p.1 #1 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
In an discussion earlier today snapsy and I had the beginnings of an exchange about AF with the Z cameras. We both thought that AF-S mode in AF did not use contrast detect to fine tune the focus. This discussion led me to dig a little deeper and see what testing had been done on AF accuracy and precision in the Z cameras. After looking I did find some useful information on Jim Kasson's blog. It looks like snapsy and I were wrong about the use of contrast detect AF in AF-S mode. Here is what Jim Kasson has to say about the issue:
"Nikon has said that, for adapted lenses, the Z7 forgoes the CDAF trim phase at the end of the PDAF phase. Looking at these results, I’m having trouble believing that. [Nikon has apparently retracted that statement. So the behavior that I’m observing matches the PDAF-with-CDAF trim that is now officially the way that adapted lenses focus on the Z7.]"
A couple of things to notice in that blog post are that: 1) One of the clear advantages for mirrorless cameras is the accurate and precise AF it provides for static subjects in all but really bad light, and 2) even with an adapted lens the AF with the Z7 is very good.
Snapsy and I were also discussing whether AF-S or pinpoint AF would be better for lenses designed for mirrorless than they are for adapted F mount lenses. Jim doesn't have a lot of data on that but we can compare across blog posts the very good AF performance of the 58 f/1.4 in the link above with the AF of the 24-70 f/4 S at Jim's blog here:
Comparing across the blog posts we see that the 58 f/1.4 has very good to excellent AF accuracy and precision on the Z7, but the performance of the 24-70 f/4 which is designed for mirrorless is even better. Given that Jim provides standard deviations for the focus errors if we assume a normal distribution of errors we can estimate the percentage of time the AF errors will be bigger than the pixel pitch of the Z7. For the 58 f/1.4, it should be a bit more than 30 percent of the time and perhaps as much of 40 percent of the time. These errors won't be massive and on average will be pretty small but they are potentially visible and in effect reduce the resolution of the camera. In contrast the 24-70 f/4 should produce focus errors that are bigger than the pixel pitch of the camera less than 10% of the time. This is of course only two lenses, but it is consistent with reasoning I presented earlier that keeping the focussing groups small in lenses designed for mirrorless should lead to AF that is both more accurate and precise and that doing so can create difference that are potentially meaningful.
p.1 #2 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Having used the Z7 with native and adapted F mount lenses, I have to say that I was under the impression that it used PDAF-with-CDAF trim due to the fact that the focus accuracy was always better than that from my D850. Not that the D850's accuracy was bad, far from it, just that the Z7 generally seemed crisper more of the time thus making me believe that there was PDAF-with-CDAF trim. It is one of the number of things that makes me like using the Z7 over the D850.
p.1 #3 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Lance B wrote:
Having used the Z7 with native and adapted F mount lenses, I have to say that I was under the impression that it used PDAF-with-CDAF trim due to the fact that the focus accuracy was always better than that from my D850. Not that the D850's accuracy was bad, far from it, just that the Z7 generally seemed crisper more of the time thus making me believe that there was PDAF-with-CDAF trim. It is one of the number of things that makes me like using the Z7 over the D850.
Me too. I actually thought this was common knowledge. However I can’t tell you why I thought that. Subliminal Nikon marketing maybe
p.1 #4 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Steve Spencer wrote:
It looks like snapsy and I were wrong about the use of contrast detect AF in AF-S mode. Here is what Jim Kasson has to say about the issue:
"Nikon has said that, for adapted lenses, the Z7 forgoes the CDAF trim phase at the end of the PDAF phase. Looking at these results, I’m having trouble believing that. [Nikon has apparently retracted that statement. So the behavior that I’m observing matches the PDAF-with-CDAF trim that is now officially the way that adapted lenses focus on the Z7.]"
A couple of things to notice in that blog post are that: 1) One of the clear advantages for mirrorless cameras is the accurate and precise AF it provides for static subjects in all but really bad light, and 2) even with an adapted lens the AF with the Z7 is very good.
Snapsy and I were also discussing whether AF-S or pinpoint AF would be better for lenses designed for mirrorless than they are for adapted F mount lenses. Jim doesn't have a lot of data on that but we can compare across blog posts the very good AF performance of the 58 f/1.4 in the link above with the AF of the 24-70 f/4 S at Jim's blog here: ...Show more →
In that blog post Jim concluded the Z must be performing a CDAF confirmation cycle in AF-S based on how much more precise and accurate the Z was vs the D850. That criteria is not valid because it's imputing the implementation of the AF system based on how well it performs rather than on its observed physical behavior, with the stated assumption there couldn't be other reason the Z is more precise and accurate than the D850. There are other reasons, principally the absence of a separate optical path to the AF system. If we're going to impute the implementation an AF based on its relative performance then that performance needs to be compared within the same system rather than on two completely disparate platforms. Fortunately the Z offers such an opportunity - comparing it's Pinpoint AF mode, which by physical observation of lens movements during AF is clearly using CDAF, to the cameras AF-S mode. Jim performed such a test a month after the above blog post (which utilized the discovery that Pinpoint AF uses full CDAF, which wasn't known at the time of Jim's original blog post) and found Pinpiont AF was more precise than AF-S, at least as measured on the 105mm. That precision differential should be impossible if AF-S was actually performing a CDAF confirmation cycle, provided a properly aligned AF target that wouldn't unfairly disadvantage the larger AF point area of AF-S. Here's the link:
p.1 #5 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
In that blog post Jim concluded the Z must be performing a CDAF confirmation cycle in AF-S based on how much more precise and accurate the Z was vs the D850. That criteria is not valid because it's imputing the implementation of the AF system based on how well it performs rather than on its observed physical behavior, with the stated assumption there couldn't be other reason the Z is more precise and accurate than the D850. There are other reasons, principally the absence of a separate optical path to the AF system. If we're going to impute the implementation an AF based on its relative performance then that performance needs to be compared within the same system rather than on two completely disparate platforms. Fortunately the Z offers such an opportunity - comparing it's Pinpoint AF mode, which by physical observation of lens movements during AF is clearly using CDAF, to the cameras AF-S mode. Jim performed such a test a month after the above blog post (which utilized the discovery that Pinpoint AF uses full CDAF, which wasn't known at the time of Jim's original blog post) and found Pinpiont AF was more precise than AF-S, at least as measured on the 105mm. That precision differential should be impossible if AF-S was actually performing a CDAF confirmation cycle, provided a properly aligned AF target that wouldn't unfairly disadvantage the larger AF point area of AF-S. Here's the link:
And in fact Jim's assessment of a CDAF confirmation cycle is a bit more measured in that more recent blog post:
"Now we’ll look at AF-S in Single mode, which uses PDAF, possibly in conjunction with CDAF. "...Show more →
Jim also reversed his earlier position and said that officially there is CDAF trim with adapted lenses. I don't think his later use of the words possibly needs to be interpreted as yet a third position. My guess is that the issue is a bit more complicated as it may depend on the lens firmware, the camera firmware, and the capabilities of the given lens. IMO, Nikon would be foolish not to use CDAF trim when it will improve AF accuracy and precision and when it will have little influence on AF speed or for lenses for which AF speed is less likely to be an important issue, but I don't think any of us know how often that is. Ideally going forward Nikon will give us the option of turning on and off CDAF trim. In my view, we ought to be able to have total CDAF (which will be slow but have the greatest accuracy; the current pinpoint AF), total PDAF (which will likely be fastest but accuracy and precision will suffer at least to some extent) and start with PDAF but add CDAF trim (for most situations this would be my preferred mode balancing speed with accuracy and precision). No reason we shouldn't have options here. The default could be lens dependent.
All of that applies to adapted lenses, however, going forward I don't think it makes any sense but to build lenses in which you can optimize both AF speed and AF accuracy and precision and the way to do that is develop lenses with small focussing groups, wide communication paths, and fast motors so that these new lenses can use both PDAF and trim the focus with CDAF at least in some situations so that we can optimize speed, accuracy, and precision.
p.1 #6 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Agree that AF-S behavior on the Z might be different across different F-mount lenses, but I would have to see evidence of that rather than assume it to be true. As it stands I haven't seen any evidence of a CDAF confirmation cycle on any F-Mount lens I have used myself, nor have I seen a test by anyone else which proves it, while I have seen counter-evidence against it (Pinpiont vs AF-S precision).
I also agree it would be preferable to give users the option to enable a CDAF confirmation cycle for AF-S. However in the absence of such an option, considering how precise the Z's AF-S behavior is without a CDAF confirmation cycle then I agree with Nikon's decision to not have one. Not only would the CDAF cycle slow down AF acquisition but it also produces a jarring focus warble when looking through the viewfinder during AF, something I always found annoying on Sony's MILCs. IMO that warble greatly interferes with photographic experience of using Sony's cameras (which I still own btw).
As to whether or not future lens designs should be specifically optimized for CDAF, I agree it would be better than not optimizing for it but I don't really see a compelling case for it. As it stands Nikon isn't using CDAF in the normal course of AF on the Z, and for the few scenarios or users who might prefer CDAF the current non-optimized designs on F-mount lenses are perfectly adequate to the task.
p.1 #7 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
Agree that AF-S behavior on the Z might be different across different F-mount lenses, but I would have to see evidence of that rather than assume it to be true. As it stands I haven't seen any evidence of a CDAF confirmation cycle on any F-Mount lens I have used myself, nor have I seen a test by anyone else which proves it, while I have seen counter-evidence against it (Pinpiont vs AF-S precision).
I also agree it would be preferable to give users the option to enable a CDAF confirmation cycle for AF-S. However in the absence of such an option, considering how precise the Z's AF-S behavior is without a CDAF confirmation cycle then I agree with Nikon's decision to not have one. Not only would the CDAF cycle slow down AF acquisition but it also produces a jarring focus warble when looking through the viewfinder during AF, something I always found annoying on Sony's MILCs. IMO that warble greatly interferes with photographic experience of using Sony's cameras (which I still own btw).
As to whether or not future lens designs should be specifically optimized for CDAF, I agree it would be better than not optimizing for it but I don't really see a compelling case for it. As it stands Nikon isn't using CDAF in the normal course of AF on the Z, and for the few scenarios or users who might prefer CDAF the current non-optimized designs on F-mount lenses are perfectly adequate to the task. ...Show more →
I am confused. You state there is no proof that Nikon is using CDAF, and then go on to state they are not using CDAF. Where is the proof they are not using it? Why start with the assumption they are not using it?
p.1 #8 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
Agree that AF-S behavior on the Z might be different across different F-mount lenses, but I would have to see evidence of that rather than assume it to be true. As it stands I haven't seen any evidence of a CDAF confirmation cycle on any F-Mount lens I have used myself, nor have I seen a test by anyone else which proves it, while I have seen counter-evidence against it (Pinpiont vs AF-S precision).
I also agree it would be preferable to give users the option to enable a CDAF confirmation cycle for AF-S. However in the absence of such an option, considering how precise the Z's AF-S behavior is without a CDAF confirmation cycle then I agree with Nikon's decision to not have one. Not only would the CDAF cycle slow down AF acquisition but it also produces a jarring focus warble when looking through the viewfinder during AF, something I always found annoying on Sony's MILCs. IMO that warble greatly interferes with photographic experience of using Sony's cameras (which I still own btw).
As to whether or not future lens designs should be specifically optimized for CDAF, I agree it would be better than not optimizing for it but I don't really see a compelling case for it. As it stands Nikon isn't using CDAF in the normal course of AF on the Z, and for the few scenarios or users who might prefer CDAF the current non-optimized designs on F-mount lenses are perfectly adequate to the task. ...Show more →
Huh? This I how Nikon describes their AF on the Z7 product page:
"Using an AF algorithm optimized for its FX-format sensor, Z 7 automatically switches between focal-plane phase-detect AF and contrast-detect AF to fine-tune focus."
Now which lenses use this, which AF modes use this, and especially whether adapted lenses use this and if so do all of them or only some of them, that we don't know. What we do know is the system is designed with phase detect AF but tuned with contrast detect AF in mind. Now whether they have implemented that in every case is beside the point. The system is designed for CDAF tuning. So everyone should expect that when they design lenses for this system (i.e., for the Z cameras) they will design the lenses to allow them to effectively use this part of the system--they will design them with small focussing groups, with lots of communication protocols and fast motors to enhance AF.
Why do you say Nikon's AF system doesn't use "CDAF tuning in the normal course of AF," when their own product description seems to say otherwise?
p.1 #9 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
gdsf2 wrote:
I am confused. You state there is no proof that Nikon is using CDAF, and then go on to state they are not using CDAF. Where is the proof they are not using it? Why start with the assumption they are not using it?
The proof is twofold:
1) You can actually see when a camera is using a CDAF confirmation , especially on a slow-focusing lens. If you've ever used a Sony MILC this manifests as a focus warble during acquisition
2) Pinpoint AF, which is clearing using CDAF by the behavior exhibited by the lens, is more precise than AF-S based on MTF measurements performed by Jim Kasson. If AF-S were using a CDAF confirmation cycle the precision would be the same.
p.1 #10 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Steve Spencer wrote:
Why do you say Nikon's AF system doesn't use "CDAF tuning in the normal course of AF," when their own product description seems to say otherwise?
Because I don't rely on marketing departments for engineering information and because I can actually observe CDAF is not being employed.
p.1 #11 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
Because I don't rely on marketing departments for engineering information and because I can actually observe CDAF is not being employed.
So, you have tested every lens that Nikon makes and actually observed that none of them use CDAF tuning on your Z camera? I doubt you have and even if you did you might consider the possibility that you camera needs servicing.
p.1 #12 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
The proof is twofold:
1) You can actually see when a camera is using a CDAF confirmation , especially on a slow-focusing lens. If you've ever used a Sony MILC this manifests as a focus warble during acquisition
2) Pinpoint AF, which is clearing using CDAF by the behavior exhibited by the lens, is more precise than AF-S based on MTF measurements performed by Jim Kasson. If AF-S were using a CDAF confirmation cycle the precision would be the same.
First of all CDAF may not be used for focus acquisition and still be used. In fact, if CDAF is used for focus tuning it won't be used for focus acquisition.
Second, you have made an inappropriate inference that if CDAF were used for fine tuning that precision would be exactly the same and if CDAF were used solely for AF. That would depend on how the algorithm were written. It is certainly possible if not probably that using CDAF for the sole AF and using CDAF for fine tuning would have code using the CDAF in different ways. For example in CDAF only AF where accuracy and precision are paramount the CDAF cycle might check final processing 3, 4, or 5 times. Whereas when CDAF is used for fine tuning the CDAF cycle might only check final focus once. If that were the case you would expect exactly the results that Jim obtained. So, your inference that because these result are different AF-S must not be using CDAF fine tuning is based on the assumption that both modes of focussing are using CDAF in the same way. I think that assumption is invalid and therefore you inference is invalid.
Third, I am skeptical that you can tell that CDAF is used to fine tune by watching the lens. It is good to know how you have arrived at your conclusions. I think it is very flimsy reasoning and I am totally unconvinced.
p.1 #13 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Steve Spencer wrote:
So, you have tested every lens that Nikon makes and actually observed that none of them use CDAF tuning on your Z camera? I doubt you have and even if you did you might consider the possibility that you camera needs servicing.
Have you tested a single F-mount lens and observed it using a CDAF confirmation cycle in AF-S, or seen a post online of anyone ever demonstrating the same? How many different model F-Mount lenses would I have to test before you would reasonably agree the camera is not employing CDAF confirmation cycles? Every single model? Is that really your position here?
p.1 #14 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
Have you tested a single F-mount lens and observed it using a CDAF confirmation cycle in AF-S, or seen a post online of anyone ever demonstrating the same? How many different model F-Mount lenses would I have to test before you would reasonably agree the camera is not employing CDAF confirmation cycles? Every single model? Is that really your position here?
You can ask questions, but you cannot provide answers it would seem. I have not tested any lens for a CDAF confirmation cycle. I have absolutely no idea how to do such a test and it seems you don't either.
No my position is simply you are making a very confident and outsized claim that contradicts what Nikon says about their AF system, and you have very flimsy evidence for this claim. I don't believe you know how to do such a test at all.
p.1 #15 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Steve Spencer wrote:
First of all CDAF may not be used for focus acquisition and still be used. In fact, if CDAF is used for focus tuning it won't be used for focus acquisition.
Second, you have made an inappropriate inference that if CDAF were used for fine tuning that precision would be exactly the same and if CDAF were used solely for AF. That would depend on how the algorithm were written. It is certainly possible if not probably that using CDAF for the sole AF and using CDAF for fine tuning would have code using the CDAF in different ways. For example in CDAF only AF where accuracy and precision are paramount the CDAF cycle might check final processing 3, 4, or 5 times. Whereas when CDAF is used for fine tuning the CDAF cycle might only check final focus once. If that were the case you would expect exactly the results that Jim obtained. So, your inference that because these result are different AF-S must not be using CDAF fine tuning is based on the assumption that both modes of focussing are using CDAF in the same way. I think that assumption is invalid and therefore you inference is invalid.
Third, I am skeptical that you can tell that CDAF is used to fine tune by watching the lens. It is good to know how you have arrived at your conclusions. I think it is very flimsy reasoning and I am totally unconvinced....Show more →
The original discussion was whether or not the Z employees a CDAF confirmation cycle. If you'd like to expand that discussion on how Nikon might be using CDAF in other ways such as auto fine-tuning then I would be receptive to that. I have some actual experience in that area with DotTune, which was implemented both as a manual technique and automated procedure in firmware.
I determined the absence of a CDAF confirmation by watching the behavior of the AF in the viewfinder, providing an example of how a camera which does perform a CDAF confirmation behaves (Sony). If you can supply an example of a camera known to use CDAF confirmation cycles (or CDAF in general) that doesn't produce the focus warble I described then I would certainly reconsider my reassessment.
p.1 #16 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Steve Spencer wrote:
You can ask questions, but you cannot provide answers it would seem. I have not tested any lens for a CDAF confirmation cycle. I have absolutely no idea how to do such a test and it seems you don't either.
No my position is simply you are making a very confident and outsized claim that contradicts what Nikon says about their AF system, and you have very flimsy evidence for this claim. I don't believe you know how to do such a test at all.
I've provided a method for you to make that observation - by observing the absence of warbling behavior of the AF and how that behavior manifests on every other camera that employs CDAF or CDAF confirmation cycle. It's the nature of CDAF algorithms that produce that warbling - the sampling of contrast at varying focus element positions necessitates it. I've also provided Jim's evidence of the precision of Pinpoint AF vs AF-S on the Z bodies. If you believe there is a more accurate and definitive method to make this assessment then I would be receptive to ideas you have on the matter.
p.1 #17 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
The original discussion was whether or not the Z employees a CDAF confirmation cycle. If you'd like to expand that discussion on how Nikon might be using CDAF in other ways such as auto fine-tuning then I would be receptive to that. I have some actual experience in that area with DotTune, which was implemented both as a manual technique and automated procedure in firmware.
I determined the absence of a CDAF confirmation by watching the behavior of the AF in the viewfinder, providing an example of how a camera which does perform a CDAF confirmation behaves (Sony). If you can supply an example of a camera known to use CDAF confirmation cycles (or CDAF in general) that doesn't produce the focus warble I described then I would certainly reconsider my reassessment....Show more →
You are making a distinction between a CDAF confirmation cycle and CDAF fine tuning that I don't think is that useful here. The original discussion was whether the Nikon Z uses CDAF at all in the normal course of AF operation. Specifically you said "As it stands Nikon isn't using CDAF in the normal course of AF," despite Nikon's claim that it does. So the claim is about whether Nikon uses CDAF not the way it uses it and your claim is that Nikon does not in the norma course of AF. The evidence you have provided for that claim is incredibly flimsy despite you deep confidence in it. I submit that it is you who needs much more compelling evidence for your claim and "I looked in the viewfinder and didn't see it" which seems to be the sum total of you evidence I don't find at all persuasive.
I also explained quite fully how your use of Jim's evidence is no evidence at all that CDAF isn't being used in the normal course of AF. It could and likely is the case that CDAF is being used differently in the two modes of AF and that would totally explain Jim's data, so you can't point to that as evidence for your statement. It is a poor and flimsy argument to do so. So we are left with you don't see it in the viewfinder and whatever small number of lenses you have tried to observe it.
p.1 #18 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Steve Spencer wrote:
You are making a distinction between a CDAF confirmation cycle and CDAF fine tuning that I don't think is that useful here. The original discussion was whether the Nikon Z uses CDAF at all in the normal course of AF operation. Specifically you said "As it stands Nikon isn't using CDAF in the normal course of AF," despite Nikon's claim that it does. So the claim is about whether Nikon uses CDAF not the way it uses it and your claim is that Nikon does not in the norma course of AF. The evidence you have provided for that claim is incredibly flimsy despite you deep confidence in it. I submit that it is you who needs much more compelling evidence for your claim and "I looked in the viewfinder and didn't see it" which seems to be the sum total of you evidence I don't find at all persuasive....Show more →
I'm making a distinction because your original post on this matter referenced CDAF trim specifically, and made no reference to other ways CDAF might be employed. Again, if the discussion has expanded to include other avenues that relate to CDAF then I'm open to such a discussion as well.
p.1 #19 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
I'm making a distinction because your original post on this matter referenced CDAF trim specifically, and made no reference to other ways CDAF might be employed. Again, if the discussion has expanded to include other avenues that relate to CDAF then I'm open to such a discussion as well.
Well you are reading in specifics in my original post that I never intended. I have been interested all along in whether Nikon's AF in the Z cameras uses CDAF and if they do in what ways. Your response was that they don't in the normal course of AF. That is where the discussion led. You made the strong claim that Nikon does not use CDAF, and I have found that claim to be based on flimsy evidence and I am unpersuaded by it. I don't think there is much point in further discussion, but if you have something else you would like to discuss that is fine by me.
p.1 #20 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Steve Spencer wrote:
Well you are reading in specifics in my original post that I never intended. I have been interested all along in whether Nikon's AF in the Z cameras uses CDAF and if they do in what ways. Your response was that they don't in the normal course of AF. That is where the discussion led. You made the strong claim that Nikon does not use CDAF, and I have found that claim to be based on flimsy evidence and I am unpersuaded by it. I don't think there is much point in further discussion, but if you have something else you would like to discuss that is fine by me....Show more →
Understood, thanks for the discussion and this thread. I always appreciate engaging in technical discussions on this forum, even if ultimately it doesn't lead to an agreement. Take care.
