The best 28mm lens for the Leica M9 is the one that can focus accurately.
Examining my Death Valley comparisons between the Leica 28mm ƒ/2 Summicron-M ASPH and the 28mm ƒ/2.8 Elmarit-M ASPH, I found that all but 2 or 3 of the sixteen meticulous A/B comparisons I shot showed the 28/2.8 backfocusing by 2-3 feet (especially bad at closer and medium distances). The two lenses were focused the same way, and the 28/2.8 was distressingly consistently backfocused.
The rental 28/2 Summicron focused perfectly in every comparison. In short, the primitive rangefinder focusing wrecked every one of my meticulous comparisons for the entire week. This is disheartening, to say the least; maybe I’ll end up using 1 or 2 of them with the appropriate caveats, but it seems to rule out the best examples I had shot.
Readers of my Guides might not realize the meticulous and time-consuming effort it takes to properly review and compare lenses. This is one case where a week’s effort is for naught. It is never fast or easy to do it right, and it is one reason that for each comparison I publish, I might have shot 3 or 4 others to cross-check my findings.
I had sent the 28/2.8 Elmarit in for adjustment at Leica back in January (and also shown the focus error as a case study in Making Sharp Images). Upon return, my check of the 28/2.8 focusing had showed it to be OK, and it absolutely has not suffered any shock or vibration or trauma.. As discussed in Shooting a New Lens — Focus, I had carefully checked the 28/2 Summicron— it was spot-on with all my comparisons shot in Death Valley.
So I am left puzzled; I can only conclude that the 28/2.8 somehow went out of adjustment quickly, and reverted to its factory-inaccurate state (it had arrived brand-new with a backfocus problem). Its focus error on my recent shooting matched its out-of-the-box focus error, suggesting some slop inherent to the lens mechanics in coupling to the rangefinder focusing mechanism. Perhaps. Because I had checked it, and found it to be OK when it returned. I did not re-check it prior to my trip— big mistake. Thus, my confidence in the 28/2.8 is shaken. Could it be sample-specific? I have no way of knowing.
Focusing errors like these make a compelling case for a Leica M camera with Live View and a screen/zoom usable for checking focus accuracy— the M9 has neither. See my Jan 23, 2012 Q&A interview with Leica (6th and 9th questions).
Not just an A/B comparison accuracy issue
The 28/2.8 focus error was/is strong enough that it was not just an issue when matching focus for lens comparisons; it’s what I’d call 2 stop error— stopping down ~2 stops is required to regain the lost sharpness, e.g., ƒ/5.6 instead of ƒ/2.8. Of course, that does not move the desired zone of sharpness to its proper location, so it’s not a real fix for anyone choosing focus carefully for the subject matter.
One crop below is soft (the 28/2.8) and one is sharp (the 28/2), both are at f/2.8. The softness of the 28/2.8 is because the lens is back-focused; it’s unquestionably a very sharp lens even wide open at f/2.8 where it is actually focused.
Dave P writes:
I am not being disbelieving - I simply don't understand how
back-focus problems can occur with manual-focus lenses. The received
wisdom on this is usually that there is some discrepancy in the
optical pathways - sensor / viewfinder - but then you would expect
the same with every lens.
Is it therefore a rangefinder / cam problem with this particular lens?
DIGLLOYD: My other M glass focuses spot-on, as did the Leica 21/1.4 and 21/3.4 I shot on the same trip on the same body. It is a lens specific issue in this case.
"same with every lens": on the Leica M9, there is a complex mechanical coupling between lens and rangefinder mechanism; as the lens focusing ring is turned, it must cause the rangefinder mechanism to move by a precise amount, and for a wide variety of lenses. This process naturally has some accuracy limitations at different focusing distances (near to far), and accuracy is less good with longer focal lengths; very high precision is required. In the perfect case, rangefinder focusing can be very accurate. One the M9, a lens can focus accurately at distance, but show more error at medium and closer range (which is the case with the 28/2.8 discussed here). One cannot rule out internal issues in a lens (e.g. a slightly loose lens element which could tweak the focus if it shifts position).
Each lens on the M9 has to couple with this mechanism. Very small deviations can easily throw the focus off several feet. An internal mechanical issue in the lens itself could also be a factor (e.g., a subtly loose lens element, which occurred recently with my Leica 21/3.4 Super-Elmar-M, though its result was a blurry right side).
In a rangefinder, there are two optical paths; one via this focusing coupling to the lens, and the other the lens to the sensor. In a DSLR, the focusing path is to the viewfinder. When I say “backfocus” I mean that the apparent focus as seen when focusing is in front of the actual focus as recorded by the sensor— the actual recorded focus is behind (back) of the apparent focus while focusing.
There can be a discrepancy in the flange focal offset (lens to mount) and/or the flange/mount to sensor distance. This is why the Zeiss ZM lenses by default, being adjusted for film, often backfocus on the Leica M9, as I show in my Guide to Leica (while Zeiss adjusts the ZM lenses very precisely, they are adjusted to the official Leica M film specification, but the digital flange to sensor distance in the M9 is at odds with this specification, slightly).
All these focusing issues go away with a Live View feature (not found on the M9), because the Live View image can be used to focus, and it is the same optical path as the recorded image— no possible discrepancy since it is the same optical path.
For a DSLR, the optical path distance to the viewfinder (ground glass) and the distance to the sensor need to be in very close agreement (they can be out of spec, but so long as the two optical paths agree, it is not a focus accuracy issue). If those two distances are at odds, then there will be a consistent back or front focus error. However, other factors overlay a variability to the focus accuracy, making any focus error more or less noticeable: focal length, max f-stop, blur characteristics interacting with the viewfinder ground glass, and especially user error (eye/hand). The latter point is easily seen by refocusing 10 times by eye, then comparing results; a rangefinder tends to be more accurate in this regard.
In terms of lens mount to sensor distance, this can be a problem with DSLRs: at one point I had two Canon 5D Mark II bodies. One would focus correctly at the infinity stop, the other would focus slightly beyond infinity, lending a hazy look to the resulting image. Focusing with Live View renders it a moot point, but Live View is not always convenient. Assuming that the hard infinity focus stop is accurate is a Bad Idea. In this case, the out of spec Canon 5D Mark II would have a consistent focus bias for manual focus unless the viewfinder optical path were also off by the same amount (and I believe it was). Also, since the effect was slight and I use Live View for any critical work, it was not an issue for me.
Luc O writes:
First of all, one “ex pluribus”, I would like to congratulate you for all the valuable information that I can read on Zeiss, Leica and others in your website. Well spent money for me ! It helped me to be a 35mm Biogon ZM…for the Zeiss Ikon ZM camera.
Regarding the backfocus issue of the Leica 28mm lens with the M9, this is reminiscent of the frontfocus with some of the Zeiss ZM lenses which you reported as being “frontfocus” with digital M9 camera but spot on with film M camera (btw Sonnar 50mm ZM focus shift could be excluded as it is a “classical” lens with by-design focus shift). The consequence of your observed issues are that the new Leica lenses that are usually “spot on” with the digital M9 should back focus with the film M cameras in a general way (or that most of the old Leica lenses should front focus on M9 camera) . However, such problems have not been reported to my knowledge (even by the website tao of leica with very picky reports). In addition, Zeiss ZM lens seem to work well with M8 camera according to your’s and other’s reports.
With my many years of scientific research, I conclude that all your observations are not explained easily and do not appeared to be explained by a simple fact as a different distance of the digital sensor. In particular, I am not convinced by a different standard for the sensor position between digital and film M Leica camera (this would be a self-killing design feature for Leica cameras) and all that makes me thing about the old dispute about the Hexar RF camera as having not exactly the same lens-film distance as the true Leica M camera (strikingly, no such dispute took place in Japan).
Hence, may be it is time to search for another explanation ? One of them could be simply industrial variability ? Possibly, a more simple but even more annoying problem !
DIGLLOYD: Setting aside confirmation of my focus findings from Zeiss which I discuss in my Guide to Leica, film has thickness, and it is also an error in logic to assume that the difference is as important for film. Film also has curl and flatness issues. Digital is unforgiving of even slight errors. Yet I am certain that the errors I see with my 28/2.8 would show up on film, but that would require a film body that was adjusted to the same specification as my digital M9P.
With the M8 and my original Leica M9 before it was adjusted for accurate focus for my lenses (back in Nov 2009 through part of 2010), the Zeiss ZM lenses were focusing just fine. It seems that the current Leica M9 digital adjustment is not in conformance with the film specification, but that current lenses and cameras are adjusted for each other. All my recent experience supports that idea.
As for focus shift with the Zeiss ZM 50/1.5 Sonnar, it happens to cancel out the focus error by around ~ƒ/2.8. Which is not much help at f/1.4.
Every Leica M body rangefinder coupling is potentially different, but I have to be very particular with mine. All my M lenses with the exception of the 28/2.8 have to focus perfectly, or I would not be able to offer what I do. So I make that effort. When a sore thumb sticks out, it’s obvious. Also, the rangefinder mechanism can require some adjustment over time.