Wide symmetrical lenses on Fuji GFX bodies: the case of the Schneider-Kreuznach APO-Digitar 35/5.6

[rdeloe]

Mirrorless medium format cameras from Fuji and Hasselblad combined with digital view cameras from manufacturers such as Arca-Swiss and Cambo are an economical way to gain tilt, swing, rise, fall and shift movements. As with every tool, there are limitations. Notably, the flange distance of all mirrorless cameras limits the kinds of wide angle lenses that can be used. As an example, modern digital technical camera lenses from Rodenstock that are wider than 60mm are either highly limited or unusable because their large rear cells clash with the mirrorless camera. The same is true of most of the wide angle Schneider-Kreuznach lenses, but there are exceptions.

These constraints are generally well understood and well documented. In this thread, I want to address a problem that is specific to using wide symmetrical lenses on Fuji GFX cameras. The issue may affect other mirrorless digital cameras and possibly some medium format backs, but I’m limiting my scope to Fuji GFX, which is where I have experience. The issue is not well understood, so I want to shine some light on it here, and also offer solutions.

The case I’ll use is the Schneider-Kreuznach APO-Digitar 35mm f/5.6 (both L-88 and XL-102). However, the issue is not limited to this lens. For example, the same problem exists with the 47mm APO-Digitar (including its earlier “Digitar” variant, and probably the Super-Angulon 47mm f/5.6 MC, which is the same lens as the Digitar versions). The issue also affects other wide angle large format lenses (e.g., I’ve seen it on a Nikkor-SW 65mm f/4.5), and rangefinder wide angle lenses (e.g., it affects my Mamiya N 43mm f/4.5 L, Mamiya G 50mm f/4 L, and Mamiya N 65mm f/4 L). It doesn’t affect the Schneider-Kreuznach APO-Digitar 60mm f/4 and variants, and it doesn’t affect any lens I’ve used that is longer than 65mm. It also doesn’t affect any retrofocus wide angle lenses in my experience, e.g., the Pentax 645 35mm lenses all work fine.

If you’re using medium format backs on technical cameras or digital view cameras, you may find this thread interesting solely from a “morbid curiosity” viewpoint because you probably don’t have this problem. However, you may still be interested in the comments on re-housing lens cells. To perk your interest, here’s something you may not have seen before: a Schneider-Kreuznach APO-Digitar 35mm f/5.6 lens in an aperture-only housing. I discuss how to do this later in the thread.

 

[rdeloe]

In a nutshell, the issue I’m focusing on here is that wide symmetrical lenses (and wide rangefinder lenses with near-symmetrical designs) that work perfectly well on film, and on digital sensors from other companies, work poorly, or not at all, on Fuji GFX bodies. Image quality in the centre will be fine, but moving away from the centre, image quality degrades, and image quality may be low to unusable even when closing down the aperture.

This set of images illustrates the problem using crops from my Siemens Star wall test setup. The images were made using the same APO-Digitar 35/5.6 XL-102 cells, but in two different housings. The first set is from the central part of the image circle at f/5.6. Both versions from the cells in two different housings look good. The second set is the right side of a shifted 7.5mm image, also at f/5.6. The Copal 0 shutter version (right hand image in each pair) is unusable as far as I'm concerned. It doesn't get a lot better closing down the aperture, whereas the version at left in my B-0 housing is good and gets a lot better.




What causes this enormous difference? I strongly suspect the culprit is the thick cover glass arrangement on GFX cameras. According to LensRentals, the sensor stack on GFX cameras has three pieces of glass that total 3.24mm, which is quite thick. The cover glass layers are in the optical path, and seem to cause problems that are not seen with other sensors when using wide symmetrical and wide rangefinder lens that are nearly symmetrical. As I mentioned earlier, the problem is not limited to the Schneider-Kreuznach APO-Digitar 35/5.6 lenses, but I’ll focus on that lens because it’s an important example.

Referring to the comparison image above, the difference between the image on the left and on the right is that the front and rear lens cells are closer together in the images in the B-0 housing (left). In my experience, positioning the cells of wide angle symmetrical and rangefinder lenses closer together than the factory specification is required when they are used on GFX cameras. Some lenses go from OK to excellent following adjustment (moving the cells closer), while others go from unusable to excellent. It depends entirely on the lens.

For instance, before I moved the cells closer together, my Mamiya G 50/4 was very good at f/8, but weak at the edges and not usable shifted much at f/5.6. I used it for a long time thinking that’s just how it was. However, after calibration, it became outstanding at f/8 and remains excellent at f/5.6 even shifted. In contrast, the Schneider-Kreuznach APO-Digitar 35/5.6 XL-102 lenses I’ve used are unusable on GFX in their original Copal 0 mechanical shutters. However, once calibrated properly, these lenses should give very good results wide open, even with some shift, and should be excellent at f/8 across the image circle that can be used on GFX cameras (~68mm); more image circle is available with an XL-102, but is not accessible on GFX outfits because the rear of the lens hits the camera’s mount at 7.5mm of shift. I have a plan for getting around this limitation, but it’s still in development; if it works, that will be another post!

 

[rdeloe]

Now we come to the heart of the matter: how do you move the cells closer together? Whether or not that is possible depends on the physical design of the lens. In the rest of this post I’ll use the APO-Digitar 35/5.6 as the example because it nicely illustrates the two main options I’ve discovered. There may be other ways to solve this problem; if someone knows one, I hope they’ll chime in on this thread so we’ll have a more complete record.

Schneider-Kreuznach APO-Digitar 35/5.6 lenses use cells designed to fit in shutters and housings that use the Copal 0 standard. This standard specifies that the mount surfaces will be 20mm apart, with a tolerance of +/- 0.025mm. This standard applies to Copal 0 mechanical shutters and electronic shutters. Compur 0 and Seiko 0 mechanical shutters also use the Copal 0 standard, which is why you can swap cells among these different shutters. Many readily-available aperture-only housings for enlarger lenses also use the Copal 0 standard. Some examples include Schneider-Kreuznach Componon-S 100/5.6, 135/5.6, and 150/5.6, and Componar 150/5.6 enlarger lenses, to name a few.

In solving this problem for myself, I learned something interesting about how Schneider-Kreuznach built the cells that were destined for use with digital sensors and electronic shutters. They clearly decided that the +/- 0.025mm tolerance that was fine for mechanical shutters and film was not tight enough for high resolution digital sensors. Therefore, they created room for shims by lowering the height of the mount surface on the front cell of lenses that were built for digital and electronic shutters. This picture shows the front cell from my L-88 (left) and XL-102 (right) APO-Digitar 35/5.6 lenses. Notice how much lower the rim is on the L-88 cell. I've placed a shim on each mount surface to help make this clear.



My L-88 cells came in a Schneider-Kreuznach Copal 0 electronic shutter, and were fitted with shims. To the best of my knowledge, the L-88 cells were only sold in electronic shutters. In contrast, the two copies of XL-102 lenses I’ve had came in Copal 0 mechanical shutters and the cells were screwed tight to the mount, with no shims; I’ve only seen pictures of XL-102 cells in an electronic shutter, and I’ve never seen one for sale.

Why is this important? I was able to install the L-88 cells in a standard Compur 0 shutter and achieve excellent calibration for GFX by using thinner shims. Conversely, both copies of XL-102 lenses I’ve owned came in Copal 0 shutters without shims. This suggests they are designed to be 20 mm apart, and that Schneider-Kreuznach assumed they would work satisfactorily even if the actual distance between the mount surface is not 20mm, but still within the tolerance range (19.975mm to 20.025mm).

If you are using Schneider-Kreuznach APO-Digitar 35/5.6 XL-102 cells that were manufactured to go into a Copal 0 shutter – which seems to be the case for most of the copies that I’ve seen in the wild – then you have a problem if you want to use the lens on GFX. As illustrated at the start of this post, image quality will be poor or unusable away from the centre, and closing down the aperture will not help much. Fortunately, for GFX users, and anyone else experiencing this issue, there are two solutions:

First, you can machine down the front mount surface. This is a job for experts who have a high risk tolerance and access to professional tools. I know people did this because the front cell of my Schneider-Kreuznach APO-Digitar 47/5.6 very obviously was machined down to make room for shims. I bought the cells in an electronic shutter. They came with a bit over 0.8 mm of shims (more than I’ve ever seen) because whoever did the machining overshot the mark. Presumably the cells originally were in a mechanical shutter, and someone wanted to use them in the electronic shutter.



Second, you can shorten the distance between the front and rear mount surfaces of the shutter or housing to make room for shims. This is very easy, and you don’t need any tools beyond a piece of very fine sandpaper. However, it is a permanent modification. In a nutshell, you’re going to carefully sand off a bit of the rear mount surface, making sure to keep it even, and to not take too much. This makes room for shims under the front cell (but rear cell works too).

I like aperture-only housings because I don’t need or want a shutter. I chose the Schneider-Kreuznach B-0 housing because it is widely available, inexpensive, and easy to modify. If anyone is interested in additional details on how that works, I can follow up later in the thread. In summary, I shaved about 0.5mm off the rear mount surface (more than needed in the end but not a problem because I have plenty of shims). That’s all that’s needed for cells like the ones from an APO-Digitar 47mm f/5.6 lens, but the front cell of an APO-Digitar 35/5.6 lens doesn’t fit in an unmodified B-0 housing because the aperture ring is too tall. Therefore, I shortened the height of the ring using files and sandpaper, and shaped the inside to fit the front cell of the lens. This design leaves me with a skinny aperture ring that still has the numbers. I can turn it wearing gloves. The last modification to the B-0 housing involved adjusting the size of the aperture so that it was correct for a 35mm lens; this too is a simple adjustment when using B-0 housings.

 

[rdeloe]

The last task to accomplish was finding the optimum spacing for the lens cells on a GFX camera. I can elaborate on how I do this later in the thread if there’s interest. In a nutshell, you use the shims to move the front cell back and forth until you achieve optimum performance at the distances that matter for your work. I calibrated mine for “normal” photography.

This calibration process confirms why Schneider-Kreuznach designed cells that were destined for digital sensors to use shims. I’ve calibrated three APO-Digitar 35/5.6 lenses now (two L-88 and one XL-102), and in each case best image quality was separated from weak image quality by as little as 0.02mm – which falls within the Copal 0 standard. Even adjustments as small as 0.01mm resulted in improvements to image quality at maximum shift.

As a benchmark, a Schneider-Kreuznach APO-Digitar 35/5.6 XL-102 should be giving you excellent image quality at f/8 across at least the first 7.5mm of shift on a 100 MP 33mm x 44mm sensor. If you are only getting acceptable image quality at f/11 on your non-Fuji setup, then your lens is not properly calibrated. If it’s in a Copal 0 shutter with no shims, the only way to get the image quality improvement that is available may be to move the cells closer together, which means modifying your shutter (or moving the cells into an aperture-only housing you’re willing to modify if you don’t need the shutter).

A final reminder to keep in mind is that optimum image quality on GFX means sub-optimal image quality on (possibly) everything else. I can easily re-calibrate my wide symmetrical lenses for a Hasselblad CVF 100C, but I can’t use the same lens on GFX and Hasselblad without recalibration.



I've posted an image that looks like this before. It's a favourite scene I use for checking final calibration of lenses I'm setting up. This one is made with my new Schneider-Kreuznach APO-Digitar 35/5.6 XL-102 in its B-0 aperture only housing. I focused on the windows in the background. The image is shot at f/8, and is a composite of 7.5mm rise in landscape and 7.5mm fall in landscape. It's nearly sharp from front to back, with just a tiny bit of softening at the extreme left and right edges, which are at the margins of depth of field. Using f/11 brings a bit more of the very closest edges of the scene into focus, but also causes noticeable overall loss of sharpness due to diffraction. I try to use f/8 whenever possible for best results. Proper calibration means the field of focus is nice and flat, so I'm able to focus in the centre and shift without having to worry about field curvature.

 

[rdeloe]

Earlier in the thread I mentioned that when calibrating a lens like the Schneider-Kreuznach APO-Digitar 35/5.6 XL-102, very small differences in cell spacing can make a big difference on image quality. It's hard to believe that 1/100th of a millimetre makes a difference, but it does. To illustrate, I provide some examples here from testing against my Siemens Star wall. When I'm calibrating a lens, I square up against the middle star, leaving room on the left and right side for as much shift as I'm hoping to get from the lens.



For this demonstration, I used combinations of shims that put the front cell 0.25mm, 0.26mm, 0.27mm and 0.28mm in front of the mount surface. Using f/5.6, I focused on the centre, and then shifted left and right 7.5mm (the maximum I can shift the lens on GFX). To make what I want you to see clear, I've zoomed in on the far right and far left Siemens Stars in the central row. Each of these crops is 255 pixels from a file that is 8,736 x 11,648 pixels.



Using 0.28mm of shims makes the best overall image. The Right side is still not quite as good as the left side, but it's still excellent for f/5.6 with shift. All it took was 1/100th of a mm.

I've never been able to get absolutely equal results over the whole image circle. There's always one side or one corner that is not as strong as the other against my Siemens Star wall with the lens at maximum aperture. It might be the lens cells, or it might be any one of dozens of other variables in the chain (e.g., I think my F-Universalis is "perfectly" aligned, but an extremely small error is all it takes to throw something off). Nonetheless, I'm very happy with this result because when I take the lens out into the world to make actual photographs, it's giving me excellent results.

Here's the key takeaway: the lens came in a Copal 0 shutter, as these normally do, with assurances from the seller that it was in good order. But this is what the same scene looks like with the cells in the Copal 0 shutter it came in with no shims. It may well be that this lens is well-calibrated for film, or for a different digital sensor. But on a GFX camera, it is unusable due to the strong field curvature that is produce (I strongly suspect) by the thick cover glass of GFX cameras. Proper calibration matters.

 

[algrove]

All good and well, but for $4k the TS Fuji lenses seem the best value today.

 

[rdeloe]

All good and well, but for $4k the TS Fuji lenses seem the best value today.

The Fuji GF 30mm is excellent value. There's nothing else at that focal length that will give you 15mm of clean shift with low distortion. But it's also large and heavy, which matters to me because I carry all my gear. Plus there's the fact that it's 30mm, which is too wide for me, and the focal lengths I use are not well covered. The S-K APO-Digitar 35/5.6 holds down the wide end. I also have excellent lenses with large image circles for my F-Universalis that cover 50mm, 65mm, 100mm and 150mm.

It's one of those "horses for courses" deals. If you like using tilt-shift lenses (which I don't) and you need 30mm, then it's an easy choice.

 

[dchew]

Rob,
As usual, great info! I may have posted this somewhere before, but I created the following simple sheet, rounded to the 0.01mm, for determining the required shim thickness. Once you are sure your cells are too close together, you tighten the cells and begin testing. Unscrew the front cell 45 degrees and see how things look. Figure out if that is too far or not far enough, then investigate rotations in between. After you find the correct rotation from tight cells, use the chart to find the required shim thickness. This chart is based on the Copal 0 pitch which is 0.5mm.

Dave

 

[Doppler9000]

The Fuji GF 30mm is excellent value. There's nothing else at that focal length that will give you 15mm of clean shift with low distortion. But it's also large and heavy, which matters to me because I carry all my gear. Plus there's the fact that it's 30mm, which is too wide for me, and the focal lengths I use are not well covered. The S-K APO-Digitar 35/5.6 holds down the wide end. I also have excellent lenses with large image circles for my F-Universalis that cover 50mm, 65mm, 100mm and 150mm.

It's one of those "horses for courses" deals. If you like using tilt-shift lenses (which I don't) and you need 30mm, then it's an easy choice.

An additional point is that once you have a Universalis and the 35mm, the extra cost to add more lenses isn’t that high, particularly in the 80-150mm range. The second GF TS lens is an additional $4K.

 

[buildbot]

This is such an excellent thread for us DIY inclined photographers, thank you!

 

[rdeloe]

Rob,
As usual, great info! I may have posted this somewhere before, but I created the following simple sheet, rounded to the 0.01mm, for determining the required shim thickness. Once you are sure your cells are too close together, you tighten the cells and begin testing. Unscrew the front cell 45 degrees and see how things look. Figure out if that is too far or not far enough, then investigate rotations in between. After you find the correct rotation from tight cells, use the chart to find the required shim thickness. This chart is based on the Copal 0 pitch which is 0.5mm.

Dave

View attachment 218839

This is super helpful. I've printed a copy for my records.

One friendly amendment though is that when you start turning the cell out, you quickly reach a point where it starts to have play, which can throw off results significantly. In my experience, a quarter turn was generally safe.

Another tricky complication is whether or not turning the cell down on the shim gets you to a clear "tight" fit. Copal 0 shims in brass are very delicate once you get thinner than 0.1mm. They're easy to deform, so you have to be careful that turning the cell down finger tight is actually all the way down. It's for this reason that I much prefer working with new steel shims from Misumi.

 

[Ben730]

Many, many thanks Rod. These are very interesting insights that you are sharing with us here.

 

[vjbelle]

Great thread, Rob! Thanks for all of the work. Even the Hasselblad V lenses act differently when used on my GFX vs. 100 CFV. Much more edge softness and curvature when used on my GFX which always steered me towards the lenses reacting with the cover glass.

Victor B.

 

[Ben730]

Interesting statement. I have the feeling that the Rodie 90 HR-SW is a little less tingly sharp with the GFX than with IQ Backs. But it's just a feeling so far. I also don't have an IQ Back with the same pixel pitch and manual focusing is error prone due to the poorer display resolution of the GFX.
The Sigma Art lenses, on the other hand, remain razor sharp, perhaps thanks to AF.

 

[rdeloe]

It's actually a very interesting question whether we should expect a lens to perform as well in an adapted setting as on its native camera. I've had good luck with most adapted Pentax 645 lenses, but I can think of two cases where I had poor or so-so results that seemed off given the reputation of the lens.

Case 1 is the latest generation 35mm 645 lens that was reworked for digital Pentax 645. I didn't think it was anything special and kept my old manual A copy instead, but the fellow I sold it to who used it where it belonged -- on a 645Z -- was over the moon.

Case 2 is even more striking. I tried the Pentax D-FA 645 25mm lens, which has a very good reputation and was (and is still) a very expensive lens. I bought a copy from MAP Camera in Japan, a very reputable seller. It was in excellent condition. However, performance on my GFX setup was awful. People who use it on 645Z bodies say it is excellent. I don't know what it was about adapting to GFX that made it a mediocre lens.

At the end of the day, I'm just grateful that adaptation is possible a lot of the time. I can remember when the choice of system was extremely consequential because you generally had to "dance with the one that brung you". If you bought Pentax 645, you had to hope that all the Pentax 645 lenses that were available met your needs, because adapting other systems' lenses really wasn't a viable option.

 

[Ben730]

I am curious whether the Fujifilm GFX Eterna will be built differently.
Here the Fujinon Premistas, the Tokina Vistas, the Arri/Zeiss DNAs, Hasselblad Hs + Vs etc. would have to work perfectly and be interchangeable,
otherwise the camera will have a hard time on the market.

 

[Ben730]

Wasn't there such a problem with Leica M lenses on Sonys in the past?

 

[usm]

This a very interesting thread. Thanks.

I am new to DIY with such lenses. Do I understand this right, that you say a SK 35XL lens in a Copal 0 mount has to be moved in a aperture-only housing to bring the lenses closer together when used with a GFX or any other Digital Back?

a Schneider-Kreuznach APO-Digitar 35mm f/5.6 lens in an aperture-only housing. I discuss how to do this later in the thread.

May I ask you for a "how to do" for this?

 

[rdeloe]

Wasn't there such a problem with Leica M lenses on Sonys in the past?

There was. I believe the issue was very similar: wide Leica M lenses did not work well on Sony cameras (smearing in the corners). People discovered auxilliary lenses they mounted on the front of the lens to correct the field curvature.

 

[rdeloe]

This a very interesting thread. Thanks.

I am new to DIY with such lenses. Do I understand this right, that you say a SK 35XL lens in a Copal 0 mount has to be moved in a aperture-only housing to bring the lenses closer together when used with a GFX or any other Digital Back?


May I ask you for a "how to do" for this?

You are most welcome.

You asked whether you have to move SK 35XL lenses in Copal 0 shutter to aperture-only housings in all cases. The answer is "No", but there are nuances.

The good news is I have never seen anyone saying they had issues using the SK 35XL on film cameras, and on other digital cameras, notably medium format backs. Many people are using the lens without issue on Phase One backs and on the Hasselblad CFV 100C back. That tells me that the SK 35XL works fine without modification on systems other than GFX.

A caveat to that general statement is there may still be room to squeeze out better image quality from the lens on other backs. For example, I've read people saying they leave the SK 35XL at f/11 to get the results they want. I wonder if that's because they need the depth of field, or because they're unsatisfied with f/8. If the latter, then their lenses may not be set up ideally because on my calibrated 35XL, f/8 is as good as it gets. Using f/11 simply adds depth of field and reduces sharpness due to diffraction. Remember, the Copal 0 standard allows +/- 0.025mm on either side of 20mm between mount surfaces. That is more than enough variation to produce photographically significant image quality degradation.

The other nuance to your question is whether you have to move the SK 35XL cells out of a Copal 0 shutter to use them on GFX. The answer is absolutely not... if you are willing to modify your Copal 0 shutter. If you need the shutter to make images, then you don't have much choice but to find a way to bring the cells a bit closer together, which unless someone has a better idea means sanding the rear mount surface and adjusting with shims. However, if you don't need the shutter, I would leave the Copal 0 shutter alone because they are rare and getting scarcer. In contrast, B-0 housings are plentiful and may be inexpensive.

I'll post some pictures and notes later about how to modify the B-0 housing to take a SK 35XL front cell.