Fast AF Rangefinder, possible?

[BobYIL]

There is an ongoing discussion about autofocus almost in all forums related to Leica or rangefinders. Autofocus, undoubtedly is a great plus; to focus a 50mm high speed lens on a moving subject with the speed and accuracy of a professional DSLR.. to keep the subject in focus while moving and being sure of a tack-sharp picture each time you pressed the shutter button.. What a great street camera would it be! Who did not dream of such a possibility?

But could there be a Leica or any rangefinder with interchangeable lens mount to have a fast autofocus? In a near future, sure, there must be, for it could be the next step of development for rangefinder.. But how? What are the possibilities and what could be the limitations? Why it could not be done until today? Here are my thoughts:


In-body motors:
Linear extension micromotor mounted in camera body to drive lens, like in some conventional, relatively older design AF lenses or the Contax G: Low AF speed; this design is to phase out, not used anymore in the recently issued designs. Although this design can be squeezed in relatively smaller lens diameter (only a worm-gear drive), I do not believe that any new RF design based on a low speed AF activated by a step-motor installed in the body could be attractive for the majority of users in our day.
(cont'd...)

 

[BobYIL]

In-lens motors:

Two alternatives:
#1. DC operated micro motors as in the M43 systems or newer type AF lenses: Fast AF.

#2. Ring type motors (around the periphery of moving lens group) like in some Canon-USM or Nikon SWM lenses: The only possible solution when the moving glass group is relatively heavy. Finds use generally in high speed and tele lenses.. Fast AF.

Although the in-lens-motor alternative provides lightning fast AF in most cases, the application may not satisfy the general rangefinder requirements; for almost all lenses with this design will protrude into the finder window considerably. As examples: The smallest AF Nikon 50/1.8 AF-D has a diameter of 63.5mm or the Canon 50/1.8 EF is 68mm whereas the same for the 50/2 Summicron is 53mm or for the 35/1.4 Summilux Asp. is 56mm -which already protrudes-. Shortly: The protrusion into the finder window may reach to objectionable levels, with wide angle lenses even more. (Even the tiniest 20/1.7 Panasonic has an outside diameter of 63mm inspite of its illumination circle of only 22mm.)



(cont'd...)

 

[BobYIL]

IMO, if there would be any full-frame RF camera to come out with AF feature in future, then the distance between the lens axis and the viewfinder axis would be at least 6-7mm larger than the one on the existing M-Leicas, for some wide angle lenses this could be more. (This would offset some of the size advantage of the RF design though..)

As for the possibility of implementing AF by moving sensor plane (like in the old Contax AX): Assuming a 90mm lens would need at least 10mm movement to go from infinity to 1 meter, then adding another 10mm depth to a Leica M-body would wipe off all its size advantages.

As for the internal focusing (IF) option for lenses: How could any manufacturer revise all their existing lenses to operate via internal focusing? Even optics do not allow this..

With the above constraints I am inclined to believe that if a rangefinder with pure optical or hybrid-type viewfinder (like the one on the X100) together with the features of AF and lens interchangeability would be introduced, then it will rather be based on the APS-C size sensor due to the above considerations. Also the requirement of having to “move” large glass groups for AF as in the case of lenses like the Noctilux or the 75 and 90mm Summicrons by the limited capacity of smaller batteries on digital rangefinders may necessitate the shift toward smaller and lighter designs. FF may not fulfill these conditions.

Insightful comments would be appreciated...

Regards,

Bob

 

[usayit]

IMO, the complexity isnt in the speed or type of motor in the lens but in the mechanism to determine focus.

SLRs with mirror boxes use phase
Mirrorless like micro 43 use contrast.

Those are the two main ones in market. At this time phase based AF is still the better of the two. Contrast focusing has a long way to go but is making some progress... driven by Olympus and Panasonic. A rangefinder With AF either by on sensor or in lens will need to solve this issue first before there is any hope of generating any serious interest from the target market.

Micro 43 will also need a better solution if it will produce a pro level camera that can go head to head with DSLRs.

 

[bensyverson]

The Contax G series seems to do pretty well with a combination of passive (TTL) and active (infrared) focusing... It's hard to imagine how you would improve on that.

 

[Pablito]

If it is has AF it is not an RF camera: the AF mechanism replaces the RF mechanism

Unless you can disable the AF and focus manually using the double-image method. I can imagine some sort of high tech camera that included BOTH an optical RF and an independent AF system with motors, etc.

A rangefinder is a DEVICE. It is PART of a rangefinder camera. Or a gun, for that matter.

http://en.wikipedia.org/wiki/Rangefinder

 

[BobYIL]

If it is has AF it is not an RF camera: the AF mechanism replaces the RF mechanism

Unless you can disable the AF and focus manually using the double-image method. I can imagine some sort of high tech camera that included BOTH an optical RF and an independent AF system with motors, etc.

A rangefinder is a DEVICE. It is PART of a rangefinder camera. Or a gun, for that matter.

http://en.wikipedia.org/wiki/Rangefinder

Hmmmm... thank you.. I was thinking Contax G1 and G2 counted as rangefinders...

 

[BobYIL]

IMO, the complexity isnt in the speed or type of motor in the lens but in the mechanism to determine focus.

SLRs with mirror boxes use phase
Mirrorless like micro 43 use contrast.

Those are the two main ones in market. At this time phase based AF is still the better of the two. Contrast focusing has a long way to go but is making some progress... driven by Olympus and Panasonic. A rangefinder With AF either by on sensor or in lens will need to solve this issue first before there is any hope of generating any serious interest from the target market.

Micro 43 will also need a better solution if it will produce a pro level camera that can go head to head with DSLRs.

No matter what process is chosen, be phase or contrast detect, at the end the physical motion mechanism for the lens group to implement the focusing action is being the key factor and there lies all the constraints and issues..

 

[BobYIL]

The Contax G series seems to do pretty well with a combination of active (TTL) and passive (infrared) focusing... It's hard to imagine how you would improve on that.

As I recall, the complaints with the AF mechanism of Contax G series were mostly on focusing error, sometimes one-third of the shots.. Also from the point of view of AF-speed it was no comparison against the ones on say the D3 or 1D3; all screw-driven AF mechanisms were relatively slow.

 

[Phil_F_NM]

The Konica Hexar is as close to perfect AF "rangefinder" you're ever going to get. It's technically an IR rangefinder, not a mechanical/optical one but that's splitting hairs. Yeah, it's a 35mm lens but it has some incredible resolving power as well.
There are two ways to actually make the lenses focus. Either use the existing M mount and move the sensor plane or develop new lenses.
Regardless, the public will almost certainly never see such an animal. A digital AF rangefinder with interchangeable lenses. There's such a small market that putting up the R&D makes recouping initial investment iffy at best. No one would want to take the risk.
So this is merely academic.

BUT, money can buy you everything so if you have enough, you can definitely get it done for a personal one-off interchangeable lens RF AF camera.

Phil Forrest

 

[bensyverson]

Theoretically it would be possible to create a camera with a co-incident rangefinder patch in the viewfinder, fast autofocus, and a manual focus ring. It would require a new system with new lenses.

Instead of being mechanically coupled, all three would have to be electronically linked. The prism or mirror in the RF would be controlled by a motor. The lens would be an internal focus type, driven by a fast motor. The focus ring would by fly-by-wire. Because there's already a RF window, the AF could utilize the rangefinder, much like the G series. I've never heard anyone complain about the G2's autofocus accuracy.

Of course, this is all highly academic, as no one would sink that kind of money into the development such an expensive micro-niche product. If this camera takes film, it has no market. If it's digital, why not just make the VF electronic like the X100, and show a magnified view when focusing? There's no obvious advantage to making it a rangefinder.

 

[usayit]

No matter what process is chosen, be phase or contrast detect, at the end the physical motion mechanism for the lens group to implement the focusing action is being the key factor and there lies all the constraints and issues..

The true bottleneck is the mechanism to determine focus. You can put the fastest focusing motor in a lens but it isnt going to be any faster nor accurate without solving this problem first.

Its been discussed to much detail in the mirrorless forums and determined to be the core issue or limit of the system. Its the reason why 4/3 AF lenses adapted to micro 4/3 with AF functioning is significantly slower (almost unusable) than the same exact lens on a mirrored 4/3 body. The difference is contrast versus phase. Google it for more info.... in summary, contrast has no idea if the focus plane is in front or behind the subject and phase does.

 

[bensyverson]

Contrast detection is basically a stupid way to focus. Any decent active AF system will beat it in terms of speed, if not accuracy.

Get rid of the TTL requirement, and you're on your way to AF nirvana.

 

[Livesteamer]

I use an M6 .85 and a 50mm Summilux to do a lot of little kid pictures under available light. My depth of field is normally very, very thin.

Until AF knows where I want to focus the mechanism is pointless. Just my opinion. Joe

 

[krötenblender]

The true bottleneck is the mechanism to determine focus. You can put the fastest focusing motor in a lens but it isnt going to be any faster nor accurate without solving this problem first.

Panasonic managed to get it's contrast af as fast as phase af in one of it's µ4/3 cameras, so there is the possibility to get this problem solved, even if the fastest phase af may be currently faster that the fastes contrast af.

Its been discussed to much detail in the mirrorless forums and determined to be the core issue or limit of the system. Its the reason why 4/3 AF lenses adapted to micro 4/3 with AF functioning is significantly slower (almost unusable) than the same exact lens on a mirrored 4/3 body. The difference is contrast versus phase.

This is not true. The reason for most 4/3s lenses to be very slow with contrast af is, that their moving parts and motor-control are not made for the different characteristics of movement:

- Phase af gives the distance (or more correctly the difference of distance) where to move the focus plane of the lens, so the lens does one continuous motion until this position is reached.
- Contrast af OTOH moves the focus plane by one tiny step and measures the max contrast, then moves again one tiny step, measures again... and so on, until max contrast is reached, what it can only know by moving a little bit behind the ideal point (where the max contrast function is behind it's maximum), so that the lens has to go back to that maximum point. This incorporates many very little and exact steps of movement, for which most phase-af-lenses have not been built.

That's why they are slow. But building lenses for this kind of movement is possible.

Google it for more info.... in summary, contrast has no idea if the focus plane is in front or behind the subject and phase does.

That is one advantage of phase af, but contrast af has the advantage of more exact focusing. Front- of back-focus does not happen with contrast af, which is especially important with wide open lenses with very thin focus planes.

 

[krötenblender]

I would say, that fast af for rf is technically possible, but the intended user group for rf-cameras won't accept it for several reasons.

And also each of them has drawbacks, that would neutralize some of the advantages of a rf. Phase-AF mostly comes with bigger bodies, contrast af needs the sensor to be active the whole time (meaning to heat it, which increases noise and battery consumption - imagine the M9-sensor to be active the whole time...😱), all active focusing methods (laser or sonic echo) are either very limited in terms of distance or speed (try to focus with echo a moving car 100m away...) or environment (try laser or echo in a smoky bar, in rain or fog).

Another point already mentioned is that most af mechanisms do technically perfect focus, but are often completely wrong when trying to guess, what you want to focus on.

Nevertheless, I think most of these problems can be solved technically. But they can not be solved economically, because nobody would by such a camera and it would be utterly panned buy the critics for not being a real rf, not being a real anything else and so on.

So, yes, I think, fast af for rf is technically possible, but it won't happen, because it is impossible for other reasons.

 

[johannielscom]

I would say we should have a yet-to-come forum for cameras that aren't here yet (and likely never will be).

The amount of threads on improving (non-existing) cameras borders on the insane.

What's the use, all this theoretical mumbo jumbo?

 

[BobYIL]

I would say we should have a yet-to-come forum for cameras that aren't here yet (and likely never will be).

The amount of threads on improving (non-existing) cameras borders on the insane.

What's the use, all this theoretical mumbo jumbo?

Not everyone agree with that... As an example the X100 has already caused quite a stir and we are discussing just one step ahead of it.. And who knows, perhaps we would be hearing about such one if the last disaster in Japan did not occur.

For a simple GF1 it takes fraction of a second to focus on a moving child whereas with the M3 it is rather luck if I could bring two images together before depressing the shutter.

 

[MartinP]

It might be more realistic to expect a focus-confirmation using the existing manual focus lenses. If that could be made to work then the mechanical rangefinder could be replaced, and we would still have the same lenses to use. What would the method be? Pinpoint active ir focussing?

 

[johannielscom]

Not everyone agree with that... As an example the X100 has already caused quite a stir and we are discussing just one step ahead of it.. And who knows, perhaps we would be hearing about such one if the last disaster in Japan did not occur.

For a simple GF1 it takes fraction of a second to focus on a moving child whereas with the M3 it is rather luck if I could bring two images together before depressing the shutter.

The solution for that was invented long before the M3 and it is called zone-focusing. That's why all the RF lenses have aperture scales on them.

Use a 35mm lens, set the aperture to f8.0 and the lens distance to 3 mtrs and you will have everything covered from approx 1.5 mtrs to 20 mtrs. Loading the light-correct film and setting the correct shutter speed does the rest.

Simple as pie.