Tech cam lens sample variation

[torger]

My recent experience with SK60XL raises a few questions about lens sample variation. With the 135 format there's lens rentals and they sometimes do measurements of several lenses so you can find some documentation of what variations to expect. In the tech cam world I don't know of any such statistics.

Anyway, I've pixel-peeped three SK60XL copies. Two tests made by myself on my own camera, and one test made by one of the American dealers. This lens has a reputation of being very sharp, and has an image circle of 120mm. As with all digitars you can't expect good sharpness at the edge though.

At f/11 all three copies show substantial left-right sharpness difference, to a varying amount. This can be seen at already at 90mm image circle. At f/16 it mostly evens out, but I don't think my first copy would. I sent back my first SK60XL to Schneider-Kreuznach, which said it performed even better than it should and sent it back without action. I haven't got it back yet but I expect the re-testing show exactly the same. To make sure it's not a camera skew error I re-mounted the lens upside down and the good side switched sides, which it wouldn't do if the skew was in camera.

Some sample variation is to be expected, but it seems to be a bit larger than I was expecting. Also it seems worrying that Schneider-Kreuznach's repair shop isn't able to detect those errors or fix them. I also asked if they regardless could take the lens apart and remount it with maximum possible precision (like a camera body recalibration or digital back recalibration), and that I would pay for it, but they just replied that they couldn't do that. Is this how it's supposed to be? If you happen to get a half-bad copy in the lottery there's no way to fix it, even if you're prepared to pay for it? Is the only way to buy a new lens, test it, send it back and repeat until you get a good copy?

I've attached the results from a loaner SK60XL which is showing the best result of the three I've pixel-peeped. It's performance is good enough for my needs (as I rarely go beyond 90mm image circle, and I generally shoot at f/16 as I'm allergic to aliasing) but it's not really what I was expecting based on the reputation this lens has. It's clearly super-sharp in the center though.

I've compared to results from a Rodenstock Apo-Sironar Digital 4.5 / 55mm, a considerably cheaper lens based on an analog design. The SK60XL is much sharper in center, but if you would shift a lot to stitch panoramas I can't say that any of the three copies of SK60XL I've pixel-peeped provides a valuable advantage, as in the outer edges of the image circle on its softer side it's not better than the 55mm, and I was not expecting that.

I've thought that one strong reason to shoot medium format is the ability to get lenses that perform close to their theoretical maximum, both due to better quality assurance, and due to the larger scale (which reduces precision requirements). My observations so far with the SK60XL makes me worry a bit though, both that sample variation is larger than it should be, mounting precision less than it should be, and more difficult (impossible?) to fix a bad copy than it should be.

What's your thoughts, experiences, tips?

[Rob C]

I don't have the lens in question, but have had an experience with bad copies of lenses.

As I live in the sticks, there was no chance of backing and forthing with various copies, such willing dealers unheard of in my part of Spain. I just traded it back in for something else I didn't really need. Even more to the point, it was the first time in my career that I'd come across such a thing as a 'bad copy' because prior to that moment (8/6/209) I had always been perfectly satisfied with whatever I'd bought. Except for a Nikon F4s that never loaded properly first go.

What I think you, I and all the others who have found themselves on the receiving end of this sort of crap are going through is nothing more and nothing less than a new system of QC. Instead of the maker paying to run the final line, we do, and better yet, if we as buyer are not bright enough to see we have a lemmon, QC becomes nobody's cost.

I don't think these firms realise that this will eventually impact on sales, because 'impulse' buys will become more rare when the happiness value becomes tainted with doubts about whether it will end up being a happy deal after all. Not all impulse buys are cheap ones - that depends on the people buying. Perhaps Leica glass understands this better than most, even though it can't do much about its camera sensors...

Incidentally, as I was looking for the invoice for that crappy lens, which at €1418.42 (wholesale) back then should not have been crappy, I couldn't help notice the pile of invoices for other camera stuff post-digital: I must have been insane parting with so much for so little. Maybe I've come to my senses. Thanks for inducing the search!

Rob C

[torger]

Actually I'm not 100% sure if this is a crappy lens, or if I'm expecting the "impossible". The blurry part corresponds to the outer edge of the sensor shifted horizontally 25mm (~100mm image circle diameter for my sensor size), so it's really pushing it. I can't complain about the performance with the sensor centered, not even with my more obviously bad copy.

I did not discover that it was bad until a fellow forum member asked me to test it for panoramic stitching. Up till then I've only used it for like 10mm shifting at f/16 and for that it was fine. The lens is very expensive though and is supposed to handle much more than that, so I was a bit surprised that it was not better than when pushing it. And then discovering that three out of three lenses I've been able to pixel peep left to right all had some sharpness issues at one side.

When demonstrating/testing huge shifts often only one side of the lens is checked, and if you happen to check the good side it looks good.

Joseph Holmes had a rant a number of years ago that tech cam lenses was not up to spec, but if I remember correctly he tested them fully open. I can give the manufacturers more slack than that and just expect the lenses to work well at the shooting aperture and smaller (generally f/11 and smaller for these lenses). However, at f/11 I was expecting the SK60XL perform better than I have observed so far. With it's claimed 120mm image circle I'd say that about 70mm is high quality for f/11 if you check the worst case side, and I think that is too small.

[Bart_van_der_Wolf]

Actually I'm not 100% sure if this is a crappy lens, or if I'm expecting the "impossible". The blurry part corresponds to the outer edge of the sensor shifted horizontally 25mm (~100mm image circle diameter for my sensor size), so it's really pushing it. I can't complain about the performance with the sensor centered, not even with my more obviously bad copy.

I did not discover that it was bad until a fellow forum member asked me to test it for panoramic stitching. Up till then I've only used it for like 10mm shifting at f/16 and for that it was fine. The lens is very expensive though and is supposed to handle much more than that, so I was a bit surprised that it was not better than when pushing it. And then discovering that three out of three lenses I've been able to pixel peep left to right all had some sharpness issues at one side.

Hi Anders,

Sharpness issues at one side, more than likely also indicates that center performance could be even better. If a lens is decentered, and that is causing it, then optical performance is affected throughout the image circle.

I always check new lenses for 3 things at least.
1. Center resolution. I use my resolution test chart for that, and expect a good lens to resolve to less than a blur diameter of 100 pixels (92 pixels is Nyquist limit) at optimum aperture. Closer to a 92 pixels diameter is better.
2. Corner resolution must be the same in each corner, to proof that there is no decentering. To avoid field curvature effects or lens mount issues (non-perpendicular optical axis), I focus and shoot each corner individually with Live View.
3. If focused on a flat plane, or by diagonal framing on the horizon, opposite corners should be equal. This will verify optical perpendicularity.

This will also make it easier to make a case for a repair, or a swap for a better copy.

Cheers,
Bart

[torger]

I have only a view camera (Linhof Techno) to test with which makes impossible/very difficult to do some of your tests. I must focus on the ground glass to start with which makes focus precision somewhat limited, and you can't really repeat the exact same focus setting.

As this camera has tilt/swing on the front standard and you have a bunch of interfaces which can cause skew, the camera itself is always the prime suspect. To rule out the camera there's one trick though, to mount the lens upside down on the lens board. That is without changing any camera setting, release the lens, losen the fastening ring turn the lens upside down and fasten it again and mount it back and shot exactly the same shot again. Of course there's a tiny risk that you disturb the camera somehow so some other setting is moved, but I made the whole procedure three times on different occasions and got the same result each time (the sharp side switched sides), so I'm as certain I can be that it's not the camera. It can still be the copal shutter though.

Although I may seem to be a very technical user, I'm quite relaxed when it comes to lens testing. I shoot something real and if I'm satisfied with the result I'm fine, and I don't expect perfection especially in the outer range of the image circle.

Test pictures really doesn't help when talking to Schneider-Kreuznach, they say "we don't look at pictures" (I've tried), the only thing they do is that they run their test procedure using their instruments and if they're satisfied with the result the lens is good and that's that. However I'm having Linhof Studio as dealer on this lens and they've really service minded (they sent me the loaner lens, even without me asking for it), I can recommend them to anyone buying this type of gear, so we'll work out something in the end. However they can't fix lenses of course, only(?) Schneider-Kreuznach can, and if they can't fix it, then the lenses aren't better than this. I can personally live with that as my shooting style is not so demanding, but I want to get to the bottom of it not only for myself but because I'm interested in how these issues are handled, makes a good addition to my ever-growing Linhof Techno review (which is about much more than just the Techno).

I don't know for sure, but as far as I understand there's no guarantee whatsoever how sharp a lens should be, it's just about trust and fulfilling expectations. And there is always at least some sample-to-sample variation, so imagine if you're a dealer -- it must be really risky to start discussing border-cases with customers. Sell the lens and hope it's good, and if not hope the customer doesn't see it or doesn't care... prior to this I thought it was always possible to fix a lemon lens, but now I'm not so sure any longer at least not with Schneider-Kreuznach. I really need to re-test the copy I sent to them (hopefully early next week) to give a final verdict of that. There's still a minor risk that I messed up the testing of that lens. I've tested this loaner lens much more thoroughly and carefully.

[torger]

By the way, anyone who knows that if you get poor lens performance if it's always some sort of mounting issue (lens element not properly spaced/centered), or if it's common that the actual glass elements aren't properly grinded? I've assumed that the lens elements are made in machines where the variation is extremely small, and that sample-to-sample variation is solely a mounting issue. Is this right?

[dchew]

Anders,
I suspect you are to some degree expecting the "impossible" based on the MTF charts from SK:
60xl MTF
They show quite a bit of astigmatism in those graphs. When they say "better than it should" I'm guessing they are comparing to these graphs.

I think so much of our perception of lens quality has to do with what we had before any specific purchase. For example, I'm happy with my 60xl because it is better than the 70hr copy I had, especially off-center. Bit I'm not sure if my 60xl is any better than yours. On top of that, most of the time I'm looking at images that have my sharpening presets just because of time constraints. And sharpened images simply look fantastic.

In regards to your mounting issue vs element quality, I suspect gross lens issues are always alignment issues. But when you get to more subtle differences the answer is "both."

Dave

[torger]

I'm no good at reading MTF charts. What I tend to do is to bring up an MTF chart for a lens I know, and then compare that to one I intend to buy and see if it's supposed to be sharper or less sharp than the known. Unfortunately the radius and spatial frequencies (and sometimes f-stops) often differ between diagrams making it hard to compare.

But indeed, at 50mm radius, 80% of that diagram, there should be quite some astigmatism. Astigmatism is seen also on the good side on the tested lens, a bit of sideways smearing. I have no idea of how to translate those MTF chart numbers into how it should look on a 6um pixel. I find it suspicious though that there is fairly large difference between left and right.

Anyway, here's the MTF chart of the SK35XL, which I also own. Last time I checked my copy it had equal sharpness left to right.
http://www.schneiderkreuznach.com/fileadmin/user_upload/bu_photo_imaging/Foto-Objektive/Fachkamera-Objektive/Digitale_Objektive/APO-DIGITAR_XL/5.6_35_XL/TD_AP_56_35_R47805_2CID.PDF
The SK35XL really only have about 75mm image circle that is high quality, the remaining 15mm up to 90mm is quite blurry. This can be seen in the MTF chart.

Let's compare SK35XL to the SK60XL. SK60XL: at ~50mm radius (my 100% crop), which is 80% in the diagram, the 60lpmm is at 75% radial and 50% tangential (f/11 infinity). The closest point on the SK35XL is ~35mm radius, where you get almost exactly the same numbers 75% radial and 50% tangential for 60lpmm.

In other words, a 100% f/11 crop from the SK60XL at 100mm image circle diameter should look equally sharp to a crop at 70mm image circle diameter from the SK35XL. Is this the case? I don't know right now, I'd need to have a look. I'm getting out now to make test shots at the same spot with the SK35XL to get some comparison material.

[dchew]

Not sure if this helps, but here is my 60xl back shifted down 18mm, so about a 98mm IC. Two 100% views are lens mounted normally then rotated 180 degrees.
f/11, focused with disto/hpf @ 35.5 ft. IQ3100, Alpa STC, std sharpening, no lcc

[torger]

Okay, comparison with the SK35XL was easier said than done. The thing with the SK35XL is that the image circle is good up to one point and then degrades very quickly. The reference point is on that slope, and a few steps left or right on the image makes a big difference, so I can't really find a reliable comparison point in the image.

From this quick test the SK35XL doesn't seem to be 100% equal left-to-right afterall at f/11 especially if you go into the breakdown part of the image circle, but within "the good" part it's fine.

From these tests I have a little theory -- the closer you get to the image circle limit the more likely it is you start to see differences left-to-right, and a lens that has a huge image circle that degrades slowly over a quite long range, like the SK60XL, it's more easily seen, than those that degrade sharply like the SK35XL (and the SK43XL too it seems). For longer lenses the image circle degrades very little up to the limit.

[torger]

Not sure if this helps, but here is my 60xl back shifted down 18mm, so about a 98mm IC. Two 100% views are lens mounted normally then rotated 180 degrees.
f/11, focused with disto/hpf @ 35.5 ft. IQ3100, Alpa STC, std sharpening, no lcc

I'd say that result look considerably better than my results from the loaner lens, or maybe rather that both your sides look about the same as the sharp side on my loaner lens. So now I've seen at least one good copy of the lens

I was about to accept the loaner lens as a good copy, but now I'm less certain. And there's always this nagging thought that maybe I'm doing something wrong afterall, that the camera is skewed or something. I think the lens rotation method is pretty fool proof though, so I can't think of anything I've done wrong...

Over at getdpi I got some info on how ALPA tests lenses. They don't reshim them or something, they just make a test shot with a 80MP back and see if it performs okay. If fine they sell it as their own, if not it goes back to the manufacturer. In that anecdotal story 1 out of 10 lenses went back to the manufacturer. When buying lens to a lens board like I do you don't get that extra quality control. I wonder if I may have had such bad luck that I've got two lenses in a row that's not performing as they should? Or maybe the SK60XL is more problematic than other lenses due to it's property with a huge image circle and a slow degradation curve?

[henrikfoto]

Not sure if this helps, but here is my 60xl back shifted down 18mm, so about a 98mm IC. Two 100% views are lens mounted normally then rotated 180 degrees.
f/11, focused with disto/hpf @ 35.5 ft. IQ3100, Alpa STC, std sharpening, no lcc

Could it be that Alpa has allready refused the bad copies before they are sold
In the Alpa mount?

Henrik

[torger]

Could it be that Alpa has allready refused the bad copies before they are sold
In the Alpa mount?

Yes, see the above post. ALPA does an extra quality control, I've heard that something like 1 out of 10 lenses goes back to the manufacturer. It's still pretty bad luck for me if I've got two lenses in a row that's not performing as they should...

That extra quality control is a pretty strong selling point for an ALPA system, but you get to pay for it...

[dchew]

I'd say that result look considerably better than my results from the loaner lens, or maybe rather that both your sides look about the same as the sharp side on my loaner lens. So now I've seen at least one good copy of the lens

I'm usually not that lucky. 

I was about to accept the loaner lens as a good copy, but now I'm less certain. And there's always this nagging thought that maybe I'm doing something wrong afterall, that the camera is skewed or something. I think the lens rotation method is pretty fool proof though, so I can't think of anything I've done wrong...

I agree the rotation test is pretty robust. If you do a focus bracket, does the soft side ever get sharp?

Yes, see the above post. ALPA does an extra quality control, I've heard that something like 1 out of 10 lenses goes back to the manufacturer. It's still pretty bad luck for me if I've got two lenses in a row that's not performing as they should...

That extra quality control is a pretty strong selling point for an ALPA system, but you get to pay for it...

I visited Alpa a year ago and was impressed with the equipment they have for lens testing. But you're right about paying for it.

[henrikfoto]

Yes, see the above post. ALPA does an extra quality control, I've heard that something like 1 out of 10 lenses goes back to the manufacturer. It's still pretty bad luck for me if I've got two lenses in a row that's not performing as they should...

That extra quality control is a pretty strong selling point for an ALPA system, but you get to pay for it...

Maybe if Alpa, Arca, Cambo, Sinar etc all test and send back lenses, most of the lenses Schneider and Rodenstock sell outside the big systems are bad copies?

I would be very suprised if they throw the refused lenses away....

Henrik

[torger]

I agree the rotation test is pretty robust. If you do a focus bracket, does the soft side ever get sharp?

I haven't had the time to make a focus bracket, but indications are that indeed it will get equally sharp on both sides but at different focus distances. The test shots have some grass in it too up to the camera and it seems like the focus plane is bending, differently on the right vs the left side. My amateur guess is then it's some decentering going on rather than badly grinded glass.

If so, shouldn't it be possible to fix a lens to become better? I mean I can send in my Linhof body and get it gone through and recalibrated, and I can do the same with my digital back, and you don't even have to have any specific alignment error. I'd like to have that possibility with my lenses too, but indications so far is that at least Schneider-Kreuznach will only fix something if it's grossly wrong. Imagine a service that could take your half-bad copy lens and upgrade it to a good copy for a few hundred euros. I'd probably service my wide angles in the blind just to make sure they're top notch without having to test the lenses myself (which is pretty boring afterall)

[dchew]

One other thought: I don't know anything about the Linhof. Can you rotate the lens in the lens board as opposed to rotating the lens board on the camera? Maybe this is what you are doing, IDK.

[torger]

One other thought: I don't know anything about the Linhof. Can you rotate the lens in the lens board as opposed to rotating the lens board on the camera? Maybe this is what you are doing, IDK.

Lens board can't be rotated, so I need to mess with losening the lens in the lens board and fasten it again turned 180 degrees. It's the better way to test anyway as it happens that lens boards are skewed too. In fact Schneider-Kreuznach did discover a skew in the lens board of the SK60XL I sent in, I still have a faint hope that that was the problem (as I haven't tried the rotation test on that yet), but the left-right skew was so small that I don't think it should be it.

[DanielStone]

Question for the OP:

Is the small "keeps it from rotating" screw installed in the back of the shutter? It's a TEENY TINY ITTY BITTY little screw, and can easily be overlooked when mounting a lens(I've only encountered these on Copal 0 shutters, and very rarely).

This little screw WILL "tilt" the shutter ever so slightly, which could cause the issue cited here in this thread.

I'd remove the lens from the board, and investigate if it is indeed installed. The screw WAS an included item on Copal 0 shutters brand new from the factory, as it was designed to keep the shutter from "spinning" when tightening the retaining ring/jam nut.

There was usually a corresponding slot that needed to be cut/ground out from the bored hole, as the screw's head was right on the edge of the hole's outer circumference.

Just a thought, but one I wanted to clear out as a possibility.
Removing the screw will not do anything bad to the shutter, as its sole reason for being there is described ^^^

-Dan

[torger]

Thanks for the tip. The Linhof boards have no dent for that screw so ot must be removed in order to mount it at all. I did remove it ☺