Wide angle lenses designed for digital sensors

[rramesh]

Advertisement (gone after registration)

Could someone explain to the less technical, what adjustments are being made in wide angle lenses to suit the design of digital sensors, to overcome the thickness of the cover glass?

 

I am intrigued after trying out the Voigtlander 15 V3 where I clearly see a marked improvement from the V2 and V1 of this lens, in eliminating colour shifts.

[250swb]

Is the problem the thickness of the cover glass? I thought it was the sensors microlenses that are designed for telecentric digital lenses and therefore not performing well with traditional lens designs. The lens manufacturer could go some way to compensate by making their lenses more telecentric. But I'm no expert in camera design.

 

Steve

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[colonel]

The thickness of the glass cover is only one factor of which there are many.

Its also worth ignoring lots of the web discussion on this as much of it is conjecture.

The fact is that manufacturers make cameras for their own lenses.

Sony A series are perfectly designed for their own lenses, the fact that certain models have corner fringing on wide angle M lenses is nothing to do with either Sony or Leica, its just a consequence of users experimenting across genre.

 

Focusing on the Leica M design, this was designed in a world of film where the thickness of the film could absorb more incident angles. Additionally film is generally (unless you expend a large amount of money in film selection, storage and custom processing) less precise and more noisy then good digital. The result of this is that M lenses are not fully corrected at the rear element for flat sensors.

 

Leica gets around this by angling the microlenses at the sides of the sensor. This solves much of the issue but not all, so some software correction is still required. Sony can't be bothered with this (although did do some for its own lens design on the A7R sensor). So its natural that M lenses for digital are best on Leica bodies.

 

However general M lenses design by Leica, Zeiss and Voigtlander since the early 2000s have taken digital into account and are becoming better at providing more corrected light rays. they will never do it completely as this would compromise the size for the M lens as you would end up with a highly corrected retrofocus design, like in a DSLR, but they are much better. Thus more recent versions are simply better with digital across the range.

 

IMHO if sharpness across the view is key you do really want to stick to more modern designs.

 

Lastly Sony also has come into this issue with the A series and has largely opted for software solutions meaning that the extreme corners in most Sony A lenses are not that good. Where they have been able to design the lens and sensor in concert we have seen spectacular results, e.g. the RX1. But this kind of design is impossible with a small interchangeable lens camera.

 

Leica's angled micro lenses are the way to go for M and in fact the M240 even has deeper pits to control the problem at the edges. Who knows what the future will bring. Sony's curved sensor patent could also be a perfect solution although even Sony says this might only have an application with fixed lens cameras for now due to the precision required between components.

 

In summary get the latest models and pay particular attention to the MTA curves at the edges both published by the manufacturers (Zeiss are particularly honest with their published MTA curves and I think Leica and Voigtlander are not bad as well). Also look at review sites like reidreviews which provide lots of real world corner samples to review. Diglloyd is qute good on the factual MTAs if you can dodge the hyperbole which he loves

 

Lastly I am bound to say by consensus rules that sharpness is only one aspect of a lens. Its arguable that extreme corner sharpness doesn't matter to 99% of photographers. but certainly micro contrast and the taste aspects of draw are very important as well. Of course as the lens goes more and more wide angle corner sharpness is more difficult to control.

Edited July 29, 2015 by colonel

[jmahto]

...... as you would end up with a highly corrected retrofocus design, like in a DSLR, but they are much better.......

Emphasis mine. I always thought that M wides are much better than Canikon's wides. Can you tell us which DSLR wides are better than let us say 28 elmarit (or take any other reference M wide) ?

 

Edit: I have found only one comparison but it is prime vs zoom (but still L). http://www.kenrockwell.com/leica/m9/sharpness-28mm.htm

Edited July 29, 2015 by jmahto

[colonel]

Emphasis mine. I always thought that M wides are much better than Canikon's wides. Can you tell us which DSLR wides are better than let us say 28 elmarit (or take any other reference M wide) ?

 

Edit: I have found only one comparison but it is prime vs zoom (but still L). http://www.kenrockwell.com/leica/m9/sharpness-28mm.htm

I think the English did not work out well in my post. I meant to say that if you wanted to produce a fully corrected M lens you could, but it would probably be unacceptably large for most M users.

 

I did not mean to say that DSLRs lenses were better.

I will blame Apple here

 

I fully agree that IMHO Canikon lenses are mostly bad wide open which is what drive me to the Leica system in the first place. In terms of wides these are difficult to produce on any system but the Leica ones are amongst the best.

[david strachan]

All the discussion has been based on FF sensors.  On the M8 the ultra wides work very well.  For example the CV 15mm V1  is a beauty. No colour shifts and nice edge definition.

 

I have 15mm through to 135mm in various makes (equivalence 20mm to 180mm).  They work very well using the "sweet spot" of the lens...the middle, not the periphery.

 

Still flying the flag for the M8...

 

cheers  Dave S

Edited July 31, 2015 by david strachan

[adan]

Advertisement (gone after registration)

The issue with digital sensors and lenses MOSTLY has to do with the microscopic structure of sensor's silicon - NOT the cover glass as such.

 

The light-sensitive-areas of sensors are at the bottom of "wells" - deep (on the microscopic scale) holes in the silicon. http://www.extremetech.com/wp-content/uploads/2013/03/Wiring-diagram-for-a-typical-front-illuminated-sensor.gif

 

Think of a huge skyscraper-lined city like New York - the sensor's actual light-sensitive areas are down at street level. At noon, the light reaches right down to street level, but at dusk or dawn or in the winter, when the sun is low, the streets are permanently in shade from the surrounding structures.

 

With a sensor, the low-angled corner rays from a rangefinder-style superwide lens are like the low sun - their angle is too flat to reach down into the light-sensitive areas, and the result is vignetting - darkened underexposed corners.

 

Ideally, a digital-friendly lens is "telecentric" - that is, it sits a long way above/away from the image plane, and directs nearly parallel light rays directly down into the light-sensitive wells, just like the noon summer sun in our hypothetical city.

 

http://www.lhup.edu/~dsimanek/3d/telece2.gif

 

As Colonel says - lenses for SLRs already do this, since the viewing mirror requires a long distance from the lens to the sensor in any case. Even superwides still must be at least 40mm or so away from the sensor, due to the mirror.

 

Another solution to the problem is to place microlenses over each sensor pixel (see first linked diagram again), to capture light from low angles and bend it down into the wells. This is what Leica did to accomodate their legacy wide-angle lenses back to 1980 or so - they couldn't change the old lenses, but they could "fix" the images by bending the light with offset microlenses, and applying post-processing to the images to lighten the remaining vignetting (thus the need for Leica's 6-bit coding - the camera's electronics must know how bad the vignetting will be with a particular wide-angle lens, and adjust the data accordingly before saving it to the SD card).

 

Camera lenses designed specifically for digital use can have "telecentricity" designed in from the beginning - but also as Colonel says, that means they will be physically bigger. Leica's newest lenses (e.g. the 16-18-21 Tri-Elmar) probably have some telecentric character (it's a lot longer than a 21 or an 18 alone). But given that "big lenses" are anathema to the Leica rangefinder ideal, and that Leica is still committed to supporting film cameras already in use - they have been conservative in going entirely to telecentric design. Counting on their microlens and post-processing technology to do the same job.

[Winedemonium]

Adan - I actually read this topic because I saw that you had posted a reply. 

 

Every time you post like this I learn a whole lot. You have a gift for explaining/teaching.

[luigi bertolotti]

....  On the M8 the ultra wides work very well.  For example the CV 15mm V1  is a beauty. No colour shifts and nice edge definition.

 

...

 

cheers  Dave S

Agree : when I used M8, I considered my CV15 V1 as "my SA 21 at film times" ... when changed to M240 I ended up selling it (and some day I'll probably get a CV15 V3... but 15 on FF is not such a priority for me)

Edited July 31, 2015 by luigi bertolotti

[jankap]

Adan, you have seen this?

kolarivision.com --- sony-a7-series-thin-filter-legacy

 

Also, Jim Kasson is experimenting with a modified Sony A7II.

Jan

Edited July 31, 2015 by jankap

[bocaburger]

Agree : when I used M8, I considered my CV15 V1 as "my SA 21 at film times" ... when changed to M240 I ended up selling it (and some day I'll probably get a CV15 V3... but 15 on FF is not such a priority for me)

Likewise.  My use of the 15 was occasional at most with film, went up exponentially with the R-D1 then M8 (I even added threads to accept an e39 UV/IR), fell back to occasional with the M9 and M240.  I kept mine though.  PP for the slight edge cast is not a bother for the small number of shots I take with it.

[rramesh]

 

Adan, you have seen this?

kolarivision.com --- sony-a7-series-thin-filter-legacy

 

Also, Jim Kasson is experimenting with a modified Sony A7II.

Jan

 

From what I read, I don't think it really works well. Apparently the issue is more than just sensor thickness. 

Edited July 31, 2015 by rramesh

[jankap]

Leica uses two methods to solve the problem. A very thin filter glass in front of the sensor and a small lens on every sensitive cell of the sensor.

Kolari has the possibility to change the filter glass only.

Jan

[hiepphotog]

From what I read, I don't think it really works well. Apparently the issue is more than just sensor thickness. 

 

As a Kolari user, I would say it works quite well. Comparing my results with the aperture series on Lloyd Chambers blog, I think it's about the same with the M240 performance.

 

What Adam is saying above only explains the cross-talk (corner color cast) and vignette issues. Cross-talk happens since not all wavelengths are going to hit their supposedly cellat such steep angles; some wavelengths will hit adjacent cell. Vignette comes from the traditional Front Side Illuminated where the sensel is buried deep beneath the signal pipeline. Any light comes from the microlenses at an angle would greatly reduce the amount of energy transferred to the sensel (imagine light zig-zag until it gets to the sensel). These two problems are largely fixed with Leica in-camera corrections (software based). And it is also the reason why the A7s has even less vignette and corner color cast than any M digital body. Much bigger cell does greatly eliminate the problem.

 

However, Sony sensor filter stack thickness is responsible for the smearing when you use WA M lens. This has to do with the intended refractive index. M lenses designed for film would not account for any additional corrective factor for any optical element after leaving the last element; any additional optical filter would induce smearing (inducing more astigmatism and changing field curvature) because the light is bent differently now before it hits the sensor. Modern M lenses (supposedly designed for digital sensors) accounted for this additional filter stack thickness (around 1mm for M240) and being more retrofocus to help with the vignette and color cast. This is well-documented on Lensrental blog by Roger Cicala and Brian Caldwell (the guy designed Coastal Optics 60 APO). This smearing effect (astigmatism+field curvature) is also documented by Zeiss by simply inserting a 2.5mm optical flat in their MTF machine while measuring the MTF of the ZM 35/1.4.

 

So yes, Kolari would help a lot, but the current thickness is 1.3mm (as opposed to 2.5mm), so expecting MM/M9 (0.8mm total) corner performance is unrealistic. Ilija from Kolari said he wouldn't want to go thinner because of possible filter breakage (seen in M9 model) and he wants to match the original UV/IR filtration as close as possible.

 

Edit: The CV 15 III is being more retrofocus to help only with the vignette and color cast. Some users show that the corner performance of the II and III is the same.

 

Edit-2: I am using the ZM 15 on my A7S Kolari.Mod. Here is a few pictures to show that you can't even achieve this with the M9 and M240: (no color cast correction)

 

 

 

Edited July 31, 2015 by hiepphotog

[rramesh]

Thanks Colonel, Adan and Heippotog for your inputs. Did also see this article on the web, specific to the 15 Voigtlander V3. 

 

http://lmlens.blogspot.sg/2015/04/voigtlander-15mm-f45-iii-v3-review-test.html

Edited August 1, 2015 by rramesh

[hiepphotog]

I have to bring this thread back cause I found this literature on the history of Sony Exmor sensor. I have some very interesting information that the OP and some of you might find useful.

 

http://www.framos.com/en/news/news/single-view/article/what-is-sonys-technology-anyway.html

 

I'll highlight some of the points that is pertinent to the topic at hand, how the current sensor tech is moving toward handling the steep incident angle of the RF WA.

 

Exmor 2nd and 3rd gen., this is how the tradition front-side illuminated CMOS designed to cope with the steep ray angle.

 

"This reduction in the metal wiring and transistor structure region was beneficial as it greatly improved the light sensitivity of the sensors using these technologies over their predecessors. In the past, photons that would come in at sharper angles or at shorter wavelengths (blue-green spectrum) would be either reflected off of the metal wiring or dissipated in the transistor region and thus would not be detected and measured in the silicon substrate. Reducing this depth increases the chances of photons reaching the pixel wells and being detected correctly by the sensor making them more sensitive to all visible light, especially in the blue and green spectrums allowing them to work with a larger selection of optics."

 

Exmor R and RS: Current Sony BSI:

 

"This new design dramatically improves the responsivity and sensitivity of these sensors. Two main reasons how this change in design improves the performance of the sensors are as follows. As I mentioned earlier, shorter wavelengths can be missed if they are not able to reach the pixel’s well and create an electron.

With the non-photon sensitive parts are out of the way, the pixels are able to see all this light that was once lost or reflected out of the sensor making them more sensitive in lower light conditions, shown in figure 4.

Secondly, light that may have hit the sensor at sharp angles now hit the sensitive area of the sensor where, in the past, it may have directed off to another area of the sensor. This is really important for color sensors as it minimizes the changes of having cross talk in the color data where, for example red light would get directed to a nearby green or blue pixel. As a result, the color fidelity of these sensors is much truer without the need to do extensive post processing to correct the color response in the captured images."

This is what I have seen as well on the A7RII. It has better color cast and vignetting control than any of the current Leica M body (or any of the current CMOS for that matter), no need for in-cam correction. So for smearing, if Sony, or the users, decides to change the sensor filter stack thickness (or Leica could switch to Sony sensor), we would have a killer M cam right there.