jhinkey wrote:
At least with Nikon systems the AF fine tune is to compensate for both the AF system mis-adjustment AND lens problems since the PDAF system calculates what it thinks the lens is supposed to do, then tells the lens and never checks back to make sure things are in focus. So with a pure PDAF lens you can still need AF fine tune, but it would be just to compensate for the lens.
Hopefully a hybrid system that uses both CDAF and PDAF together is what they are doing.
It was a long held belief that PDAF systems worked like this, but it's been demonstrated that Canon AF systems don't and never have done.
It's easy to test this yourself, go to a window, wind the focus all the way out to infinity, activate the centre AF point, line it up with the window frame and half press the shutter. As soon as you hear the lens start to move, swing the camera around to point at the background.
Does the lens carry on to focus close up where the window frame was, or does it grab focus on the background?
I'm curious what everyone is referring to when they talk about Sony Alpha bodies not holding their value. Is it a reference to the used market and some Sony employees dumping stock into Buy & Sell?
Obviously the A7 and A7R dropped in price over time, but as far as I can tell, the more refined bodies of they A7S and A7II are still retailing at their launch price. Where is the expectation coming from that you'll be able to snatch up an A7RII for $1000 less six months from now?
I have a hunch that they A7S(II) and A7RII will not be so quick to drop.
I'm gonna wait, there's a few very important releases this year (5DS/R and A7rII) so i'm gonna wait it out and see what happens.
I've not got too much into the sony line-up at the moment and hedging my bets by using Canon lenses on Metabones (I did come from Canon 5D2 so have a lot of canon and Zeiss ZE glass still).
To be honest I like the sounds of the promised AF improvement if it can also carry through to 3rd party, but I'm still getting too much metabones grief such as communication errors where you cant select the aperture and also some weird adapter flare from the metabones mk4.
Firstly I didn't see a place in the forums for introductions. I found FredMiranda thanks to a friend at POTN who told me of a larger Sony presence here. Having had (and still have) several Sony bodies and lenses I should have found this place much earlier. Truthfully I was looking for a larger audience to sell some Sonys too to help fund the A7Rii. That said I need a Sony home, moreso here the knowledge and share than POTN....
I recently took a gamble on a Canon 400mm DO II. There were precious few real world reviews of that lens when I ordered and took possession. It was a successful gamble. Which is to say I'm quite willing to gamble on the A7Rii.
My game plan is basically the same but soon to be simplified.
For BIFs I'll still be using a 7Dii + either a 300 F2.8 or the 400 DOii.
For macro I'll still be using my beloved 70D which also is sweet for astro via the swivel screen and augmented by a modded T2i.
The Sony A7R was my landscape machine but it was sold today....hopefully soon my EOS 5Diii can follow suit. For my shooting the 5Diii does nothing the A7R would do concerning landscape, my 6D as well and if the A7Rii is as promised the 5Diii will see little use.
For my active lifestyle (yep old people still backpack, paddle, cycle) I've switched to m4/3s and will also see the RX100iii go soon, me preferring the Lumix GM-5 with interchangeable lenses.
I mentioned simplified---the hope that the A7Rii can do A7R and A7S duties.
So in my mind it will replace the two mentioned Sony FFs and the 5Diii.
I have a Metabones III and have used several EOS lenses on the A7R=I have NO illusions to the prospective AF until it is in my hands and proven. G.Laing is highly respected reviewer, in fact one of my favorites and I've bought him several cups of coffee! But that was with one EOS lens and one doesn't make a good 'n'.
This camera checked all the boxes on my personal wish list. It also checked a lot more boxes than I imagined that it would, and costs about $600 more than I figured it would. I need to see how well it does some of those unexpected things, specifically AF EOS lenses. If it does that even OK, like 5Dii OK, then it will be hard to resist.
The thing is, I will miss, due to the August ship date, much of the summer travel for work season when I do much of my shooting, so I will probably be patient into Xmas or next Spring and wait for prices to drop. If they don't, its value to me depends on how well it does the things I did not expect it to do.
If Sony employees resell them like they did the other A7x bodies at comparable discounts, that will be hard to resist.
MedicineMan404 wrote:
Firstly I didn't see a place in the forums for introductions. I found FredMiranda thanks to a friend at POTN who told me of a larger Sony presence here. Having had (and still have) several Sony bodies and lenses I should have found this place much earlier. Truthfully I was looking for a larger audience to sell some Sonys too to help fund the A7Rii. That said I need a Sony home, moreso here the knowledge and share than POTN....
I recently took a gamble on a Canon 400mm DO II. There were precious few real world reviews of that lens when I ordered and took possession. It was a successful gamble. Which is to say I'm quite willing to gamble on the A7Rii.
My game plan is basically the same but soon to be simplified.
For BIFs I'll still be using a 7Dii + either a 300 F2.8 or the 400 DOii.
For macro I'll still be using my beloved 70D which also is sweet for astro via the swivel screen and augmented by a modded T2i.
The Sony A7R was my landscape machine but it was sold today....hopefully soon my EOS 5Diii can follow suit. For my shooting the 5Diii does nothing the A7R would do concerning landscape, my 6D as well and if the A7Rii is as promised the 5Diii will see little use.
For my active lifestyle (yep old people still backpack, paddle, cycle) I've switched to m4/3s and will also see the RX100iii go soon, me preferring the Lumix GM-5 with interchangeable lenses.
I mentioned simplified---the hope that the A7Rii can do A7R and A7S duties.
So in my mind it will replace the two mentioned Sony FFs and the 5Diii.
I have a Metabones III and have used several EOS lenses on the A7R=I have NO illusions to the prospective AF until it is in my hands and proven. G.Laing is highly respected reviewer, in fact one of my favorites and I've bought him several cups of coffee! But that was with one EOS lens and one doesn't make a good 'n'....Show more →
Welcome to FM!
I also think along the lines that the A7r II is a unification of the variants in the A7 line, with a very few exceptions, like ISO 400K for the A7s and low weight for the A7r.
Alundeb thanks for the welcome
Well I said months ago if they'd put the AF of the A6000 (which I use mostly for digiscoping) into the A7R I'd be happy...not only did they do that but 5-axis too. My better half has an Oly M5 mk II and I've seen her hand hold 2 second shots=amazing, and a brilliant move by Sony at least marrying into that part of Olympus.
There is and seemingly has always been a disconnect with Sony and those of us who bracket frequently, well not just Sony...this past weekend we backpacked for 3 days and I carried a Lumix GM-5, bracket a cave shot and still have to use an external (in this case my phone) release. So I wonder if in this grandious creation called A7Rii if the countdown timer has been finally removed from the same menu sub-tree as the bracketing functions? If so I think I'll actually write Sony and give a dose of laudable encomium But I'm not holding my breath.
So we speak of the A7Rii as a melding of the A7R and A7S but the sad reality is that those 'lowly' 12 million pixels of the A7S have a light collecting ability that 42 million pixels can never have, its just the physics of the much larger photosite on the A7S, so I would also not get my hopes up that the A7Rii can replace (or possibly even compete) with the sensor of the A7S when we are truly talking low light. I again hope Sony proves me wrong.
Now a question for you kind folks about buffer size. I did read 5fps x 42 frames in fine jpg. Did anyone see/read how many in raw before the buffer fills? One could extrapolate.....
How much fun silent shutter will be! One thing I did enjoy with my A7S was strapping the camera on my chest and walking through a crowded street scene with the camera clicking away; then going through the frames later and seeing what I got by happenstance.
I am investing in glass.. So far I have the 35mm and 50mm Loxias ..plus I have several really nice A mount Zeiss and Legacy Minoltas that are worth adapting in the future. .. After the A7rii comes out and goes through a couple of iterations .and the inevitable flaws are exposed/fixed .. I may pick up one or see if they start to sell LNIB Used on FM. Meanwhile, I am quite happy with my Loxias on a basic A7 body.
MedicineMan404 wrote:
So we speak of the A7Rii as a melding of the A7R and A7S but the sad reality is that those 'lowly' 12 million pixels of the A7S have a light collecting ability that 42 million pixels can never have, its just the physics of the much larger photosite on the A7S, so I would also not get my hopes up that the A7Rii can replace (or possibly even compete) with the sensor of the A7S when we are truly talking low light. I again hope Sony proves me wrong.
Well your hopes might just come true.
Whilst a larger photosite will capture more light and thus have less noise on it's own, the α7ʀ has the advantage of more photosites in total and when you downsample an α7ʀ image to match the α7s dimensions the noise levels are very similar at ISO 6400. See the SNR 18% tab here (make sure it's set to "print" not "screen"):
The α7s only starts to pull away at ISO 12800 and above, and that's due to the electronics not the fact it has larger pixels.
The α7ʀ II will also have the advantage of back side illumination which will give it a small sensitivity boost. Sony claim it's cleaner than the α7ʀ, so when resampled it could actually beat the α7s in the middle ISOs.
MedicineMan404 wrote:
How much fun silent shutter will be! One thing I did enjoy with my A7S was strapping the camera on my chest and walking through a crowded street scene with the camera clicking away; then going through the frames later and seeing what I got by happenstance.
Completely silent shutter is a huge deal. I've photographed classical concerts professionally where photography is only allowed during the applause at the end of each piece. Some photographers spend thousands on "sound blimps" so they can shoot concerts, and on film sets during shooting etc.
Many people will be able to benefit from more discrete street photography. I recently shot a time lapse during a four hour car journey and the sound of the shutter firing every 2 seconds starts to drive you insane after a while...
snapsy wrote: alundeb wrotesnapsy wrotealundeb wrotesnapsy wroteThere is no way to introduce a single movement bias into a closed-loop "sense+movement" system without that movement bias being eliminated by the recursion of the system itself. It would be possible if there were two AF tune values - the existing in-band sensing bias plus an additional out-of-band movement bias. And the proof is in the pudding - I've never handled a lens that required engaging AF system movement (lens motor or body) to tune properly, out of about 50 lenses I've tuned for myself and others. In fact, it's the requirement that an AF acquisition occur that makes other tuning methods prone to error because they introduce AF motor mechanical variability into what is strictly an optical issue. A lens can be out-of-spec with respect to its positioning mechanism but that can't be solved reliably by existing AF tune values because again that is only used as an offset into the sensing. Those lenses have to be calibrated mechanically using the adjustment points provided in the design.
The reason individual lenses need AFMA (in addition to body calibration issues like mirror misalignment) is due to lens-specific aberrations, specifically decentered or tilted elements that affect the two optical paths to the phase sensors.What difference does it make to the loop if:
A: The AF drive system sets the focus distance to 10.1 m nominally and the real focus distance becomes 1.02 m because of miscalibrated AF drive, or
B: The AF drive system sets the focus distance to 10.1 m nominally and the real focus distance becomes 1.02 m because of misaligned elements
The difference is that AF tune as designed today can account for B but not A, for reasons I listed in earlier posts. To account for A in its current design the AF tune mechanism would would have to ignore its last sensing operation since that sensing would show that focus is off by the phase equivalent of .08m if the last movement operation included a +.08m bias.Now I see what you mean. AFMA cannot compensate for a mismatch between The AF drive and sense if implemented as you describe. I was thinking of the AF drive and AF sense as one unit.
What I really ment to point out, more specifically, was:
What difference does it make to the loop if:
A: The AF drive system sets the focus distance to 10.1 m nominally, and the AF sense reports 10.1 m, and the real focus distance becomes 10.2 m because of miscalibrated AF both drive and sense, or
B: The AF drive system sets the focus distance to 10.1 m nominally, and the AF sense reports 10.1 m, and the real focus distance becomes 1.02 m because of misaligned elementsNot sure if for B you mean misalignment for sensing or movement. I don't think misaligned elements can cause misfocusing in terms of movement; they could cause misfocus in terms of sensing, esp for the peripheral AF points that would be more sensitive to the misalignment. At least for decentered/tilted misalignments. If instead by misaligned you mean the focusable element within the lens is misaligned such that its position is wrong relative to where the AF drive though it put the lens. I believe most USM systems have positional sensing mechanisms that account for that and so they don't rely on absolute element position assumptions.
Lastly, if by misaligned you mean as detected by phase sensing afterwards (which would be correct and what I think you meant), then there would be no difference between A and B as both would be detected as out-of-phase by the sensing mechanism. But in both cases there may not be a final AF movement to correct that differential for a given PDAF acquisition cycle; as the experiments in this article demonstrated, the system is not entirely closed loop; sometimes the system will perform a final AF movement and either not re-sense the phase or not correct if the phase is slightly off, likely because it thinks it's already "close enough" to not warrant the performance penalty of another movement cycle....Show more →
I think we can agree that since lens specific AFMA compensates for some optical differences between lenses, it does not only compensate for the difference between the PDAF sensor plane and the image sensor plane. If AFMA is necessary with a perfectly aligned but separate PDAF sensor, it is also necessary with a PDAF sensor integrated in the image sensor.
alundeb wrote: snapsy wrotealundeb wrotesnapsy wrotealundeb wrotesnapsy wroteThere is no way to introduce a single movement bias into a closed-loop "sense+movement" system without that movement bias being eliminated by the recursion of the system itself. It would be possible if there were two AF tune values - the existing in-band sensing bias plus an additional out-of-band movement bias. And the proof is in the pudding - I've never handled a lens that required engaging AF system movement (lens motor or body) to tune properly, out of about 50 lenses I've tuned for myself and others. In fact, it's the requirement that an AF acquisition occur that makes other tuning methods prone to error because they introduce AF motor mechanical variability into what is strictly an optical issue. A lens can be out-of-spec with respect to its positioning mechanism but that can't be solved reliably by existing AF tune values because again that is only used as an offset into the sensing. Those lenses have to be calibrated mechanically using the adjustment points provided in the design.
The reason individual lenses need AFMA (in addition to body calibration issues like mirror misalignment) is due to lens-specific aberrations, specifically decentered or tilted elements that affect the two optical paths to the phase sensors.What difference does it make to the loop if:
A: The AF drive system sets the focus distance to 10.1 m nominally and the real focus distance becomes 1.02 m because of miscalibrated AF drive, or
B: The AF drive system sets the focus distance to 10.1 m nominally and the real focus distance becomes 1.02 m because of misaligned elements
The difference is that AF tune as designed today can account for B but not A, for reasons I listed in earlier posts. To account for A in its current design the AF tune mechanism would would have to ignore its last sensing operation since that sensing would show that focus is off by the phase equivalent of .08m if the last movement operation included a +.08m bias.Now I see what you mean. AFMA cannot compensate for a mismatch between The AF drive and sense if implemented as you describe. I was thinking of the AF drive and AF sense as one unit.
What I really ment to point out, more specifically, was:
What difference does it make to the loop if:
A: The AF drive system sets the focus distance to 10.1 m nominally, and the AF sense reports 10.1 m, and the real focus distance becomes 10.2 m because of miscalibrated AF both drive and sense, or
B: The AF drive system sets the focus distance to 10.1 m nominally, and the AF sense reports 10.1 m, and the real focus distance becomes 1.02 m because of misaligned elementsNot sure if for B you mean misalignment for sensing or movement. I don't think misaligned elements can cause misfocusing in terms of movement; they could cause misfocus in terms of sensing, esp for the peripheral AF points that would be more sensitive to the misalignment. At least for decentered/tilted misalignments. If instead by misaligned you mean the focusable element within the lens is misaligned such that its position is wrong relative to where the AF drive though it put the lens. I believe most USM systems have positional sensing mechanisms that account for that and so they don't rely on absolute element position assumptions.
Lastly, if by misaligned you mean as detected by phase sensing afterwards (which would be correct and what I think you meant), then there would be no difference between A and B as both would be detected as out-of-phase by the sensing mechanism. But in both cases there may not be a final AF movement to correct that differential for a given PDAF acquisition cycle; as the experiments in this article demonstrated, the system is not entirely closed loop; sometimes the system will perform a final AF movement and either not re-sense the phase or not correct if the phase is slightly off, likely because it thinks it's already "close enough" to not warrant the performance penalty of another movement cycle.I think we can agree that since lens specific AFMA compensates for some optical differences between lenses, it does not only compensate for the difference between the PDAF sensor plane and the image sensor plane. If AFMA is necessary with a perfectly aligned but separate PDAF sensor, it is also necessary with a PDAF sensor integrated in the image sensor.
All of the sensor-based PDAF systems I've read about use phase detect for the initial focus acquisition and then refine that focus using contrast detect. The same appears to be true on the A7rII based on this interview, which is likely faster than the A7/A7II due to the A7rII's faster sensor readout (faster CDAF sampling). This means AF tune wont be required. From the interview:
DE: What about the low-light limit when you have phase-detect pixels? It seems to me that the separate phase-detect sensors like in an SLR, they can have big pixels, but on the sensor chip, we have, I mean they're limited to the size of the actual pixels.
KM: That's a good question. So, when we use... in terms of autofocus, we are using a hybrid focus, so first we use the phase-detection focus to get close to the object. Then near the object, we start to use contrast AF to get a clear peak. Then we adjust the focus. Then in order to make very precise focus (because 42-megapixel means very very small pixels), and also the focusing point could be a very small one, therefore we phase-detection together with contrast AF, contrast focus.
snapsy wrote: alundeb wrotesnapsy wrotealundeb wrotesnapsy wrotealundeb wroteWhat difference does it make to the loop if:
A: The AF drive system sets the focus distance to 10.1 m nominally and the real focus distance becomes 1.02 m because of miscalibrated AF drive, or
B: The AF drive system sets the focus distance to 10.1 m nominally and the real focus distance becomes 1.02 m because of misaligned elements
The difference is that AF tune as designed today can account for B but not A, for reasons I listed in earlier posts. To account for A in its current design the AF tune mechanism would would have to ignore its last sensing operation since that sensing would show that focus is off by the phase equivalent of .08m if the last movement operation included a +.08m bias.Now I see what you mean. AFMA cannot compensate for a mismatch between The AF drive and sense if implemented as you describe. I was thinking of the AF drive and AF sense as one unit.
What I really ment to point out, more specifically, was:
What difference does it make to the loop if:
A: The AF drive system sets the focus distance to 10.1 m nominally, and the AF sense reports 10.1 m, and the real focus distance becomes 10.2 m because of miscalibrated AF both drive and sense, or
B: The AF drive system sets the focus distance to 10.1 m nominally, and the AF sense reports 10.1 m, and the real focus distance becomes 1.02 m because of misaligned elementsNot sure if for B you mean misalignment for sensing or movement. I don't think misaligned elements can cause misfocusing in terms of movement; they could cause misfocus in terms of sensing, esp for the peripheral AF points that would be more sensitive to the misalignment. At least for decentered/tilted misalignments. If instead by misaligned you mean the focusable element within the lens is misaligned such that its position is wrong relative to where the AF drive though it put the lens. I believe most USM systems have positional sensing mechanisms that account for that and so they don't rely on absolute element position assumptions.
Lastly, if by misaligned you mean as detected by phase sensing afterwards (which would be correct and what I think you meant), then there would be no difference between A and B as both would be detected as out-of-phase by the sensing mechanism. But in both cases there may not be a final AF movement to correct that differential for a given PDAF acquisition cycle; as the experiments in this article demonstrated, the system is not entirely closed loop; sometimes the system will perform a final AF movement and either not re-sense the phase or not correct if the phase is slightly off, likely because it thinks it's already "close enough" to not warrant the performance penalty of another movement cycle.I think we can agree that since lens specific AFMA compensates for some optical differences between lenses, it does not only compensate for the difference between the PDAF sensor plane and the image sensor plane. If AFMA is necessary with a perfectly aligned but separate PDAF sensor, it is also necessary with a PDAF sensor integrated in the image sensor.All of the sensor-based PDAF systems I've read about use phase detect for the initial focus acquisition and then refine that focus using contrast detect. The same appears to be true on the A7rII based on this interview, which is likely faster than the A7/A7II due to the A7rII's faster sensor readout (faster CDAF sampling). This means AF tune wont be required. From the interview:
DE: What about the low-light limit when you have phase-detect pixels? It seems to me that the separate phase-detect sensors like in an SLR, they can have big pixels, but on the sensor chip, we have, I mean they're limited to the size of the actual pixels.
KM: That's a good question. So, when we use... in terms of autofocus, we are using a hybrid focus, so first we use the phase-detection focus to get close to the object. Then near the object, we start to use contrast AF to get a clear peak. Then we adjust the focus. Then in order to make very precise focus (because 42-megapixel means very very small pixels), and also the focusing point could be a very small one, therefore we phase-detection together with contrast AF, contrast focus....Show more →
This discussion derived from the context of CDAF being disabled with some types of lenses via adapter. If PDAF is not accurate, and CDAF is disabled, Canon lens owners are screwed with the A7r II. AFMA is their only hope to get accurate AF with EF lenses if Sony don't turn around and enable CDAF. Which in turn makes the AF much slower because of the slow protocol.
Matt Grum wrote:
The α7s only starts to pull away at ISO 12800 and above, and that's due to the electronics not the fact it has larger pixels.
The A7s's Ultra High ISO advantage is indirectly the result of its large pixels because bigger pixels means fewer pixels for the same sized sensor and fewer pixels means lower aggregate read noise for low-conversion gain modes utilized for High ISO. A good description of this can be found in Aptina's DR-Pix whitepaper (link).
alundeb wrote: snapsy wrotealundeb wrotesnapsy wrotealundeb wrotesnapsy wroteThe difference is that AF tune as designed today can account for B but not A, for reasons I listed in earlier posts. To account for A in its current design the AF tune mechanism would would have to ignore its last sensing operation since that sensing would show that focus is off by the phase equivalent of .08m if the last movement operation included a +.08m bias.Now I see what you mean. AFMA cannot compensate for a mismatch between The AF drive and sense if implemented as you describe. I was thinking of the AF drive and AF sense as one unit.
What I really ment to point out, more specifically, was:
What difference does it make to the loop if:
A: The AF drive system sets the focus distance to 10.1 m nominally, and the AF sense reports 10.1 m, and the real focus distance becomes 10.2 m because of miscalibrated AF both drive and sense, or
B: The AF drive system sets the focus distance to 10.1 m nominally, and the AF sense reports 10.1 m, and the real focus distance becomes 1.02 m because of misaligned elementsNot sure if for B you mean misalignment for sensing or movement. I don't think misaligned elements can cause misfocusing in terms of movement; they could cause misfocus in terms of sensing, esp for the peripheral AF points that would be more sensitive to the misalignment. At least for decentered/tilted misalignments. If instead by misaligned you mean the focusable element within the lens is misaligned such that its position is wrong relative to where the AF drive though it put the lens. I believe most USM systems have positional sensing mechanisms that account for that and so they don't rely on absolute element position assumptions.
Lastly, if by misaligned you mean as detected by phase sensing afterwards (which would be correct and what I think you meant), then there would be no difference between A and B as both would be detected as out-of-phase by the sensing mechanism. But in both cases there may not be a final AF movement to correct that differential for a given PDAF acquisition cycle; as the experiments in this article demonstrated, the system is not entirely closed loop; sometimes the system will perform a final AF movement and either not re-sense the phase or not correct if the phase is slightly off, likely because it thinks it's already "close enough" to not warrant the performance penalty of another movement cycle.I think we can agree that since lens specific AFMA compensates for some optical differences between lenses, it does not only compensate for the difference between the PDAF sensor plane and the image sensor plane. If AFMA is necessary with a perfectly aligned but separate PDAF sensor, it is also necessary with a PDAF sensor integrated in the image sensor.All of the sensor-based PDAF systems I've read about use phase detect for the initial focus acquisition and then refine that focus using contrast detect. The same appears to be true on the A7rII based on this interview, which is likely faster than the A7/A7II due to the A7rII's faster sensor readout (faster CDAF sampling). This means AF tune wont be required. From the interview:
DE: What about the low-light limit when you have phase-detect pixels? It seems to me that the separate phase-detect sensors like in an SLR, they can have big pixels, but on the sensor chip, we have, I mean they're limited to the size of the actual pixels.
KM: That's a good question. So, when we use... in terms of autofocus, we are using a hybrid focus, so first we use the phase-detection focus to get close to the object. Then near the object, we start to use contrast AF to get a clear peak. Then we adjust the focus. Then in order to make very precise focus (because 42-megapixel means very very small pixels), and also the focusing point could be a very small one, therefore we phase-detection together with contrast AF, contrast focus.This discussion derived from the context of CDAF being disabled with some types of lenses via adapter. If PDAF is not accurate, and CDAF is disabled, Canon lens owners are screwed with the A7r II. AFMA is their only hope to get accurate AF with EF lenses if Sony don't turn around and enable CDAF. Which in turn makes the AF much slower because of the slow protocol.
There was mention that Sony worked with Metabones on a firmware update to the IV EF adapter during the A7rII development to improve AF performance. Perhaps part of that effort was to address any CDAF issues.
alundeb wrote: snapsy wrotealundeb wrotesnapsy wrotealundeb wrotesnapsy wroteThe difference is that AF tune as designed today can account for B but not A, for reasons I listed in earlier posts. To account for A in its current design the AF tune mechanism would would have to ignore its last sensing operation since that sensing would show that focus is off by the phase equivalent of .08m if the last movement operation included a +.08m bias.Now I see what you mean. AFMA cannot compensate for a mismatch between The AF drive and sense if implemented as you describe. I was thinking of the AF drive and AF sense as one unit.
What I really ment to point out, more specifically, was:
What difference does it make to the loop if:
A: The AF drive system sets the focus distance to 10.1 m nominally, and the AF sense reports 10.1 m, and the real focus distance becomes 10.2 m because of miscalibrated AF both drive and sense, or
B: The AF drive system sets the focus distance to 10.1 m nominally, and the AF sense reports 10.1 m, and the real focus distance becomes 1.02 m because of misaligned elementsNot sure if for B you mean misalignment for sensing or movement. I don't think misaligned elements can cause misfocusing in terms of movement; they could cause misfocus in terms of sensing, esp for the peripheral AF points that would be more sensitive to the misalignment. At least for decentered/tilted misalignments. If instead by misaligned you mean the focusable element within the lens is misaligned such that its position is wrong relative to where the AF drive though it put the lens. I believe most USM systems have positional sensing mechanisms that account for that and so they don't rely on absolute element position assumptions.
Lastly, if by misaligned you mean as detected by phase sensing afterwards (which would be correct and what I think you meant), then there would be no difference between A and B as both would be detected as out-of-phase by the sensing mechanism. But in both cases there may not be a final AF movement to correct that differential for a given PDAF acquisition cycle; as the experiments in this article demonstrated, the system is not entirely closed loop; sometimes the system will perform a final AF movement and either not re-sense the phase or not correct if the phase is slightly off, likely because it thinks it's already "close enough" to not warrant the performance penalty of another movement cycle.I think we can agree that since lens specific AFMA compensates for some optical differences between lenses, it does not only compensate for the difference between the PDAF sensor plane and the image sensor plane. If AFMA is necessary with a perfectly aligned but separate PDAF sensor, it is also necessary with a PDAF sensor integrated in the image sensor.All of the sensor-based PDAF systems I've read about use phase detect for the initial focus acquisition and then refine that focus using contrast detect. The same appears to be true on the A7rII based on this interview, which is likely faster than the A7/A7II due to the A7rII's faster sensor readout (faster CDAF sampling). This means AF tune wont be required. From the interview:
DE: What about the low-light limit when you have phase-detect pixels? It seems to me that the separate phase-detect sensors like in an SLR, they can have big pixels, but on the sensor chip, we have, I mean they're limited to the size of the actual pixels.
KM: That's a good question. So, when we use... in terms of autofocus, we are using a hybrid focus, so first we use the phase-detection focus to get close to the object. Then near the object, we start to use contrast AF to get a clear peak. Then we adjust the focus. Then in order to make very precise focus (because 42-megapixel means very very small pixels), and also the focusing point could be a very small one, therefore we phase-detection together with contrast AF, contrast focus.This discussion derived from the context of CDAF being disabled with some types of lenses via adapter. If PDAF is not accurate, and CDAF is disabled, Canon lens owners are screwed with the A7r II. AFMA is their only hope to get accurate AF with EF lenses if Sony don't turn around and enable CDAF. Which in turn makes the AF much slower because of the slow protocol.
As I said previously, however, there would still be room for metabones or another adapter maker to add AFMA. Bo Ming who works for metabones even has some experience developing AFMA for Contax N lenses converted to Canon mount (his conversions, at least the later ones, did offer AFMA). So, there is some reasonable hope, IMO, that even if Sony does nothing, the adapter makers (and metabones in particular) could develop AFMA.
Should Sony and 3rd parties adapters manufacturer actually fully reverse-engineer the communication protocols between Sony vs. Canon&Nikon, they have created the single mount camera to rule all lenses out there; AF and MF.
It will be interesting to see how well Canon lenses work. Clearly it is better than the existing AF on the A7R. Frankly, I'd rather have accurate AF than fast and somewhat accurate. Of course, both would be excellent.
Steve Spencer wrote: alundeb wrotesnapsy wrotealundeb wrotesnapsy wrotealundeb wroteNow I see what you mean. AFMA cannot compensate for a mismatch between The AF drive and sense if implemented as you describe. I was thinking of the AF drive and AF sense as one unit.
What I really ment to point out, more specifically, was:
What difference does it make to the loop if:
A: The AF drive system sets the focus distance to 10.1 m nominally, and the AF sense reports 10.1 m, and the real focus distance becomes 10.2 m because of miscalibrated AF both drive and sense, or
B: The AF drive system sets the focus distance to 10.1 m nominally, and the AF sense reports 10.1 m, and the real focus distance becomes 1.02 m because of misaligned elementsNot sure if for B you mean misalignment for sensing or movement. I don't think misaligned elements can cause misfocusing in terms of movement; they could cause misfocus in terms of sensing, esp for the peripheral AF points that would be more sensitive to the misalignment. At least for decentered/tilted misalignments. If instead by misaligned you mean the focusable element within the lens is misaligned such that its position is wrong relative to where the AF drive though it put the lens. I believe most USM systems have positional sensing mechanisms that account for that and so they don't rely on absolute element position assumptions.
Lastly, if by misaligned you mean as detected by phase sensing afterwards (which would be correct and what I think you meant), then there would be no difference between A and B as both would be detected as out-of-phase by the sensing mechanism. But in both cases there may not be a final AF movement to correct that differential for a given PDAF acquisition cycle; as the experiments in this article demonstrated, the system is not entirely closed loop; sometimes the system will perform a final AF movement and either not re-sense the phase or not correct if the phase is slightly off, likely because it thinks it's already "close enough" to not warrant the performance penalty of another movement cycle.I think we can agree that since lens specific AFMA compensates for some optical differences between lenses, it does not only compensate for the difference between the PDAF sensor plane and the image sensor plane. If AFMA is necessary with a perfectly aligned but separate PDAF sensor, it is also necessary with a PDAF sensor integrated in the image sensor.All of the sensor-based PDAF systems I've read about use phase detect for the initial focus acquisition and then refine that focus using contrast detect. The same appears to be true on the A7rII based on this interview, which is likely faster than the A7/A7II due to the A7rII's faster sensor readout (faster CDAF sampling). This means AF tune wont be required. From the interview:
DE: What about the low-light limit when you have phase-detect pixels? It seems to me that the separate phase-detect sensors like in an SLR, they can have big pixels, but on the sensor chip, we have, I mean they're limited to the size of the actual pixels.
KM: That's a good question. So, when we use... in terms of autofocus, we are using a hybrid focus, so first we use the phase-detection focus to get close to the object. Then near the object, we start to use contrast AF to get a clear peak. Then we adjust the focus. Then in order to make very precise focus (because 42-megapixel means very very small pixels), and also the focusing point could be a very small one, therefore we phase-detection together with contrast AF, contrast focus.This discussion derived from the context of CDAF being disabled with some types of lenses via adapter. If PDAF is not accurate, and CDAF is disabled, Canon lens owners are screwed with the A7r II. AFMA is their only hope to get accurate AF with EF lenses if Sony don't turn around and enable CDAF. Which in turn makes the AF much slower because of the slow protocol.As I said previously, however, there would still be room for metabones or another adapter maker to add AFMA. Bo Ming who works for metabones even has some experience developing AFMA for Contax N lenses converted to Canon mount (his conversions, at least the later ones, did offer AFMA). So, there is some reasonable hope, IMO, that even if Sony does nothing, the adapter makers (and metabones in particular) could develop AFMA.
Unless the E-Mount protocol has a mechanism to notify the lens/adapter of the "final AF movement", I'm not sure how the adapter could implement a reliable AFMA.