Improved Digital Rangefinder for the M?

[pico]

Advertisement (gone after registration)

Another crazy thought?

 

A different approach to digitally assisted focus confirmation for the conventional Leica M.

• It would not change the way one focuses the M
  
• no changes to any M (or LTM adapted) camera lens - perfect backwards compatibility
  
• no change to the camera's appearance
  
• would not require live-view so therefore the M could still use a CCD
  

Briefly: It would employ a tiny lens layered over a small sensor that fits inside the body, behind the camera lens. It would view the subject area coming from the camera lens that matches the patch area + an area surrounding the focus-patch coverage. (If it cannot be put aside like the BTL sensors are, then perhaps a flip-in such as the CLE used.)

 

A couple methods might be possible to make the assist work. In one method, contrast of the focus patch area plus surroundings found by the BTL sensor would be employed and the RF patch would light up slightly when the sensor has a profound difference. The photographer can then elect to accept the confirmation or not - press the shutter button or not. IOW, in ambivalent situations the photographer can elect to use only the current method of rangefinding. (FWIW we have ambivalent focusing situations with the current system anyway, and we cope beautifully.)

 

Another approach would have an additional sensor in the rangefinder mechanism. The photographer would focus in the traditional manner and when the BTL sensor detects concordance of the two sensors, there would be an indicator in the user's view window. I would prefer the patch to brighten slightly, but other LED indicators would do. An advantage to this method would be the possibility of having a pre-focus option so that once set to a particular distance, the camera could fire the shutter once something comes into concordance. This is possible with some DSLRs now.

 

Of course a lot of engineering has to be noodled-out, and I have to believe there are other approaches.

 

What I dearly want is what is in the bullet list above. Any way to reach that, but nothing can be removed from the list.

 

Collective wisdom is welcomed.

[Advertisement]

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

[Michael Geschlecht]

Hello pico,

 

While the M combined range/viewfiner mechanism has been 1 of the shining stars of an M it has also always been its Achillies Heel.

 

Range/viewfinder mechanisms have both +'s such as: Ability to see outside the angle of view of the lens in use & -'s such as: Complexity @ 1 : 1.

 

Another "-" is cost. Part of the cost problem was solved when Leitz went from the later M3 range/viewfinder to that in the M2 which was the beginning of the range/viewfinders used today. The majority of the cost differential between M2's & M3's (substantial) was not button rewind as opposed to lever or self timer or not: It was the range/viewfinder mechanism.

 

Another "-" is: Putting aside the M3 for the moment: The M2 range/viewfinder is really good @ what it does & it is really hard to improve on. That is as much a part of the Achillies Heel as the cost. Maybe more so.

 

Partially because:

 

It costs a lot to make & it is part of a receding technology.

 

@ the same time:

 

There is nothing out there competing w/ it within its parameters that comes close to equalling it.

 

Therefore:

 

How do you replace it w/ something equally good or better & save a Euro?

 

Answer:

 

Replace it w/ something less costly to make which meets or exceeds current parameters & which is preferably based on a developing technology which will be here for a long time. Not a receding technology.

 

Or:

 

Make it irrelevant or less relevant.

 

Or:

 

Some of both.

 

Keep in mind: The M2 rangeviewfinder has been around in pretty much the same form in a number of different cameras since 1958.

 

Best Regards,

 

Michael

[jaapv]

I proposed the second method a few years ago. The objection to this method is that it can never be more accurate than the existing RF mechanism, so it would have little added value.

[pico]

I proposed the second method a few years ago. The objection to this method is that it can never be more accurate than the existing RF mechanism, so it would have little added value.

 

How or why would it not be adequate?

[marknorton]

If you put framing and focussing in the same bucket, my concern with the existing design has always been the limited sweet-spot of operation. Going long or wide or if you have imperfect vision means you may want to buy into a range of expensive and fiddly accessories to frame and focus accurately. Leica sells at least 6 lenses which the camera cannot handle natively and there's no doubt I am more reluctant to use my wide angle lenses because of the then need to mess about with the clunky Frankenfinder.

 

In an ideal world, the M viewfinder would offer variable magnification, built-in diopter correction and a bigger eye-piece diameter to improve eye comfort. The FF may be clunky, but the improved eye comfort makes it easy to use. All of these suggestions relate to the viewfinder optics and it may just not be possible to include them without a radical change to the M form factor which is where the M5 landed Leica in trouble. The fact remains though that Leica has done next to nothing to improve things over 50 years as the focal length range and lens speeds have increased.

 

I think the viewfinder/rangefinder is the M's defining feature and is well overdue for a redesign.

 

As for focussing accuracy, examining what is coming through the lens is always going to be hampered by the fact the lens is stopped down. A stopped down lens will make determining focus more difficult both because of reduced light and increased depth of field. A separate sensor behind the lens would add massively to the complexity. However in a world where a future M might use a CMOS sensor supporting live view and a clip-on EVF, it might be possible to use software to analyse the image area represented by the focus patch and make some sort of focus determination to provide a focus confirmation in the viewfinder.

 

The problem is that contrast-based focussing systems rely on watching how the image changes as the lens is focussed and P&S cameras focus sweep to find best focus which wouldn't necessarily happen with a manual focus lens. I don't think therefore that we could ever have a simple focus confirmation - the camera doesn't know what it's looking at and therefore whether it is in focus.

 

Instead, I think the viewfinder focus conformation could take the form of a bar graph showing the level of detail resolved in the focus patch area and it would be up to the photographer to do the focus sweep to maximise the detail. It would be possible to take the image samples from the focus patch area and put them through a Fourier Transform which yields a frequency spectrum and the bar graph could represent the spectrum content. The more high frequency content, the further to the right the bar graph and the more in focus the image.

 

In use, you'd turn the focussing ring back and forth to maximise the bar graph setting. The absolute value of the bar graph would depend on the subject matter and working aperture but accurate focus would come with the bar graph peaked. It wouldn't work in all lights and for all subjects but then again, neither does the existing rangefinder if there's no detail in the focus patch area.

[lars_bergquist]

Pico, I have been over that terrain, in my mind. Here are a couple of technical problems:

 

The focal plane problem: Any sensor that measures contrast OR phase through the lens would have to be in the exact focal plane of the lens in order to be accurate. The swinging meter paddle of the M5 and the CL (NOT the Minolta CLE which measured light from the first shutter curtain) moved in front of not only the focal plane, but also in front of the shutter. The accuracy was sufficient for light metering, but it cannot suffice for focusing. The alternative would be a reading off the live view sensor. But you don't like that. So a mirror arrangement would be necessary, producing a focal plane in two different places. A SLR, in other words. Here, the image sensor and the AF sensor/s are in different places, but at an identical distance from the lens. But I gather that a SLR camera is not what you want.

 

The accuracy problem: So the focusing has to be done by a separate arrangement, like the M does today. Now when I look at the RF patch, I can se a lot of things, some of them not only centimeters but often meters distant from each other – and from the camera. But because I can see them all, I can select one detail – the right eye of a half-profile portrait instead of the left eye, for example – and precisely focus on that. No electronic sensor can attain to that selectivity because it has to have a minimum physical extension. What it measures is the average contrast or phase coincidence across a certain and not inconsiderable area of the subject.

 

The presentation problem: What is more, both a main sensor and a separate electronic rangefinder suffer from the same fundamental problem of presentation. I must have an exact indication in the finder of which point on the subject that the focus sensor mechainsm reads. And that indication has to be accurate on the order of plus/minus 1 second of arc, because that is close to the limit of the eye's resolution. Now remember that the optical finder and the electronic focus confirmation unit are two entirely separate things. So is the focus of the lens and the M rangefinder, but the problem is in one dimension only, that of depth. Here, it is in two dimensions, and aggravated by the necessary parallax. Again, this is no big deal in the M because you are actually, directly, visually seeing the separate image that is formed by the optical system of the rangefinder, basically a Kepler telescope. But we would have instead someting like when I am using the optical finder of my X-100. That indication in the finder just points to the general area where the lens will be focused, if all goes well. This is all right because that camera has a semi-wide lens of no great speed. Usually, the focusing is acceptable. At least if I stop the lens down a bit. But if you try to focus a 90mm f:2 lens that way, your failure rate will likely be unacceptable. – The indication of the AF spot on the screen of a SLR camera also has this problem, but it is alleviated by the fact that there is no parallax here; the image on the main sensor and on the AF cross sensor is the same optical image, generated by the same lens.

 

So the problem is maybe more complicated than many people think, and it cannot be made to go away by vague appeals to the Technology Fairy (common enough, though you are not guilty of that, Pico).

 

The dour old spoilsport from the Rangefinder Age

[lars_bergquist]

Advertisement (gone after registration)

By the way, I buy most of Mark's technical reasoning, but I am not convinced by his proposal for a focus confirmation display.

 

The dour old spoilsport from the Rangefinder Age

[giordano]

Briefly: It would employ a tiny lens layered over a small sensor that fits inside the body, behind the camera lens. It would view the subject area coming from the camera lens that matches the patch area + an area surrounding the focus-patch coverage. (If it cannot be put aside like the BTL sensors are, then perhaps a flip-in such as the CLE used.)

 

1. Remember that the size of the rangefinder patch relative to the sensor varies greatly with the lens you are using. With a 28mm lens it's fair to call it tiny; with a 135mm lens the height and width of the patch are approaching a third of those of the sensor. Your "tiny" lens and sensor would actually have to be quite big (or the rangefinder patch could be made much smaller).

 

2. What you're proposing is to mount a positive lens behind the actual M lens to form a real image on a secondary sensor in front of the shutter curtain. In other words it has to shorten the focal length. It's a bit like a teleconverter except with negative magnification. Can there possibly be room for those optics and a sensor in the 15mm free space between the rear element of the M lens and the shutter blades? (Less than 15mm if you want to be able to use old-fashioned 21mm and shorter designs.)

A couple methods might be possible to make the assist work. In one method, contrast of the focus patch area plus surroundings found by the BTL sensor would be employed and the RF patch would light up slightly when the sensor has a profound difference.

I agree with what Lars described as "the accuracy problem": maximising contrast across an area as big as or (as Pico suggests) larger than the rangefinder patch cannot possibly match the discrimination possible with the current rangefinder or even with current DSLRs. It would be like counting on Contax G autofocus to to beat the M rangefinder ... and we know how that ended.

 

Another approach would have an additional sensor in the rangefinder mechanism.

This is much harder than it sounds, because the only place in the current range/viewfinder where there's a real image whose focus could be sensed is on the user's retina.

 

I think I agree with just about everything both Mark and Lars have said. FWIW I quite like Mark's "bar graph".

[marknorton]

It would of course be possible to calibrate the camera so that the area used for focus analysis precisely corresponds to the focus patch in the viewfinder depending on the focal length but there is the more than slight inconvenience that this will also very with focussing distance. In the end, the best way to evaluate focus is through the live view image of the sensor itself.

[rosuna]

I had the chance of playing with the Olympus OM-D this week. I was shocked. The general response of the camera is incredible and the viewfinder is comfortable. The best EVF system I have seen so far. Fast, accurate...

 

The technology is mature. The reflex cameras are escaping from the flames. The Nikon D600 (a cheap full frame camera) will be presented soon. The APS-C territory is being cleared for the mirrorless systems. It is done.

 

Leica has to set the base for the long term development and survival of the M system.

[oronet commander]

Nice to hear that, rosuna. I'll hopefully try one tomorrow -together with an X-Pro 1- and would like to see real advancements from old m4/3.

[pico]

This is all wonderful feedback. Many thanks.

 

Mark's idea to provide focus feedback via a single line graph is interesting to me because it involves the photographer in the decision using feedback that is not perfectly binary. It's a very Leica-like principle.

 

Regarding the determination of focus using firmware, some promising algorithims are coming about through psychophysics.

 

abstract; Optimal defocus estimation in individual natural images

full text: http://www.pnas.org/content/108/40/16849.full

methodology: http://www.pnas.org/content/suppl/2011/09/19/1108491108.DCSupplemental/pnas.201108491SI.pdf

 

Regarding Lars' observations - I simplified (or ignored) the fact that the sensor behind the lens was not at the focal plane when I suggested a lens with sensor which would correct the focus plane for the sensor.

 

Reading the articles linked above (not from the Technology Fairy), if one considers focus detection as more than just contrast sensing as the article does, perhaps it is possible to provide that 1 second of arc.

 

I am compelled to agree that the device would be an assist and rely upon a photographer's decision to accept or not: the original rangefinder would still be there for those with more intention than the assist might offer.

 

My prejudice for a CCD is duly noted. I just cannot get into live-view. I have it on a Panasonic and it is just not for me. Using CMOS not for viewing but for reading the image to support, among other things, focus assist might be promising.

[thighslapper]

Well.....

 

This discussion has basically confirmed my conviction that the M10 will retain all the original virtues of the M9 and add an optional Visoflex EVF that will do all that the benefits of a CMOS sensor allow with a bonus of additional resolution, dynamic range and high ISO performance.

 

There is no other conceivable solution retains backward compatability and makes the current lenses usable.

 

Why else would the X2 viewfinder be called a Visoflex if a derivative is not ultimately intended for use in the M line of cameras ????

 

Leica sets great store by tradition and names are not given out randomly and without good reason.......

[AlanG]

The Nikon One, the new Canon 650 have phase detect AF right on the sensor. This technology will make for faster focusing of mirrorless cameras.

 

Even with the contrast detect method, an M with live view could have focus confirmation in the optical viewfinder that would be fast enough for its manual focus lenses. At that point one could eliminate the rangefinder and just put a "see through LCD" overlay in the viewfinder that would indicate the focus confirmation location. You could choose anywhere in the field of view for focus confirmation.

 

This could be slicker than aligning the M's rangefinder spot and a lot more accurate. It would eliminate all calibration and focus shift issues. Lots of third party lenses would work on it too. There are all kinds of ways they could implement it... a rectangle that turns from red to green when focused. They could even simulate the movement of the current mechanism if they wanted to.

[thighslapper]

The Nikon One, the new Canon 650 have phase detect AF right on the sensor. This technology will make for faster focusing of mirrorless cameras.

 

Even with the contrast detect method, an M with live view could have focus confirmation in the optical viewfinder that would be fast enough for its manual focus lenses. At that point one could eliminate the rangefinder and just put a "see through LCD" overlay in the viewfinder that would indicate the focus confirmation location. You could choose anywhere in the field of view for focus confirmation.

 

This could be slicker than aligning the M's rangefinder spot and a lot more accurate. It would eliminate all calibration and focus shift issues. Lots of third party lenses would work on it too. There are all kinds of ways they could implement it... a rectangle that turns from red to green when focused. They could even simulate the movement of the current mechanism if they wanted to.

 

Yes, but moving about the point of focus on an LCD screen as in current cameras is a real pain.

In a fixed central position it would work but still cannot match the pinpoint accuracy you can achieve manually.

The ONLY real advantage to non-manual focussing is speed..... for moving subjects and fast snapshots ..... and as the M series has no Autofocus lenses the whole point of the exercise is rendered redundant.....

Stick an old manual Nikon lens on a D700 and you will soon see how pointless most of this technology is.......

[stevelap]

The Nikon One, the new Canon 650 have phase detect AF right on the sensor.........

 

 

Yes, although as far as I am aware this solution hasn't been successfully implemented on a full-frame sensor yet. But if Leica are moving to CMOS anyway, and if they and their chip provider can pull it off without affecting underlying sensor performance, it might just be the answer. It doesn't have to be used for autofocus (and couldn't anyway with existing M lenses), but it could provide accurate focus confirmation when needed and perhaps facilitate some of the viewfinder developments discussed earlier in the thread.

[AlanG]

Yes, but moving about the point of focus on an LCD screen as in current cameras is a real pain.

In a fixed central position it would work but still cannot match the pinpoint accuracy you can achieve manually.

The ONLY real advantage to non-manual focussing is speed..... for moving subjects and fast snapshots ..... and as the M series has no Autofocus lenses the whole point of the exercise is rendered redundant.....

Stick an old manual Nikon lens on a D700 and you will soon see how pointless most of this technology is.......

 

I was not talking about AF. Just using the sensor contrast detection as a focus confirmation in place of a rangefinder. It is very easy to select the focus point via a touch sensitive LCD or using a joystick or other control. It isn't as if you will have to continually change this. Maybe put it in one position when shooting a vertical portrait and then move it to another spot if you turn the camera to horizontal.

 

I use focus confirmation all of the time on Canons with TSE and other manual focus lenses. No problem.

[thighslapper]

More than happy to concede the point Alan......

 

but on an M9 I can see no practical advantage over the existing optical system using a human as the processor and rather than a semiconductor .....

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[jaapv]

Yes, but moving about the point of focus on an LCD screen as in current cameras is a real pain.

In a fixed central position it would work but still cannot match the pinpoint accuracy you can achieve manually.

The ONLY real advantage to non-manual focussing is speed..... for moving subjects and fast snapshots ..... and as the M series has no Autofocus lenses the whole point of the exercise is rendered redundant.....

Stick an old manual Nikon lens on a D700 and you will soon see how pointless most of this technology is.......

The system of locking focus on a certain part of the image, so that the focal point follows it when moving the camera is quite practical though.

[AlanG]

More than happy to concede the point Alan......

 

but on an M9 I can see no practical advantage over the existing optical system using a human as the processor and rather than a semiconductor .....

 

You wouldn't have the issue of inaccuracy from focusing in the center and re-framing or the time delay from doing that. Also it would compensate for the focus shifting when stopping down. The "rangefinder" mechanism and lenses would never need focus calibration.

 

The human would still do the "processing" as you'd have to monitor the focus confirmation indicator or beep. That is no different than trying to decide when two images are superimposed accurately.

 

It would have a hybrid optical/EVF that could even allow for an electronic superimposed image over just the focus confirmation patch and still have the optical view on the rest. This could then be switched to full frame EVF when that is needed. I really don't see any choice but Leica evolving something like this someday if they decide to go with AF lenses and EVF and still keep their old MF lenses useful while also retaining an optical viewfinder. A single hybrid viewfinder is much smaller and more elegant than an optical viewfinder/rangefinder with a clip on EVF.

 

I see this technology in a new body as a possible bridge between the current M lenses and future AF lenses. And they could take their time coming out with AF lenses for it. I see this as pretty radical and doubt if Leica has the resources to do this right now and I'm not sure if all of the required technology is available yet. But maybe they just want to keep the rangefinder/viewfinder the way it is and have no plans for AF either.