p.30 #1 · A thinner sensor stack may be possible after all!
Another SLR lens that seems to have improved after conversion is the Contax Carl Zeiss Distagon 28/2.8 - to my great disappointment this lens was never as sharp as it could be particularly in the corners and had an odd mid zone dip before the conversion. It now is very good at f8 across the frame and has that amazing Zeiss microcontrast - the link below is a series of infinity images taken at all apertures
p.30 #4 · A thinner sensor stack may be possible after all!
nicoimages wrote:
Another SLR lens that seems to have improved after conversion is the Contax Carl Zeiss Distagon 28/2.8 - to my great disappointment this lens was never as sharp as it could be particularly in the corners and had an odd mid zone dip before the conversion. It now is very good at f8 across the frame and has that amazing Zeiss microcontrast - the link below is a series of infinity images taken at all apertures
p.30 #6 · A thinner sensor stack may be possible after all!
nicoimages wrote:
Thanks for your comments everyone. Following on from Charle's Canon 35 TS test (love that building picture) have tested my Canon 24 TS-E L MkII on the Sony A7R_M
First impressions are that shifted to the extremes it is a tiny bit sharper in the corners than it used to be and behaves similarly to what I remember it on the Canon 5MkII. Had forgotten how average the lens is overall if not shifted. To demonstrate this the link below includes test pictures of all apertures with no movements applied and then two images with max shift applied at f8 and f11
Wow, does the TS-E 24 MkII really perform that poorly, or was it accidentally tilted? Without any foreground area on the right side, it's difficult to tell if it's massive field curvature, or a tilted lens. But that doesn't look like I'd expect such a lens to perform. FWIW, I have the TS-E 17 and previous version of the 24. There has been some discussion about the 17 needing to be focused off-centre to optimize across the frame performance but your 24mm images look extreme...
p.30 #7 · A thinner sensor stack may be possible after all!
rscheffler wrote:
Wow, does the TS-E 24 MkII really perform that poorly, or was it accidentally tilted? Without any foreground area on the right side, it's difficult to tell if it's massive field curvature, or a tilted lens. But that doesn't look like I'd expect such a lens to perform. FWIW, I have the TS-E 17 and previous version of the 24. There has been some discussion about the 17 needing to be focused off-centre to optimize across the frame performance but your 24mm images look extreme...
No tilting unfortunately it does seem to have so quiet severe field curvature wide open and lacks in micro contrast compared to the Zeiss. Was testing it using live view tonight and you can more consistent results if you focus in the corners when not shifted and stopping down.
However when shifted the issue seems to disappear and it works quiet well even at f8 which makes me think that it has been optimised to be used with movements and that it has to be focused carefully if using apertures wider than f5.6.
p.30 #8 · A thinner sensor stack may be possible after all!
Could this possibly be because these Mark 2 T-S lenses were designed with the much thicker filter stack in mind? Severe field curvature is typically the tell-tale of mis-match filter thickness so far. Looking at the sample shots (I didn't check them before cause I have no interest in the lens), I am quite sure that's the problem. The field curvature seems to curve inward, toward the camera on both sides. Tilting would not bow the plane of focus like that.
p.30 #9 · A thinner sensor stack may be possible after all!
hiepphotog wrote:
Could this possibly be because these Mark 2 T-S lenses were designed with the much thicker filter stack in mind? Severe field curvature is typically the tell-tale of mis-match filter thickness so far. Looking at the sample shots (I didn't check them before cause I have no interest in the lens), I am quite sure that's the problem. The field curvature seems to curve inward, toward the camera on both sides. Tilting would not bow the plane of focus like that.
I have been through my archive and can confirm that this lens is just as bad prior to the thin sensor modification unless stopped down to f11 - if anything it seems to have improved after the modification when shifted to the extremes.
The image available via the link was taken a few months ago at f7.1 with no tilt/shift applied and shows clear signs of field curvature. I had not used this lens for a while as I was rather disappointed by the performance and correcting verticals in photoshop seemed to get more consistent results in terms of clarity. Shifts seem to be better now though so I may start using it again.
p.30 #10 · A thinner sensor stack may be possible after all!
hiepphotog wrote:
Could this possibly be because these Mark 2 T-S lenses were designed with the much thicker filter stack in mind? Severe field curvature is typically the tell-tale of mis-match filter thickness so far. Looking at the sample shots (I didn't check them before cause I have no interest in the lens), I am quite sure that's the problem. The field curvature seems to curve inward, toward the camera on both sides. Tilting would not bow the plane of focus like that.
I'm trying to get my head clear on how the filter stack thickness and field curvature relate. I am convinced they do, since the light path is bent inward (more vertical), which seems like it should reduce the path length progressively as you move toward the edges of the sensor.
Let's start with a perfect lens that has no field curvature on film. My mental model suggests that adding a stack to the optical path would decrease the effective path length near the edges (at higher incident ray angles) relative to the path length in the center by refracting the ray as is enters the glass with a higher refractive index than air. The shorter path near the edges (vs. that with no stack) would curve the plane of focus (in the subject domain) away from the camera position toward the edges.
Conversely, if our perfect lens was designed for a 2mm stack and you replaced it with a 1mm stack (all else being equal) the plane of focus would curve nearer to the camera (inward).
This is the case you are making here, yes? I know Lloyd C has a detailed field curvature analysis of the 24mm TS-II on his site, but I don't subscribe to that section. Can you point me to a reference that documents its behavior in this regard?
Interesting (to me), looking at Roger's database of sensor cover thickness for Canon shows:
Canon 10D 2.7 mm
Canon 20D 2.75 mm
Canon 30D 2.45 mm
Canon 1Ds Mk II 1.0 mm
Canon 5D 1.5 mm
Canon 70D 1.2 mm
Canon 6D 2.0 mm
Canon T41 1.2 mm
...a fairly wide variability, but certainly thicker (in general) than what is in the A7rM that Nico is shooting.
So, the difficulty is that most lenses don't have flat fields and we don't generally know what stack they are designed for - especially true of DSLR lenses that have been introduced over the course of many years. (By the way, Nikon has been more variable in its stacks than Canon, so this is not uniquely a Canon issue.) So whether making the stack thinner will hurt or help the field curvature behavior of any given lens, other than those designed for film or Leica is hard to predict.
Thanks to all for doing tests with a variety of lenses and sharing. It's probably the only way to really know.
p.30 #11 · A thinner sensor stack may be possible after all!
According to Zeiss's Dr. Nasse in his publication "From the series of articles on lens names: Distagon, Biogon and Hologon" the curvature/degradation induced by the sensor plate thickness is primarily in the tangential direction. If it was both sagittal and tangential, then I think we'd see a curved plane of focus, and one that is relatively sharp. Instead what we're seeing is smearing. I'm not sure how that equates if the glass thickness is thinner than for which the lens is designed. I'd guess it will still result in a somewhat undesirable under correction in the tangential direction.
From Nasse:
"Lenses with a very large beam tilt react in a much more sensitive manner to a change of refractive index in the image space caused by filter plates in front of the sensor (such as low pass and IR-blocking filters). If the filter plate is not considered in the design of the lens, the edge definition will suffer. The effect of the additional path through the glass grows exponentially with the beam inclination. A Distagon which never achieves more than 20° beam tilt in the corner of the image reacts more tolerantly than a symmetrical wide- angle lens, which might reach a 45° tilt. This is why filters in digital Leicas are very thin – to remain compatible with older optics. If the filter is significantly thicker, the contrast transfer for the image edge becomes worse for tangential structures. In the graph of the curves, this looks like the old retrofocus lenses but is caused by astigmatism rather than lateral chromatic aberration. The focus is shifted to greater distances for tangential structures by the additional path through the glass. If the best edge definition is to be achieved, then all that can be done is to stop down further.
In Roger's lens exit pupil database, the TS-E 24 is 86mm, which is still at least 2x better than most M lenses (he lists). Assuming an M lens like the 50/2 Summicron at ~50mm is similar to the ZM50/2, and the ZM performs well on the a7 cameras (based on my tests), you'd think a lens with 86mm exit pupil distance wouldn't be much affected by sensor thickness variations. And I think that's what Nicholas is saying when he only sees minor improvement with that lens when shifted, which is when light rays will reach the sensor at the most acute angles.
That said, the CV40/1.4 (and I think also the 35/1.4) has very strong field curvature on a digital M, but when used on an a7, actually is much better... and almost like any other 'normal' lens....
p.30 #12 · A thinner sensor stack may be possible after all!
Well, I just based on the Zeiss measured MTF chart of the ZM 35 with native thickness and Sony thickness. Indeed the tangential direction suffers severely, but that's just part of the problem. The shape of the sagittal curve changes too. And the closer it is toward the corner, the worse the effect would become (a steep drop in both curves). So the smearing and change in field curvature depend on the incident angle and the refractive index of the glass.
One thing that most wouldn't care to explain is why the center performance on the thicker glass tend to be better than on the thinner glass.
Michael: the difference in thickness of the Canon cameras is actually physical thickness. But I would say they all have the same optical thickness (or overall refractive index). That's why some folks did a Full Spectrum conversion before and didn't see any improvement in RF WA performance. They tend to replace the cover glass with a blank one of the same "optical thickness" to avoid moving the sensor. Case in point:
p.30 #13 · A thinner sensor stack may be possible after all!
hiepphotog wrote:
One thing that most wouldn't care to explain is why the center performance on the thicker glass tend to be better than on the thinner glass.
Michael: the difference in thickness of the Canon cameras is actually physical thickness. But I would say they all have the same optical thickness (or overall refractive index). That's why some folks did a Full Spectrum conversion before and didn't see any improvement in RF WA performance. They tend to replace the cover glass with a blank one of the same "optical thickness" to avoid moving the sensor. Case in point:
To the first point, I don't think I have seen any well-documented evidence of degradations in the center with film lenses, but generally there is little or no improvement, either. This is as-expected since the center of the image is the best-case with regard to ray angle. Perhaps very fast lens (like a Noctilux) would show a difference, but I have not seen any tests like that done.
On the article you reference regarding Full Spectrum conversion, unless I missed it, he specifically states that they replaced the filter stack with an equivalent optical thickness of clear glass to avoid having to move the sensor. If that's the case, I see no reason why that particular full-spectrum conversion would improve the corners, which is about what he shows.
Moving to a thinner stack does requiring moving the sensor, as he notes, and which KolariVision does on these conversions. The author also doesn't state that he ever tried a thinner sensor glass and tested it for edge performance. Instead he states that, "If you think about the amount of aberrations introduced by a typical clear glass or UV filter, that is the same difference that you can gain. I.e. has anyone ever seen CA or blurring caused by a good filter?"
This comment suggests (at least to me) that he is comparing a filter in front of the lens to one in the optical path between the lens and the sensor. Those are completely different situations.
p.30 #14 · A thinner sensor stack may be possible after all!
hiepphotog wrote:
Well, I just based on the Zeiss measured MTF chart of the ZM 35 with native thickness and Sony thickness. Indeed the tangential direction suffers severely, but that's just part of the problem. The shape of the sagittal curve changes too. And the closer it is toward the corner, the worse the effect would become (a steep drop in both curves). So the smearing and change in field curvature depend on the incident angle and the refractive index of the glass.
One thing that most wouldn't care to explain is why the center performance on the thicker glass tend to be better than on the thinner glass. ...Show more →
I agree about smearing, but not sure about general field curvature. Smearing is field curvature, but in just the tangential direction, according to Nasse. If thicker sensor glass also caused sagittal drops, similar to tangential, you'd expect to see something more like normal field curvature. Where I'm a bit confused about this is this graph posted by Roger:
It would somewhat contradict Nasse's statement that it's primarily a degradation in the tangential direction. But we're not seeing the above with rangefinder lenses, which have short exit pupil distances, on the a7 cameras. We're seeing a lot of smearing, as stated by Nasse.
I'd be curious to see the MTF for the ZM 35 on Sony. Is there a link to that info, or can you post a screen grab, etc. (of both would be ideal)? Do you have a link to where center performance is better with thicker glass, than thinner? Is it for digitally optimized lenses, or film era lenses?
I found reference to this in Roger's post when the 55mm Otus was tested with no glass, 2mm and 4mm. It would seem to be optimized for 2mm, which one would expect considering it's a recent lens and mostly will see use on digital cameras. Note the Otus also has a 78mm exit pupil, so not terribly long relative to a traditional 50mm design. But at that focal length, I'd expect it would be better at handling potential edge degradation if used on a thicker cover glass. I'm not sure why central performance changes - it's beyond my understanding of this topic. Perhaps the lens was measured wide open and there are effects relevant to the greater divergence of light rays from a wide aperture vs. a smaller one when stopped down that interact negatively without the glass layer for which the lens was designed? Splitting the difference between performance without glass and 2mm, it would seem the Otus would still perform very well. Maybe a slight falloff, but would it be noticeable? That seems to remain the question with the a7 conversions, but my guess is whatever tradeoff there is with digitally optimized lenses will be minor compared to the gains we've seen with rangefinder lenses.
In the same link it's interesting to note that the C/Y 28/2.8 that Nicholas posted images from above, is 54mm exit pupil distance, which apparently may be short enough, for a SLR wide angle, to cause problems on the Sony cameras, and is more or less resolved with the modification.
As Michael mentioned, it would seem the only way to really know how lenses will perform after the modification is to try them. Interpretation of the numbers will remain educated guesses.
p.30 #15 · A thinner sensor stack may be possible after all!
Michael, we are actually saying the same thing regarding the "optical thickness" and full spectrum mod part. I wanted to point out that the different in filter stack thicknesses on different Canon bodies reported on Roger's blog are actually physical thickness. I believe they all have the same optical thickness.
Ron, I believe that MTF chart was only a theoretical one to illustrate Caldwell's point (hence the "perfect lens" part). I can't share the ZM 35 MTF charts on mirrorless from Zeiss out of respect for the Lloyd's subscription system. But essentially, Zeiss did their measurement at all apertures on a 2.5mm cover glass and a native (assuming 0.8mm cover glass) thickness. Actually, rereading that very page, Lloyd suggested the same thing: both astigmatism and change in field curvature. He also quoted Zeiss explanation on this topic, and they said there would be a shift in focus point (i.e. change in field curvature) due to the change in glass index.
On the topic of on-axis performance from the page you linked, this was what Roger noted:
"The MTF is better now higher in the center, but there is more astigmatism off-axis. (I was surprised at the on-axis effect, but Brian tells me that the amount of glass in the path creates on-axis spherical aberration that could affect center MTF on wide-aperture lenses. At least that's what I think he said. Any errors of interpretation are mine.)"
Again, Roger was not sure if his interpretation of the explanation was right, but it's still interesting. This also corroborates with the supplied Zeiss MTF of the ZM 35.
And I agree that most of what we are discussing here will remain educated guesses. However, I believe we can accumulate an overwhelming amount of evidences to support the idea. For instance, from my experience with the ZM 15 and G 21, I know the curvature toward the edges and corner bowing away from the camera. By simply focusing for the corners part, I can get them in focus (but still fuzzy because of the astigmatism). The center focus would be shifted closer to the camera at this point.
I don't have any native Leica camera, but I believe that you (1) don't need to refocus as much to get sharp corners and (2) the in-focus corners would not be as fuzzy (or just plain low in contrast) on a Leica M body.
p.30 #16 · A thinner sensor stack may be possible after all!
rscheffler wrote:
I agree about smearing, but not sure about general field curvature. Smearing is field curvature, but in just the tangential direction, according to Nasse. If thicker sensor glass also caused sagittal drops, similar to tangential, you'd expect to see something more like normal field curvature. Where I'm a bit confused about this is this graph posted by Roger:
It would somewhat contradict Nasse's statement that it's primarily a degradation in the tangential direction. But we're not seeing the above with rangefinder lenses, which have short exit pupil distances, on the a7 cameras. We're seeing a lot of smearing, as stated by Nasse....Show more →
This is just a guess, but being that this is Brian Cadwell's calculations for a 'perfect' lense, part of the answer might be that such a perfect lens does not exist. We're not dealing with lenses that would otherwise resolve 40lp/mm so effectively well at f/2. Perhaps smearing is the effect of the loss of quality at 20lp/mm or 10lp/mm?
It would somewhat contradict Nasse's statement that it's primarily a degradation in the tangential direction. But we're not seeing the above with rangefinder lenses, which have short exit pupil distances, on the a7 cameras. We're seeing a lot of smearing, as stated by Nasse.
Ron,
I think Nasse's paper and the chart above agree - both show larger decreases in the Tangential MTF from a mismatch between the lens design and the cover glass thickness.
If the spreading results from varying ray angle of the rays that should all focus on a single point, it makes sense that the rays coming in tangentially (i.e. from the line that runs through the center of the lens) have the widest range of angles and, hence, the most significant effect.
Have done some further testing and the problem below was due to my adaptor being too thin after the modification which explains a lot - after shimming it so that the Zeiss reached infinity at the right point it is perfect again
Have been doing a few further tests with SLR lenses after the modification and there is a small detectable change with regards to field curvature with some lenses and no change with others - not sure why though as the exit pupil distance argument may not stand up here.
For example the Zeiss 21/2.8 ZF.2 has always performed flawlessly after f5.6. On the modified Sony field curvature seems to be an issue now. I managed to find an old test image with this lens before the conversion and was very surprised at the difference. The Olympus 24/2.8 however behaves just as it used to and the Contax Zeiss 28/2.8 has actually improved.
Have included a reference image with the Voigtlander 21/1.8 which at f4 now is even sharper than the Zeiss ever was so I am not complaining :-) :-)
Despite this rather unexpected behaviour with some SLR lenses the Kolarivision modification was absolutely worth it - I have started enjoying using the camera again and using Leica M lenses with it makes all the difference.
Steve Spencer Offline • • • • • • • Upload & Sell: On
p.30 #20 · A thinner sensor stack may be possible after all!
nicoimages wrote:
Have been doing a few further tests with SLR lenses after the modification and there is a small detectable change with regards to field curvature with some lenses and no change with others - not sure why though as the exit pupil distance argument may not stand up here.
For example the Zeiss 21/2.8 ZF.2 has always performed flawlessly after f5.6. On the modified Sony field curvature seems to be an issue now. I managed to find an old test image with this lens before the conversion and was very surprised at the difference. The Olympus 24/2.8 however behaves just as it used to and the Contax Zeiss 28/2.8 has actually improved.
Have included a reference image with the Voigtlander 21/1.8 which at f4 now is even sharper than the Zeiss ever was so I am not complaining :-) :-)
Despite this rather unexpected behaviour with some SLR lenses the Kolarivision modification was absolutely worth it - I have started enjoying using the camera again and using Leica M lenses with it makes all the difference....Show more →
Nico,
How sure are you that your adapters isn't too thin? The Z* 21 f/2.8 seems to under perform when the sensor is too thin (i.e., when the lens is too close to the sensor). Non-floating element lenses like the Olympus OM 24 f/2.8 and the Zeiss C/Y 28 f/2.8 aren't nearly as affected by a thin sensor. In this case i suppose the conversion might have brought the sensor a smudge too close and that is affecting the 21 (which is very sensitive to this issue) and not the OM 24 or Zeiss 28.
