p.2 #1 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Just recorded this video demonstrating the CDAF warble in Pinpoint AF vs absence of it in AF-S, for both a short and long focus-racked acquisition. This was with the Nikon Z 35mm f/1.8.
I also pinged Jim with a link to this thread and we're bouncing ideas off each other about possible experiments that might be more revealing.
Aug 17, 2019 at 07:33 PM
proam Offline • [X]
p.2 #2 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Was pretty obvious to me that the FTZ was using hybrid phase + contrast detect from a pure performance standpoint. Didn't realize people were under the impression otherwise.
Perhaps you got confused regarding the contrast detect only pinpoint mode?
p.2 #3 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Same warble comparison but with the 50mm f/1.8G on the FTZ. This combo focuses slower for CDAF than the native 35mm so the difference between CDAF and PDAF is more obvious:
p.2 #5 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
Adding AF-C to the test, which should definitely be PDAF only and appears the same as AF-S. Nikon 50mm f/1.8G + FTZ:
Snapsy I think what this demonstrates is that pinpoint AF is not the same as AF-S AF, but didn't we know that already? And even though we know that AF-S and AF-C are undoubtedly not the same, we can't see that in this demonstration. That raises for me anyway what seems to be a very plausible account that your viewfinder test here is only sensitive to certain types of CDAF, but not sensitive to all uses of CDAF. Said another way, if AF-S uses a CDAF for a short duration to fine tune AF, as Nikon's literature suggests, then I am not at all sure you would see it in this viewfinder test. Do you have any evidence that you would? You have shown evidence that if you do the whole AF with CDAF you can see it, but that does not at all convince me that if you use CDAF to fine tune AF you will see it in the viewfinder.
p.2 #6 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Steve Spencer wrote:
Snapsy I think what this demonstrates is that pinpoint AF is not the same as AF-S AF, but didn't we know that already? And even though we know that AF-S and AF-C are undoubtedly not the same, we can't see that in this demonstration. That raises for me anyway what seems to be a very plausible account that your viewfinder test here is only sensitive to certain types of CDAF, but not sensitive to all uses of CDAF. Said another way, if AF-S uses a CDAF for a short duration to fine tune AF, as Nikon's literature suggests, then I am not at all sure you would see it in this viewfinder test. Do you have any evidence that you would? You have shown evidence that if you do the whole AF with CDAF you can seat, but that does not at all convince me that if you use CDAF to fine tune AF you will see it in the viewfinder....Show more →
I thought about the possibility of automatic/dynamic AF fine tuning on the Z today and I couldn't come up with many plausible scenarios of how it would work or help. This goes into the theory and operation behind my DotTune method. The AF tuning mechanism built into DSLRs is designed to correct phase differentials introduced by misalignments in the optical path to the PDAF sensors, relative to the optical path to the sensor. As such, the AFMA values a user enters into the camera are simply offsets that feed into the PDAF system's interpretation of phase differentials. On a body+lens that don't require AFMA, a well-focused subject yields an in-phase measurement at the PDAF sensor. On a body+lens that require AFMA, a well-focused subject yields an out-of-phase measurement at the PDAF sensor - the AFMA value entered into the camera then corrects this by adjusting that phase valuation so that the out-of-phase measurement is reinterpreted as in-phase. This entire process is entirely static and is not influenced or affected by the motorized elements of the AF system (ie, focus motor in lens or body). This is why you can tune a lens using DotTune without ever performing an AF acquisition. Before DotTune, most photographers assumed that the focus motor had a material impact into the need for AFMA - DotTune proved that to not be the case. That's not to say AF focus motor positioning errors never happen - but if you look at some of the best PDAF implementations, such as the improvements Canon made starting with the 5D III and newer lenses (see Roger Cicala's great three-part series on this), it's clear that positioning errors on modern lenses is relatively uncommon.
On a MILC, most of the elements that can cause the misalignments that necessitate AFMA are gone, specifically the mirror, sub-mirror, and PDAF sensor array. In fact the entire separate optical path to the AF system is gone - replaced by the direct path to the PDAF array on the image sensor itself. The only remaining elements in the system that can cause misalignments are the lens (decentered/tilting) and the mount (tilted/improperly machined). It's presumed the manufacturing processing that lays out the metal masks for the on-sensor PDAF sensors aren't subject to much variability.
For those remaining sources of possible misalignment I suppose it's possible Nikon implemented an automatic tuning mechanism, similar to what me and ML team first did with the automated version of DotTune (see my video here), which Nikon in turn implemented starting with the D810. Canon has also patented a similar method (link), although it's not clear they've implemented it on a camera yet. But such an automatic method should only need to be employed a few times, to cover the range of focal lengths (zooms) and also possibly variance at different focus differences. I don't see the practical way they could employ such a system on a continuous basis, both because it would be slow and also unnecessary because the misalginemnts would be static in nature and not benefit from continuous adjustment. So the question becomes - if Nikon did implement an automatic tuning mechanism, how often would it actually be employed? What would trigger its re-calibration?
The other case where an automatic AFMA could be employed is to correct for spherical aberration (focus shift). In fact in my DotTune procedure I listed this as a possible auxiliary use for AFMA - to use an AFMA value to bake-in SA-correction for those who regularly shoot thin DOF at near-wide open apertures. It's not a very flexible solution since the AFMA value would be specific to certain apertures. An automatic system would be much more flexible...but since the Z's always focus stopped-down anyway (up through f/5.6), there really isn't a need for an AFMA-based SA solution.
I have a lot more thoughts on this but this post is already too long so I'll stop here for now.
p.2 #7 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
I thought about the possibility of automatic/dynamic AF fine tuning on the Z today and I couldn't come up with many plausible scenarios of how it would work or help. This goes into the theory and operation behind my DotTune method. The AF tuning mechanism built into DSLRs is designed to correct phase differentials introduced by misalignments in the optical path to the PDAF sensors, relative to the optical path to the sensor. As such, the AFMA values a user enters into the camera are simply offsets that feed into the PDAF system's interpretation of phase differentials. On a body+lens that don't require AFMA, a well-focused subject yields an in-phase measurement at the PDAF sensor. On a body+lens that require AFMA, a well-focused subject yields an out-of-phase measurement at the PDAF sensor - the AFMA value entered into the camera then corrects this by adjusting that phase valuation so that the out-of-phase measurement is reinterpreted as in-phase. This entire process is entirely static and is not influenced or affected by the motorized elements of the AF system (ie, focus motor in lens or body). This is why you can tune a lens using DotTune without ever performing an AF acquisition. Before DotTune, most photographers assumed that the focus motor had a material impact into the need for AFMA - DotTune proved that to not be the case. That's not to say AF focus motor positioning errors never happen - but if you look at some of the best PDAF implementations, such as the improvements Canon made starting with the 5D III and newer lenses (see Roger Cicala's great three-part series on this), it's clear that positioning errors on modern lenses is relatively uncommon.
On a MILC, most of the elements that can cause the misalignments that necessitate AFMA are gone, specifically the mirror, sub-mirror, and PDAF sensor array. In fact the entire separate optical path to the AF system is gone - replaced by the direct path to the PDAF array on the image sensor itself. The only remaining elements in the system that can cause misalignments are the lens (decentered/tilting) and the mount (tilted/improperly machined). It's presumed the manufacturing processing that lays out the metal masks for the on-sensor PDAF sensors aren't subject to much variability.
For those remaining sources of possible misalignment I suppose it's possible Nikon implemented an automatic tuning mechanism, similar to what me and ML team first did with the automated version of DotTune (see my video here), which Nikon in turn implemented starting with the D810. Canon has also patented a similar method (link), although it's not clear they've implemented it on a camera yet. But such an automatic method should only need to be employed a few times, to cover the range of focal lengths (zooms) and also possibly variance at different focus differences. I don't see the practical way they could employ such a system on a continuous basis, both because it would be slow and also unnecessary because the misalginemnts would be static in nature and not benefit from continuous adjustment. So the question becomes - if Nikon did implement an automatic tuning mechanism, how often would it actually be employed? What would trigger its re-calibration?
The other case where an automatic AFMA could be employed is to correct for spherical aberration (focus shift). In fact in my DotTune procedure I listed this as a possible auxiliary use for AFMA - to use an AFMA value to bake-in SA-correction for those who regularly shoot thin DOF at near-wide open apertures. It's not a very flexible solution since the AFMA value would be specific to certain apertures. An automatic system would be much more flexible...but since the Z's always focus stopped-down anyway (up through f/5.6), there really isn't a need for an AFMA-based SA solution.
I have a lot more thoughts on this but this post is already too long so I'll stop here for now....Show more →
I was thinking of something a lot more simple. In pinpoint AF it seems almost certain that the camera goes through multiple CDAF cycles and the first of those cycles often are not very close to what the final focus will be. In this situation you see the warble in the viewfinder because there are lots of cycles and/or because the cycles require more movement of the focus groups. In contrast if AF-S does the majority of focusing with PDAF, but does just one CDAF cycle after PDAF has gotten the focus very close to optimal, then I am not at all sure you would see any warble in the viewfinder from the one cycle (or perhaps two) that only tunes the focus to be more precise. Then AF-C doesn't use CDAF at all. If this reasoning is correct, then you can't see the difference between AF-S in the viewfinder and AF-C in the viewfinder because the CDAF happens quickly and is a very small adjustment. This is why I asked what evidence you have that you can see all CDAF in your viewfinder test. Just because you can see CDAF when it is the only AF being used, happens over a relatively long period of time, and involves big changes in focus, does not mean you can see CDAF when it is used minimally and happens very quickly.
p.2 #8 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Steve Spencer wrote:
I was thinking of something a lot more simple. In pinpoint AF it seems almost certain that the camera goes through multiple CDAF cycles and the first of those cycles often are not very close to what the final focus will be. In this situation you see the warble in the viewfinder because there are lots of cycles and/or because the cycles require more movement of the focus groups. In contrast if AF-S does the majority of focusing with PDAF, but does just one CDAF cycle after PDAF has gotten the focus very close to optimal, then I am not at all sure you would see any warble in the viewfinder from the one cycle (or perhaps two) that only tunes the focus to be more precise. Then AF-C doesn't use CDAF at all. If this reasoning is correct, then you can't see the difference between AF-S in the viewfinder and AF-C in the viewfinder because the CDAF happens quickly and is a very small adjustment. This is why I asked what evidence you have that you can see all CDAF in your viewfinder test. Just because you can see CDAF when it is the only AF being used, happens over a relatively long period of time, and involves big changes in focus, does not mean you can see CDAF when it is used minimally and happens very quickly....Show more →
In a hybrid AF implementation (PDAF+CDAF confirmation), the initial PDAF acquisition would be expected to arrive at a focus position that was near to the correct focus position - the purpose of that is to allow the subsequent CDAF confirmation to limit the range of focus positions it is required to cycle through to arrive at peak contrast (ie, to avoid most of the performance penalty of CDAF). Failure of PDAF to arrive within a certain range of near-correct focus positions would result in either the CDAF confirmation arriving at the wrong focus (since it was limited to a range of focus positions that didn't include peak contrast) or a performance penalty from having to reset the scan and do a full CDAF acquisition. Assuming the PDAF acquisition did its job then I agree the range of CDAF movements would be relatively small. This is the expected behavior of Sony's Hybrid AF implementation, and on that platform the warble from the CDAF is still visible, particularly on lenses with slow CDAF focusing characteristics. The lens I chose for the Z6 warble test video I posted yesterday was the 50mm f/1.8G, which is also a slow-CDAF focusing lens - and still no warble was visible in regular AF-S mode.
p.2 #9 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
In a hybrid AF implementation (PDAF+CDAF confirmation), the initial PDAF acquisition would be expected to arrive at a focus position that was near to the correct focus position - the purpose of that is to allow the subsequent CDAF confirmation to limit the range of focus positions it is required to cycle through to arrive at peak contrast (ie, to avoid most of the performance penalty of CDAF). Failure of PDAF to arrive within a certain range of near-correct focus positions would result in either the CDAF confirmation arriving at the wrong focus (since it was limited to a range of focus positions that didn't include peak contrast) or a performance penalty from having to reset the scan and do a full CDAF acquisition. This is the expected behavior of Sony's Hybrid AF implementation, and on that platform the warble from the CDAF is still visible, particularly on lenses with slow CDAF focusing characteristics. The lens I chose for the Z6 warble demonstration video I posted yesterday was the 50mm f/1.8G, which is also a slow-CDAF focusing lens....Show more →
Yeah, but can you see that in Sony all the time? You can see it some of the time, but all the time? Again you don't know because you haven't shown your method can always pick up CDAF or at all how sensitive your test is. And what if Nikon has developed really good PDAF so that the AF-confirmation is easier and takes less times/cycles and less adjustments, than Sony would you see that? Let's propose for a minute that your are right that the current AF-S has such good PDAF that you get the excellent AF accuracy and precision that Jim has shown in his tests or something close to it. What if Nikon added a CDAF confirmation cycle to that already very good focus, would you be able to see that in your viewfinder test? You haven't shown that you could, so I can't see how you can rule out that it is happening. What you need to show is that your warble in the viewfinder can pick up a known very short duration CDAF with very minimal adjustment. What you have shown us is that your warble in the viewfinder test can pick up is very long and very big adjustments. You say that it can also pick up Sony's CDAF confirmation, but we don't know how long those adjustments take or how big the adjustments are. I still think it is totally possible that Nikon is adding a CDAF fine tune adjustment after PDAF in AF-S and your viewfinder test just isn't seeing it.
It also is possible that Nikon is using the CDAF fine tuning (i.e., confirmation) only with lenses that have fast CDAF, and that would make sense because if CDAF is slow you might want to forgo it because the improvement might not be worth the hit you would take in AF speed. If that is the case, then trying to see it with a slow CDAF lens would not be the right strategy. Nikon isn't using it there, but they might well be using it for lenses in which it works better. If this reasoning is correct, then the difference between Sony and Nikon is simply that Sony always uses CDAF fine tuning, but Nikon only uses it for lenses with fast CDAF. Both strategies would have their advantages and disadvantages. My larger point, however, is that you can't use your test to rule out CDAF until we know how sensitive it is to picking up CDAF and we basically know nothing about that.
p.2 #10 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Steve Spencer wrote:
Yeah, but can you see that in Sony all the time? You can see it some of the time, but all the time? Again you don't know because you haven't shown your method can always pick up CDAF or at all how sensitive your test is. And what if Nikon has developed really good PDAF so that the AF-confirmation is easier and takes less times/cycles and less adjustments, than Sony would you see that? Let's propose for a minute that your are right that the current AF-S has such good PDAF that you get the excellent AF accuracy and precision that Jim has shown in his tests or something close to it. What if Nikon added a CDAF confirmation cycle to that already very good focus, would you be able to see that in your viewfinder test? You haven't shown that you could, so I can't see how you can rule out that it is happening. What you need to show is that your warble in the viewfinder can pick up a known very short duration CDAF with very minimal adjustment. What you have shown us is that your warble in the viewfinder test can pick up is very long and very big adjustments. You say that it can also pick up Sony's CDAF confirmation, but we don't know how long those adjustments take or how big the adjustments are. I still think it is totally possible that Nikon is adding a CDAF fine tune adjustment after PDAF in AF-S and your viewfinder test just isn't seeing it.
It also is possible that Nikon is using the CDAF fine tuning (i.e., confirmation) only with lenses that have fast CDAF, and that would make sense because if CDAF is slow you might want to forgo it because the improvement might not be worth the hit you would take in AF speed. If that is the case, then trying to see it with a slow CDAF lens would not be the right strategy. Nikon isn't using it there, but they might well be using it for lenses in which it works better. If this reasoning is correct, then the difference between Sony and Nikon is simply that Sony always uses CDAF fine tuning, but Nikon only uses it for lenses with fast CDAF. Both strategies would have their advantages and disadvantages. My larger point, however, is that you can't use your test to rule out CDAF until we know how sensitive it is to picking up CDAF and we basically know nothing about that....Show more →
When you mentioned CDAF fine tuning yesterday I was thinking you were discussing AFMA, which is why I went into historical detail behind the need for AFMA and the possible cases Nikon might employ an automatic method on the Z. If by "fine tuning" you mean CDAF trim / confirmation cycle then you can discard what I wrote on that topic since it's mostly not relevant. From now on let's refer to this as simply CDAF trim to avoid confusion on my part.
If Nikon's PDAF is more precise than Sony's then I agree, its CDAF trim phase would require fewer focus position movements, and thus the visual manifestations of those movements would be less obvious. My contention though is they would still be visible. If you watch videos of very fast CDAF implementations, particularly on Panasonic MFT bodies where DFD serves the function of PDAF, you will still see a focus warble for the CDAF trim phase. I just rechecked this on my GH5 with a DFD-supported Panasonic lens and the warble is quite noticeable.
Another way to approach this is to ask if the Z's even need CDAF trim for maximum precision. You stated a possible CDAF trim might not be visible on the Z because its PDAF acquisition is arriving so close to its correct position. If so, that might argue its PDAF might be good enough to not even require CDAF trim. On that point I just did a search of Jim's AF articles on the Z to see if he's ever compared AF-S to AF-C, the latter of which I think we both agree would never employ a CDAF trim. Here's a few I found:
p.2 #12 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
gdsf2 wrote:
Maybe somebody should just ask a Nikon NPS rep?
I wouldn't have high expectations of getting an accurate answer. I've spoke at length with Nikon USA reps at CES and NAB and have gotten wrong answers on subjects I already knew the correct answers to.
p.2 #14 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
When you mentioned CDAF fine tuning yesterday I was thinking you were discussing AFMA, which is why I went into historical detail behind the need for AFMA and the possible cases Nikon might employ an automatic method on the Z. If by "fine tuning" you mean CDAF trim / confirmation cycle then you can discard what I wrote on that topic since it's mostly not relevant. From now on let's refer to this as simply CDAF trim to avoid confusion on my part.
If Nikon's PDAF is more precise than Sony's then I agree, its CDAF trim phase would require fewer focus position movements, and thus the visual manifestations of those movements would be less obvious. My contention though is they would still be visible. If you watch videos of very fast CDAF implementations, particularly on Panasonic MFT bodies where DFD serves the function of PDAF, you will still see a focus warble for the CDAF trim phase. I just rechecked this on my GH5 with a DFD-supported Panasonic lens and the warble is quite noticeable.
Another way to approach this is to ask if the Z's even need CDAF trim for maximum precision. You stated a possible CDAF trim might not be visible on the Z because its PDAF acquisition is arriving so close to its correct position. If so, that might argue its PDAF might be good enough to not even require CDAF trim. On that point I just did a search of Jim's AF articles on the Z to see if he's ever compared AF-S to AF-C, the latter of which I think we both agree would never employ a CDAF trim. Here's a few I found:
Notice that Jim doesn't say anything like prove or disprove. Instead he says things like: "This isn’t quite a smoking gun . . ." and "It will take some more testing to find that out." Let me be clear. I think Jim's tests here are interesting. I think your viewfinder test is interesting, but I don't think they justify concluding with any confidence that AF-S doesn't use CDAF at all. It might use it in a feeble way. It might use it for some lenses (e.g., those designed for CDAF) and not with others (e.g., adapted DSLR lenses with big focussing group like the 105 f/1.4). It might use it in a feeble way for those lenses designed for CDAF. To me this is all interesting, but the bottom line is that I think that new lenses for the Z cameras should be built to allow fast, accurate, and precise CDAF, that is they should be built with small focussing groups, large communication pathways, and with fast and silent motors. Even if the Z cameras aren't using that capability now, there is every reason to suspect that it will be useful in the future as better and more complex algorithms are developed. We can certainly expect new AI based algorithms to be a big part of future AF, why wouldn't Nikon develop lenses that have the best shot of taking advantage of those developments whatever they turn out to be?
p.2 #15 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
Steve Spencer wrote:
Notice that Jim doesn't say anything like prove or disprove. Instead he says things like: "This isn’t quite a smoking gun . . ." and "It will take some more testing to find that out." Let me be clear. I think Jim's tests here are interesting. I think your viewfinder test is interesting, but I don't think they justify concluding with any confidence that AF-S doesn't use CDAF at all. It might use it in a feeble way. It might use it for some lenses (e.g., those designed for CDAF) and not with others (e.g., adapted DSLR lenses with big focussing group like the 105 f/1.4). It might use it in a feeble way for those lenses designed for CDAF. To me this is all interesting, but the bottom line is that I think that new lenses for the Z cameras should be built to allow fast, accurate, and precise CDAF, that is they should be built with small focussing groups, large communication pathways, and with fast and silent motors. Even if the Z cameras aren't using that capability now, there is every reason to suspect that it will be useful in the future as better and more complex algorithms are developed. We can certainly expect new AI based algorithms to be a big part of future AF, why wouldn't Nikon develop lenses that have the best shot of taking advantage of those developments whatever they turn out to be?...Show more →
Naturally none of these experiments and observations are definitive. All we can do is draw from the evidence we have and make reasonable guesses as to how the Z's AF-S operates. In my view the evidence weights to the absence of CDAF trim but this view is open to revision based on future evidence and discoveries.
I still agree with you that it would be better than not for Nikon to optimize their lenses for CDAF. The question is whether such designs are more complicated or costly for Nikon and if so, whether that cost is justified on a system that doesn't appear to currently use CDAF very often.
p.2 #16 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
snapsy wrote:
Naturally none of these experiments and observations are definitive. All we can do is draw from the evidence we have and make reasonable guesses as to how the Z's AF-S operates. In my view the evidence weights to the absence of CDAF trim but this view is open to revision based on future evidence and discoveries.
I still agree with you that it would be better than not for Nikon to optimize their lenses for CDAF. The question is whether such designs are more complicated or costly for Nikon and if so, whether that cost is justified on a system that doesn't appear to currently use CDAF very often. ...Show more →
Well, I think that 105 f/1.4 test actually suggest that if Nikon isn't using CDAF fine tuning for the AF, they better get off their butt and find a way of doing it--at least for lenses designed for CDAF. That test showed some pretty serious focus errors when you rack focus and more front focus that isn't ideal even when you don't. That tells me that Nikon could very well improve their AF by utilizing CDAF at least in some modes of focussing and developing a true hybrid AF using both PDAF and CDAF if they haven't already. One of the difficulties of adapting F lenses is they aren't designed for CDAF--they have big focussing groups, they have somewhat limited communication pathways, and some have pretty slow motors, with such lenses your AF options are limited. You have to make painful and unfortunate trade off decisions. Should I use CDAF trim which will help prevent front focus but slow focussing down or do I forego CDAF and use only PDAF and live with the front focus so that I have better AF speed. Not an easy decision to make. If you design a lens for using CDAF you don't have to make these difficult tradeoff decisions nearly so much. You can add CDAF trim preventing front focus and still have good AF speed.
Further I would submit it would be a mistake to design new lenses, now that we know that CDAF can be used and that future algorithms might take good advantage of it, that one knows are going to be bad at CDAF. Seems incredibly short sighted to do that to save a little money on lens development. Certainly at this point it looks like future AF is a lot more likely to use CDAF in all kinds of different ways than it is even used today, so it isn't just the system of today that matter but also the system of five years from now.
p.2 #17 · Contrast Detect AF in AF-S mode with Z lenses and adapted lenses
I've been working on devising more experiments. One idea was to find AF scenarios where PDAF can successfully focus but CDAF can't (or vice-versa), and use that scenario to infer when either PDAF or CDAF is being used based on the success/failure of an AF acquisition, and use that information to infer whether CDAF trim could be employed.
Putting that aside for now, I've observed to interesting behaviors of the Pinpoint AF mode:
1) It's only supported when fully zoomed-out in LV (ie, default view). If you magnify LV at all then the camera reverts to standard AF-S mode. You can tell this is occurring from the absence of warble while focusing while in magnified LV mode. This kind of defeats one the purposes of Pintpiont AF, which is to limit the AF area to a smaller portion of the subject - magnifying in helps to further limit the AF area but doing so switches back to normal AF-S, which means you lose the CDAF ability.
2) CDAF scan behavior is different based on the starting position of focus. For the AF target setup I have, I tried three different starting focus position scenarios at the start of Pinpoint AF:
a) Focus position set to in-focus (ie, target already focused)
b) Focus position set to slightly in front of target (ie, front-focused)
c) Focus position set to slightly to rear of target (ie, back-focused)
For cases 'a' and 'c' the camera performed a full bi-directional CDAF scan (expected). For case 'b' the camera performed only a rearward scan, which resulted in faster CDAF acquisition. This might just be a quirk of the initial contrast detection sample and how it relates to the balance of DOF for this target....or it might indicate an optimization in the camera where it starts scanning/remembering contrast for case 'b', allowing it to avoid a redundant scan forward of that direction, and doesn't perform that optimization for the case 'c'. Anyway, here's a slow-motion video I shot demonstrating these three cases:
