Focussing Issues

[archi4]

Advertisement (gone after registration)

http://www.l-camera-forum.com/leica-forum/leica-m9-forum/122176-m9-coincidence-infinity.html

Here is a really beautiful explanation. Look at the posts by Julian Thomson

[Advertisement]

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

[Guest mc_k]

http://www.l-camera-forum.com/leica-forum/leica-m9-forum/122176-m9-coincidence-infinity.html

Here is a really beautiful explanation. Look at the posts by Julian Thomson

 

Yes, his diagrams are beautiful--are they too beautiful? And why? This is what I was asking earlier in the thread.

 

Here I have plotted the distance to focus plane vs the distance to subject in the same fashion, representing the usual two adjustment points. The dotted line is an ideal rangefinder and lens.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[Lindolfi]

It is not correct to have linear graphs in a plot of object distance versus object distance based on errors in a rangefinder mechanism.

 

The plots by Julian Thompson are plotted in image distances and not object distances, in which case you do get linear graphs.

 

These errors in offset and gain (roller position and arm length) have to do with changes in image distance, which translate into a non-linear change in object distance. To show this (although it remains awkward to use plots in object distances), I've calculated the effect of 0.3 mm roller error (both ways) and 1% error in rangefinder arm length (both ways) as well as a situation in which a gain error is compensated at infinity setting. The rangefinder focus distance starts at 0.7 meter and runs on to 1000 meter.

 

 

Both the offset and gain errors give an increased error in focus with growing object distance, also when expressed as percentage of that distance. When infinity is set correctly (even when there is a gain error), the percentage drops when going towards infinity. That is the reason why infinity setting is so important. Luckily it is also the most accessible one to adjust in the camera.

 

(the calculations were based on a focal length of 50 mm and standard formulae used in optics)

[thighslapper]

My one solitary recalcitrant lens is off to be recalibrated today (all the others focus perfectly with my replaced M9 body) by a well known UK Leica specialist....

 

There was a distinct snort of amusement on the phone when I told him I was using an M9.... clearly the tolerance demands placed by this camera on the lenses is providing him with with plenty of work....... and he bemoaned the use of test charts which he said were the bane of his life with users sending him lenses with miniscule 'errors' that were well within tolerance for everyday use.

 

The offender is a used (mint) 50/1.4 that has obviously lain in a drawer in california unused for a year or so.... I couldn't face waiting forever for a new one....

 

Apparently even dried out/gummed up lubricant can provide the few 100ths of a mm error that is needed to screw up 100% focussing.

 

We will see what he finds....

[Guest mc_k]

...

The plots by Julian Thompson are plotted in image distances and not object distances, in which case you do get linear graphs.

...

 

Thanks, this is great info.

 

It was not clear what was being plotted, actually. Hopefully the picture below is correct; the range is for a 50mm lens, calculated from the lens equation. Dotted line represents a perfect body and lens. But how do you know it's still linear when it's off...can you see this from looking at the parts involved? Or do you have a source for this?

 

Thanks again.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[Lindolfi]

The linearity simply follows from the mechanism of the rangefinder arm. The position inward of the rangefinder cam on the lens is directly coupled to the choice of the optical focal point relative to the sensor/film plane. That position is translated by the rangefinder arm into an angle of the small prism that throws the light from the small window next to the "M9" logo into your viewfinder patch. That rotation is almost linearly related in angle with the pushing distance of the cam, since the arm is on average at right angles to the pushing direction. (To see this, mount a lens on a film M; open the back and inspect the motion of the rangefinder arm). The travel distance is almost 3 mm, while the arm itself has a length of 14 mm and so the maximal angle the rangefinder arm is outside 90 degrees is only 6.1 degrees (asin(1.5/14)) and so the error in nonlinearity over that angle is about 0.2%.

 

According to your second graph in the last post, you have played with a gain error of 23%, meaning that your rangefinder arm is not 14 mm, but 10.8 mm in length, so you have simulated an arm length error of more than 3 mm! That is well outside the adjustment range of the length of the arm, so has no real value as a simulation.

 

The graphs in image distance that combine with the plots I gave in the object distances can be seen here below

 

[Guest mc_k]

Advertisement (gone after registration)

...According to your second graph in the last post, you have played with a gain error of 23%, meaning that your rangefinder arm is not 14 mm, but 10.8 mm in length, so you have simulated an arm length error of more than 3 mm! That is well outside the adjustment range of the length of the arm, so has no real value as a simulation...

 

I don't think you're supposed to read anything off the graph. it's a picture that tells you what direction to do the adjustment, when properly labeled. I exaggerated the error so you can see the difference between the lines.

 

Now I'm confused, where is the "fixed point" in your graph of the gain error? (I know in real life it's not quite fixed, but you know what I mean.) Thanks.

[Guest mc_k]

I think you changed the second graph?

 

EDIT: I think the original graph said 1% gain error and had roughly parallel lines

[Guest mc_k]

The following is your first graph; don't take the numbers too seriously. What are you using to plot these? (pretty)

 

In your second graph, you should take the 0.7m as from the focal plane to get the correct image distance (for the Leica, and other small cameras). It doesn't matter here, but it may somewhere else.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[Guest mc_k]

...That position is translated by the rangefinder arm into an angle of the small prism...

 

this is the part I'm unsure about; in the only Leica diagram I've ever seen, the arm is connected to another arm, which the prism sits on. But I'm not even sure it was an M rangefinder. Someone referred me to Kisselbach, but I couldn't find anything in that book at all. Thanks for the details...I will think on it.

[Lindolfi]

Yes, I have changed the graphs, because originally I calculated the gain to be a fraction of total image distance and that is wrong: If the gain is to be caused by changes in rangefinder arm length, it should have zero effect when the arm is at right angles to the optical axis. So I implemented that. Now that the effect is so much less, I increased the gain to 7% to make the effect more visible.

 

The plots are made in MatLab. If you have access to MatLab, I can send you the m-file that calculates the curves and plots the graphs.

 

My description of the rangefinder mechanism does not do justice to it, so here a more complete one:

 

1] The cam on the lens moves inward when focus is turned from close by to infinity

 

2] The roller of the rangefinder arm, which is in contact with the cam, is pushed inward

 

3] The rangefinder arm is turned around its axis and the angle of turn is determined in offset by the eccentric axis setting of the roller (offset) and the length of the arm (gain)

 

4] On the same axis as the rangefinder arm, a second arm turns with it over the same angle inside the rangefinder compartment.

 

5] This second arm pushes against an element rotating around its own axis, which carries a tiny lens. This element is spring loaded to avoid backlash or play.

 

6] By turning this lens, light rays from different directions, coming in through a small prism that is fixed behind the small window next to the "M9" logo on the body, are thrown into the rangefinder patch of your viewfinder

 

If you look at the turning angle of the element with the tiny lens, it is less than the turning angle of the rangefinder arm. That has to do with the longer lever arm on the element with the tiny lens than the one on the rangefinder arm axis, that is in contact with that longer lever arm.

 

There is a second action going on, which is pushing of the framelines up and backward in the rangefinder housing as the rangefinder roller is pushed in, in order to correct for parallax that is produced by viewfinder axis and optical axis of the lens not coinciding. This pushing action is performed by a notch directly connected to the rangefinder arm.

 

Erwin Puts has published some images on his site of the mechanism: click

[Lindolfi]

Here is an additional gif animation I made of the rangefinder mechanism in action:

 

 

And here the numbered parts

 

 

[1] roller of rangefinder arm

 

[2] arm that is on the same axis as the rangefinder arm

 

[3] longer lever arm on the axis of the tiny lens element

 

[4] tiny lens element

 

[5] prism behind small window next to "M9" logo

 

[6] framelines frame

 

(This mechanism is that of an M4-P, but the M9 has the same mechanism, which I leave closed for now )

[Guest mc_k]

...

 

I don't have Matlab, but I would love to know what function you plotted (before and after).

 

Thanks for the "action photo" and additional info.

[Lindolfi]

The MatLab code is annotated heavily, so should be quite easy to read. I've published it on a normal web page, which can be found here: click

 

This code is simple and quite easy to transfer to your own environment, so that MatLab is not needed. Have fun

[Guest mc_k]

...

This code is simple and quite easy to transfer to your own environment..

...that would be "pencil and paper"

 

o.k., "image distance at which rangefinder arm is perpendicular to optical axis

bcenter=0.0513" ... so this is the fixed point for the gain. This is something you measured on your own camera? It just about corresponds to an object distance of 2m.

 

That is pretty wild what you showed originally--how the gain adjustment almost doesn't matter. I guess that's why some cameras only have the one adjustment.

 

Thanks.

[Lindolfi]

I measured it in the M4-P through the back when the shutter was open, by turning the focus tab of the lens until the arm was perpendicular, measuring the focus distance with a tape and back calculate to image distance.

[bocaburger]

When looking at the slotted eccentric ring on the gain pivot, after unlocking it with the screw, turn the slotted ring clockwise to correct back-focusing or counterclockwise to correct front-focusing. That's from hands-on experience adjusting my M9. (But the graphs and mathematical explanations are neat ) NOTE: first be certain infinity is properly adjusted, and re-check and re-adjust if necessary after adjusting the gain.

 

The adjustment isn't difficult, but the big issue is whether it needs doing. I had 11/12 lenses front-focusing exactly the same amount, so I was 99.9% certain the rangefinder was the culprit. (After gain adjustment the 12th lens was then back-focusing by that same amount and is now gone for re-collimation). The fewer lenses one has, the less confidence one ought to have that the rangefinder should be adjusted. The coincidental odds are against 11 lenses all being mis-collimated by exactly the same amount, but with 2-3 lenses not so much. It's possible to adjust the rangefinder to perform adequately with mis-collimated lenses, but not the preferable method unless you never plan on buying or using any further lenses.