 70-300L with DPAF, or are you just imagining that 70-300 II would be faster?

until proven otherwise, i go with the assumption and expectation, that new lenses (presumably in new mount) will have superior AF performance on future Canon mirrorfree FF cameras compared to "legacy" EF lenses. not because of a hollow tube adapter, but because of limitations of (most?) EF lenses AF drive system with regards to mirrorfree operation.

"until proven otherwise, i go with the assumption and expectation, that the laws of physics do not apply to Canon's FF mirrorless"

 

[fullstop]

Why do you think there should be no difference if DPAF were a worse focusing system?

? don't understand what you're after here? I did never state "DP-AF to be a worse focusing system". All I am hearing/reading, is that EF lenses on DSLRs do not have equal/as good AF performance in LiveView mode compared to viewfinder mode with separate phase AF sensor - which is what those EF lenses were designed for. It is very well possible and even likely, that this AF performance delta is less pronounced with latest DP-AF implementations compared to previous CD-AF live view implementations.

Have you actually compared focusing speeds of 70-300 II and 70-300L with DPAF, or are you just imagining that 70-300 II would be faster?

nope. I don't own any of the 70-300 lenses and am not inclined to run such tests myself. But those 2 lenses, EF 70-300 IS [Ring USM] and 70-300 IS II [Nano USM] would certainly be a good pair to include in such testing.

I am not "imagining" [never] or assuming anything. Call it a "hypothesis" to be tested. I just don't want to rule out, that the EF lens with Nano USM was designed by Canon with LiveView/mirrorfree DP-AF operation in mind and MAY have better AF performance, respectively "LESS of an AF penalty in LiveView" mode compared to viewfinder DSLR mode than "regular/older/legacy" EF glass.

"until proven otherwise, i go with the assumption and expectation, that the laws of physics do not apply to Canon's FF mirrorless"

dumb attempt at personal attack, ignored. Cut it out.

 

[3kramd5]

All I am hearing/reading, is that EF lenses on DSLRs do not have equal/as good AF performance in LiveView mode compared to viewfinder mode with separate phase AF sensor - which is what those EF lenses were designed for.

Right, and I asked why you don’t account for any of the inherent weaknesses of on sensor AF in making that comparison. You’re only looking at the lens, not the fundamentally different set of sensors. The separate AF sensor unit is tremendously advantaged relative to color imaging pixels.

At the risk of a car analogy, this is like putting a set of Corvette tires on a Malibu, and concluding that the reason the Corvette beats the Malibu in a drag race is that the borrowed tires weren’t designed for it.

until proven otherwise, i go with the assumption and expectation, that new lenses (presumably in new mount) will have superior AF performance on future Canon mirrorfree FF cameras compared to "legacy" EF lenses.

That’s probably fair, but it’s an entirely different assumption to the one you’re making above (namely, that DPAF and off-sensor AF are equivalent, and the only thing which would impact performance is lens design).

don't understand what you're after here? I did never state "DP-AF to be a worse focusing system".

I won’t use the words better or worse; they’re subjective. What off sensor PDAF is that DPAF is not, is dedicated and purpose-designed.

While DPAF is so far the most elegant solution I’ve seen to share resources, the focusing sensors are clearly impaired relative to an off-sensor unit (although less so than other on sensor methods which not only filter out all but a very narrow color band, but then blind half the filtered pixel).

 

[neuroanatomist]

There is undeniably a difference in AF performance for all/most EF lenses when used in live view mode on a DSLR (including DP-AF) vs. viewfinder operation. if it was not (also) lens related but solely "a camera thing" there should be no difference. But there is.

Your assumption is predicated on DPAF delivering equivalent performance as a dedicated PDAF system. Canon's marketing hype aside, there's no data to support that.

Not that data ever informs your opinions anyway.

 

[Ozarker]

===

My European sources say that the ONLY thing that's coming to Photokina 2018 from Canon is an announcement about a Canon Cinema EOS XC-15-style video camera with an interchangeble lens mount and a one inch sensor BUT full 4K imagery at decent bit rates useable for budget filmmakers. My sources are not sure whether this is just a simple product engineering announcement for an eventual full rollout at CES 2019, or whether this is the full camera introduction for sale in October/November. They are simply just not sure!

LOL! The same sources are laughing their butts off. They told me you are the easiest guy in the world to troll... and then you run with it.

 

[3kramd5]

He must be in Orange County, California. Lotsa know-it-all nuts there.

They tend to be pretty level-headed, Santa Ana notwithstanding. The know-it-allism increases as you go north.

 

[neuroanatomist]

They tend to be pretty level-headed, Santa Ana notwithstanding. The know-it-allism increases as you go north.

North or south, it's all the land of fruits and nuts.

 

[Don Haines]

? don't understand what you're after here? I did never state "DP-AF to be a worse focusing system". All I am hearing/reading, is that EF lenses on DSLRs do not have equal/as good AF performance in LiveView mode compared to viewfinder mode with separate phase AF sensor - which is what those EF lenses were designed for. It is very well possible and even likely, that this AF performance delta is less pronounced with latest DP-AF implementations compared to previous CD-AF live view implementations.

From what I am led to believe, the difference between systems is that DPAF can tell which direction you are out of focus, while the AF module in a mirrored camera can tell how far you are out of focus..... this gives two different methods for achieving focus....

With DPAF, you keep moving the lens element until you get focus, and then you stop. There may be a slight overshoot and you may have to reverse direction and return a few steps...

With an AF module, you know how far out of focus you are, and if you have the appropriate lens data, you can calculate how long you have to accelerate the lens elements and how long it will take to stop them, and then “jump” to the right focus, as opposed to moving to correct focus at a steady pace....... incidentally, this is also why native lenses AF faster than third party lenses.....

DPAF is slower, but because there is no need to align the AF sensor and calibrate all the lenses, it is more accurate and you don’t have to AFMA lenses....

There is no universal A is better than B here..... it all depends on what you are doing.

 

[3kramd5]

North or south, it's all the land of fruits and nuts.

Hah, fair statement given who we elect.

 

[fullstop]

Your assumption is predicated on DPAF delivering equivalent performance as a dedicated PDAF system. Canon's marketing hype aside, there's no data to support that.

Would be interesting to see AF tests for STM lenses. Whether or not 40/2.8 or 50/1.8 have pretty much equal focus performance on say an 5D4 in viewfinder mode with separate Phase AF sensor and in DP-AF liveview ... or not. To find out whether lenses built to handle both camera-side AF methods, deliver equal performance, or are better or worse in one of those modes. If about equal and "regular" EF lenses not, it would indicate that it is - partly or mainly or exclusively - dependant on lenses' AF drive and communications capabilities.

Have not really searched for such tests/data, but not seen anything float by either.

 

[3kramd5]

Would be interesting to see AF tests for STM lenses. Whether or not 40/2.8 or 50/1.8 have pretty much equal focus performance on say an 5D4 in viewfinder mode with separate Phase AF sensor and in DP-AF liveview ... or not. To find out whether lenses built to handle both camera-side AF methods, deliver equal performance, or are better or worse in one of those modes. If about equal and "regular" EF lenses not, it would indicate that it is - partly or mainly or exclusively - dependant on lenses' AF drive and communications capabilities.

Have not really searched for such tests/data, but not seen anything float by either.

The camera doesn't drive the lenses with different commands or voltage when one switches from off sensor to on sensor. A USM lens may struggle to do the back-and-forth of CDAF as readily as a STM lens, but that doesn't really play in phase detect, where the distance is known by the spacing between where rays intersect the linepair.

While the lens electronics don't change, the AF system does. Occam's razor, my man. It's obviously not a sound deductive method, but in this case the simplest solution, that on-sensor AF is different from off-sensor AF, is also predictable by the nature of the sensors used. Forget their size, spacing, etc, and just consider the filters in front of the image sensor. In order to re-construct color, every pixel has bandpass filters which only allow through a narrow band of the spectrum (reg, green, or blue). Off-sensor AF doesn't have a color filter array; they see the entire spectrum and thus have significantly more light to work with.

From what I am led to believe, the difference between systems is that DPAF can tell which direction you are out of focus, while the AF module in a mirrored camera can tell how far you are out of focus..... this gives two different methods for achieving focus....

Both off- and on-sensor PDAF know direction.

 

[fullstop]

The camera doesn't drive the lenses with different commands or voltage when one switches from off sensor to on sensor. A USM lens may struggle to do the back-and-forth of CDAF as readily as a STM lens, but that doesn't really play in phase detect, where the distance is known by the spacing between where rays intersect the linepair.

While the lens electronics don't change, the AF system does. Occam's razor, my man. It's obviously not a sound deductive method, but in this case the simplest solution, that on-sensor AF is different from off-sensor AF, is also predictable by the nature of the sensors used. Forget their size, spacing, etc, and just consider the filters in front of the image sensor. In order to re-construct color, every pixel has bandpass filters which only allow through a narrow band of the spectrum (reg, green, or blue). Off-sensor AF doesn't have a color filter array; they see the entire spectrum and thus have significantly more light to work with.



Both off- and on-sensor PDAF know direction.