New ways to autofocus manual lenses

Should work with any camera, even old film cameras:
Not quite sure how one calibrates the follow-focus movements to the separate distances, but sure, that can sort-of work. However, something like using a TechArt Pro LM-EA7 to drive a follow-focus motor (as opposed to unit-focusing the lens) would work a lot better than using a separate focus sensor....
 
Should work with any camera, even old film cameras:
Not quite sure how one calibrates the follow-focus movements to the separate distances, but sure, that can sort-of work.
The process is shown in the video. Or do I just not understand how you mean?
However, something like using a TechArt Pro LM-EA7 to drive a follow-focus motor (as opposed to unit-focusing the lens) would work a lot better than using a separate focus sensor....
I can see as you point out advantages with the Techart way to do it.

The focusing directly on the lens also works on lenses with floating elements but it is not why they made this rig I guess.

This rig that DJI has developed is very similar to what is common in the film/video world where focusing motors turning directly on the lenses is how they do it, steered from different add on controllers, but so far only via manual control or preset focus points with ramping functions for the focus movement speed sync to where the actors have their stopping/standing points.

On film productions they have a separate guy with the title "focus puller" whose job is to work the focus on the lens.

The cog band that can be seen in the video sitting around the lens is a strap that can be used on any MF lens. Though it is not the preferred method, they prefer lenses that already has those cogs, so called Cine lenses.

In the video world they also seem to like vintage MF lenses, which might be one reason why the prices goes up. They like old vintage lenses so much that there is a small business with companies that do rehousing of lenses to work better for these cog wheel motors. Either by modifying the existing lens bodys to get permanent cogs on the aperture and focus rings (like any cine lens has) or by making totally new lens bodys with new helicoids and apertures and lifting over the glass only. You can get for instance almost all Russian lenses in different cine conversions and not that expensive either for the amount of work that has been done on them.

So for the video/film world to get reliable autofocus this way is quite a big invention.

For a stills camera only, one can remove a lot of the items that the rig in the YT video shows, like the gimbal and the whole tracking head for automatic subject following (which in itself is extremely fascination how well it works). And except giving autofocus it is there this invention really flies since one now can film oneself automatically with a MF lens with both camera and focus following. Seems to work fairly well in the video I think.

Next step should be a motor also for turning the zoom and then it is fully automated even in depth following on existing MF lenses.

Lidar can focus in pitch dark conditions (self driving cars use Lidar). Many mobiles already have Lidar sensors, like the later Apple ones with image sensor and Lidar from Sony and I would guess it soon will come built in also on free standing cameras. For better autofocus and also a precision depth map. Lightroom can already now use depth maps for masks in the edits but normal cameras don't provide them (Apple mobiles do).

DJI who makes this rig is the worlds largest drone manufacturer, they also make different gimbals for film/video and they own Hasselblad since a few years back, so it will be interesting to see if their new technologies also comes into that product line. Hasselblad seems to try to integrate their past with the future, the new digital databack fits the old 500 series of cameras for instance.

--
Best regards
/Anders
----------------------------------------------------
Mirrorless, mirrorless on the wall, say which is the best camera of them all?
Some images:
https://www.dpreview.com/forums/post/64169208
https://www.dpreview.com/forums/post/64221482
 
Last edited:
Should work with any camera, even old film cameras:
Not quite sure how one calibrates the follow-focus movements to the separate distances, but sure, that can sort-of work.
The process is shown in the video. Or do I just not understand how you mean?
That suggests it simply marks the infinity and closest focus stops. I would not expect the log interpolation from that to be accurate enough; maybe it is over the limited range time-of-flight sensors work in? I also didn't see a very precise way to tell the unit what distance the closest focus was at.
However, something like using a TechArt Pro LM-EA7 to drive a follow-focus motor (as opposed to unit-focusing the lens) would work a lot better than using a separate focus sensor....
I can see as you point out advantages with the Techart way to do it.

The focusing directly on the lens also works on lenses with floating elements but it is not why they made this rig I guess.
DJI did something easy for them that is entirely camera make/model independent -- follow focus hardware is pretty well standardized (if rather bulky).

This is actually much easier than the interfacing and drive problems the LM-EA7 has to deal with. In fact, I could probably hack-together a unit like this in a week or two using 3D-printed parts with an Arduino Pro Micro, 28BYJ-48 stepper, and an ultrasonic distance sensor. (No, I'm not kidding. In fact, once I complete my current projects, especially my Lafodis 4x5 scanning camera, I very well might do this.)
 
Should work with any camera, even old film cameras:
Not quite sure how one calibrates the follow-focus movements to the separate distances, but sure, that can sort-of work.
The process is shown in the video. Or do I just not understand how you mean?
That suggests it simply marks the infinity and closest focus stops. I would not expect the log interpolation from that to be accurate enough; maybe it is over the limited range time-of-flight sensors work in? I also didn't see a very precise way to tell the unit what distance the closest focus was at.
The Lidar they use does not work at longer distances (due to no ray bouncing, suggesting it is rather weak) so they calibrate at 1 and 4 meters. Seems to work well, he uses a f0.95 lens wide open in the video I linked to and it follows him automatically.
However, something like using a TechArt Pro LM-EA7 to drive a follow-focus motor (as opposed to unit-focusing the lens) would work a lot better than using a separate focus sensor....
I can see as you point out advantages with the Techart way to do it.

The focusing directly on the lens also works on lenses with floating elements but it is not why they made this rig I guess.
DJI did something easy for them that is entirely camera make/model independent -- follow focus hardware is pretty well standardized (if rather bulky).
Yes and by doing a separate stand alone system it breaths life in older cameras that has bad autofocus which many older ones have especially in video. Or those which have no autofocus at all.
This is actually much easier than the interfacing and drive problems the LM-EA7 has to deal with. In fact, I could probably hack-together a unit like this in a week or two using 3D-printed parts with an Arduino Pro Micro, 28BYJ-48 stepper, and an ultrasonic distance sensor. (No, I'm not kidding. In fact, once I complete my current projects, especially my Lafodis 4x5 scanning camera, I very well might do this.)
I agree that the tech isn't that impressive in itself on a per piece level and that much of it has been around for quite some time, but in combination with the tracking that controls the gimbal mount to follow the set target automatically and for the price and the nice overall system thinking I think they put together quite a nice package.

Personally I'd love something that could autofocus my manual vintage lenses correctly with Eye-AF at f1.2 over the whole image surface but the LM-EA7 doesn't seem to be up to that and sofar we haven't had any alternatives so all development in this direction makes me happy.

--
Best regards
/Anders
----------------------------------------------------
Mirrorless, mirrorless on the wall, say which is the best camera of them all?
Some images:
https://www.dpreview.com/forums/post/64169208
https://www.dpreview.com/forums/post/64221482
 
Last edited:
Should work with any camera, even old film cameras:
Not quite sure how one calibrates the follow-focus movements to the separate distances, but sure, that can sort-of work.
The process is shown in the video. Or do I just not understand how you mean?
That suggests it simply marks the infinity and closest focus stops. I would not expect the log interpolation from that to be accurate enough; maybe it is over the limited range time-of-flight sensors work in? I also didn't see a very precise way to tell the unit what distance the closest focus was at.
The Lidar they use does not work at longer distances (due to no ray bouncing, suggesting it is rather weak) so they calibrate at 1 and 4 meters. Seems to work well, he uses a f0.95 lens wide open in the video I linked to and it follows him automatically.
Hm. I see they are saying LiDAR rather than time-of-flight imaging... somehow I hadn't noticed that. LiDAR typically requires physical scanning of the scene, whereas time-of-flight is usually implemented by timing relative numbers of photon arriving from a precisely timed interval from a precisely timed pulse. Both (and also structured light) are usually done using NIR because the FCC doesn't regulate that, and there's getting to be a huge amount of NIR light pollution these days because of that.

Of the active focus technologies, I actually favor either RF or "sonar" (really ultrasonic) because they don't spew NIR all over the place.

Of course, if they could simply decode the lens control protocol, like the LM-EA7 does, they could be completely passive....
 

Keyboard shortcuts

Back
Top