Testing Focusing Accuracy

[Englander]

Advertisement (gone after registration)

I haven't gone through all the posts about the M rangefinder and its lenses focusing inaccurately, but I have read enough of them--especially regarding the 35/1.4 ASPH-- and of the techniques used to come to their author's conclusions, that I am motivated to point out what looks like specious technique. The conclusions and their explanations may be correct but since the method seems suspect, correct conclusions would be more by accident than accuracy. I am not saying that M lenses are not sometimes in need of adjustment, nor that M rangefinders do not go out of whack, nor that there are not focus shifts inherent in various designs, only that one of the most popular techniques for testing focusing accuracy is probably not appropriate.

 

In order to test a precision apparatus, there ought to be an appropriate, precise and repeatable testing technique. Using a target stretching away from a tilted camera is not such a technique for rangefinder cameras, especially if you want to be precise. Such a method may be appropriate for testing focusing through the lens but the rangefinder itself is neither at the same height nor at the same position as the lens being tested.

 

The finder window is above and to the side; this is obvious, but its importance for testing focus seems to be overlooked. Distances are calculated from the finder window and it is not centered above the lens. The M compensates framing for parallax but does not shift the focusing mechanics in the same way for closer subjects. If the camera is tilted 45 degrees toward a ruler, the rangefinder itself is now CLOSER by a measurable amount to the target than the center of the sensor. This difference in distance may be enough to influence any critical testing of focusing. The rangefinder cannot be accurate both for its position and for a position that is behind it, eg, the center of the sensor. When people complain about 10 or 20 mm of inaccuracy and I can see that the rangefinder of a titled camera is that much forward of the imaging plane, it seems likely that the angle and not the rangefinder is the problem.

 

The same phenomenon would hold for tilting the target where one side of the target is closer to one side of the camera than the other: the viewing & range finding window is not the same distance from the target as the lens and the center of its sensor.

 

The only technique I know that avoids this problem and that is available to users for testing the near-distance focusing accuracy of M lenses and rangefinders is to have a target that is absolutely parallel to the imaging plane, to focus on that target and then to change the distance between the target and the camera by small distances without changing the lens' focus. Evaluation of the accuracy of focus is based on comparison of the resulting images.

 

The best method I have found for assuring absolute parallelism--and it must be absolute, not "sort of"--is to use two mirrors, one attached to the camera lens and one attached to the target, one of which has a hole in it for viewing (with rangefinders this one is on the target). While peering through the mirror attached to the target, the target is adjusted until an infinite reflection is seen in the tripod-mounted camera's mirror.

 

Either the target or the camera can then be moved closer or farther. But since parallelism must be maintained, I have found that mounting the camera on a micrometer macro focusing rail is easiest. The camera is then moved back and forth in increments as small as you like. Your lens and camera may be best at a zero position or it may be better closer or farther from the target. There may be shifts of focus based on aperture or they may not be appreciable, after all. Ultimately, I have always been amazed at how accurate the Leica equipment is and how often I have found that any focusing inaccuracies were frequently due to operator error.

 

When there have been inaccuracies, I have been much more confident of my conclusions using this technique because it is repeatable and the results are comparable. I can compare my results with others'; with the tilted camera method, that is unlikely and probably explains a lot of the variation in forum member's results. If I get different results at different times, the problem is most likely my technique similar to making pronouncements after miss-focusing in the field. If the best results are -10 mm on the first and subsequent iterations with one lens but 0 with another, I am certain the problem is the lens and can evaluate whether I think that at 1 meter I will ever be able to focus within that error while both my subject and I are breathing. Clearly, if it is -20mm with all lenses, there is a problem with the camera. I always hope to find it is my error because it is easier for me to correct my behavior than to send my equipment away.

 

Joe

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[Jamie Roberts]

The M compensates framing for parallax but does not shift the focusing mechanics in the same way for closer subjects. If the camera is tilted 45 degrees toward a ruler, the rangefinder itself is now CLOSER by a measurable amount to the target than the center of the sensor. This difference in distance may be enough to influence any critical testing of focusing. The rangefinder cannot be accurate both for its position and for a position that is behind it, eg, the center of the sensor. When people complain about 10 or 20 mm of inaccuracy and I can see that the rangefinder of a titled camera is that much forward of the imaging plane, it seems likely that the angle and not the rangefinder is the problem.

 

Joe--

 

In theory this is correct, I guess, but I actually haven't noticed this as a practical effect. When I shoot a ruler @ 45 degree angle, what I focus on is in focus, regardless of aperture.

 

More, of course, is in focus with wider angle lenses, and the longest I've got is a 50 right now, but I'll check the 75 when I get it too.

[jonoslack]

HI Joe

Excellently put - and of course, you are right to an extent. But a receding target is a much better way of evaluating what's happening, and we don't always shoot things which are parallel with the sensor. Clearly when shooting a receding target one should make sure that the actual point of focus is a point, and that the camera is not tilting downwards (or upwards). However, if your target is at the centre of the rangefinder this isn't likely to be the case.

 

In the case of the 35mm f1.4 in question, at f5.6 at 30 metres it was out by a factor of 10 feet or so - if that's a function of not being parallel to the plane then we should all give up rangefinder focusing!

[tashley]

Thanks Joe,

 

I understand what you're saying here and I agree in principle but I have to say that with my two examples of the 35 lux, I was only prompted to make these tests as a result of consistently misfocessed real world results which I was not getting with other lenses, even ones which are of similarly wide aperture and longer focal length. If I do the 'ruler' test with any of my other lenses, the results are spot on. Further, the guy at Solms himself said he had tested my two samples and some others and that they all did, under his lab conditions, exactly what mine did in my not so cool home 'lab'.

 

The methodology might be flawed but it did concur with Solms' own findings and with an awful lot of real world shots. My focus success rate with the 35 lux is very significantly lower than with the 50 lux and I am satisfied that this is because the 35 has no floating element and is therefore more prone to backfocus when stopped down.

 

The mystery that remains is why some people appear to have examples that don't do this when Solms now says they all do. Maybe I should try to find a mildly miscalibrated example!

 

Best

 

Tim

[Englander]

Well, the 35/1.4 ASPH seems to be particular problem for Leica. I mentioned on this forum previously that I had had to return mine AFTER it had been coded and suggested that others should pay attention to their lenses performing differently after they had been updated with a new coded mounting ring. I have previously had problems with the Noct/1 as well. I did not mean to suggest that there are not problems nor that your problems weren't real, but rather that the ruler-test was suspect and liable to suggest problems when there are none and that it was highly susceptible to operator error. Frequently people panic when they hear about someone else's problem, then they do a sloppy test and like all hypochondriacs, they become certain they have the same problem.

 

Jono, the problem is that even though we ALWAYS shoot things that are parallel to the sensor (because that is what a plane of focus is), we typically are not aware of the necessity to be careful about it. Depth of Field is an illusion, there is only one plane of focus and as a plane it has no depth; our task is making that plane coincide with what we want to be in focus and we rightly expect Leica to aid us in doing it.

 

Joe

[tashley]

Well, the 35/1.4 ASPH seems to be particular problem for Leica. I mentioned on this forum previously that I had had to return mine AFTER it had been coded and suggested that others should pay attention to their lenses performing differently after they had been updated with a new coded mounting ring. I have previously had problems with the Noct/1 as well. I did not mean to suggest that there are not problems nor that your problems weren't real, but rather that the ruler-test was suspect and liable to suggest problems when there are none and that it was highly susceptible to operator error. Frequently people panic when they hear about someone else's problem, then they do a sloppy test and like all hypochondriacs, they become certain they have the same problem.

 

Jono, the problem is that even though we ALWAYS shoot things that are parallel to the sensor (because that is what a plane of focus is), we typically are not aware of the necessity to be careful about it. Depth of Field is an illusion, there is only one plane of focus and as a plane it has no depth; our task is making that plane coincide with what we want to be in focus and we rightly expect Leica to aid us in doing it.

 

Joe

 

 

So what did you eventually do Joe? Did you get a working example or try something else? I'm keen to have a bash at the 35 Biogon or the CV Nokton and would love to hear of anyone else's experience with them - particularly WRT self-coding, the cyans, etc etc.

 

Best

 

Tim

[Englander]

Advertisement (gone after registration)

Tim,

I returned the lens to Leica for adjustment and have not received it back yet. I bought the lens new not long after it was first issued and used it successfully for many years. Fortunately I had done a focusing test with it (as described above) and the M8 prior to sending it for coding, so I knew it was working properly before they messed with it. Can't help you with using other brands on Leica; I have never done that but there are plenty on this forum who have.

 

Joe

[Shootist]

I haven't gone through all the posts about the M rangefinder and its lenses focusing inaccurately, but I have read enough of them--especially regarding the 35/1.4 ASPH-- and of the techniques used to come to their author's conclusions, that I am motivated to point out what looks like specious technique. The conclusions and their explanations may be correct but since the method seems suspect, correct conclusions would be more by accident than accuracy. I am not saying that M lenses are not sometimes in need of adjustment, nor that M rangefinders do not go out of whack, nor that there are not focus shifts inherent in various designs, only that one of the most popular techniques for testing focusing accuracy is probably not appropriate.

 

In order to test a precision apparatus, there ought to be an appropriate, precise and repeatable testing technique. Using a target stretching away from a tilted camera is not such a technique for rangefinder cameras, especially if you want to be precise. Such a method may be appropriate for testing focusing through the lens but the rangefinder itself is neither at the same height nor at the same position as the lens being tested.

 

The finder window is above and to the side; this is obvious, but its importance for testing focus seems to be overlooked. Distances are calculated from the finder window and it is not centered above the lens. The M compensates framing for parallax but does not shift the focusing mechanics in the same way for closer subjects. If the camera is tilted 45 degrees toward a ruler, the rangefinder itself is now CLOSER by a measurable amount to the target than the center of the sensor. This difference in distance may be enough to influence any critical testing of focusing. The rangefinder cannot be accurate both for its position and for a position that is behind it, eg, the center of the sensor. When people complain about 10 or 20 mm of inaccuracy and I can see that the rangefinder of a titled camera is that much forward of the imaging plane, it seems likely that the angle and not the rangefinder is the problem.

 

If you take some measurements you will find that the center of the range finder window and the center of the viewfinder window is equaidistant from the center of the lens mount. It has to be for the rangefinder to work properly. Yes it is above the center line of the lense but that is compensated for as is the focus distance. Set the camera on a tripod and focus on something a medium distance away. Mark on your subject the approximate center point of the RF patch then rotate the focus ring to the nearest focusing distance and to infinity and you will see the RF patch shift up and down and to the right and left of your original center point.

 

 

The same phenomenon would hold for tilting the target where one side of the target is closer to one side of the camera than the other: the viewing & range finding window is not the same distance from the target as the lens and the center of its sensor.

 

The only technique I know that avoids this problem and that is available to users for testing the near-distance focusing accuracy of M lenses and rangefinders is to have a target that is absolutely parallel to the imaging plane, to focus on that target and then to change the distance between the target and the camera by small distances without changing the lens' focus. Evaluation of the accuracy of focus is based on comparison of the resulting images.

 

The best method I have found for assuring absolute parallelism--and it must be absolute, not "sort of"--is to use two mirrors, one attached to the camera lens and one attached to the target, one of which has a hole in it for viewing (with rangefinders this one is on the target). While peering through the mirror attached to the target, the target is adjusted until an infinite reflection is seen in the tripod-mounted camera's mirror.

 

Either the target or the camera can then be moved closer or farther. But since parallelism must be maintained, I have found that mounting the camera on a micrometer macro focusing rail is easiest. The camera is then moved back and forth in increments as small as you like. Your lens and camera may be best at a zero position or it may be better closer or farther from the target. There may be shifts of focus based on aperture or they may not be appreciable, after all. Ultimately, I have always been amazed at how accurate the Leica equipment is and how often I have found that any focusing inaccuracies were frequently due to operator error.

 

When there have been inaccuracies, I have been much more confident of my conclusions using this technique because it is repeatable and the results are comparable. I can compare my results with others'; with the tilted camera method, that is unlikely and probably explains a lot of the variation in forum member's results. If I get different results at different times, the problem is most likely my technique similar to making pronouncements after miss-focusing in the field. If the best results are -10 mm on the first and subsequent iterations with one lens but 0 with another, I am certain the problem is the lens and can evaluate whether I think that at 1 meter I will ever be able to focus within that error while both my subject and I are breathing. Clearly, if it is -20mm with all lenses, there is a problem with the camera. I always hope to find it is my error because it is easier for me to correct my behavior than to send my equipment away.

 

Joe

 

Your method seem to be good but not easy for the normal user, Pro or amateur, to setup.

The shooting of a subject at a 45° angle away from the camera is good enough to check the focusing. If that is off then your method will show it is off also. If that test says it is OK then your method might say that it isn't. But if what is in the center of the rangefinder patch is in focus through the viewfinder and in focus when you look at the image then it's good enough.

[ho_co]

Joe--your analysis is extremely good, but I differ on three points.

 

First is the point Ed just made, viz that the lens is indeed at the center of the rangefinder base.

 

Second: You are right that the distance between the rangefinder and the lens are different when using a focusing target such as the proffered one for the D70, but only if the camera is in a 'landscape' orientation. Flip it to 'portrait' (easier to focus with the horizontal line anyway), and again the rangefinder and lens are the same distance from the target.

 

Third: It doesn't seem to me that it should be difficult for the rangefinder to be accurate for the focus plane a certain distance behind it. Couldn't that distance simply be calculated into the lens focus cam or into the rangefinder geometry? In other words, if you focus at 1 m, couldn't the camera mechanism simply be set to focus at, say, 1007 mm?

 

But again we are back to the question of rangefinder accuracy, which has to consider the subject contrast, mechanical slop in the system, resolution of the eye etc.

 

I think Ed is correct that for practical purposes the earlier threads offer adequate accuracy, while your procedure takes the accuracy to the max that could be done at home without Leica's testing equipment. Very well done, raising several interesting points!

 

--HC

[lars_bergquist]

One: If you want to accuse someone else of sloppiness, then your testing procedures shouldn't be even sloppier.

 

Two: This is about infinity focus only. 'Optical infinity' for any lens is usually deemed to be, in practice, 1000 X focal length. Here however we are testing not a lens, but a rangefinder. So I wouldn't dream of testing infinity focus on any object closer than about one kilometer.

 

Three: Even sceptically computed depths of field are paper thin with high speed lenses at close range. The practical difference between something nearly invisible in focus, and nothing in focus, is usually merely theological. Ever heard about stopping down? Never, during nearly sixty years of photography, have I imagined it to be a practical technique to use any lens wide open at one meter.

 

Four: The only affliction known to man more deadly than Rangefinder Angst is Meter Paranoia. It takes only a special mindset to acquire one of them, or even both.

 

The horrible old man from the Age of Guessing Focus

[Englander]

Wow! I am well aware that there are many schools and traditions of photography within the Leica M camp and that some of those traditions enjoy different aspects of the equipment. Different levels of precision are demanded by different users. "Good enough" for some is barely tolerable for others. The tradition of zone-focusing, which by definition doesn't really require precise focusing at all, is very strong among Leica users; but at the same time, there is also a tradition that favors precise focusing of lenses wide open at close distances.

 

Certainly the angled-ruler test of focusing accuracy is sufficient for some, but I thought it might not be precise enough for many. It is definitely easily executed. If the method I suggested seems to be more trouble than its worth or more accurate than necessary, don't use it. I suppose the part that sounds most complicated is the mirror adjustment, but such a technique is frequently used for adjusting the standards of view cameras and the mirrors are sold by several companies including Zig-align. Many people use focusing rails for macro work, so it seemed not to be too daunting a suggestion.

 

As a reply to some of the remarks made in response to my posting:

- the fact is that because the rangefinder viewing window itself is not centered in any normal orientation directly above the lens, the ruler test can easily indicate problems of precision where there are none. Coming to such an incorrect conclusion can lead to adjusting a camera or lens that does not need it or sending equipment away only to be told that it is not defective.

 

-you do not view the subject through a window that is directly above the lens; that the center of the rangefinder base is centered on the lens doesn't change where you view the object and that is why there is always a parallax problem. Even if the camera is turned to "portrait position," the window is still to the side and above the lens; but in the portrait orientation, at least, the rangefinder and the sensor are in the same plane so the results with the ruler-method can be more accurate than in the "landscape mode." If I were to use the ruler-method, this would certainly be the orientation I would use.

 

-even though the finder patch and the fame lines move for parallax correction, the focusing mechanism itself does not change its position by tilting downward or sideways when focusing on close objects.

 

-Howard, the rangefinder can only be accurate in one place at a time because focus, by definition, is on a plane. That's why a rangefinder, especially with shorter lenses, can be so much better than any other focusing system: it is not fooled by the apparent depth of field of wide lenses. The rangefinder is set up to find focus at the film plane (sensor position) when that plane is directly below the apparatus (in landscape mode); but when the camera is tilted forward, the sensor position is no longer directly beneath the apparatus. That shift of the sensor position is small and the resulting error may be small and insignificant to some or it may be small and significant to others. At close distance the tilt has far more effect than at greater distances.

 

-the question addressed is not "infinity focus," but rather near focus where the rangefinder system is most stressed and most accuracy is required. Finding out if the infinity stop on the lens and rangefinder coincidence at infinity are the same is fairly easy and another matter.

 

-Lars, if the technique I have suggested is "sloppy," I would be interested to understand how it can practically be improved. There are many portraits posted on this forum made at f/1 and f/1.4 at close distances so it would seem clear, even if it isn't to your style, many people like to do it and it is likely that all of them have "heard of stopping down." Rangefinder focusing requires placing faith in a mechanical system that is known to fail periodically; testing it may be paranoia to you, but to others it is simply "evaluating tools." I would also suggest that finding out if your meter is consistent, let alone accurate, is important to photographers who rely on one for exposure information. As I said earlier, I know there are traditions of "zone focusing" and "exposure guess-timating" that have produced wonderful images just as there are precisely focused and exposed images that are equally resonant. While concerns about focusing precision may be a deadly affliction, I would prefer it to suffering the depression of bringing back a lot of out of focus images as a result of not having evaluated my equipment.

 

Joe

[Shootist]

I shoot my tests with the camera mounted to a tripod and leveled and then tape the target to a wall or my refrigerator horizontally and position the camera at a angle of 45° to the target. So I guess you could say the camera is in portrait orientation compared to the target/subject.

[ken_tanaka]

This is all good, clean fun. But beyond taking a few minutes to get a general feeling for how your own camera and lenses behave it's not worth becoming computer-geek-style obsessive about this. Realize that the rangefinder focusing mechanism is an artifact of the 20th century's industrial age; it's a marvel of mechanical engineering, precision machining, and assembly. But it is, at its best, an instrument of approximation. Its accuracy will likely fluctuate with each lens, with the age of the camera, and probably with the camera's temperature.

 

If want extremely accurate focusing the rangefinder is not the right instrument. Otherwise just relax and enjoy the M camera (including the M8) for what it was always designed to be; a small be-there camera that takes pretty good pictures.

[Jamie Roberts]

Ken--

 

You're probably right about absolute accuracy varying with each lens. But that's not what I see in actual practice.

 

There's no problem at all with any of the Elmarits wide open; so far, both my Summiluxes also close focus, wide open, with about the same degree of accuracy as my DMR.

 

Is it as razor fast as an AF system? No. But it works when the (current) AF system doesn't, reliably and predictably.

 

No doubt it will need to be realigned with age, but I don't know any system that doesn't need that (perhaps the 1d3 is the first one--I don't know yet).

 

I really don't know why Tim, for example, is having such problems focussing a 35mm lens (not at close focus, either) at f4 or f5.6.

 

That's just nuts--think about it for a bit. That *is* stopped down. The system shouldn't be out by large fractions of a meter or two, 20th century focussing technology notwithstanding.

[tashley]

Ken--

 

You're probably right about absolute accuracy varying with each lens. But that's not what I see in actual practice.

 

There's no problem at all with any of the Elmarits wide open; so far, both my Summiluxes also close focus, wide open, with about the same degree of accuracy as my DMR.

 

Is it as razor fast as an AF system? No. But it works when the (current) AF system doesn't, reliably and predictably.

 

No doubt it will need to be realigned with age, but I don't know any system that doesn't need that (perhaps the 1d3 is the first one--I don't know yet).

 

I really don't know why Tim, for example, is having such problems focussing a 35mm lens (not at close focus, either) at f4 or f5.6.

 

That's just nuts--think about it for a bit. That *is* stopped down. The system shouldn't be out by large fractions of a meter or two, 20th century focussing technology notwithstanding.

 

Neither do I but I am sure that it's not me, as a result of not having any issues with pther, longer, more fussy lenses... and don't g forget that Solms say they get exactly the same results as me...

 

Best

 

Tim

[carstenw]

Well, I had a million theories about why my 35/2 Asph shots were sharp and my 50/1.4 Asph shots were not, but today I adjusted my M8 to be dead-on from 1-2m for both lenses, and now infinity is off. I guess my M8 needs adjusting, but the lenses seem okay.

[ho_co]

I don't have an M8 (still waiting) and haven't tested these lenses as many you have, but I respectfully offer this:

 

The fact that some users of the 35/1.4 have no problems with its focus after adjusting their own rangefinders and that Leica can explain why the described problem is likely to occur is fascinating.

 

But I think there may be nothing new in the discussion. E Puts some while ago (before the advent of plane sensors in Leica cameras), while comparing Leica's and others' lenses, commented that Leica's computations tended to assume that film would buckle forward from the 'film plane' as described by the camera. The particular other brand's designers in the comparison had made other assumptions.

 

If the lens is designed to bring the image to focus just in front of the sensor, then the most nearly in-focus point on the sensor would be behind the point of attempted rangefinder focus (if the rangefinder is accurate, of course).

 

The rangefinder is set up to find focus at the film plane (sensor position) when that plane is directly below the apparatus (in landscape mode)

Joe--agreed. That was my contention, though my expression of that point was obviously unclear. I had misunderstood your original post to indicate that you felt the rangefinder mechanism could only offer approximate accuracy because the film plane was behind the plane of the rangefinder.

 

--HC

[Jamie Roberts]

Neither do I but I am sure that it's not me, as a result of not having any issues with pther, longer, more fussy lenses... and don't g forget that Solms say they get exactly the same results as me...

 

Best

 

Tim

 

Tim, which leaves me in a quandry. I just can't reproduce your findings, is all--and I'm both glad and sorry for that.

 

I'm now not even sure how best to test to make sure I'm really trying to put the lens through its paces!

 

But it's certainly not off anything like noticable at 10 feet, let alone close up.

 

So how to proceed? I'm not going to do the mirror set up for the simple fact I'm not into cutting mirrors around here.

 

I'm also not sure why Leica can reproduce this and I can't; I'm sure there are others with 35 1.4s that feel the same way I do.

 

So if you have any ideas at all, I'd still like to get to the bottom of this.

 

If we're left with the chrome lens body as the only difference--and they are somehow different, because this thing weighs more than my friend's 35--then I find that even more disturbing, in a way than all M8s being off with the 35 1.4...

 

@ Carsten--I'm sorry your infinity setting is off. Perhaps it's the close focus on your rangefinder that really needs adjusting.

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[tashley]

Hi Jamie,

 

I'm still trying to get to the bottom of this since I really want a usable 35 lux. I don't have a single rational theory now, other than the black/chrome thing. If your version was in some way happily mis-adjusted to frontfocus accidentally, your wide open shots wouldn't work.

 

What we really need is a wider sample, which is what I'd hoped to garner here.

 

And I'm with you on cutting up mirrors. If my ruler test gave the same results as my real-world experience and as Solms' lab test, it's 'within tolerance' for my purposes!

 

Just looked at your galleries again. Gorgeous. That lens sure does focus!

 

;-)

Tim

[jaapv]

What you would want to look at is if the arm that carries the connecting wheel of the camera is not slightly bent. It has been know to happen. The wheel must run exactly on the edge of the helix. If it is slighly out, it will be tilted by a fraction, producing exactly the kind of problems described here.