Sony's new curved sensor - this is a big deal

Started 3 months ago | Discussions
alanr0
Senior MemberPosts: 1,249
Like?
Believe it or not
In reply to peevee1, 3 months ago

peevee1 wrote:

webrunner5 wrote:

There was a thread somewhere on this site that had a link to some testing site that pretty much stated that most "Film Era" legacy lenses were pretty bad performers on modern day digital cameras. And they had tons of example pictures and lenses they tested. And most has to do to to the rear lens element on old stuff. Even coating and amazingly the thickness of the glass in front of the sensor. I remember a Leica sensor has a 1mm thick glass and m4/3 has 4mm glass in front. The thicker the glass the worse old lenses do. YMMV.

4mm? It is VERY hard to believe, especially on cameras with IBIS. Maybe Pana. IBIS must be really strong to handle that weight.

http://www.lensrentals.com/blog/2014/06/the-glass-in-the-path-sensor-stacks-and-adapted-lenses

Believe it or not, this is what Roger Cicala's LensRentals blog published

Cheers.

-- hide signature --

Alan Robinson

Reply   Reply with quote   Complain
alanr0
Senior MemberPosts: 1,249
Like?
Re: How curved?
In reply to stevo23, 3 months ago

stevo23 wrote:

alanr0 wrote:

stevo23 wrote:

eyeswideshut wrote:

Afraid not. Curvature of the sensor will allow simpler lens designs but wreak havoc with legacy lenses designed for a flat field sensor.

I don't know why - legacy lenses are anything but flat. We don't know how much curve is in this new sensor and we don't know how flat lenses really are other than the fact that they aren't flat. Flat is the goal, but flat they are not.

If we refer back to the photoblographer article you linked to earlier we can make an estimate.

One of the touted advantages is that rays from the centre of the lens strike the sensor at close to normal incidence. For a simple lens, that puts the centre of curvature at the rear nodal point of the lens. The radius of curvature of the sensor must be close to 35 mm for the 35 mm f:1.8 lens mentioned.

? How is this known?

One rationale for the curved sensor is that it makes possible a simpler lens design with low aberrations at relatively wide aperture. A feature of such lenses is that the rear nodal point and exit pupil are both reasonably close to the physical centre of the lens.

A second advantage of the curved focal surface is that light from the lens strikes the sensor at close to normal incidence even at the edges.

So, to a first approximation, light from the lens appears to radiate from a region on-axis and at a distance roughly equal to the focal length of the lens. For light to strike the sensor at normal incidence, the centre of curvature of the sensor must be located at the same point. It follows that for a 35 mm lens, the radius of curvature of the sensor must also be approximately 35 mm.

This will not be exact, but this image from the photoblographer article is broadly consistent with a radius of curvature comparable with the focal length of the lens

Sony 35mm f1.8 lens for curved sensor patent illustration via Egami

At the edge of the 24 x 36 mm frame, 18 mm off-axis, the height of the sensor will be 3.9 mm above the centre.

? What am I missing?

Simple trigonometry tells us that at the edge of the field, curvature lifts the sensor surface 3.9 mm above a flat plane through the centre of the sensor if the radius of curvature is 35 mm.

Alternatively, if the figure above is to scale (which it may not be), the sensor height at the edge of the field is about 2.6 mm.

With a conventional flat field f/1.8 lens focused at the centre of the field, the diameter of the circle of confusion at the edge of the sensor will be (2.6 mm / 1.8) = 1.4 mm.

Even with this lower curvature estimate, it would be completely impractical to use a conventional lens with such a strongly curved sensor. For an 'acceptable' circle of confusion of 0.036 mm, only a small central area of less than 3 mm radius will be in focus.

HTH

-- hide signature --

Alan Robinson

Reply   Reply with quote   Complain
peevee1
Senior MemberPosts: 5,729Gear list
Like?
Re: Believe it or not
In reply to alanr0, 3 months ago

alanr0 wrote:

peevee1 wrote:

webrunner5 wrote:

There was a thread somewhere on this site that had a link to some testing site that pretty much stated that most "Film Era" legacy lenses were pretty bad performers on modern day digital cameras. And they had tons of example pictures and lenses they tested. And most has to do to to the rear lens element on old stuff. Even coating and amazingly the thickness of the glass in front of the sensor. I remember a Leica sensor has a 1mm thick glass and m4/3 has 4mm glass in front. The thicker the glass the worse old lenses do. YMMV.

4mm? It is VERY hard to believe, especially on cameras with IBIS. Maybe Pana. IBIS must be really strong to handle that weight.

http://www.lensrentals.com/blog/2014/06/the-glass-in-the-path-sensor-stacks-and-adapted-lenses

Believe it or not, this is what Roger Cicala's LensRentals blog published

Cheers.

-- hide signature --

Alan Robinson

I have read it. This is from GX1. Maybe other m43 cameras also do it. The graphs show that within image circle of m43 the glass increases image quality.

Reply   Reply with quote   Complain
sportyaccordy
Senior MemberPosts: 1,147Gear list
Like?
Re: Sony's new curved sensor - this is a big deal
In reply to jcharding, 3 months ago

jcharding wrote:

Well, if implemented fully (by Sony and whomever Sony sells the sensor too) it could also mean that all legacy lenses become useless. I'm not sure any increased performance is worth that trade off.

Another way of looking at is another Sony plot to make people buy more Sony stuff due to Sony's unique approach. Not for the first time.

-- hide signature --

This is the most infuriating thing about it for me. This would be the ultimate platform abandonment by Sony. Ditching the flat sensor altogether.

Still, this would be offset by the significantly higher IQ and cheaper, simpler, lighter lenses. Overall I'm excited.

 sportyaccordy's gear list:sportyaccordy's gear list
Sony Alpha NEX-C3 +4 more
Reply   Reply with quote   Complain
stevo23
Senior MemberPosts: 3,621Gear list
Like?
Re: How curved?
In reply to alanr0, 3 months ago

alanr0 wrote:

stevo23 wrote:

alanr0 wrote:

stevo23 wrote:

eyeswideshut wrote:

Afraid not. Curvature of the sensor will allow simpler lens designs but wreak havoc with legacy lenses designed for a flat field sensor.

I don't know why - legacy lenses are anything but flat. We don't know how much curve is in this new sensor and we don't know how flat lenses really are other than the fact that they aren't flat. Flat is the goal, but flat they are not.

If we refer back to the photoblographer article you linked to earlier we can make an estimate.

One of the touted advantages is that rays from the centre of the lens strike the sensor at close to normal incidence. For a simple lens, that puts the centre of curvature at the rear nodal point of the lens. The radius of curvature of the sensor must be close to 35 mm for the 35 mm f:1.8 lens mentioned.

? How is this known?

One rationale for the curved sensor is that it makes possible a simpler lens design with low aberrations at relatively wide aperture. A feature of such lenses is that the rear nodal point and exit pupil are both reasonably close to the physical centre of the lens.

A second advantage of the curved focal surface is that light from the lens strikes the sensor at close to normal incidence even at the edges.

So, to a first approximation, light from the lens appears to radiate from a region on-axis and at a distance roughly equal to the focal length of the lens. For light to strike the sensor at normal incidence, the centre of curvature of the sensor must be located at the same point. It follows that for a 35 mm lens, the radius of curvature of the sensor must also be approximately 35 mm.

This will not be exact, but this image from the photoblographer article is broadly consistent with a radius of curvature comparable with the focal length of the lens

Sony 35mm f1.8 lens for curved sensor patent illustration via Egami

At the edge of the 24 x 36 mm frame, 18 mm off-axis, the height of the sensor will be 3.9 mm above the centre.

? What am I missing?

Alternatively, if the figure above is to scale (which it may not be), the sensor height at the edge of the field is about 2.6 mm.

I thought you might be doing this. Probably not worth making this assumption.

With a conventional flat field f/1.8 lens focused at the centre of the field, the diameter of the circle of confusion at the edge of the sensor will be (2.6 mm / 1.8) = 1.4 mm.

Even with this lower curvature estimate, it would be completely impractical to use a conventional lens with such a strongly curved sensor. For an 'acceptable' circle of confusion of 0.036 mm, only a small central area of less than 3 mm radius will be in focus.

You're making a lot of assumptions here that probably can't be made just yet.

Reply   Reply with quote   Complain
Joseph S Wisniewski
Forum ProPosts: 33,979Gear list
Like?
Actually, it's not.
In reply to brownie314, 3 months ago

brownie314 wrote:

If Sony can get it right (make it manufacturer at a "reasonable" price) this could be a game changer bigger than all of this mirrorless stuff (actually, many of will argue that mirrorless wasn't really a game changer - but that is another topic).

This has the potential to make it possible to put really small, simple lens designs on bodies and have IQ on par with the most expensive glass available.

The curved sensor myth gets debunked here every few months.

Read the papers that accompany Sony's work. The sensor is curved because they stressed the substrate to alter electron mobility. Now that they've proved that this is a good thing to do, they're working on ways of inducing stress while retaining a more useful flat sensor. In short, the curvature was an unpleasant side-effect of a useful research path.

Some PR guys got ahold of this and went all psycho, the way PR guys do.

Optically, the curved sensor counters one aberration, curvature of field. It does nothing about spherical aberration (the primary resolution limiting factor for fast lenses), astigmatism, chromatic aberration, or distortion. It simply makes lens design more difficult.

-- hide signature --

Rahon Klavanian 1912-2008.
Armenian genocide survivor, amazing cook, scrabble master, and loving grandmother. You will be missed.
Ciao! Joseph
www.swissarmyfork.com

 Joseph S Wisniewski's gear list:Joseph S Wisniewski's gear list
Canon EOS 5D Mark II Nikon D90 Nikon D2X Nikon D3 Nikon D100 +43 more
Reply   Reply with quote   Complain
akustykmagmanetpl
Contributing MemberPosts: 600Gear list
Like?
so... increase body to decrease lens cost - i.e. peeing against the wind, right?
In reply to brownie314, 3 months ago

so... while other ICL camera manufacturers try to push prosumer (and FF is prosumer) body price down, in order to cash in on the lenses (because this is where money lies) Sony is trying to do it exactly the other way around?

and this in a situation when they already have too many lens mounts and most of people have no idea what works with which camera?

hardy a game breaker to me. more like: just breaker!

 akustykmagmanetpl's gear list:akustykmagmanetpl's gear list
Fujifilm X-E1 Canon EOS 6D Canon EF 17-40mm f/4.0L USM Canon EF 70-300mm f/4-5.6L IS USM Canon EF 24-105mm f/4L IS USM +3 more
Reply   Reply with quote   Complain
alanr0
Senior MemberPosts: 1,249
Like?
Re: How curved?
In reply to stevo23, 3 months ago

stevo23 wrote:

alanr0 wrote:

Alternatively, if the figure above is to scale (which it may not be), the sensor height at the edge of the field is about 2.6 mm.

I thought you might be doing this. Probably not worth making this assumption.

With a conventional flat field f/1.8 lens focused at the centre of the field, the diameter of the circle of confusion at the edge of the sensor will be (2.6 mm / 1.8) = 1.4 mm.

Even with this lower curvature estimate, it would be completely impractical to use a conventional lens with such a strongly curved sensor. For an 'acceptable' circle of confusion of 0.036 mm, only a small central area of less than 3 mm radius will be in focus.

You're making a lot of assumptions here that probably can't be made just yet.

How about starting with some facts.

At f/1.8, the centre of the sensor is illuminated by a cone of half-angle 15 degrees.  For the curvature of the sensor to make even a small difference to the sensor efficiency, the tilt at the edge must be at least 5 degrees.  For uniform spherical curvature, that would lift the sensor edge is 0.8 mm above the centre.

If the curvature is much less than this, then the "improved efficiency at the edge" claim fails.

If the curvature is as large as this, the sensor will be useless with existing lenses.  For example the Pentax FA 50 mm f1.4 is rather soft wide open, but still resolves 800 line widths/picture height across the frame at f/2.  For a generous circle of confusion of 16/800 = 0.02 mm the focus error at the sensor must be less than 0.04 mm at f/2.

A reasonable conclusion is that existing large aperture lenses of moderate quality or better have a field which is flat to much better than 0.1 mm, and probably less than 0.05 mm.

If the curvature is low enough to be useful with existing lenses, there is no advantage in terms of vignetting, and minimal advantage in terms of lens design.

Joseph Wisniewski knows something about lens design, and has an entirely plausible explanation.

Cheers

-- hide signature --

Alan Robinson

Reply   Reply with quote   Complain
stevo23
Senior MemberPosts: 3,621Gear list
Like?
Re: How curved?
In reply to alanr0, 3 months ago

alanr0 wrote:

stevo23 wrote:

alanr0 wrote:

Alternatively, if the figure above is to scale (which it may not be), the sensor height at the edge of the field is about 2.6 mm.

I thought you might be doing this. Probably not worth making this assumption.

With a conventional flat field f/1.8 lens focused at the centre of the field, the diameter of the circle of confusion at the edge of the sensor will be (2.6 mm / 1.8) = 1.4 mm.

Even with this lower curvature estimate, it would be completely impractical to use a conventional lens with such a strongly curved sensor. For an 'acceptable' circle of confusion of 0.036 mm, only a small central area of less than 3 mm radius will be in focus.

You're making a lot of assumptions here that probably can't be made just yet.

How about starting with some facts.

At f/1.8, the centre of the sensor is illuminated by a cone of half-angle 15 degrees. For the curvature of the sensor to make even a small difference to the sensor efficiency, the tilt at the edge must be at least 5 degrees. For uniform spherical curvature, that would lift the sensor edge is 0.8 mm above the centre.

You have no idea what the curvature is yet. I don't believe the sensor will have a simple curve.

If the curvature is much less than this, then the "improved efficiency at the edge" claim fails.

I'm not sure why. You have no idea what the exit angle of the light "rays" are and you have no idea what the curvature of the lens element or the sensor are.

If the curvature is as large as this, the sensor will be useless with existing lenses. For example the Pentax FA 50 mm f1.4 is rather soft wide open, but still resolves 800 line widths/picture height across the frame at f/2. For a generous circle of confusion of 16/800 = 0.02 mm the focus error at the sensor must be less than 0.04 mm at f/2.

A lot of assumptions, no facts yet to back it up.

A reasonable conclusion is that existing large aperture lenses of moderate quality or better have a field which is flat to much better than 0.1 mm, and probably less than 0.05 mm.

Facts - that's what we need. What is the curvature of existing lenses? For sure they vary quite a bit. You have no idea, you're only speculating. You could be right, but this is only speculation.

If the curvature is low enough to be useful with existing lenses, there is no advantage in terms of vignetting, and minimal advantage in terms of lens design.

I don't really see any plausible science to back that up. We would need to know what the curvature of existing lenses is and then the change in angle of incidence at various places on the sensor to know how this differs from today in terms of performance.

Reply   Reply with quote   Complain
webrunner5
Contributing MemberPosts: 645Gear list
Like?
Re: No it's not
In reply to meland, 3 months ago

meland wrote:

webrunner5 wrote:

There was a thread somewhere on this site that had a link to some testing site that pretty much stated that most "Film Era" legacy lenses were pretty bad performers on modern day digital cameras. And they had tons of example pictures and lenses they tested. And most has to do to to the rear lens element on old stuff. Even coating and amazingly the thickness of the glass in front of the sensor. I remember a Leica sensor has a 1mm thick glass and m4/3 has 4mm glass in front. The thicker the glass the worse old lenses do. YMMV.

One of the issues is that digital sensors (or the IR / anti aliasing filters mounted on the surface of the sensor) tend to be very reflective. So lenses designed for digital, as opposed to film, often have a rear element and/or baffling which is designed to prevent reflected light from the sensor being directed back into the lens which causes flare or a loss of contrast.

But I think you may have got your measurements wrong (a missing decimal point perhaps?) in your comment "I remember a Leica sensor has a 1mm thick glass and m4/3 has 4mm glass in front."

No I am not missing a decimal point. The guy had a shot of each of the glass between his fingers, the 1mm one and the 4mm one and I just could not believe my eyes how thick the glass is in front of m4/3 sensors. I mean it IS thick.

Found the link. http://www.lensrentals.com/blog/2014/06/the-glass-in-the-path-sensor-stacks-and-adapted-lenses  Very interesting read.

 webrunner5's gear list:webrunner5's gear list
Sigma DP2 Merrill Canon EOS 5D Mark II Fujifilm X-A1
Reply   Reply with quote   Complain
TrojMacReady
Senior MemberPosts: 8,455
Like?
Re: so... increase body to decrease lens cost - i.e. peeing against the wind, right?
In reply to akustykmagmanetpl, 3 months ago

2 mounts for still cameras, A and E. How many does Canon have?

This whole concept is for fixed lens cameras, even according to their own propositions.

Reply   Reply with quote   Complain
peevee1
Senior MemberPosts: 5,729Gear list
Like?
Re: so... increase body to decrease lens cost - i.e. peeing against the wind, right?
In reply to TrojMacReady, 3 months ago

TrojMacReady wrote:

2 mounts for still cameras, A and E. How many does Canon have?

This whole concept is for fixed lens cameras, even according to their own propositions.

APS-C lenses are useless on FF, and FF lenses don't fit well with APS-C in terms of prices, sizes, weights and often focal lengths... The fact is Sony already has 5 proprietary lens systems. Canon has 3, with 2 of them each making several times more sales than all 5 of Sony taken together, allowing them to amortize cost of their development better. Some restraint in developing lens systems would make a lot of sense both for Sony and Sony's customers. For example, just 1 FF (A-mount) and one APS-C (E-mount) (with both simple converter and a speed booster) would cover all technical requirements and advantages of both formats, while providing compatibility where it matters.

Reply   Reply with quote   Complain
steelhead3
Veteran MemberPosts: 4,179Gear list
Like?
Re: so... increase body to decrease lens cost - i.e. peeing against the wind, right?
In reply to peevee1, 3 months ago

With no R&D, it is all profit for Canon...Sony has 2 mounts and unlike Canon EF EFs, is fully interchagale between FF and apc.

 steelhead3's gear list:steelhead3's gear list
Sony a77 II
Reply   Reply with quote   Complain
peevee1
Senior MemberPosts: 5,729Gear list
Like?
Re: so... increase body to decrease lens cost - i.e. peeing against the wind, right?
In reply to steelhead3, 3 months ago

steelhead3 wrote:

With no R&D, it is all profit for Canon...Sony has 2 mounts and unlike Canon EF EFs, is fully interchagale between FF and apc.

"Interchangeable" if you like your circles or even rectangles in the middle of dark frames, or turning your uber-expensive FF camera into low-res very mediocre APS-C camera by cropping. In other words - purely theoretically.

Reply   Reply with quote   Complain
steelhead3
Veteran MemberPosts: 4,179Gear list
Like?
Re: so... increase body to decrease lens cost - i.e. peeing against the wind, right?
In reply to peevee1, 3 months ago

We in the Sony world don't have those problems because of a lot of R&D spent, our cameras either, FF or apc work seamlessly and when you have sensors with high resolution and Dynamic range, you know you are getting the best photo possible.

 steelhead3's gear list:steelhead3's gear list
Sony a77 II
Reply   Reply with quote   Complain
akustykmagmanetpl
Contributing MemberPosts: 600Gear list
Like?
Re: so... increase body to decrease lens cost - i.e. peeing against the wind, right?
In reply to TrojMacReady, 3 months ago

TrojMacReady wrote:

2 mounts for still cameras, A and E. How many does Canon have?

This whole concept is for fixed lens cameras, even according to their own propositions.

that isn't so obvious to me.

Canon has technically speaking 3 mounts, but if I'm not mistaken they are all physically (dimensions and electronic contacts layout) the same. what changes is flange (EF-M against the rest) and in case of Canon proprietary EF-S lens - artificial mechanical obstruction - if you take any third party APS-C lens for Canon it fits normal EF camera and works on them (i.e. Tokina 12-24 or Sigma 18-125 to name the lens I tested myself).

Sony as such has 4 different options: FF or APS-C and mirror(ed) or mirrorless. although they are to some degree interchangable

but my point wasn't as much about the total amount but about the fact that Sony had introduced quite a lot of incompatibility issues in a short period. if you make a timeline of Canon's AF system mounts then you get:

EF - 1987

EF-S - 2003

EF-M - 2013

since the takeover of  Konica-Minolta in 2006 Sony has added 3 new mounts. that's 3 in 8 years compared to 3 in 27 years and explains quite well why people are confused with Sony offering and where the "give us another mount, Sony" jokes come from.

 akustykmagmanetpl's gear list:akustykmagmanetpl's gear list
Fujifilm X-E1 Canon EOS 6D Canon EF 17-40mm f/4.0L USM Canon EF 70-300mm f/4-5.6L IS USM Canon EF 24-105mm f/4L IS USM +3 more
Reply   Reply with quote   Complain
steelhead3
Veteran MemberPosts: 4,179Gear list
Like?
Re: so... increase body to decrease lens cost - i.e. peeing against the wind, right?
In reply to akustykmagmanetpl, 3 months ago

Again, Sony has two mounts A and E each fully compatible within their respective mounts.

 steelhead3's gear list:steelhead3's gear list
Sony a77 II
Reply   Reply with quote   Complain
akustykmagmanetpl
Contributing MemberPosts: 600Gear list
Like?
Re: so... increase body to decrease lens cost - i.e. peeing against the wind, right?
In reply to steelhead3, 3 months ago

steelhead3 wrote:

Again, Sony has two mounts A and E each fully compatible within their respective mounts.

meaning - I can put APS-C A lens on FF A body and use all of the pixels of the camera?

or do we understand something different by compatibility?

 akustykmagmanetpl's gear list:akustykmagmanetpl's gear list
Fujifilm X-E1 Canon EOS 6D Canon EF 17-40mm f/4.0L USM Canon EF 70-300mm f/4-5.6L IS USM Canon EF 24-105mm f/4L IS USM +3 more
Reply   Reply with quote   Complain
steelhead3
Veteran MemberPosts: 4,179Gear list
Like?
Re: so... increase body to decrease lens cost - i.e. peeing against the wind, right?
In reply to akustykmagmanetpl, 3 months ago

What your talking about is completely different; within the 2 mounts you have apc and FF lenses (still fully compatible).  Even though the mount is the same, the lenses have different image circles.

 steelhead3's gear list:steelhead3's gear list
Sony a77 II
Reply   Reply with quote   Complain
BasiliskPhoto
Contributing MemberPosts: 675
Like?
Image stabilisation...
In reply to Dave Lively, 3 months ago

Would be perfectly possible, you just rotate the lens and sensor together, rather than moving either the lens or the sensor.

Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum MMy threads