Is it time for Mirrorless / DSLR Z camera for low light AF?

[MJ_Photo38]

This is physically impossible to do

first because there is simply no mirror in Z cameras

second, because it has currently the shortest flange distance of any full frame mirrorless camera, and if you're able to fit a mirrorbox anbd an AF sensor in 16mm instead of 40, I'm sure any camera company would be interested to hire you.

 

[SKersting66]

And the DSLR type PDAF system is physically throwing away performance to do that. There's minimal performance degradation from defocus because the effective aperture of the PDAF system is very high, so DOF is consistently huge and sensitivity consistent across apertures and lens to lens.

What performance is being discarded by the dedicated system? Seems like a typo?

 

[SKersting66]

Nikon says it doesn't see the need for X-Y sensing given all of the other image analysis tech at its disposal, ...

I think this is largely true... I'm pretty convinced that all of the advanced AF (subject/eye recognition) is AI processing of the video stream, and that the PDAF itself is much coarser than we'd like to think it is. I know The AI system (and thus the PDAF) is not using only the information/area I specify. I think PDAF has always been coarser than we like to think; it's just that as resolution increases (lens and sensor) its' shortcomings become more apparent.
I also wonder if the shorter flange distance isn't something of a contributor here...
--
https://www.flickr.com/skersting

 

[mosswings]

And the DSLR type PDAF system is physically throwing away performance to do that. There's minimal performance degradation from defocus because the effective aperture of the PDAF system is very high, so DOF is consistently huge and sensitivity consistent across apertures and lens to lens.

What performance is being discarded by the dedicated system? Seems like a typo?

DSLR PDAF systems could, and in fact, some do, take advantage of the larger apertures of a fast lens. It does require a longer sensing array and IIRC additional separation optics to do so. Some high-end Canons (I'm not sure about Nikons) had f/2.8 AF sensors. Note that this didn't increase light sensitivity, but accommodated the wider separation angle of the f/2.8 image pairs which provided additional resolution to handle the tighter AF precision the f/2.8 lens demanded.
In this sense most implementations of DSLR PDAF throw away some potential gains with high-end lenses in order to achieve consistent performance across the entire line of lenses - at least those that have f/5.6 or wider max apertures. For teleconverted telephotos, the manufacturers offered f/8 AF sensing otherwise those lenses wouldn't AF reliably.
With mirrorless, you get a bump in light sensitivity with wider aperture lenses because the AF pixels directly sense the incoming light.

 

[primeshooter]

** Mirror wouldn't have flip down. It could flip up or slide across. All that matters is mirror method of DSLR's be switched to when there is too little light for mirrorless AF.

What does it matter how the mirror is deployed? The bottom line is that the mirror has to reflect light somewhere; up, down, left, or right. In order to do this it needs the mirror needs to orient itself to a 45 degree angle. That takes up space and brings us right back to the SLR with lenses that are mounted further away from the film/sensor plane to make room for the mirror.

Perhaps I'm missing something here. Did you have a different idea of how this would work and still keep the Z mount?

I think this is simply a suggestion based on casual knowledge of the technologies, not something worked through in any detail.
The thing that sticks out in my mind about this thread and that seems to have triggered astirusty is that astrophotogs have complained about the pulling capability of the Z8 and Z9 resulting in poorer images than the D850. This suggests that there is justification in the market for a non-stacked body optimized for still image quality at the cost of maximal AF performance, but it would not be at the price point desired by the target buyer.

I think that, technology will always find the answer. The answer is not to bring back the mirror and the phase detect AF sensor hiding below it, but I understand the sentiment, sure. It was me that brought this up, and my point was, that there is a penalty for the stacked sensor. As I said, the Z 8 is a wonderful camera, just has some more noise going on because of this. Even despite that, I still prefer it in every way to the Z7ii that it succeeds.

This morphed into mirrorless AF that varies in performance with aperture, affecting AF at moderately-low light levels and medium apertures as employed by natural light portraitists and wedding photographers.
The complaints driving astirusty are mostly ones of technological maturity, not fundamental inferiority - although there are some aspects that might be, but nothing requiring abandonment of mirrorless technologies.
From a physical technology standpoint, Canon seems to hold a basic advantage over its competitors; besides the Company Formerly Known as Olympus, it's the only one using X-Y AF sensing in its high end ILCs, which with shaped microlensing can be done without significant effect on imaging sensitivity or quality (DPAF II).
What it really boils down to - again - is comparing one manufacturer's immature technology to competitors' more mature and what may be more workable offerings.

There should be an option to focus at the widest aperture regardless of the set aperture for low light purposes, selectable in the menu by the user. Oh, and a 4:5 crop mode...

 

[mosswings]

I think this is simply a suggestion based on casual knowledge of the technologies, not something worked through in any detail.
The thing that sticks out in my mind about this thread and that seems to have triggered astirusty is that astrophotogs have complained about the pulling capability of the Z8 and Z9 resulting in poorer images than the D850. This suggests that there is justification in the market for a non-stacked body optimized for still image quality at the cost of maximal AF performance, but it would not be at the price point desired by the target buyer.

I think that, technology will always find the answer. The answer is not to bring back the mirror and the phase detect AF sensor hiding below it, but I understand the sentiment, sure. It was me that brought this up, and my point was, that there is a penalty for the stacked sensor. As I said, the Z 8 is a wonderful camera, just has some more noise going on because of this. Even despite that, I still prefer it in every way to the Z7ii that it succeeds.

If I misattributed, I apologize.

I would say this, though: stacked sensor does not necessarily mean noise/image performance degradation. Designing a camera sensor for maximum AF speed probably does. I'd point a finger first at the Z9/8's shutterless design.

A stacked sensor enables a lot of processing possibilities that require huge data bandwidth and different readout configurations. A somewhat slower implementation, still stacked, could avoid the disadvantages of a shutterless design but still deliver top-class AF. It will be interesting to see if a Z6iii or Z7iii demonstrates this IQ optimization in a stacked architecture.

 

[Leonard Shepherd]

DSLR PDAF systems could, and in fact, some do, take advantage of the larger apertures of a fast lens.

Are you sure?

Some who are well respected claim the aperture in the AF unit in the mirror box restricts AF used brightness to about f5.6 aperture.

This is similar limitation to a DSLR viewfinder restricting image brightness and the depth of field impression to no wider than (depending on the camera model) between about f2.2 and f4.

You cannot see the full narrowness of depth of field at wider than f2.2 with a DSLR viewfinder - you can see it right up to f.095 with Nikon ML.

Some lens designs have been know for focus shift as they are stopped down. This can lead to some focus error shooting stopped down.

Focus shift generally takes place at wider than f5.6 - and is compensated for in a system like Nikon ML that focus at the stopped down aperture.

It seems a few Nikon Z lenses have enough focus shift to be a problem with a DSLR system - with the potential trade off of higher optical quality than a lens designed without the focus shift.

Some high-end Canons (I'm not sure about Nikons) had f/2.8 AF sensors.

Yes - it worked with just a few EOS lenses.

Moving away from the topic heading - while ML (except some Olympus) cannot take advantage of cross type sensors - my hands on experience is overall ML is increasingly better because

1/ the viewfinder starts by being several stops brighter than a DSLR

2/ actual depth of field is displayed wider than about f2.2 (the Nikon DSLR limit) with faster apertures.

3/ in reasonable daylight ML viewfinders at even f16 are quite bright compared to dim on a DSLR when using a button to stop down to f16 to check depth of field.

4/ Lenses with focus shift are not an AF challenge on Nikon ML.

5/ The dramatically improved electric communication with the Z mounts= allows lens distortion to be easily corrected electronically - this is difficult optically without other optical trade offs.

6/ Subject detection is constantly developing with ML - and is not available on Nikon DSLRs.

7/ Others can list their preferred ML advantages.

8/ Nikon still sells the D850 and D780 new and is not forcing anybody to go ML - though few would go back to no longer made Nikon film bodies for action photography

--
Leonard Shepherd
In lots of ways good photography is similar to learning to play a piano - it takes practice to develop skill in either activity.

 

[Vince P]

The low light focusing of my Z8 beats my D850, D500 etc. So there isn’t any point bringing over any technology from the DSLR era.

If someone can’t focus their recent Z camera in low light I’d suggest that they need to learn the camera better

I'm a huge fan of the Z9, but also: tell me you're not a wedding photographer without telling me.

There are award winning wedding photographers using Z6 and Z7, let alone the Z8 and Z9. Now, I have no relevant experience with the latest Canons but I have Sony and Nikon and for me, neither is any better or any worse in low light focusing, on people, just different. One of the bitterest pills to swallow, for us all, is if even one person can get the equipment to work, then it's not the equipment.

However if different equipment works better for you then you should definitely use that.

 

[astirusty]

** Mirror wouldn't have flip down. It could flip up or slide across. All that matters is mirror method of DSLR's be switched to when there is too little light for mirrorless AF.

It has been pointed out small flange area (space of lens to image sensor) wouldn't allow a standard DSLR mirror to flip up/down.

Being this would a new design Z body, what if mirror slid in from side (as originally mentioned), eliminating space required to flip mirror up/down?

Going by previously posted drawing of DSLR camera, only small sub-mirror is required for AF. Meaning less space is needed to side, than would be required for entire main mirror.

Another possibility would be an IRIS with mirrored surface. Little flange space required and could replace imaging sensor protective cover, as has been previously suggested by others.

Depending on patents, partial mirrored sub-mirror or IRIS might work for AF; similar to Sony's hybrid AF system which passed some light thru to imaging sensor.

 

[SKersting66]

And the DSLR type PDAF system is physically throwing away performance to do that. There's minimal performance degradation from defocus because the effective aperture of the PDAF system is very high, so DOF is consistently huge and sensitivity consistent across apertures and lens to lens.

What performance is being discarded by the dedicated system? Seems like a typo?

DSLR PDAF systems could, and in fact, some do, take advantage of the larger apertures of a fast lens. It does require a longer sensing array and IIRC additional separation optics to do so. Some high-end Canons (I'm not sure about Nikons) had f/2.8 AF sensors. Note that this didn't increase light sensitivity, but accommodated the wider separation angle of the f/2.8 image pairs which provided additional resolution to handle the tighter AF precision the f/2.8 lens demanded.
In this sense most implementations of DSLR PDAF throw away some potential gains with high-end lenses in order to achieve consistent performance across the entire line of lenses - at least those that have f/5.6 or wider max apertures. For teleconverted telephotos, the manufacturers offered f/8 AF sensing otherwise those lenses wouldn't AF reliably.
With mirrorless, you get a bump in light sensitivity with wider aperture lenses because the AF pixels directly sense the incoming light.

I am aware of Canon's f/2.8 AF points and how that all works. But I don't really see that as "discarded performance;" just another set of design consideration tradeoffs. E.g. most optical aberrations originate from the wide aperture regions...

I've long considered why TC's reduced AF points, and I do not believe it is because of point obscuration/masking. I've put a 2x TC on an f/5.6 lens and was still able to focus using most of the points with a high contrast target (D850)... but it was definitely very slow and less reliable.

 

[primeshooter]

I think this is simply a suggestion based on casual knowledge of the technologies, not something worked through in any detail.
The thing that sticks out in my mind about this thread and that seems to have triggered astirusty is that astrophotogs have complained about the pulling capability of the Z8 and Z9 resulting in poorer images than the D850. This suggests that there is justification in the market for a non-stacked body optimized for still image quality at the cost of maximal AF performance, but it would not be at the price point desired by the target buyer.

I think that, technology will always find the answer. The answer is not to bring back the mirror and the phase detect AF sensor hiding below it, but I understand the sentiment, sure. It was me that brought this up, and my point was, that there is a penalty for the stacked sensor. As I said, the Z 8 is a wonderful camera, just has some more noise going on because of this. Even despite that, I still prefer it in every way to the Z7ii that it succeeds.

If I misattributed, I apologize.
I would say this, though: stacked sensor does not necessarily mean noise/image performance degradation. Designing a camera sensor for maximum AF speed probably does. I'd point a finger first at the Z9/8's shutterless design.
A stacked sensor enables a lot of processing possibilities that require huge data bandwidth and different readout configurations. A somewhat slower implementation, still stacked, could avoid the disadvantages of a shutterless design but still deliver top-class AF. It will be interesting to see if a Z6iii or Z7iii demonstrates this IQ optimization in a stacked architecture.

No apology required. What I would say is that the stacked sensor's readout speed being so fast; induces a noise penalty in the electronics. I don't think there is an issue or problem to be solved here as such, or at least not a major one that requires something like the OPs suggestion. We aren't bringing back the horse and cart to solve a problem with motorcars are we? (And I love dslrs). We will likely see another camera without this aimed purely at landscape photography with exceptional dynamic range with perhaps a global shutter in it or other cameras.

Sony's stacked sensor cameras all have a noise penalty over their non stacked versions.

 

[Droster]

Yes, we're not moving away from normal sensors as the norm and they will probably continue to be for the foreseeable future. At least not until global sensors become mainstream enough to push stacked sensors down. But even then, it's also not a given because normal sensors aren't exactly lacking and we're not exactly demanding normal sensors to be replaced either.

 

[mosswings]

And the DSLR type PDAF system is physically throwing away performance to do that. There's minimal performance degradation from defocus because the effective aperture of the PDAF system is very high, so DOF is consistently huge and sensitivity consistent across apertures and lens to lens.

What performance is being discarded by the dedicated system? Seems like a typo?

DSLR PDAF systems could, and in fact, some do, take advantage of the larger apertures of a fast lens. It does require a longer sensing array and IIRC additional separation optics to do so. Some high-end Canons (I'm not sure about Nikons) had f/2.8 AF sensors. Note that this didn't increase light sensitivity, but accommodated the wider separation angle of the f/2.8 image pairs which provided additional resolution to handle the tighter AF precision the f/2.8 lens demanded.
In this sense most implementations of DSLR PDAF throw away some potential gains with high-end lenses in order to achieve consistent performance across the entire line of lenses - at least those that have f/5.6 or wider max apertures. For teleconverted telephotos, the manufacturers offered f/8 AF sensing otherwise those lenses wouldn't AF reliably.
With mirrorless, you get a bump in light sensitivity with wider aperture lenses because the AF pixels directly sense the incoming light.

I am aware of Canon's f/2.8 AF points and how that all works. But I don't really see that as "discarded performance;" just another set of design consideration tradeoffs. E.g. most optical aberrations originate from the wide aperture regions...

I've long considered why TC's reduced AF points, and I do not believe it is because of point obscuration/masking. I've put a 2x TC on an f/5.6 lens and was still able to focus using most of the points with a high contrast target (D850)... but it was definitely very slow and less reliable.

Yes, that's because the Nikon AF system at least does not start obscuring the f/5.6 AF points until close to f/8. It's possible for the central points to retain function even higher than this. Engineering and manufacturing tolerancing and margins that guarantee full function at f/5.6 and all that.

In case you're wondering where I'm getting all this info, visit Photons to Photos. Marianne Oelund in 2014 completely dissected and analyzed a D300 AF system in a multipart opus that still stands today as definitive. Photons to Photos rehomed it just before the whole DPreview closure debacle so as to preserve it.

 

[mosswings]

DSLR PDAF systems could, and in fact, some do, take advantage of the larger apertures of a fast lens.

Are you sure?

Heck yes, but not in the way I think you're taking it, and that may be poor word choice on my part. You concur with my mention below of f/2.8 AF sensing in certain high end Canon bodies. That's what I'm talking about here.

Some who are well respected claim the aperture in the AF unit in the mirror box restricts AF used brightness to about f5.6 aperture.

Not sure how to parse this comment, but the bog standard f/5.6 Nikon AF system from the D300 can usually focus, at least centrally, with lenses having an f/8 widest aperture owing to design margin and tolerances. See Oelund. f/5.6 is the widest for full performance.

This is similar limitation to a DSLR viewfinder restricting image brightness and the depth of field impression to no wider than (depending on the camera model) between about f2.2 and f4.

And the Nikon DSLR PDAF system at least had an effective aperture of around f/22-24. This very small aperture provided tremendous DOF (of the AF system only) and contributed to the system's confident performance even when strongly defocused.

You cannot see the full narrowness of depth of field at wider than f2.2 with a DSLR viewfinder - you can see it right up to f.095 with Nikon ML.

Yes, but why is this relevant?

Some high-end Canons (I'm not sure about Nikons) had f/2.8 AF sensors.

Yes - it worked with just a few EOS lenses.

And Nikon had f/8 AF sensors for TC'd telephotos.

My comment, "DSLR PDAF systems could, and in fact, some do, take advantage of the larger apertures of a fast lens.", would be better said as "DSLR PDAF systems could, and in fact, some do, adjust to the focusing demands of fast lenses". This more accurately describes what the F/2.8 focusing points I was talking about actually are doing.
The system isn't getting additional light from those large apertures - it's simply broadening the separation angle of the image pairs - generated by the small apertures of the AF optics - to improve focusing precision.
Contrast this with mirrorless, which has no separation optics other than the masking/microlensing on the imaging sensor itself.

 

[skyrunr]

There are award winning wedding photographers using Z6 and Z7, let alone the Z8 and Z9. However if different equipment works better for you then you should definitely use that.

You can great photos from Nikon, and I have. Nikon's AF-C can not be as good as the competition's. They are not mutually exclusive. My hit rates are about 10% lower with Nikon. However, those 10% were sometimes key shots using the same technique and settings. Why jump through hoops with AF settings when you can use a system that just works better? I still prefer Nikon and that is why I'm interested in them improving.

From your gear list, I don't see that you own any Sony full-frame ILC gear or lenses.

 

[Vince P]

There are award winning wedding photographers using Z6 and Z7, let alone the Z8 and Z9. However if different equipment works better for you then you should definitely use that.

You can great photos from Nikon, and I have. Nikon's AF-C can not be as good as the competition's. They are not mutually exclusive. My hit rates are about 10% lower with Nikon. However, those 10% were sometimes key shots using the same technique and settings. Why jump through hoops with AF settings when you can use a system that just works better? I still prefer Nikon and that is why I'm interested in them improving.

From your gear list, I don't see that you own any Sony full-frame ILC gear or lenses.

??? FX3 FX6 FX9 are all full frame as are the 24, 35, 50, 85 GMs along with the 16-35, 28-135 & 70-200.

Your hit rate being lower doesn't mean that the camera is 10% "Worse" just that for you i doesn't work as well. Better depends on what works for the user.

 

[mosswings]

I think this is simply a suggestion based on casual knowledge of the technologies, not something worked through in any detail.
The thing that sticks out in my mind about this thread and that seems to have triggered astirusty is that astrophotogs have complained about the pulling capability of the Z8 and Z9 resulting in poorer images than the D850. This suggests that there is justification in the market for a non-stacked body optimized for still image quality at the cost of maximal AF performance, but it would not be at the price point desired by the target buyer.

I think that, technology will always find the answer. The answer is not to bring back the mirror and the phase detect AF sensor hiding below it, but I understand the sentiment, sure. It was me that brought this up, and my point was, that there is a penalty for the stacked sensor. As I said, the Z 8 is a wonderful camera, just has some more noise going on because of this. Even despite that, I still prefer it in every way to the Z7ii that it succeeds.

If I misattributed, I apologize.
I would say this, though: stacked sensor does not necessarily mean noise/image performance degradation. Designing a camera sensor for maximum AF speed probably does. I'd point a finger first at the Z9/8's shutterless design.
A stacked sensor enables a lot of processing possibilities that require huge data bandwidth and different readout configurations. A somewhat slower implementation, still stacked, could avoid the disadvantages of a shutterless design but still deliver top-class AF. It will be interesting to see if a Z6iii or Z7iii demonstrates this IQ optimization in a stacked architecture.

No apology required. What I would say is that the stacked sensor's readout speed being so fast; induces a noise penalty in the electronics. I don't think there is an issue or problem to be solved here as such, or at least not a major one that requires something like the OPs suggestion. We aren't bringing back the horse and cart to solve a problem with motorcars are we? (And I love dslrs). We will likely see another camera without this aimed purely at landscape photography with exceptional dynamic range with perhaps a global shutter in it or other cameras.

Sony's stacked sensor cameras all have a noise penalty over their non stacked versions.

All I'm saying is that we be careful in associating "stacked sensor" with "noise penalty". That shouldn't cause a noise penalty per se, but stacked sensors' killer app is that they can process and offload data very fast, so they're showing up in speed cameras with hefty UHD video capability first.

The faster we go on the same road the more the car rattles. Same thing for processing datapaths. We can smooth out the road, we can improve the suspension of the car, we can beef up the car's structure, we can slow down. Slowing down is the simplest thing to do but compromises tracking and focus. Stacked sensors can deliver speed when needed and precision when it's needed, but at a cost.

Ultimately, stacked sensors will provide an integration size benefit as well as an isolation benefit between noise sensitive pixels and noisy digital processing circuits even if their speediness is not exploited. But that's a ways down the pike.

In any regard, stacked sensors are still very young, and I think they're not irretrievably bad for highest IQ. About global shutters, on the other hand, I'm less optimistic.

 

[skyrunr]

Sorry, I missed your "other gear" FX bodies. Personal observations and 10% aside, even IF Nikon matched AF performance of Canon and Sony, the form factor and cost/value is just not as competitive. Not to mention how much they have been depreciating and are being discounted after less than a year. That isn't something to ignore.

I'm glad you're happy with your Nikon gear as it is.

 

[ghostfox_1]

Sorry, I missed your "other gear" FX bodies. Personal observations and 10% aside, even IF Nikon matched AF performance of Canon and Sony, the form factor and cost/value is just not as competitive. Not to mention how much they have been depreciating and are being discounted after less than a year. That isn't something to ignore.

I'm glad you're happy with your Nikon gear as it is.

u wot m8

The z8/9 are better value than the equivalent canon/sony bodies were (at the time of release).

A lot of us also like the form factor. If people want tiny bodies, by all means, canon and sony are cramped and uncomfortable.

 

[Droster]

Sorry, I missed your "other gear" FX bodies. Personal observations and 10% aside, even IF Nikon matched AF performance of Canon and Sony, the form factor and cost/value is just not as competitive. Not to mention how much they have been depreciating and are being discounted after less than a year. That isn't something to ignore.

I'm glad you're happy with your Nikon gear as it is.

I'm sorry but you really can't say the cost/value is not competitive when

1) The Z9 is the cheapest flagship

2) The Z8 is the cheapest second stacked sensor camera

Sony's a1 is $6.5k and the a9ii was $4.5k before the global sensor came in and pushed it up to $6k. And if you want to talk about depreciation we have the Canons. The R3 release MSRP was $6k. Now B&H has it listed as $4.5k. The R5 was overpriced to begin with at $4k, now it's at $3.4k where a non-stacked sensor resolution body should be.