Is rangefinder self-calibration via the infinity adjustment really a workable solution?

No!  No!  No!

I see a lot of talk on Leica user sites about how it is "easy" to use a 2mm hex key to adjust the infinity setting, but that is a hell of a lot harder to do correctly than anyone lets on. You can make things vastly better with one lens at a time, but operationally, there are a few difficulties with multiple lenses - as I learned when I had to make an emergency field adjustment (to get a 90 Summicron to work). The point of this discussion is that it might take a few minutes to adjust your camera for a single lens but that you are unlikely to succeed at achieving perfect focus across multiple lenses. In other words, throwing a hex key in the bottom of your bag is not going to lead to any kind of meaningful field fix.

Theory

Start from the proposition that all Leica-mount lenses have a tolerance in the brass ring (cam) that contacts the RF roller in the body. Continue along with the assumption that all Leica lenses have the RF cam timed to simulate a nominal 50mm lens (it’s actually a little longer than that, but that’s a different discussion). Next consider that Leica makes a finite number of focusing mounts for each type lens (which to some degree cancels out the fact that there is a tolerance in focal length between optical units). In a completely unforgiving environment like 18x27mm digital, you are going to be lucky to have the focus perfect with one lens, much less a collection of three or more lenses.

Praxis

With this in the background, a lot of people will “tweak” their M8 body to work "perfectly" with their favorite lens. Whatever the surface appeal of this, the calibration process is both miserable and deleterious to the focus of other lenses.

First, the 2mm hex nut has a ton of inertia in it - enough to cause a 50mm long steel hex key to flex before the nut actually moves. This causes the nut to "snap" from one position to the next and by the time you can feel the hex nut move, it has already moved too far. By the way, the proper adjustment is often much, much less than the "2-3 degrees" I see recited here. The ideal way to adjust the nut is to tap on the hex key, but this threatens to pop the key out of the socket and damage the socket. It is also exceedingly dificult to do when you are trying to keep the arm immobile (see below).

Footnote: you need a very, very good hex key with a short arm that is perfectly hexagonal. Cheap hex keys often have damaged short ends as a result of how they are manufactured.

Second, it is hard to adequately immobilize the RF lever arm when making adjustments that induce nearer focus (i.e. moving the hex key toward the RF window to make the camera focus "nearer"). It's a tough mechanism, but even as someone who works on Compur shutters occasionally, it makes me nervous.

Third, you have to check every distance with every lens (and multiple apertures) every time you turn that hex nut. What makes your Summilux 75mm work unbelievably well can do incredible violence to the near-focusing accuracy of your other lenses - like a 35mm Summilux ASPH. The problem is that with Leica's acceptable tolerances in RF cam position, fast lenses will actually be all over the place. Then you have to account for focus shifts in lenses. Not fun.

Fourth, you will see variations in how your lenses line up at infinity. This is due to variations (within tolerance) in RF cam position. Most of the time this is not a problem. But be aware that RF errors work a much more severe effect as the distances from the camera increase. The idea is to keep RF error low enough that it is caught up in the depth of field of the lens as the distance increases.

Fifth, "balancing" your lenses will result in the DOF being strange with some. In the normal case, DOF is 1/3 in front and 2/3 in back of the focused point. When you attempt to reconcile various lenses, you will end up that some are 2/3 in front of the focus point and others 1/2 in front. The more of your DOF is in front of the subject, the less DOF increase you will get from stopping down.

At the end of the day, if you have a lot of lenses, you are really going to have a headache. This is not a process that takes a couple of minutes - settle in for several hours if you want to do it right. I ended up being forced to do it in the field - which beats 5 weeks at Leica - but it was unpleasant.

Tips

If you have to undertake this project (I only did by being forced to), I would recommend the following NOTE: We are all adults. I am not responsible for any negative results you get from applying these tips.:

0.  Pick your infinity.  We can talk about Samsara any time you like, but for the purposes of RF adjustment, just assume that we're in it.  Pick a suitably distant object, preferably 2km or more from where you are working. 

1. Understand your sharpness “goal.” What is the minimum level of sharpness you are willing to accept from each of your lenses on a flat subject at minimum aperture? One measure might be picking up the grain on a piece of satin aluminum at 1m. Or picking up the thread pattern in a dress shirt at 2.5m. Or the aggregate in cement at 5m.

2. Understand your DOF goals. What is your most important lens? How do you want it to behave?

3. Buy good tools. A good 2mm hex wrench costs a few dollars and is much less likely to damage your camera than one from the dollar (EURO) store.

4. Do your verification with a magnifier AND without one. Sometimes the magnifier impacts results. Likewise, if you use contacts and glasses, try your final adjustment with each.

5. Use a variety of subjects. You don’t always shoot flat objects; don’t test on flat objects. You are part of the system. How you focus will always influence the results.

6. Use flash. It might not be the best light outside when you try to fix your RF. Don’t kid yourself that a tripod is very useful for checking the RF. You will hand hold the camera in real life, and you need to check the camera against numerous real-world objects to make sure it works. Flash helps you move around – and because it creates a high effective shutter speed, it eliminates motion blur. I used a Nikon SB-20 with gaffer tape over 3⁄4 of the head, in manual-power mode, with the head set to bounce off the ceiling. At 1m with a 75mm Summilux at f/1.4 and ISO 160, the power factor is 1/16.

7. Bracket focus. This will help you determine what is going on with flat subjects. Initially focus the subject. Move the focus a little bit nearer. Shoot. Put the RF spot on. Shoot. Nudge focus farther away. Shoot. Then check all three at the focus spot at 1:1 by holding down play as you press the arrow buttons to switch shots.

8. Do not depend on the viewscreen. The LCD can help you understand whether something is more in focus than something else, but the quick-and-dirty demosaic-ing algorithm in the camera can get screwed up by certain object textures. Double-check on your computer. In general, you are getting close to perfect when the camera screen at 1:1 produces a moire on something like a finely-textured fabric.

9. Watch the eyestrain. It may take several hundred shots to align a rangefinder – so give your eyes time to rest now and then.

10. Understand that ultimately, you may need to use a special technique with one or more lenses. For example, with a fast telephoto, plan to focus at the nearest distance where the RF looks like it lines up. Or you might want to do your precision focusing from far to near (or vice versa).

Prospects for coupled RF

The thing I see from this is that the mechanical rangefinder is a dead-end technology with fast lenses, particularly if sensor resolution goes up. With digital, the tolerance for error is almost less than the practical unit of rangefinder adjustment - and acceptable manufacturing tolerances in the population of manufactured Leica lenses (even in the last 20 years) lead to inconsistent results on sensor. This helps explain why Leica is bringing out slower Summarit lenses.

Leica USA seems to be resistant to adjusting the cams on lenses that are “within spec,” meaning that you will always have one or two “special” lenses that require a special technique. I can see their point. Some error has to be tolerated; otherwise, the company could be sucking up warranty costs related to 30 to 50 year-old lenses rather than the M8 itself.

The solution to all of this is probably to replace the expensive RF system with some kind of passive focusing indicator/confirmation similar to that in a DSLR. This would have three principal advantages. First, it would allow the elimination of the mechanical RF coupling and thus eliminate a tolerance that could cause errors. Second, it would remove play in the RF system’s mechanical components. Finally, it would eliminate vagaries of human eyesight and eye position as sources of focusing error. Where you would detect the image is the hard part; maybe off a mirror on the front shutter curtain, with the system calibrated to assume an image plane sightly behind that.

Alternatively, Leica could offer a remanufacturing service whereby it would profile the lens cams by focusing them at multiple points and then computing the correct offsets along a curve - like a modernized version of Linhof RF cams for specific examples of lenses.

Or Leica could sell M8s with "best match" sets of 3 lenses. Or adjust your M8 to be a best match to 3 of your existing lenses.

Now where is that aspirin?