Left/Right Zone of Focus Skew: Not Always the Lens, the Sensor and Lens Mount Might not be Plane-Parallel
Get Canon 5DS R at B&H Photo.
See Case Study: Severe Lens Skew (D800E, Samyang 14mm f/2.8) for an example of clear-cut lens skew on a known-good camera body.
Symmetry issues are an issue at all price levels; see Asymmetry with Leica 180mm f/3.5 APO-Elmar-S.
The Canon 11-24mm f/4L (about $2999) is a must-have lens for the Canon shooter. It’s so much fun in the field! The 11-14mm range is a wonderfully useful extension versus the more limited range of the 14-24mm Nikkor. The 11-24mm shows its limits on the 5DS R, but short of Zeiss Otus, virtually all lenses show obvious limits, and even the Otii can be nitpicked more easily.
The Sigma DG HSM Art lenses are absolutely not in the same league as Zeiss Otus. Camera skew or not, they need a lot of stopping down at 50MP to get to quality that the Otii deliver wide open. Still, they are much better than the Nikon or Canon equivalents and are a superb value.
About 10 days ago, I sent in my Canon 24-70mm f/2.8L II to be adjusted because I was seeing a skew in focus: left side focused to the distance, right side to the foreground (this was from back in mid June). Canon returned the lens as “performing to specifications”.
The “performing to specifications” response caused a sinking feeling in my gut, so I just finished examining many images from the lenses mentioned above, all shot on the same Canon 5DS R on my recent trip.
All of the lenses are consistent: while not always obvious, any shot at distance shows a clear bias of focus to the distance on the left side, and focus to the foreground on the right side. In other words, I either have four bad lenses, or a misaligned camera body. That is to say that the lens mount is not plane parallel to the sensor.
That said, the Canon 11-24mm f/4L is much more subtle; wide open the periphery is not fully sharp, and there is significant depth of field at f/4 so it’s harder to see without the right scene. In most cases one can be forgiven for thinking there is no issue and with near/far scenes shot at any kind of angle and with closer focus (not infinity), most of my work looks perfectly fine, so I have a sense of relief there. Examining the f/1.4 Sigma lenses, the issue is more prominent (much wider apertures). And lens design can influence performance too (amount of telecentricity). Finally, closer focus reduces the issue; it is the highly sensitive infinity focus that is “touchy”. A suitable grassy or rock slope with distant focus is where the issue rears its head most obviously (50 feet out and farther for focus).
Given the evidence (consistency among four lenses), I think it’s fair to say that error is in the camera body. Which aside from damaging a great deal of field work, puts me in a bind as within 12 hours I have to head to an cycling event in the Sierra, and I was planning on shooting the 5DS R before and after. Now my shooting plans are in chaos, so I may just shoot Nikon and Otus and a few other things while I await a replacement 5DS R. I do have the new Sigma dp0 Quattro coming, though the thought of using Sigma Photo Pro to process more than a few images makes my stomach churn.
I intend to document the sensor/mount skew by example, as I think this is something everyone should be aware of when buying ultra high resolution digital cameras. Perfect symmetry is hard to come by, but at 22-24 MP, a slight skew was harder to detect. At 50 megapixels, the tolerances are now problematic and anything and everything shows up, and a slight error in lens added to an error in the camera can add up and combine badly. But in this case it appears to be just the camera at fault.
I shot the Zeiss Otus 55mm f/1.4 APO-Distagon in my backyard on a house and deck scene at about 80 feet distant. I cannot detect any skew with the Otus 55/1.4, even wide open.
Then I shot the Canon 24-70mm f/2.8L II on the same scene as the Otus 55/1.4, at 24mm and 28mm. It looks just like the bad results prior to Canon service: the left side shows reasonably good definition and the right side shows a strongly blurred area that hardly improves from f/2.8 to f/5.6. Camera or lens? It sure looks like the lens to me, and yet Canon service stated it is “performing to expectations”. And then those Sigma Art lenses aren’t symmetric either.
Next I tested the Zeiss ZE 21mm f/2.8 Distagon. I could not detect any lens skew. With these mixed results, I am forced to say I don’t really know for sure what is wrong. It could be a combination of factors. The right way to do this is to have at least two identical camera bodies and lenses and hold one variable constant. But dual 5DS R bodies are hard to come by.
Finally I tested the Sigma 24mm f/1.4 DG HSM Art. It shows a very mild right-side focus bias to the foreground, as seen in the field at greater distance (closer range may be harder to see and lens focusing can change things). It’s not a big skew just as I found in the field, but it is visible.
This is a hard one: I’ve dealt with skew issues for years, seeing it fairly often with various gear (see Sony 35/1.4 and Tamron 15-35 most recently), but this one has me tearing my hair out. Maybe it’s just bad odds on the lenses PLUS minor tolerance differences that happen to affect the 24-70, 24/1.4 and 50/1.4 lenses. Except for the Canon 24-70mm f/2.8L II (“performing to specifications”), I would characterize the skew as mild. Moreover it’s hard to detect at closer range (examining other field images). It appears to me that focus distance is involved, infinity focus being the most demanding and least forgiving of even the slightest issue, but also the lens design (perhaps the degree of telecentricity).
I once tested four Zeiss 21/2.8 Distagons on a Nikon DSLR; each was different but in the end the camera body proved to be the culprit (Nikon found it off slightly and it was perfect when it came back as many subsequent shoots proved out). Those 21/2.8 Distagons were sent back to Zeiss and were all MTF tested at the factory after my shoot, and only one was out of tolerance and not by much. A small error in the camera alignment can combine with a small error in the lens to make things additively worse or slightly better. Moreover things can change with focusing distance as elements change position and orientation internally.
Up in the Sierra Nevada today June 17, I shot the Zeiss Otus 55/1.4 APO-Distagon on the side of an old building. No skew.
The Otii are held to very tight tolerances, best in the industry IMO. I’m wondering if the camera is fine or just a tiny bit off and what I’m seeing is tolerances (“performing to specifications”) on multiple lenses: maybe “in spec” has not changed much since the days of 12-24 megapixel cameras, yet we are now at 50MP. It might even be the case that the manufacturing process for many lenses cannot be reliably implemented to the tolerances needed for 50MP.
Further examination of images from various lenses leads me to conclude that the camera is not likely to be at fault. I’d love to have a 2nd camera body, but the 5DS R remains in short supply. But in a week or so I’ll send it to Canon, and Canon has the sophisticated gear to test sensor/flange parallelism.