fbpx
Resolution Tests

A Thinner Sensor Stack

Published January 7, 2015

A few months ago, before my hiatus from blogging, we did a series of articles showing the effect that a thick sensor stack (the glass above the sensor) had when we use adapted lenses designed for a thin sensor stack. The first one was mostly about theory, the second about when it was likely to actually be noticeable, and the third gave a general summary of when you might expect problems. All of that theory and prediction is good and useful.

The articles generated a fair amount of discussion about removing filter stacks on cameras so that they would perform better with legacy or film lenses designed for little or no sensor stack. In theory, that would make a big difference, but changing the stack has its own set of issues: focus is changed, infrared filtering can become inadequate, etc. Recently, though, our friends at Kolarivision, who had contributed a lot to our database on filter stack thickness, asked us to do some independent resolution testing for them.

They had modified some Sony a7R cameras, removing most of the thick sensor stack and replacing it with thinner Schott BG39 glass (I do not know the exact thickness of the replacement glass, but it is described as ‘significantly thinner’). The replacement glass closely matched the original IR transmission, maintaining accurate colors, but, in theory at least, should improve resolution on wide-angle, wide-aperture, short-backfocus distance lenses. In other words, it should improve the performance of the a7R using wide-angle Leica and other M-mount lenses designed for film.  

What We Did

This was a simple screening test to see if thinning the sensor stack actually made a difference with some better Leica lenses mounted via adapter. We compared the same lenses on a stock a7R with the Kolarivision modified a7R. The same lens-adapter combination was used on each camera, after screening several adapters to get the best lens-adapter combination we could. It’s not an absolutely perfect test, no test using adapters can be, but it should be a practical test, reflecting what a person using the camera would be likely to see.

We compared results with four lenses. To see if the modification improved M-mount lenses designed for thin sensor stacks, we tested the Leica 35mm f/1.4 Summilux M ASPH and Leica 24mm f/1.4 Summilux M ASPH. To see if the thinner stack reduced performance on native Sony lenses, we also tested the Sony 35mm f/2.8 ZA Sonnar and Sony 55mm f1.8 ZA Sonnar lenses. Our standard Imatest lab set up was used with shooting distances of 9 to 18 feet depending on focal length. Best magnified manual focus was found for each combination, then multiple focus-bracketed shots taken, with the best result used for comparison.

Results with Leica Lenses

The $64 question was, of course, if thinning the filter stack made a measurable difference with M-mount lenses designed for a thinner stack. If things worked, we expected center resolution to not change much, but off-axis resolution to improve significantly.

That’s exactly what we saw with the Leica 35mm f/1.4 Summilux M ASPH.

 Sony A7rKolarivision A7r
Center MTF 5015901610
Average MTF 5011201200
Average Corner MTF50550790

Looking at all of the numbers, the most noticeable difference was in the outer 1/3 of the image area. The modified and unmodified cameras were just about identical in the inner 1/3 of the image with the greatest difference in the corners.

Results with the Leica 24mm f/1.4 Summilux ASPH were slightly different. With this lens the centers are again identical, and the second 1/3 (area from 1/3 to 2/3 of the way to the corners) improved, giving us a higher average MTF50. The absolute corners, however, hadn’t changed a great deal. Whether this is because the sensor-stack change isn’t helping enough at extreme angles, or because of other factors (micro-lens alignment on the sensors for one, perhaps raw data manipulation in-camera, or stuff I haven’t even thought of), I can’t say.

 Sony A7rKolarivision A7r
Center MTF 5018001780
Average MTF 5010651130
Average Corner MTF50660680

In summary, though, we seem to be seeing a very real improvement. It’s not massive of course, and this is simply a preliminary screening test. Multiple lenses will need to be tested both in the lab and in real-world shooting. Effects on astigmatism, color fringing, and a host of other things will need to be evaluated. But from the point of just resolution using Leica or other M-mount lenses on an a7R, thinning the sensor stack seems to result in a noticeable improvement.

Results with Native-mount Sony Lenses

A lot of people are like me: they would like some improvement with adapted lenses, but don’t want to sacrifice using the native lenses for times when autofocus is important. So we wanted to make certain the thinner stack wasn’t going to mess up performance of some native lenses. (Kolarivision has reshimmed the camera to correct for the change in focus that the new sensor stack creates.)

We didn’t expect to see much change with the 35mm f/2.8 ZA Sonnar, given its narrower maximum aperture, and indeed, there really wasn’t much difference between the two cameras. (Remember, please, these are being tested at f/2.8, not f/1.4 like the Leica lenses, so don’t make yourself look foolish and say the Sonnar is sharper than the Leica, OK?)

 Sony A7rKolarivision A7r
Center MTF 5021002080
Average MTF 5014001390
Average Corner MTF50660640

As expected, there was no significant change in the resolution of the 35mm f/2.8 lens with or without the sensor stack change. We did some walk-around comparison tests and autofocus, once the camera was properly shimmed, was just as quick and accurate with the modified camera as with the stock version.

The Sony 55mm f/1.8 ZA Sonnar showed similar results. There may have been a slight softening with the modified camera in the extreme corners, but it was very minor if it was real at all.

 Sony A7rKolarivision A7r
Center MTF 5017001720
Average MTF 5013601360
Average Corner MTF50760720

These aren’t ideal tests for comparison, of course. Ideally we’d be comparing to Sony 24mm f/1.4 and 35mm f/1.4 lenses; but there aren’t any. These are simply the only native-mount primes we had available, so it’s probably the most practical test. The modification doesn’t seem to affect resolution with these native lenses significantly.

I should also point out that there is a slight decrease in corner resolution with the modified camera compared to stock. A difference of 20 lp/IH is almost certainly not significant. The difference of 40 lp/IH might be, that’s about the range where we see a consistent difference. A larger sample size will be necessary to clarify that.

Summary

Don’t take this as more than what it is — a simple feasibility experiment to see if a thinner sensor stack might improve a7R body performance with M-mount lenses designed for thinner or no sensor stack. Theory would suggest it should help improve performance, and this preliminary study suggests that it does, indeed. It’s not a complete cure-all, of course, there are multiple other factors that affect lens performance when shooting on a different body. But this certainly seems worth further investigation.

For those who are are excited about it, I understand Kolarivision is already performing this modification HERE in limited quantities. As these modified cameras get out into the field, we’ll get a much clearer idea of just how much improvement they give to actual photographs.

 

Roger Cicala and Aaron Closz

Lensrentals.com

January, 2015

Author: Roger Cicala

I’m Roger and I am the founder of Lensrentals.com. Hailed as one of the optic nerds here, I enjoy shooting collimated light through 30X microscope objectives in my spare time. When I do take real pictures I like using something different: a Medium format, or Pentax K1, or a Sony RX1R.

Posted in Resolution Tests
  • Edward Leskin

    Boy this looks interesting. I own a LARGE Leica M lens collection and am considering this modification for my A7r. I want to know if this modified camera has been tested with the Leica f1.0 Noctilux at f1.0 and with the Voigtlander Nokton at f1.2. I have read that at wider f stops, there could be a problem. I also own a 35mm Summicron version 4, a 28mm Elmarit M from the 1970’s and the second version of the 15mm Voigtlander Super Wide. Any information and testing with these lenses would be greatly appreciated before I make a decision.

  • Ron

    What are the implications of this phenomenon for third party lens manufactures that make a single lens design for multiple body brands? The Sigms, Tamrons, etc. can they adjust for sensor stack thickness or average it out over different mounts/body brands?

  • John

    Do filter stacks optically work the same as rear glass filters?

    I’m wondering if it makes any sense to remove clear glass filters from lenses made for film when using them on digital (replacing them with glassless seals).

  • Ray Yee

    Thanks very much for trying it out. It’s a pity, I really like the lens.

  • Michael Demeyer

    Ray,

    We have not seen significant changes to the color shading issues with this modification. Fortunately, they can be dealt with in post, unlike the smearing for which this is the only solution of which I am aware.

    Michael

  • Eric

    Thanks for that follow up on sensor thickness.
    As you already said, more tests with other lenses are needed for conclusion.
    Another angle for the analysis is the Zeiss Loxia lenses that seem to adapt classic ZM designs to compensate for sensor thickness. How Zeiss did the compensation is still a mystery. Maybe you can find out?

  • Ray Yee

    Hi, could you do a test with the Voigtlander 15mm f4.5 Super Wide Heliar to see if this eliminates the purple fringing which is this lens is known to cause on the A7/R?

    Would be interesting to know. Thanks!

    Ray

  • intrnst

    [Off-topic]

    Hey, Roger.
    Hope you’re rested and well, after the holidays, and ready to write some of the good old backstabbing historical papers once again.

    Pretty please?

  • Michael Demeyer

    Lloyd Chambers just finished a comparison on 6 lenses:

    Leica 18/3.8
    Leica 21/3.4
    Zeiss 21/4.5
    Leica 24/3.8
    Leica 28/2.8
    Leica 50/2 APO

    Improvements on all, including the 50mm APO Summicron.

    Details at http://www.diglloyd.com

    Michael

  • Charlie Webster

    I have my own A7 on the way in for this mod. I’ve made some test images with the SEM21 at multiple apertures “before”. I’m a big fan of Roger and his insights on this site, however I just wish these tests had been made with “problem”, like the skopar mentioned, wides and as many note above, at infinity with distant objects in the corners. That’s where the A7 and A7r will make you cry with UWA M onboard. But it’s preliminary as you point out, so fingers crossed the A7K will rock!! Of course trusty M9 is ready to set the “standard” 🙂

  • Jim

    I think it was fairly well known that the sensor glass in the a7r had a measureable effect due to ray-angle issues, but kudos for testing this out and putting numbers against the theory. Rather than the native Sony FE glass though, it would be much more interesting to see the results with the Zeiss Loxia 50mm and 35mm, which have been advertised as having optics specifically designed to address this issue. Do you plan on testing these lenses, as I expect they will be much more telling?

  • Makes sense!

  • Michael Demeyer

    …and thicker glass (colored) probably makes it easier to get better rejection of IR, UV, etc.

    “Our two weapons are surprise, fear, …”

  • Michael Demeyer

    One reason for a thicker stack is that it moves dust and dirt further from the sensor, reducing the impact on images. Not arguing it’s a good trade-off but, if the native lenses are designed for it, it’s something.

    I wondered about that, too…

  • If a thinner glass sensor stack does not reduce the sony lens quality, or better, if a thick glass does not improve it’s definition, what is the sense of using it? To avoid the use of 3party lenses? Reading the tests I cant figure any advantage in using it. I have a a7 converted to IR and it works very well with the stack changed to a IR filter, thinner, even with a 14mm lens. Wich design decision make every system use a different solution?

  • Michael Demeyer

    Horst,

    We used the 21mm Skopar F4 as a primary test lens in the development of the modification and the improvements with it are substantial. We chose it because it is in the class of lenses that are seriously compromised by this issue.

    I should be able to post some infinity focus images with that lens and a few others next weekend on a modified A7.

    Michael

  • Horst Tobart

    Again I am astonished that you are right into some problem that really bothers be (as well as many others that adapt lenses).
    Thank you very much for going that route of clearifying with real experiments what to expect from currently available sensors!
    That said I might suggest to try some of the really problem lenses like the short Zeiss Biogons or the Voigtlaender Coloar Skopars for further tests at infinity.

  • I think a lot of us went down over the holidays, I am just recovering from the flu myself and I very seldom get sick.

    I just got my a7R a few days ago and I have not really had a chance to get out there. Looking forward to more on this. Thanks Roger.

  • Roger Cicala

    CarVac,
    I’ve been out sick for several weeks – we’re so backed up that I’ll need to just get the backlog taken care of for the rest of January.

    Roger

  • CarVac

    I have seen on the blog The Last Word (article # 6329, I bet spam filtering would eat the url) that the Leica 28/2.8 is particularly horrible with corner smearing even on the A7s, and way better with the M240.

    Can you test that?

  • Someone

    Roger, it’s good to have you back at your game!

  • Michael Demeyer

    Although Schott data sheets specify performance at 1mm thickness (generally), glass is available in various thicknesses so this is not an indication of what might be used in this particular application.

    I have a modified A7 on the way back and will post some ‘real world’ comparisons in a week or two when time and weather permit.

    Michael

  • Toshik
  • A lot has been made about the sensor stack on the A7 models, but almost all the talk is about using leica/rangefinder lenses. Of course Sony doesn’t care about pandering to this user group, and I suppose rightly so.

    Incidentally Sony has recognised alternative glass shooters in their marketing material for the A7ii.

  • Roger Cicala

    Lee, that’s a really good point and exposes one of the weaknesses with using Imatest for testing. I think until we have these cameras in the field and see what happens at various focus distances we won’t know what it does for sure. There’s every possibility that improvement would be greater at infinity than what we saw close up. Or not, of course.

    Roger

  • L.P.O.

    Lee Saxon: I think Sony is at the moment saved by its meager primer selection. In Roger’s test, the wide (35mm) Sony lens wasn’t fast (f/2.8), and the faster (f/1.8) Sony lens wasn’t wide (55mm). If I have understood Roger’s earlier articles correctly, thickness of the stack shouldn’t make much difference for these lenses, which is exactly what happened. If Sony were to release a fast and wide lens, like a 24/1.4, or an even wider 14/2.8, that would be the real test.

    PS.
    As a Finn, reading this article with a serious face was a bit difficult. I don’t know where Kolarivision got their name from, but in Finnish the word “kolari” means (traffic) collision. So, every time I see that name, I unintentionally read it as “crash-vision”.

  • Lee Saxon

    Ron is definitely right. I tested a bunch of 20-25 & 35 M-mount lenses and the smearing (and especially color shift) were always far, far worse at/near infinity.

    I’m really surprised to see that making this change did not degrade performance on native lenses as much as it improved it for adapted lenses. That’s really interesting and potentially exciting news (insofar as a lot of people have been wishing like crazy that Sony would thin the stack in the next generation of FE cameras and these results suggest that it’s at least possible).

  • All this tells me is if Sony wants the A7 system to go fully mainstream they need to spend more time and money on rolling out an entire line of lenses that are optimized to their camera bodies. Until then, the A7 is still an experimental system forcing those you have made the switch to the Sony platform to search for suitable lenses from third parties that may give suitable results.

    This also tells me a lot about what Leica has achieved over the past few years in adapting their existing lens line to their digital cameras as they have a complete system, i.e, Nikon and Canon too.

  • Roger Cicala

    Hi Ron,

    That is a very pertinent point – the closer shooting distances may be part of what we’re seeing. This was just a feasibility study, to see if the idea was of possible benefit. A lot more investigation is going to be needed, but these initial results make me think further investigation is warranted.

    Roger

  • Ron

    Curious what the shooting distances were for the Leica lens tests? One reason the 24 might not show such a significant difference between cameras is if it was done at a relatively close focusing distance to fill the frame with the test chart. My understanding is edge sharpness degradation is most critical at infinity focus. Intuitively, this seems to be where a thinner sensor should make a bigger difference.

    This is perhaps outside the scope of the test, but IMO, an additional benchmark to keep as reference is how these lenses perform on a Leica M. There are sensor resolution factors to work around, but while it is interesting to see how much better the modified a7R is over the stock version, the question of most interest (at least to me) is how much closer it is to matching how these lenses perform on the Leica, with respect to edge smearing.

    The numbers from the modified camera show there is some improvement with the Leica lenses, but what we don’t really know is how much of the gap it bridges between a stock Sony and a Leica M.

    In terms of lenses to test, I’d be very curious about the Leica 21/3.4 Super Elmar. This lens is already very sharp across the frame from wide open on a Leica M (though admittedly not a very fast lens), yet on the Sony a7 cameras, image details towards the edges are quite smeared (at infinity).

Follow on Feedly