50MP Canon 5DS/R spec leaked

[bcooter]

It's the same thing every 18-24 months.

It's obvious Doug.

Nothing is ever talked about in digital camera forums more than a gazillion pixels and DeeeRRRR.  Maybe corner sharpness (fill in one of those perplexed little yellow faces).

But be fair, you guys sold those two phrases for 10 years, so it's kind of funny your tired of it.

Honestly, I have no dog in this fight, because I would be very surprised to find myself standing in line with the thought I "had" to buy a new still only camera.

I might buy one because I wanted to, but not because I felt I needed to.    Big difference.

But what would be kind of cool is if some real modern useable tech was added to all these pixels.  

PDAF focusing, modular systems, a sensor that had image stabilization.

Ever pick up an olympus em-1.  Obviously it's a tiny sensor and seems like no threat to you guys,  but that floating sensor is amazing.  

You can shoot something static, hand held at unheard of shutter speeds.

For video it's like a built in steady-cam.  

And modular, olympus even makes a viewfinder that clips on top of the viewfinder.  

Think about it.  You could have an optical finder and a clip on evf in one camera.

That right angle grip thing that everybody sticks below a camera could also be modular.  

One for stills that has more power and . . .  I dunno . . .  faster frame rates or holds your sunglasses.

How bout' an optional grip that allowed for better video encoding, xlr inputs, even storage options.

Even an extra port for a touch screen finder that worked like the cheapo Canon 70d.

Selectable formats like . . . the em-1.   4:3 for vertical stills, 2:3 for horizontal stills and 16:9 and 2:1 for video, because the super 35mm format works well for motion capture.

After all there a two billion PL mount lenses floating around.

Or how about a global shutter?

Then different lens mounts.   Obviously the mirrorless cameras do it, even my Leica S2 allow me to put on lenses from three makers, with full functionality.

Now software.    I can't imagine the user interface of lightroom that actually did keying, face tracking and timeline stacking like Di Vinci, without the arcachic PCness of Di Vinci.

I understand you guys don't wanna sell little cameras because there is no margin in them and I'm not talking about a whiz bang version of a 1dc for twenty bucks.

Something that really went forward, was long term useable without tossing out a lens set or learning how to become an ebay seller.

But honestly, It's kind of hard for me to think about springing for a still camera that lists at the same price as a real movie camera.

At that point you'd have a camera and a system  that really would bridge both worlds of motion and stills and something worth talking about.

Something that would put me at the counter.
.

IMO

BC

[Jim Kasson]

With digital, small sensors are faster and easier to design, produce and sell...

I can understand that small sensors are easier to produce -- yield and defect density. I guess they're faster to produce since you can get more of them on a wafer. I can't understand why they're easier or faster to design. We're not using drafting boards, X-acto knives and Rubylith any more. A bigger chip and more pixels is just a matter of specs to the CAD software, isn't it?

Jim

[jduncan]

No one ever really talked about the D3x, the Sony A900 had the same sensor for a third of the price, and the 5D2 was the hot topic with live view and video. I personally know two photographers who dumped their MF/LF film kits over the 5D2.

Pretty big shoes to fill, seeing as it's Canon after all, it needs to be able to match the Nikon D8xx series let alone the 645 systems.

In the olden days, the only factor was film size, as every camera had every type of film available to it. With digital, small sensors are faster and easier to design, produce and sell, while larger sizes need proven technology that can produce a good enough yield for the required dimensions, resulting in them getting stuck with decades old base technology, like if development on CRT displays continued to this day. The Sony 33x44mm sensor is the very first medium format sensor that's actually based on the same modern technological base as smaller cameras, and is a good indicator of how things would have been if large-size CMOS wasn't prohibitive to manufacture.

"Matching resolution" is a pretty bold statement if you're only counting pixels - the 36mp D800 in some ways had let down many early adopters who discovered that Nikon glass is woefully inadequate to resolve the sensor, and began adapting other brands...

In my opinion, MF needs to, and may well soon go mirrorless too.. and I don't mean tech cams, just regular cameras with autofocus. If Sony has any sense about them, and they continue developing large sensors, it may no longer be that MF will have to make do with old tech.

All it's going to take is a relatively inexpensive MF camera, like a big Sony mirrorless, to introduce more people to the concept and it's going to snowball from there. The ice will get broken sooner or later.
 

I'm happy there's a Sony Exmor 33x44mm sensor in the first place, never mind size. Sony says it's difficult to manufacture as it is, so give it time, maybe give them money too. I would much rather have 51 million pixels that are over 5 microns big each.

As I said, using proper DR and Canon in the same sentence is dicey given their history, but the problem of resolution is also one of file size. Do I really want to shoot the 5D R knowing i'm using up 60mb+ each time for detail I may not even get? People who will buy this camera will almost certainly be those who have TS-Es, Otuses, and the new Sigma Art lenses and have been waiting for so long for something just like this.

Basically I don't see your comments as negating what I say,  more an extension. But 4 notes:
1) I say mirror-less commodities, not just mirror-less.  What I am conveying is that I believe mirror-less cameras will commodity full frame sensors (of the 135 kind). No complex by hand manufacturing.  Nikon issues with both the  D600 and D750 show that assembly and tolerances are an issue when you are trying to build at the lower costs possible. Will be far easier to address with a fully digital camera (even robots could assembly the camera)
2) I was happy to see the Sony sensor too, but was concern about size.
3) I am fully aware of Canon dynamic range, my DSLR is a Nikon, if  I new it was a Sony sensor I would not have mention DR.
4) It could be that the camera is a hoax we don't have Canon confirmation yet.
Best regards,

[ErikKaffehr]

Well,

When image quality is good enough for ones needs other thins matter a lot. Obviously other things matter to different people.

Many of the things BC mentions are desirable. But again, not all people want/need the same features. I would agree that 4/3 cameras have a lot very good features.

As a landscape shooter making stills I don't need video that often, but when presenting stills a short video can really enhance a presentation. So I feel video is important also for landscape shooters. In that context I would prefer 4K video.

I still feel that more pixels are better when printing large, say 1m (24") or so. I am also quite positive about the small pixels. Good lenses, and there are plenty of them, deserve a sensor that gives a correct image.

Best regards
Erik

It's obvious Doug.

Nothing is ever talked about in digital camera forums more than a gazillion pixels and DeeeRRRR.  Maybe corner sharpness (fill in one of those perplexed little yellow faces).

But be fair, you guys sold those two phrases for 10 years, so it's kind of funny your tired of it.

Honestly, I have no dog in this fight, because I would be very surprised to find myself standing in line with the thought I "had" to buy a new still only camera.

I might buy one because I wanted to, but not because I felt I needed to.    Big difference.

But what would be kind of cool is if some real modern useable tech was added to all these pixels.  

PDAF focusing, modular systems, a sensor that had image stabilization.

Ever pick up an olympus em-1.  Obviously it's a tiny sensor and seems like no threat to you guys,  but that floating sensor is amazing.  

You can shoot something static, hand held at unheard of shutter speeds.

For video it's like a built in steady-cam.  

And modular, olympus even makes a viewfinder that clips on top of the viewfinder.  

Think about it.  You could have an optical finder and a clip on evf in one camera.

That right angle grip thing that everybody sticks below a camera could also be modular.  

One for stills that has more power and . . .  I dunno . . .  faster frame rates or holds your sunglasses.

How bout' an optional grip that allowed for better video encoding, xlr inputs, even storage options.

Even an extra port for a touch screen finder that worked like the cheapo Canon 70d.

Selectable formats like . . . the em-1.   4:3 for vertical stills, 2:3 for horizontal stills and 16:9 and 2:1 for video, because the super 35mm format works well for motion capture.

After all there a two billion PL mount lenses floating around.

Or how about a global shutter?

Then different lens mounts.   Obviously the mirrorless cameras do it, even my Leica S2 allow me to put on lenses from three makers, with full functionality.

Now software.    I can't imagine the user interface of lightroom that actually did keying, face tracking and timeline stacking like Di Vinci, without the arcachic PCness of Di Vinci.

I understand you guys don't wanna sell little cameras because there is no margin in them and I'm not talking about a whiz bang version of a 1dc for twenty bucks.

Something that really went forward, was long term useable without tossing out a lens set or learning how to become an ebay seller.

But honestly, It's kind of hard for me to think about springing for a still camera that lists at the same price as a real movie camera.

At that point you'd have a camera and a system  that really would bridge both worlds of motion and stills and something worth talking about.

Something that would put me at the counter.
.

IMO

BC

[Gel]

The main draw for me with Medium format is a larger sensor. If it wasn't for the larger sensor look and subsequent increase in image quality I'd be 35mm all the way.

The current innovation we're getting now means that sooner or later there will be another one or two manufacturers offering a medium format cam. It's not a case of it, but when. It's important to recognize that businesses are out to make money. Usually, any way they can and if a manufacturer felt there was a gap in the market for them to explore then you betcha they'd exploit it.

[hjulenissen]

I can understand that small sensors are easier to produce -- yield and defect density. I guess they're faster to produce since you can get more of them on a wafer. I can't understand why they're easier or faster to design. We're not using drafting boards, X-acto knives and Rubylith any more. A bigger chip and more pixels is just a matter of specs to the CAD software, isn't it?

Jim

Maybe not, but it seems that designing sensors for a small niche (even if those customers have deep pockets) is comparatively "harder" than designing for the mass market.

I don't know much about sensor design, but I am guessing that starting with the excellent APS-C (and smaller) sensel designs, there is a significant amount of work and risk before you have scaled this to a larger sensor that can be manufactured economically. Not merely doing "copy & paste" of a file representing the physical layout of a single sensel.

-h

[Steve Hendrix]

Maybe not, but it seems that designing sensors for a small niche (even if those customers have deep pockets) is comparatively "harder" than designing for the mass market.

I don't know much about sensor design, but I am guessing that starting with the excellent APS-C (and smaller) sensel designs, there is a significant amount of work and risk before you have scaled this to a larger sensor that can be manufactured economically. Not merely doing "copy & paste" of a file representing the physical layout of a single sensel.

-h

Yes, it's inherently difficult and expensive to scale chip technology, not to mention a completely new lineup of (also larger) optics. So the yield has to be lucrative to make the effort worth the trouble. When ventures like that don't pan out, it isn't pretty. Oh yeah, there's an opportunity, but what's the risk/reward, how much time and effort is it going to take, and what will the potential gains be vs putting that time, effort, and capitol somewhere else.


Steve Hendrix
Capture Integration

[Kolor-Pikker]

Basically I don't see your comments as negating what I say,  more an extension.

Yeah I wasn't arguing against anything, just clarifying some points that I figured didn't tell the whole story.

I can understand that small sensors are easier to produce -- yield and defect density. I guess they're faster to produce since you can get more of them on a wafer. I can't understand why they're easier or faster to design. We're not using drafting boards, X-acto knives and Rubylith any more. A bigger chip and more pixels is just a matter of specs to the CAD software, isn't it?

Jim

Basically, all computer chips, whether it be the processor in your PC or the sensor in your camera, are made using a process called photolithography. A silicon base is used as the insulator, and then layers of conductive material are layed on top and the excess parts are burnt away with a laser through a stencil. The stencil is a 2D mask of what the entire surface of the chip will look like at a given layer, and many masks are used to etch in the nano-scale wires, channels, and other bits that make up the chip.

The problem is that focusing a mask with sufficient detail for small features and have it cover a large suface area at the same time doesn't always work. Imagine it as comparing a 1080p projector vs. a 4K projector, both can cover the same screen area, but one can put more dense detail per unit of measure and/or cover a larger area, however it also costs several times as much. When a lithography lab may cost billions (honestly don't know), a significant increase in scale may not be economically viable, especially when the product it'll manufacture will be for a niche market.

The fact is that there are very few (if any) chips that actually need to be as large as photographic sensors, so it's not possible to adapt hardware from other manufacturing methods. To make a full-size 645 sensor in one pass, someone would have to take the plunge to develop a lithographic system designed specifically for this.

[BJL]

I can understand that small sensors are easier to produce -- yield and defect density. I guess they're faster to produce since you can get more of them on a wafer. I can't understand why they're easier or faster to design.

Production challenges are probably the main ones, like the need for on-wafer stitching when a sensor is larger than the industry-standard 33x26mm limit of all suitable fab. equipment.  (If I were a camera-maker investigating a new format needing new lenses, it would not be 54x42mm or any other flavor of medium format; it would be about 32x24mm, to be as big as possible without the need for that on-wafer stitching: "medium format" in the new sense of being between "APS-C" and 35mm.)

But the longer signal paths and great number of pixels per row and per column probably add some design challenges, in both speed and synchronization, especially with the more complicated wiring and read-out systems of modern active pixel CMOS sensors, which involves addressing individual pixels and reading some or all pixels in a row simultaneously.

[jduncan]

Production challenges are probably the main ones, like the need for on-wafer stitching when a sensor is larger than the industry-standard 33x26mm limit of all suitable fab. equipment.  (If I were a camera-maker investigating a new format needing new lenses, it would not be 54x42mm or any other flavor of medium format; it would be about 32x24mm, to be as big as possible without the need for that on-wafer stitching: "medium format" in the new sense of being between "APS-C" and 35mm.)

But the longer signal paths and great number of pixels per row and per column probably add some design challenges, in both speed and synchronization, especially with the more complicated wiring and read-out systems of modern active pixel CMOS sensors, which involves addressing individual pixels and reading some or all pixels in a row simultaneously.

I agree, but at 44 x33 mm you are already stitching or using a very particular stepper.
It have been done :
http://www.dpreview.com/articles/6300767092/leaf-afiii-aptus-10-7-6

Using standard steppers an unusual 52 x 33 mm sensor could be created[1] (60MP chip using the same pixel pitch as the sony sensor). It will be interesting, and the sensor will fit within MF lenses image circle.

The size you listed is the typical one, but Canon has some steppers with a greater "Field Size"  that will allow a 52 x39mm sensor with two "exposures".  Maybe Sony is using a similar stepper for the chip, but I have no idea (my guess is that they are using a Normal stepper two exposures).

Best regards,


[1] The interesting part will be the handling of the viewfinder by example, so maybe a 48 x36 mm sensor will do, but it's more complex than two stitches with common steppers

[John Koerner]

"Matching resolution" is a pretty bold statement if you're only counting pixels - the 36mp D800 in some ways had let down many early adopters who discovered that Nikon glass is woefully inadequate to resolve the sensor, and began adapting other brands...

Lost in the debate is the fact Canon has been dramatically overhauling virtually all of their key lenses first ... could it be because they realize that, without an upgrade in lenses, having an uber-sensor is next to pointless?

Canon is coming out with an 11-24, they already have a new 24-70 II, a 70-200 II, have a brand new 100-400 and 200-400, etc.

"Now" would make the most sense to begin to introduce new ultra-sensor cameras ... not "before" these lens upgrades were made.

The ever-increasing new lens lineup is what will allow any uber-sensor to be maximized, I would think

[BJL]

I agree, but at 44 x33 mm you are already stitching or using a very particular stepper.
It have been done :
http://www.dpreview.com/articles/6300767092/leaf-afiii-aptus-10-7-6

I think you missed my point: of course sensors bigger than 33x26mm exist, but they all need the more expensive, lower-yield process of on-wafer stitching (etching part of the sensor, then moving the wafer to a new position to etch another part of it, moving with sub-micron position accuracy, and so on).

The size you listed is the typical one, but Canon has some steppers with a greater "Field Size"  that will allow a 52 x39mm sensor with two "exposures".  Maybe Sony is using a similar stepper for the chip, but I have no idea (my guess is that they are using a Normal stepper two exposures).

Two is still be stitching, but anyway, note my use of the word "suitable" when I mentioned the "33x26mm limit of all suitable fab. equipment."
I know of only two models of stepper from any maker with a field size bigger than 33x24mm:
- The discontinued Canon FPA-3000 iW with 50x50mm field size and rather large 0.8 micron minimum feature size: http://www.ventexcorp.com/downloads/FPA-3000iW+Specification.pdf
- The newer FPA-5510 iV with 52x34mm field size, but an even coarser 1.5 micron minimum feature size: http://www.usa.canon.com/cusa/semiconductor/products/semiconductor_equipment/steppers/fpa_5510iv_stepper#Specifications

As far as I can tell, even the smaller 0.8 micron minimum feature size of the FPA-3000 iW mean is unsuitable for making the pixel sizes that the photographic market uses (putting aside special lower resolution uses like X-rays, machine vision, big telescopes and so on.) Canon itself has explained in several white-papers that its needs to use stitching to make its 36x24mm sensors, and that the "APS-H" CMOS sensors of some 1D models was the largest size that it could make without stitching, even though that FPA-3000 iW already existed when Canon made its first 36x24mm sensors.

These steppers might be used to fab. some large, lower resolution sensor for X-rays, machine vision, and such.  For example the KAF-4320 (now sold by Onsemi) with its 24 micron pixels and a 50x50mm size that fits the field size of the FPA-3000 iW perfectly: http://www.onsemi.com/pub_link/Collateral/KAF-4320-D.PDF

[Gary Ferguson]

Lost in the debate is the fact Canon has been dramatically overhauling virtually all of their key lenses first ... could it be because they realize that, without an upgrade in lenses, having an uber-sensor is next to pointless?

Canon is coming out with an 11-24, they already have a new 24-70 II, a 70-200 II, have a brand new 100-400 and 200-400, etc.

"Now" would make the most sense to begin to introduce new ultra-sensor cameras ... not "before" these lens upgrades were made.

The ever-increasing new lens lineup is what will allow any uber-sensor to be maximized, I would think

Maybe there's another motivation. With 35mm DSLR sales declining it might be time to try and sell more lenses (and more expensive lenses!) per camera body. And if you need a consumer rational to encourage the upgrade then adding pixels takes some beating!

[kers]

About lenses; when cameras were only 6mp and the Canons were 11mp i was suprised to hear that according to some Nikon tech their lensens could do  25MP or more...
It sounded incredible back then.... ( not so long ago)
I have found out ( by using the v1 camera ) that at the moment most of the Nikkor lenses can do more than 70MP in the center but not in the corners... ( as you can see the problems already with 36MP)
I think it will be about the same with the Canon lenses...  difficult to get the corners at 50mp before diffraction sets in.

[BJL]

About lenses; when cameras were only 6mp and the Canons were 11mp i was suprised to hear that according to some Nikon tech their lensens could do  25MP or more...
It sounded incredible back then.... ( not so long ago)
I have found out ( by using the v1 camera ) that at the moment most of the Nikkor lenses can do more than 70MP in the center but not in the corners... ( as you can see the problems already with 36MP)

I will say it again: sensors in 36x24mm format and larger have not yet caught up with the useful resolution and detail-handling of some fine-grained low-speed monochrome films, so if some photographers had a legitimate use for those films with suitable lenses, there should also be a use (with the right lenses) for some further increases in sensor resolution.  For example, T-MAX 100 has MTF better than 50% to well beyond the 100p/mm Nyquist limit of the Sony-Nikon 36MP sensor: http://www.kodak.com/global/en/professional/support/techPubs/f4016/f4016.pdf

And fans of shallow DOF will often be happy with great resolution in the small part of the image that is actually in focus, and is usually not near the corners.

[ErikKaffehr]

Hi,

Diffraction is just about aperture and has absolutely nothing to do with corners. Another point, it is not so that corners are the most important part of the image just because resolution is improved.

The simple truth is that higher resolution gives you a better image. If you use a cooke bottle instead of a real lens, improvement will be modest, with high quality glass it will be significant. Now, it may be that the improvement will not be noticable. For instance, I have been shooting MFD 39MP for one and a half year and I still feel that those 39MP with Zeiss primes show zero benefits over my Zeiss zooms on my Sony Alpha 99 at 24 MP when printed at A2 size. But, print large and the benefits will be there.

3.8 micron pixels are nothing new, we had it on 24 MP APS-C for a long time. Clearly, 24 MP on full frame is sharper than 24 MP on APS-C, but it needs pretty heavy OLP-filtering to avoid aliasing artefacts. Small pixels need far less OLP-filtering.

A lens that works on the 7D will also work on a 5Ds, corners may of course suffer. But, the 5Ds will offer benefits over 5DII in almost all aspects in image quality, a small loss of DR may result, though.

Best regards
Erik

About lenses; when cameras were only 6mp and the Canons were 11mp i was suprised to hear that according to some Nikon tech their lensens could do  25MP or more...
It sounded incredible back then.... ( not so long ago)
I have found out ( by using the v1 camera ) that at the moment most of the Nikkor lenses can do more than 70MP in the center but not in the corners... ( as you can see the problems already with 36MP)
I think it will be about the same with the Canon lenses...  difficult to get the corners at 50mp before diffraction sets in.

[Gary Ferguson]

If you use a cooke bottle instead of a real lens

hey, don't go maligning Cooke lenses, they're a great British company!

http://www.cookeoptics.co.uk

 

[satybhat]

This has been said every time a higher res dSLR camera has been released. Anyone on this forum for a long enough time will recall it from the release of at least a dozen cameras over the year. It fundamentally misunderstands what medium format is and does; pixels and dr are just two of the many many reasons why our clients choose medium format over a 35mm alternative.

Team Phase One has been profitable and growing since the financial crises. In my evaluation it's not "despite" the 5DII, D800, A7R, 645D, 645Z etc, but because of them. Anytime a product has been released that gets people talking about, thinking about, and evaluating image quality Team Phase One has benefited.

If this camera has broad compatibility with technical cameras and view cameras, syncs up to 1/1600th, offers a waist level viewfinder, has an optical viewfinder as large and bright as a full frame 645 body, offers vertical composition without rotating the camera, provides instant focus mask on-LCD, customizable exposure warning, automatic correction of horizon and perspective based on the internal levels, a color look developed via close collaboration between the hardware and software team, and comes with a suite of lenses already proven to provide fantastic quality, at any aperture, at 80mp (let alone 50mp)... then maybe it will threaten medium format.

Otherwise it's just another 35mm SLR with a similar resolution to one of the lower end medium format options on the market. Surely a great camera for many applications, and undoubtedly a very good value for the dollar, but, fundamentally, still just another 35mm SLR.

+1.
Basically, not many camera systems out there which will give all of the above AND ISO 35 or ISO 50 when you want it.

[eronald]

Canon have traditionally benefitted heavily as a company from their CMOS tech, in cameras, printers , scanners, copiers, and everything electronic. I'd be very surprised if they ditched the sensor part - it would be a major slap in the face for them, and de facto an indication that they have lost the ability to do R&D in their core business.

Edmund

I think you missed my point: of course sensors bigger than 33x26mm exist, but they all need the more expensive, lower-yield process of on-wafer stitching (etching part of the sensor, then moving the wafer to a new position to etch another part of it, moving with sub-micron position accuracy, and so on).

Two is still be stitching, but anyway, note my use of the word "suitable" when I mentioned the "33x26mm limit of all suitable fab. equipment."
I know of only two models of stepper from any maker with a field size bigger than 33x24mm:
- The discontinued Canon FPA-3000 iW with 50x50mm field size and rather large 0.8 micron minimum feature size: http://www.ventexcorp.com/downloads/FPA-3000iW+Specification.pdf
- The newer FPA-5510 iV with 52x34mm field size, but an even coarser 1.5 micron minimum feature size: http://www.usa.canon.com/cusa/semiconductor/products/semiconductor_equipment/steppers/fpa_5510iv_stepper#Specifications

As far as I can tell, even the smaller 0.8 micron minimum feature size of the FPA-3000 iW mean is unsuitable for making the pixel sizes that the photographic market uses (putting aside special lower resolution uses like X-rays, machine vision, big telescopes and so on.) Canon itself has explained in several white-papers that its needs to use stitching to make its 36x24mm sensors, and that the "APS-H" CMOS sensors of some 1D models was the largest size that it could make without stitching, even though that FPA-3000 iW already existed when Canon made its first 36x24mm sensors.

These steppers might be used to fab. some large, lower resolution sensor for X-rays, machine vision, and such.  For example the KAF-4320 (now sold by Onsemi) with its 24 micron pixels and a 50x50mm size that fits the field size of the FPA-3000 iW perfectly: http://www.onsemi.com/pub_link/Collateral/KAF-4320-D.PDF

[hjulenissen]

And fans of shallow DOF will often be happy with great resolution in the small part of the image that is actually in focus, and is usually not near the corners.

When using my large aperture lenses for shallow DOF, I seldom feel the need for more than my current 18MP. The 85 f/1.8 is a good lens, but it is not razor sharp at f/1.8.

Now, at moderate apertures (where my lenses tends to perform better), I might have some benefit from bumping the pixel count.

To appreciate the benefits of high sensel count sensors the most, you would want to use lenses that are diffraction limited at relatively large apertures (but this will probably not be their maximum aperture), i.e. expensive ones. This affects the DOF and the center vs corner performance that can be had.

Something like the Otus 55 f/1.4 seems to have peak center sharpness at f/4.0 and peak corner sharpness at f/7.1 or so (at least in one particular measurement on one particular camera). This gives you a quite narrow range of apertures to choose from if you want to maximize sharpness on a 50MP camera.
http://www.lenstip.com/390.4-Lens_review-Carl_Zeiss_Otus_55_mm_f_1.4_ZE_ZF.2_Image_resolution.html

Luckily, photography is about more than maximizing sharpness/detail.

-h