Full Frame Mirrorless: Sideline or Replacement of dSLR?

[3kramd5]

Yes, as you need to split your "electron well" into two: one to capture those photoelectrons when "the shutter is open" and one to store them for digitization when "the shutter is closed". For each pixel.

Thanks, that makes general sense. Still, if you’re looking for high sync speeds, you’re by definition lighting the scene so DR /and native ISO aren’t substantially concerning, so the trade may be worthwhile.

Also, I just went down a rabbit hole:

IEEE has a paper titled “A 100dB dynamic range CMOS image sensor with global shutter.” The prototype has fairly big pixels (10micron^2), but uses sample-and-hold with correlated calibration to significantly reduce noise.

It does use separate per-pixel wells like you’ve described (one for the PD, which is amplified and then stored in a sample and hold well before readout). It mentions another implementation: skimming (a term introduced in another paper, “CMOS Image Sensor with NMOS-only Global Shutter and Enhanced Responsivity”). The latter paper is stretching my understanding of the technology (I’m not an EE), but it seems to improve efficiency by decoupling the capacitance of the PD from the conversion. The shutter never entirely closes, and when reset closes, the voltage across the PD remains while discharge occurs to the sense node, allowing a gain in responsivity.

 

[Kit.]

Also, I just went down a rabbit hole:

IEEE has a paper titled “A 100dB dynamic range CMOS image sensor with global shutter.” The prototype has fairly big pixels (10micron^2), but uses sample-and-hold with correlated calibration to significantly reduce noise.

"Logarithmic architecture pixel"? A nice idea for a continuous light source, but may lead to unpredictable color shifts when using flash.

And the pixels, yes, are quite big.

 

[Michael Clark]

Canon have long put a lot of faith in what the professionals tell them that they (the pros) need in a camera - IMO they have at times done that to the seeming detriment of the general user but overall it has been successful. I see no reason for them to shift dramatically from that approach when it has served them so well - unless the twitterati complaints about the 6D2 has psychologically scarred them. And Canon have already publicly acknowledged that they cannot ignore the 135million EF lenses out there - and the pros do not want to see their EF lenses looking like they will be write-offs in 3-4 years. A simple adapter looks like a band aid solution and will send the message that the new mount is the future and all those EF lenses are on the way out.

I agree with ahsanford's reference to a "one-time bolus of sales" - it is easy to forget that a company like Canon needs to consider what happens once that pent up demand has gone, the dust has settled and people return to that mundane 'let's get out and take photographs' mentality. They then subconciously absorb what it all means - rather like those who went to Sony with all the excitement of the new technology then realised months later that what they actually miss is the simple interface and the fact their old Canon gear 'just works'.

So when a company like Canon has built their strategy on what pros would like to see, the proposal of some that it will be a new mount with EF as a design afterthought will be a massive leap in the dark. If it fails, is it because it is not EF or because they did not get the mirrorless component right? And unravelling that will be a nightmare amongst the self-important shitstorm that social media tends to generate.

The problem Canon has with building their strategy on what pros want is that the number of true full-time pros is dwindling dramatically and swiftly into a number so small that it will not be sufficient to support their past business model in this respect. The well-to-do amateur enthusiast market has been propping up Canon's lines of pro bodies and lenses for years. As long as those customers accept/agree with what the pros say they want it will continue to work.

But a point will eventually come when what happens in the blog/social media/rumors/twitter/reviewers more interested in clicks than in quality content (TN... cough cough... Fro)/IG sphere will outweigh, in the minds of buyers, what the 'Explorers of Light', highly respected PJs, and their ilk think.

 

[Michael Clark]

You seem treat this decision as if Canon was a startup with a clean slate. It is not.

You forgot that EF has one epic advantage over EF-X. EF lenses already exist and folks already own them. So there absolutely is value in a full EF mount mirrorless: it would be seamless to their SLR experience, not require an adaptor, and not require additional spending on lenses.

I'm not saying full EF is better than thin mount. Either will work. But to state that there's no point to come out with a full EF mirrorless body is nonsense, IMHO.

- A

Or to put it more succinctly: If Canon indicates that EF is a lame duck mount, most of their customers will have absolutely zero reason to stick with Canon when choosing their next mount.

Canon rolled the dice once back in 1987, and ultimately it proved to be the right move as the new EF mount and lens based USM allowed performance doing Sports/Action that Nikon's all-mechanical connection at the time could not even pretend to come close to matching.

If Canon chooses to so much as hint that they will 'abandon' the EF mount, they better have something else just as revolutionary up their sleeve as the EF mount+USM was in the early 1990s or it will be a disaster of epic proportions.

 

[Michael Clark]

In short, every slow zoom could get smaller because the mirror has gone poof. Heretofore f/5.6 was the slowest max aperture EF allowed (third parties broke this rule, but not Canon) so that every lens worked on every EF/EF-S body. With the mirror gone, that rule has been tempered to f/6.3, which we've seen on EF-M and the betting man might expect to see that for an FF lens or two.

It's also possible that Canon could the AF to work (in general, not just at a few AF points) at f/8 max aperture, which could be a very big deal for miniaturizing really long FL zooms: imagine a budget plasticky 200-600mm f/6.3 - 8.0 IS STM lens that you could still screw 77mm filters into!

- A

Canon has broken the f/5.6 rule plenty of times, they just didn't/don't call those lenses what they actually were/are:

EF 28-300mm f/3.5-5.6 L IS slower than f/5.6 at every focal length longer than about 85mm. Every focal length past 200mm is f/6.7 or slower.

EF 100-400mm f/4.5-5.6 IS II is slower than f/5.6 anywhere past around 235mm and tops out at f/6.3 at 400mm.

EF 100-300mm f/4.5-5.6 passed f/5.6 at 140mm and was f/6.4 at 300mm.

There are plenty of other examples. Maybe the "f/5.6 rule" was really always the "less than f/8 rule?"

 

[3kramd5]

Canon has broken the f/5.6 rule plenty of times, they just didn't/don't call those lenses what they actually were/are:

EF 28-300mm f/3.5-5.6 L IS slower than f/5.6 at every focal length longer than about 85mm. Every focal length past 200mm is f/6.7 or slower.

EF 100-400mm f/4.5-5.6 IS II is slower than f/5.6 anywhere past around 235mm and tops out at f/6.3 at 400mm.

EF 100-300mm f/4.5-5.6 passed f/5.6 at 140mm and was f/6.4 at 300mm.

There are plenty of other examples. Maybe the "f/5.6 rule" was really always the "less than f/8 rule?"

From an AF perspective, geometry matters, not transmissibility. When you say “slower,” speed is implied, so do you mean t-stop? Or did you actually measure the aperture diameter?

 

[neuroanatomist]

From an AF perspective, geometry matters, not transmissibility. When you say “slower,” speed is implied, so do you mean t-stop? Or did you actually measure the aperture diameter?

My question also. Most people seem to erroneously believe that the f-number limitation for AF systems/points is about the amount of light available, rather than the necessary baseline for phase separation.

 

[docsmith]

Agreed. From a light gathering perspective, very few lenses have T-stops exactly equal to their F-stop. That would mean 100% optical transmission. Elements do absorb/reflect/etc a very small amount of light. So, very few (any?) f/5.6 lenses are letting the same amount of light through as if there was no glass in the lens, just the iris.

I have assumed that Canon bakes in a certain inefficiency. This is really about how much light the PDAF system needs to efficiently do its job. We know from third party lenses that it can function with less light than f/5.6. But, it really does get down to, my understanding, speed, accuracy, etc.

It is a pity, as I do not think they even do them any more, but Popular Photography magazine used to test PDAF speed at different EV levels of light. Speed really falls off as less light is hitting the sensors. I have to assume less light allowed by the lens has a similar effect, perhaps a few others as well.

 

[neuroanatomist]

The AF point requirement (e.g. an f/5.6 AF point vs. an f/2.8 AF point) really has nothing to do with the amount of light. Pumping 4x or 40x more light through an f/5.6 lens won’t activate an f/2.8 AF point, and you could illuminate the scene in front of an f/11 lens (e.g. 400/5.6 with a 2x TC) with the light towers from Fenway Park placed 5 meters from your subject, and the f/5.6 or f/8 AF still won’t work.

 

[ahsanford]

The question, of course, is when Canon might change it's 'EF/EF-S lens mandate' for f/5.6 now that AF technology has evolved a bit. I believe every current-gen SLR other than the 2000D has f/8 AF points today.

Any chance Canon risks p---ing off folks with older bodies to make a 150-600 f/6.3 IS STM happen at a far cheaper cost than an f/5.6 version? That could be the critical difference between a competitive $1500 lens and a 'there goes Canon again' $2500 lens.

- A

 

[docsmith]

The AF point requirement (e.g. an f/5.6 AF point vs. an f/2.8 AF point) really has nothing to do with the amount of light. Pumping 4x or 40x more light through an f/5.6 lens won’t activate an f/2.8 AF point, and you could illuminate the scene in front of an f/11 lens (e.g. 400/5.6 with a 2x TC) with the light towers from Fenway Park placed 5 meters from your subject, and the f/5.6 or f/8 AF still won’t work.

I hadn't been thinking of it in that way, triggering the f/2.8 sensors. I had more been thinking of it in terms of that there is a measurable consequence of having less light hit the same AF sensor whether it be f/2.8 or F/5.6.

But, now that you mention it, any thoughts on differences other than light hitting the AF sensor that would trigger the f/2.8 sensor over a f/5.6 sensor? Quantity and direction of light are the physical things I can think of, but direction would be more a function of focal length than aperture leaving light intensity as the primary difference off the top of my head. So, what does trigger the f/2.8 sensor if it isn't light intensity? Genuinely curious (thesis not required ).

 

[docsmith]

The question, of course, is when Canon might change it's 'EF/EF-S lens mandate' for f/5.6 now that AF technology has evolved a bit. I believe every current-gen SLR other than the 2000D has f/8 AF points today.

Any chance Canon risks p---ing off folks with older bodies to make a 150-600 f/6.3 IS STM happen at a far cheaper cost than an f/5.6 version? That could be the critical difference between a competitive $1500 lens and a 'there goes Canon again' $2500 lens.

- A

They already have EF-m lenses that go to f/6.3. So the easy answer would be yes, I can see it. Especially for the EF-s lenses. I wonder how far Canon would take it? EF-m: already there. EF-s, sure, not that different than EF-m; EF (non-L)---maybe, this is where it gets interesting, but I am going to say yes, for a super tele zoom; EF "L" lens--I am thinking this is where the line gets drawn. I would be surprised to see a EF "L" lens above f/5.6 max aperture.

 

[ahsanford]

Genuinely curious (thesis not required ).

Neuro dips his toes into this in his 'thesis' here:
https://www.the-digital-picture.com/Photography-Tips/canon-eos-dslr-autofocus-explained.aspx

​

Generally, an aperture value is associated with an AF line sensor. The terminology usually used is "f/number-sensitive", e.g, you may have an f/5.6-sensitive line sensor, or an f/2.8-sensitive line sensor. The f/number refers to the maximum aperture of the lens, because AF is performed with the lens wide open (i.e. the aperture you choose for the shot does not matter, only the max aperture of the lens). The use of 'sensitivity' in this context implies that light levels matter, because that's what we think of when we normally use f/numbers. In this case, though, a wider aperture simply means a wider baseline for the rangefinder system is required for that line to function. Personally, I think better terminology might be to use threshold instead of sensitivity, so an f/2.8-threshold line would require an f/2.8 lens to function, and if you mounted an f/4 lens, that sensor line would not operate. An f/5.6-threshold sensor would work with any lens having a max aperture of f/5.6 or wider.​

​

Note that these thresholds are not absolute - a lens with a narrower aperture than the threshold might still work, but at reduced effectiveness, accuracy, and speed. Thus, Canon limits the functionality to the rated aperture for a given AF sensor. However, some third party lenses (e.g. Tamron and Sigma zooms with a max aperture of f/6.3 at the long end) effectively trick the AF system into thinking there's an f/5.6 lens attached. Likewise, although not condoned by Canon, it is possible to use tape to block some of the contacts on a Canon 1.4x extender used with an f/5.6 lens, resulting in the camera attempting to autofocus with an f/8 lens on bodies which are limited to f/5.6. Sometimes, it even works...​

​

All EOS bodies have f/5.6-sensitive sensors, and thus will work with any Canon EF or EF-S lens. Some 1-series bodies have an f/8-sensitive sensor at the center AF point, enabling them to autofocus (properly, and with official support) with an f/5.6 lens plus 1.4x extender or an f/4 lens plus 2x extender - a significant benefit for users of supertelephoto lenses. Notably, that feature is not included in the specification for the 1D X, which is limited to f/5.6 lenses for autofocus.​

​

An f/2.8 sensor line is more accurate than an f/5.6 sensor line - the wider the aperture threshold, the wider the rangefinder baseline for triangulation and thus, the more accurate the measurement of focus. However, the wider the aperture, the fewer lenses that work with that aperture (and the more expensive those lenses are), and also, the detection range of f/2.8 sensors is narrower, meaning it may take longer for an f/2.8 line to achieve a focus lock when a subject is well out of focus. As a result, AF systems will usually focus in two steps when possible - 'coarse' focus with an f/5.6 line, then 'fine' focus with an f/2.8 line.​

- A

 

[Mikehit]

The AF point requirement (e.g. an f/5.6 AF point vs. an f/2.8 AF point) really has nothing to do with the amount of light. Pumping 4x or 40x more light through an f/5.6 lens won’t activate an f/2.8 AF point, and you could illuminate the scene in front of an f/11 lens (e.g. 400/5.6 with a 2x TC) with the light towers from Fenway Park placed 5 meters from your subject, and the f/5.6 or f/8 AF still won’t work.

Or is that because the camera tells the lens it is set at f2.8 and the camera is programmed to say 'I don't care howmuch light there is I ain't doing it'. You only need to see older bodies that AF max a f5.6 - put a third party 1.4x tc on a Canon f5.6 lens and it will AF (or at least try), put a Canon 1.4xtc on the same lens and it will refuse to even try.
In other words Canon want the AF to work in all cases, not 'it will work if there is enough light' because the 'enough light'becomes a wide variable. If Canon has defined 'normal'lighting conditions then the user know that if it is at f2.8 it will AF.

 

[neuroanatomist]

Or is that because the camera tells the lens it is set at f2.8 and the camera is programmed to say 'I don't care howmuch light there is I ain't doing it'. You only need to see older bodies that AF max a f5.6 - put a third party 1.4x tc on a Canon f5.6 lens and it will AF (or at least try), put a Canon 1.4xtc on the same lens and it will refuse to even try.
In other words Canon want the AF to work in all cases, not 'it will work if there is enough light' because the 'enough light'becomes a wide variable. If Canon has defined 'normal'lighting conditions then the user know that if it is at f2.8 it will AF.

Once again, it's not about the amount of light. Say it with me..."It's not about the amount of light."

Yes, Canon limits it by firmware to f/5.6 for the appropriate 'width' of the light cone falling on the PDAF sensor. 3rd party/non-reporting TCs (and Canon TCs with pins taped) often fail to deliver reliable AF.

 

[justaCanonuser]

I came late to this thread and didn't have time to read all postings. Turning to the original post, I'd say that DSLR sales will smoothly decline and ML rise as we can see today. But there will be a certain threshold when the elaborate mechanics of an advanced SLR will be so costly to produce that the prize of such cameras will substantially rise and they may end as a niche product such as Leica's rangefinders. So I agree with your vinyl market picture.

That said, DSLRs still have advances for some applications: no lag and real world depiction in the viewfinder is one aspect, the other is that you can use the viewfinder without emptying the battery. That's e.g. important for wildlife photography, because you have to observe e.g. a bird's nest many times before something interesting happens and you start shooting. With ML you would have to carry loads of exchange batteries with you, with DSLR much less.

Another advantage is for those who change between digital and chemical photography with classical cameras, because your eye and your imagination keeps to be trained for optical viewfinders.

The size is getting lesser important, since Sony and Leica seem keen to blow up their ML FF models to finally beat the dimensions of SLRs with the next generations. Well, e.g. in Arabian countries people are more impressed by bigger cameras (and cars), so it may also be market driven, not only by in-built 10-axis gyro satellite link stabilization stuff ...

 

[Michael Clark]

The AF point requirement (e.g. an f/5.6 AF point vs. an f/2.8 AF point) really has nothing to do with the amount of light. Pumping 4x or 40x more light through an f/5.6 lens won’t activate an f/2.8 AF point, and you could illuminate the scene in front of an f/11 lens (e.g. 400/5.6 with a 2x TC) with the light towers from Fenway Park placed 5 meters from your subject, and the f/5.6 or f/8 AF still won’t work.

You are correct that the geometry has nothing to do with the amount of light, but you can bet your sweet you-know-what that the S/N ratio on the PDAF sensor is affected by the amount of light for any given geometry-limited AF line pair.

Thus Chuck Westfall's often cited quotes that 'One Shot' AF using a longer sampling period than 'AI Servo' as the reason 'One Shot' can AF in up to two stops lower light than 'AI Servo' on the same camera with the same lens can. In dwindling light eventually noise overcomes signal and the PDAF sensor no longer provides any meaningful information, no matter how wide the baseline is. If there's a wider than f/2.8 lens on the camera but there's not enough light to be able to differentiate the signal from the noise on the PDAF sensor, the geometry doesn't do anything.

As has been oft discussed in the past, Canon's "f/5.6" AF points often work, sometimes even rather well, with lenses having maximum apertures somewhere between f/5.6 and f/8. Particularly if those lenses are used in bright light. Which might lead one to conclude that the "f/5.6 geometry" is really much closer to "almost but not quite f/8 geometry."

 

[Michael Clark]

From an AF perspective, geometry matters, not transmissibility. When you say “slower,” speed is implied, so do you mean t-stop? Or did you actually measure the aperture diameter?

Since the difference between f-number and T-stop with those lenses often varies widely from one focal length to another, the difference between f-number and T-stop can not be fully explained by transmission loss. It seems that the actual entrance pupil size does not grow at the same rate with the 28-300mm from, for example, 70mm to 100mm as it does from 35mm to 70mm.



As with all variable aperture zoom lenses, the size of the entrance pupil, as magnified by the optical elements between the aperture diaphragm and the front of the lens, is not keeping up with the focal length as it is zoomed to longer focal lengths. Yet at certain points in the zoom range of these lenses they seem to "make up" some of the loss. If it was all transmission loss one would expect the difference between f-number and T-stop to be constant and the above graph would be flat lines for each of the lenses. Link to lens comparison at DxO Mark. As always, ignore the "number scores" and go to the actual measured data by clicking on 'Measurements' and then 'Transmission'.

 

[neuroanatomist]

You are correct that the geometry has nothing to do with the amount of light, but you can bet your sweet you-know-what that the S/N ratio on the PDAF sensor is affected by the amount of light for any given geometry-limited AF line pair.

Thus Chuck Westfall's often cited quotes that 'One Shot' AF using a longer sampling period than 'AI Servo' as the reason 'One Shot' can AF in up to two stops lower light than 'AI Servo' on the same camera with the same lens can. In dwindling light eventually noise overcomes signal and the PDAF sensor no longer provides any meaningful information, no matter how wide the baseline is. If there's a wider than f/2.8 lens on the camera but there's not enough light to be able to differentiate the signal from the noise on the PDAF sensor, the geometry doesn't do anything.

As has been oft discussed in the past, Canon's "f/5.6" AF points often work, sometimes even rather well, with lenses having maximum apertures somewhere between f/5.6 and f/8. Particularly if those lenses are used in bright light. Which might lead one to conclude that the "f/5.6 geometry" is really much closer to "almost but not quite f/8 geometry."

Of course the light level ultimately limits AF functionality. But the line sensors are quite sensitive. Past discussions made much of some cameras' ability to AF at -3 EV vs. -2 EV, but an example of that difference is f/2.8, 1/15 s, ISO 51200 vs. 102400. Even with older/cheaper AF sensors only sensitive down to -0.5 EV, the illumination at which light sensitivity becomes limiting is still so dim as to yield images of poor (for many, unacceptable) technical quality.

Also, in reference to the earlier discussion point, although the camera will prevent AF based on the geometry limitation of aperture, it will not do so when light levels are limiting. In other words, a camera with f/5.6 AF points will not AF when an f/8 lens is mounted, but if you try focusing with an f/5.6 lens in an unlit coal mine it will hunt for focus in the darkness.

 

[neuroanatomist]

Since the difference between f-number and T-stop with those lenses often varies widely from one focal length to another, the difference between f-number and T-stop can not be fully explained by transmission loss. It seems that the actual entrance pupil size does not grow at the same rate with the 28-300mm from, for example, 70mm to 100mm as it does from 35mm to 70mm.

View attachment 179591

As with all variable aperture zoom lenses, the size of the entrance pupil, as magnified by the optical elements between the aperture diaphragm and the front of the lens, is not keeping up with the focal length as it is zoomed to longer focal lengths. Yet at certain points in the zoom range of these lenses they seem to "make up" some of the loss. If it was all transmission loss one would expect the difference between f-number and T-stop to be constant and the above graph would be flat lines for each of the lenses. Link to lens comparison at DxO Mark. As always, ignore the "number scores" and go to the actual measured data by clicking on 'Measurements' and then 'Transmission'.

At their worst, the deltas are less than 2/3-stop, and generally less than 1/2-stop. Keep in mind that lens specifications are rounded to the nearest incremental value, and that rounding is almost always in favor of the manufacturer. For example, the 28-300mm f/3.5-5.6 might actually be something like 29.6-291.3mm f/3.72-5.87.