Patent: Fast prime lenses for curved image sensors

Del Paso · Mar 19, 2021

F 1,0 lenses will certainly be quite affordable.
For Rolls Royce drivers...

DrToast · Mar 19, 2021

It seems like a curved sensor is a perfect match for a fixed lens camera. Get it done, Canon!

BeenThere · Mar 19, 2021

Can’t say I would be interested in another set of lenses, but maybe some specialty could find it worth while. Seems as though the advantages would have to be very compelling.

vladk · Mar 19, 2021

I would expect this approach to be applied to extra compact industrial applications first (surveilance, monitoring, etc.)

Stuart · Mar 19, 2021

Why would it need a new Lens mount? Isn't the curved idea about getting wavelengths to fall more evenly at sensor edges the same physical mount will still work for the centre of the sensor.

Or could movable elements within the lens focus the light according to the body attached.

Or would curved sensors be used on APS-c bodies and in that way be a whole new smaller mount system.

Josh Leavitt · Mar 19, 2021

I'd love a trio of fixed full-frame curved sensor cameras: 24mm f/1.2, 50mm f/1.2, and 100mm f/1.2 would be my choice. The idea of curved sensors has been around for several years now, but to my knowledge the technical hurdle has been the brittleness of the silicon substrate when attempting to bend it under pneumatic or hydraulic pressure. The larger the sensor, the more likely it is to break. Hopefully Canon will find a workaround for it.

Antono Refa · Mar 19, 2021

Stuart said:
Or would curved sensors be used on APS-c bodies and in that way be a whole new smaller mount system.

I'd think the smaller the sensor, the lesser the problem of lenses having a curved image plane.

Dragon · Mar 19, 2021

With a set of (relatively) small primes that fast, I can see quite a few folks going for a system like that. The 85 f/1 and the 100 f/1.2 would be a dream come true for portrait shooters.

melgross · Mar 20, 2021

Mmm. Two series of lenses, two series of bodies. Yes, I can see by the speed of those lenses that this would be interesting, but that just seems to be breaking the line up further.

the secret of film lenses is that they (at least, almost all of them) weren’t designed quite with a flat field in mind. Pretty close, not not completely. The problem is that film is never truly flat. The corners and edges are curled a tiny bit towards the front of the film plane. Unless you have a vacuum holder, of course. Not everyone did, did they? Mostly just specialized work, and then, usually specialized lenses. I had a copy machine in my lab for reproduction work, you know, halftone dots. Pretty critical everything matches up. So vacuum holders for the original and the four big film halftone repro. True apo flatfield f9 lenses. But otherwise, nah.

use most of those lenses on digital, and often the corners, and possibly the edges, are out. Maybe not by much, but you’ll see it.

lens design is easier if flat fields aren’t required. Hence all of those f1.0 and f1.2 lenses there.

chik0240 · Mar 20, 2021

I don't think curved sensor would work for a changeable lens camera, it should be better suited to fixed lens camera for great image quality with insane aperture or insane wide aperture long zoom lens with reasonable lens size.

Goran56 · Mar 20, 2021

melgross said:
Mmm. Two series of lenses, two series of bodies. Yes, I can see by the speed of those lenses that this would be interesting, but that just seems to be breaking the line up further.

the secret of film lenses is that they (at least, almost all of them) weren’t designed quite with a flat field in mind. Pretty close, not not completely. The problem is that film is never truly flat. The corners and edges are curled a tiny bit towards the front of the film plane. Unless you have a vacuum holder, of course. Not everyone did, did they? Mostly just specialized work, and then, usually specialized lenses. I had a copy machine in my lab for reproduction work, you know, halftone dots. Pretty critical everything matches up. So vacuum holders for the original and the four big film halftone repro. True apo flatfield f9 lenses. But otherwise, nah.

use most of those lenses on digital, and often the corners, and possibly the edges, are out. Maybe not by much, but you’ll see it.

lens design is easier if flat fields aren’t required. Hence all of those f1.0 and f1.2 lenses there.

I guess the idea is great! probably it will be easier to reach light and the same amount to the corners as well as in the center and sharpness will also bas as good at the corners. Maybe the problem with broken pixels could be helped by extra layer between the bendable back and the sensor layer, a soft one which might help not to brake the pixels.

Dragon · Mar 20, 2021

To achieve the degree of curvature indicated on the sensor it would have to be formed when very hot (i.e. soft). I wonder if Canon has figured out how to do this, or if the lens patents are just a hedge against the time when someone does figure out how to make a curved sensor.

fox40phil · Mar 20, 2021

No new mount... but I want those f1.0 lenses and the 100mm 1.2!!

melgross · Mar 21, 2021

Dragon said:
To achieve the degree of curvature indicated on the sensor it would have to be formed when very hot (i.e. soft). I wonder if Canon has figured out how to do this, or if the lens patents are just a hedge against the time when someone does figure out how to make a curved sensor.

They could just be grinding the material into the final shape before the other operations take place. After all, lens grinding is a pretty mature technology.

Dragon · Mar 22, 2021

melgross said:
They could just be grinding the material into the final shape before the other operations take place. After all, lens grinding is a pretty mature technology.

That would translate to a huge use of purified silicon, not to mention a whole new lithography process to properly image on the curved surface. Theoretically possible, but not even remotely cost effective.

David - Sydney · Mar 22, 2021

Del Paso said:
F 1,0 lenses will certainly be quite affordable.
For Rolls Royce drivers...

The drivers don't get paid that much.... RR owners of the other hand!

Jerryrigged · Mar 23, 2021

Dragon said:
That would translate to a huge use of purified silicon, not to mention a whole new lithography process to properly image on the curved surface. Theoretically possible, but not even remotely cost effective.

Yeah, right?! Seems like a curved sensor had lots of hurdles to overcome to be economical!

Del Paso · Mar 23, 2021

David - Sydney said:
The drivers don't get paid that much.... RR owners of the other hand!

All wrong.
You've got no idea how much mine costs me !

SteveC · Mar 23, 2021

Del Paso said:
All wrong.
You've got no idea how much mine costs me !

It's possible you misunderstood his joke...I can't be sure whether you did or not. He's assuming the driver is a chauffeur (who probably can't afford an R5 unless he eats Ramen and has no family) and the owner of the car (not the Rolls Royce company) is in the back seat.

However, I am sure a lot of Rolls Royce owners do drive their own vehicles.