Is a native EF mount coming to a Canon full frame mirrorless camera? [CR1]

[melgross]

Honestly, I haven't seen the hard numbers, but my understanding that it is the kind of price you can expect from foundries that do CMOS sensors for you. R&D and having a luxury of running your own fab grossly under its capacity are extra, but once the sensor technology matures, they won't be a big deal either.


You claimed (or at least suggested) that the cameras for such sensors (using EF optics) must be too big, too heavy and too expensive. Which is not the case.


No, they are generally harder to sell, in particular, because of a very small choice of compatible modern lenses, which are also very heavy and expensive, and that's what makes the sensor expensive (R&D and fab setup costs split between a tiny amount of sensors).

If a sensor flawlessly works with almost all big (and small; at least everything that accepts a teleconverter) whites, that would not be a problem.

And you’re making the statement that they’re not, which is not the case. fF sensors are about the largest chips made. So few are on a wafer, and that brings the cost up. There are still millions of cameras, for several years of sales. So it’s not a tiny number of sensors. Intel sells many different chips. Many of them only sell in the high hundreds of thousands. Millions isn’t a small number.

 

[stevelee]

This is interesting to me because the reason for buying a Sony camera I was told was that IBIS meant that you didn’t have to pay for IS every time you bought a lens. So Sony lenses would be smaller, lighter, and cheaper. Sounds like it didn’t turn out that way.

 

[melgross]

Sure, the component cost of the sensor is much higher with FF than crop, but the decision to use a FF sensor means that you need:

• Larger pentaprism / OVF
• Larger, potentially more robust/complicated mirrorbox assembly
• Larger, potentially more robust/complicated shutter
• (possibly) A larger, more complicated / expensive AF setup, though it may be that it may not be that different between a 5-series and (say) a 7-series for that; a 7-series covers more of the frame than a 5-series, so it may just be a similar setup covering more of a crop sensor's real estate than a FF sensor's real estate. I defer to the AF scholars here.

So yes, the sensor figures prominently in cost, but it brings in a lot of additional cost along for the ride to make use of that larger sensor.

- A

Well, we just need to compare the top DSLR price to what the price of the top SLR. Yes, they stopped making them. But when they were making both, for years, the top DSLR model was over twice the price of the SLR model. Much of the mechanical shutter system was the same, and a much less complex system for sensor mounting and memory cards vs. a mechanical film wind system, etc.

The aps-c model cost as much as the top mechanical model.

At one point, I read that the sensor was well over half the cost of the camera. That ratio has come down no doubt, but it’s still pretty high.

 

[denstore]

And what purpose would that extra half stop serve? Focus would be poor. Even now, auto focus struggles with f 1.4. Manual focus these days is a joke.

Strange, I rarely have had problem with focus at f/1.4. Not with f/1.2 either when I took time to micro adjust the cameras. But maybe I’m not as picky. But I do like nice background blur, and I do find the 50/1.2 to be nicer than the 50/1.4. Is it worth the 2.5x higher price? It is to me, but your mileage may vary.

 

[ahsanford]

This is interesting to me because the reason for buying a Sony camera I was told was that IBIS meant that you didn’t have to pay for IS every time you bought a lens. So Sony lenses would be smaller, lighter, and cheaper. Sounds like it didn’t turn out that way.

Not smaller. Not at all. Physics is a cruel mistress this way.

Somewhat lighter is possible without IS, but (a) mount differences can mask some weight differences and (b) Lens IS is more effective than IBIS for longer focal lengths so Sony may put lens IS on a lens anyway (70-200 GM as one example). It's only if a company has both an IS and non-IS version of the same lens that you can appreciate how little the IS weighs, but if you compared a Nikon 24-70 IS and Sony 24-70, there's certainly going to be more going on than just the presence of IS or not.

Cheaper is again possible but not at Sony's paltry marketshare, which lower production volumes raise cost compared to Canon.

But don't throw the baby out with the bathwater. IBIS can deliver some of what it promises.

- A

 

[3kramd5]

Well, we just need to compare the top DSLR price to what the price of the top SLR. Yes, they stopped making them. But when they were making both, for years, the top DSLR model was over twice the price of the SLR model. Much of the mechanical shutter system was the same, and a much less complex system for sensor mounting and memory cards vs. a mechanical film wind system, etc.

The aps-c model cost as much as the top mechanical model.

At one point, I read that the sensor was well over half the cost of the camera. That ratio has come down no doubt, but it’s still pretty high.

I expect that is very much an “it depends” situation. If you’re buying sensors from a semicon powerhouse by the hundreds of thousands, you probably stand to see them cost less than if you’re developing and producing your own sensors for small production run cameras, and maintaining a line for the purpose. I wouldn’t be surprised if the sensor cost in a d850 is lower than in a 5Div, for example.

 

[ahsanford]

I expect that is very much an “it depends” situation. If you’re buying sensors from a semicon powerhouse by the hundreds of thousands, you probably stand to see them cost less than if you’re developing and producing your own sensors for small production run cameras, and maintaining a line for the purpose. I wouldn’t be surprised if the sensor cost in a d850 is lower than in a 5Div, for example.

I would be a little surprised, actually.

5D4 volumes are higher than D850 volumes (presumed based on market share).

Further, the 5D4 sensor is designed and made in-house while the D850 sensor is made by Sony. Unless Sony has some economies of scale from other sensor fab that makes all sensors cheaper in their shop to produce, Nikon is likely paying more than if they did what Canon did and made those sensors themselves. Sony may be hitting them with licensing fees for using Sony IP in their sensor design, or possibly Nikon outsourced the design of the entire sensor itself to Sony (which would cost quite a bit).

We'll never know, but my guess is that relying on a competitor to make your sensors is not cheaper than doing it yourself unless making sensors fundamentally isn't your core competency and learning how to do that might cost too much. But as we know, Nikon still makes its own sensors, so I don't think that's a consideration here. My guess is that Nikon is paying more than they need to in order to get access to those lovely high-DXO-score sensors.

- A

 

[3kramd5]

Unless Sony has some economies of scale from other sensor fab that makes all sensors cheaper in their shop to produce

Right, that’s something I postulate may be. Sony Semicon makes a tremendous volume of sensors. Not these ones, of course, but they may have processes and equipment which give them a competitive advantage in fab.

We'll never know, but my guess is that relying on a competitor to make your sensors is not cheaper than doing it yourself unless making sensors fundamentally isn't your core competency and learning how to do that might cost too much. But as we know, Nikon still makes its own sensors, so I don't think that's a consideration here.

- A

They do (still make sensors)?

Maybe they went to Sony for scale, similar to Samsung on the galaxy phones (where Samsung fabbed some and Sony fabbed some). Regarding my business (airborne C4ISR electronics), in few situations are we able to beat supplier costs in house even with their margins. Our labor is too expensive. Perhaps my experience non-applicable to this industry.

 

[ahsanford]

They do (still make sensors)?

Believe so, esp. for their highest end sensors for sports rigs, but even that is changing. Of late, 'designed by Nikon and produced by Sony' is apparently becoming common for them:

https://petapixel.com/2018/07/17/yes-nikon-designs-its-own-sensors/

Meanwhile, PP also claimed that the D5, D500 and D850 were all made by Sony.

- A

 

[BillB]

Believe so, esp. for their highest end sensors for sports rigs, but even that is changing. Of late, 'designed by Nikon and produced by Sony' is apparently becoming common for them:

https://petapixel.com/2018/07/17/yes-nikon-designs-its-own-sensors/

Meanwhile, PP also claimed that the D5, D500 and D850 were all made by Sony.

- A

Now and then, I wonder whether Nikon is on a trajectory to end up as part of Sony. Of course, that would depend on Sony being willing to work out a deal with Nikon to take over their photography operation.

 

[neonlight]

Nature of the AF drive in lens and its control makes the difference. On-sensor Phase-Af like Canon DP-AF gives a different signal compared to separate DSLR Phase AF sensor. Modern lenses designed for mirrorfree camera systems with on-sensor AF typically are focus-by-wire and have linear electromagnetic drives, rather than rotational AF drives plus mechanically coupled manual focussing gear as in almost all EF / L lenses.

I was really thinking of the optics. Canon changes its drive electronics all the time. The phase AF and DPAF are derived from diferent chips but the control processor DIGIC xxx will presumably talk to these and send commands down the same wires (I assume) therefore EF lens will autofocus with whatever motor it has. You have not said anything which rules out EF on FF MILC, but EF might work faster with different motors in future - but it seems Canon will ensure older EF's will still work with at least one of their FF MILC bodies.

 

[3kramd5]

Believe so, esp. for their highest end sensors for sports rigs, but even that is changing. Of late, 'designed by Nikon and produced by Sony' is apparently becoming common for them:

https://petapixel.com/2018/07/17/yes-nikon-designs-its-own-sensors/

Meanwhile, PP also claimed that the D5, D500 and D850 were all made by Sony.

- A

I can’t find any indication Nikon owns any fab capacity (not that my web search skills are authoritative ).

http://image-sensors-world.blogspot.com/p/image-sensor-companies-list.html

I believe nikon does sensor design work, though they likely leverage much of the IP from the fab (e.g. the BSI process from Sony). Fabricating in-house at this point would require a huge capital expense that would ripple into the unit sensor costs.

But indeed maybe I muddied the waters too much comparing canon fab to Sony fab rather than hypothetical nikon fab to Sony fab.

 

[ken]

I would be a little surprised, actually.

...<snip>... My guess is that Nikon is paying more than they need to in order to get access to those lovely high-DXO-score sensors.

- A

Or perhaps they've cross-licensed some of their own camera / lens patents to Sony as part of the compensation.

 

[melgross]

Strange, I rarely have had problem with focus at f/1.4. Not with f/1.2 either when I took time to micro adjust the cameras. But maybe I’m not as picky. But I do like nice background blur, and I do find the 50/1.2 to be nicer than the 50/1.4. Is it worth the 2.5x higher price? It is to me, but your mileage may vary.

Autofocus for high speed lenses is a known problem. All of the tolerances in a camera make the razor thin focus zone extremely difficult to hit, and the slightest movement can knock it out. My main problem starts with f 1.2, but really it’s the call for f 1, and even faster, lenses that has me shaking my head.

 

[scyrene]

I can't speak for that poster, but some folks see mirrorless as a chance to smash the limits/orthodoxy/constraints we live with in our current systems. Some folk look to mirrorless not for size or the potential upsides of mirrorless, but because it's their company's one chance in 20-30 years to try something new: f/0.95 lenses, medium format mirrorless, etc. comes to mind.

In short, while the mount / sensor size / etc. are unknown, anything could be possible. Some people are off the races with that notion. Let them dream -- no harm there.

- A

No, but I want lenses as fast or faster than the ones available for EF today. It’s a part of lens development as important as ever.
If Nikon brings out a new system, with really fast glass available and good ergonomics, and Canon at the same time decides to make small bodies, with few and mostly slow lenses, I will probably go with the Nikon. I’m not that bound to any specific brand. But I do dislike the trend of small cameras, with retro style flat and angular bodies.

Well if the Canon mirrorless camera is native EF, then it already has native f/1.2 lenses (and indeed, although discontinued, a native f/1.0 lens). If they choose a new mount, I'd be astonished if one of the first lenses to be released was an ultrawide aperture - and that goes for Nikon as well - even if the new mount theoretically supports it. While everyone is motivated by different features, I'd contend that the market for lenses wider than f/1.2 is vanishingly small.

Edit: sorry, I replied before catching up on the thread, I see many good points have been made and mine are a bit superluous now.

 

[melgross]

I expect that is very much an “it depends” situation. If you’re buying sensors from a semicon powerhouse by the hundreds of thousands, you probably stand to see them cost less than if you’re developing and producing your own sensors for small production run cameras, and maintaining a line for the purpose. I wouldn’t be surprised if the sensor cost in a d850 is lower than in a 5Div, for example.

I wouldn’t be, because it’s very likely that the 5 sells in substantially higher numbers. A lot of companies go in-house because of lowered costs. Tesla is doing that with their computer, dropping Nvidia for their own in-house developed unit, though they aren’t actually making them. We’re just talking about a very few hundred thousand.

Canon has a large chip manufacturing business, and have for many years. They aren’t a regular foundry, but also make their own camera processing chips in the millions, as well as others. They also make chips for LEDs, etc.

 

[melgross]

Believe so, esp. for their highest end sensors for sports rigs, but even that is changing. Of late, 'designed by Nikon and produced by Sony' is apparently becoming common for them:

https://petapixel.com/2018/07/17/yes-nikon-designs-its-own-sensors/

Meanwhile, PP also claimed that the D5, D500 and D850 were all made by Sony.

- A

It’s believed that Nikon has a licensing deal with Sony that allows Nikon to modify their sensors. Nikon claims, publicly, that they design their own sensors. That Nikon modified Sony sensor is probably the reality.

I haven’t heard anything about Sony making cameras for them. That would shock me!

 

[neuroanatomist]

Autofocus for high speed lenses is a known problem. All of the tolerances in a camera make the razor thin focus zone extremely difficult to hit, and the slightest movement can knock it out.

"A known problem." Seems like I need to pull out my handy-dandy academic publication translator.

"It is known...," really means, "I think."

"It is widely known...," really means, "Me and a few of my friends think..."

You're referring to motion. Yes, the DoF at, for example 85mm f/1.2, is razor thin. But if you or the subject move after AF is locked, that's not an AF problem. Have you tried AI Servo? FYI, Canon specifies AF precision as either within one depth of focus at max aperture for f/5.6 AF points, or within 1/3 of the depth of focus for f/2.8 AF points, at the lens' max aperture regardless of that f/number. Relative to their max aperture, fast lenses are no less precise than slow ones (and can be more precise with f/2.8 AF points).

 

[ahsanford]

Autofocus for high speed lenses is a known problem.

"A known problem..."

I can't call it a known problem, but I personally have had night and day difference in AF accuracy/consistency from the 50 f/1.2L and 85 f/1.4L IS on my 5D3.

I think it's more fair to say that 'autofocus for some large aperture lenses can be problematic on the wide open end'.

The 50L has been a finnicky diva for me over the span of two rentals. Even after AFMA, with ruthless technique (stationary subjects, 1/60 or faster shutter speed, no focus and recompose, single AF point etc.), the lens seemed to simply whiff with the AF 10-20% of the time unless I was stopped down to f/2.8 or narrower. It drove me nuts, if I'm honest.

The 85 f/1.4L IS, on the other hand, was so so so much better with the same careful shooting approach. The AF was simply automatic for me -- it was a joy to use and not have to worry about the gear letting me down. I shot that thing wide open without fear and without disappointment.

I rarely blame my tools as much more often than not I am the reason something didn't go as planned. But I believe in this case I was doing everything humanly possible to succeed with a wide aperture lens and the 50L let me down while the 85 f/1.4L IS was lights out.

So as much as I agree with Neuro that the lens itself is no more/less precise with AF than slower lenses, the actual consistency of nailing the focus in practice does seem to vary with certain lenses. LensTip and others who have attempted to publish AF hit rates would tend to agree with this, but they don't publish all their methods, the test cameras change over time (making lens AF comparisons challenging), etc. so it's hard to rely on them as a truly useful source of information on the subject.

- A

 

[Kit.]

And you’re making the statement that they’re not, which is not the case.

That's not true. You asked me how big and expensive I would want a camera with such a sensor to be. I referred you to a pretty big and a pretty expensive camera line, $1500 more expensive than another camera of the same manufacturer with the same sensor format.

I am not asking for a small or a cheap camera.

fF sensors are about the largest chips made. So few are on a wafer, and that brings the cost up.

The MF sensors are bigger (Sony, for example, produces IMX211 sensor, which is bigger by area than the square for the supposed round sensor would be). Still, it's not the cost of silicon that drives their price up. Silicon is not that expensive.

There are still millions of cameras, for several years of sales. So it’s not a tiny number of sensors.

Are they selling millions of FF sensors of the same model?

They definitely are not selling millions of MF sensors of the same model.

Intel sells many different chips. Many of them only sell in the high hundreds of thousands. Millions isn’t a small number.

If you are about Itaniums, they are very expensive. More expensive than some FF cameras, despite a smaller die size than an FF sensor has.