Matching Primes for 3D Work

[Phil Bates]

I've been testing Nikon primes to see if I can get any two to match for a side-by-side rig. It's a challenge because if the focus is a tiny bit different, there is a zoom effect. I am concerned that in the field I won't be able to see the difference on small monitors, and the slight change will spoil the 3D effect.

It makes it impossible to compare distortion between lenses to try and find the best match (I am purchasing 6 lenses at a time to give me a pool to pick from)

Is there anyone out there whose figured out how to successfully use DSLR primes? Is the only other option Zeiss, Red Pro primes or other Cinema glass?

Thanks,
Phil

[Mark L. Pederson]

Is there anyone out there whose figured out how to successfully use DSLR primes?
Thanks,
Phil

Very, very tough Phil. Reality is - you get as close as you can and fix as much in post as possible. We have shot some time lapse 3D stuff using DSLRs + Canon lenses - obviously at a fixed focal length - and it came out just fine.

[KETCH ROSSi]

In my experience and all the testing done so far, which we tried the Canon's and Nikons, for size reason, it is not easy at all to use Still lenses, so we give up!

We had great luck and results with the RPP's seemed to be very good in matching with other sets, in some cases we tried and tested four set of lenses coming form other manufacturers, but I just simply can't wait to get our Cine Leicas, and also with a rumored Mini Set of Primes PAIRS for S3D specifically.

I just love working with Primes, even so more setting up is necessary as you need changing lenses more often and recalibrate the rig, but we will test the RPZ 17-50's, even so we know that Angenioux are the best.

Truly look forward to our ET Pulsar rig with Epic and Leicas, hehe.

Also Phil, if you do mostly Parallel, shooting its less of an issue.

[Jim Geduldick]

Phil

There are tools for fixing mono ocular disparities between lenses if you are using non Cine lenses its hard to match those being cranked off an assembly line.
And what monitors are you using that you say small?
I have a few methods for shooting distortion grids pre stereo shoots to match up for finishing.

[Deanan]

Phil

There are tools for fixing mono ocular disparities between lenses if you are using non Cine lenses its hard to match those being cranked off an assembly line.
And what monitors are you using that you say small?
I have a few methods for shooting distortion grids pre stereo shoots to match up for finishing.

That's true for cine primes also. The focal lengths and image size do vary from lens to lens. Zooms are much more flexible if you can compensate/correlate two together.

[jimhare]

We recently completed a 3D long form project. Decided to use dual Master Primes in the end and they were great!

[Phil Bates]

Phil

And what monitors are you using that you say small?


A monitor you would use in the field for wildlife and nature shooting, 12 inch diagonal or less.



Phil

I have a few methods for shooting distortion grids pre stereo shoots to match up for finishing.

I think that's a great idea and I came to a similar conclusion. The problem comes in when you want to focus on something distant like clouds. If the focus itself is a bump different between lenses there is a slight image size difference. I guess you point the camera first at the distant mountain (or something with clearly defined edges) and shoot that as reference before pointing at the sky, then as you suggest fix that in post.

If setting the lens on the hard stop of Infinity was an option, it would make things easier, but many lenses focus on infinity BEFORE you hit the stop, so again it makes it hard to match the two with a field monitor.

Thanks Jim,
Phil

[Jeff Coatney]

This is where huge resolution really pays off (can't wait for 5K). Cinema glass is best due to stringent QC-- dump the Nikons and Canons altogether because lens to lens QC is practically non-existant.

[Philip Powell]

Zooms are much more flexible if you can compensate/correlate two together.

That's something I've been curious about. I know that many use zooms for the reasons mentioned. But don't zooms have the potential to "flatten" the 3D effect? Are most shooters working at the wide end of the lens then to avoid that?

[Riku Naskali]

I can only say that I prefer to work at the wide end as much as possible when shooting stereo...

Using long lenses is hard because they flatten the image by itself, and also you'll usually need to use smaller IAD's to keep the positive parallax in check... So one usually ends up with double flattening effect.

On the other hand, for some shots the cardboard-look really fits :-)

[Jacek Zakowicz]

I've been testing Nikon primes to see if I can get any two to match for a side-by-side rig. It's a challenge because if the focus is a tiny bit different, there is a zoom effect. I am concerned that in the field I won't be able to see the difference on small monitors, and the slight change will spoil the 3D effect.

It makes it impossible to compare distortion between lenses to try and find the best match (I am purchasing 6 lenses at a time to give me a pool to pick from)

Is there anyone out there whose figured out how to successfully use DSLR primes? Is the only other option Zeiss, Red Pro primes or other Cinema glass?

Thanks,
Phil
Honestly I don't see it as a big problem:
Wide lenses, deep stop for 3D and far distance means that DOF should take care of the focus for you as long as you are within hyperfocal distance so what your primary concern should be is magnification. As long as it's matched while you pay attention to hyperfocal setting your distortion should be pretty close as well. I think it's just a matter of switching priorities from focus to magnification while you work. This is my take, try testing and let us know...:smiley:

[Eric Lange]

I've been testing Nikon primes to see if I can get any two to match for a side-by-side rig. It's a challenge because if the focus is a tiny bit different, there is a zoom effect. I am concerned that in the field I won't be able to see the difference on small monitors, and the slight change will spoil the 3D effect.

It makes it impossible to compare distortion between lenses to try and find the best match (I am purchasing 6 lenses at a time to give me a pool to pick from)

Is there anyone out there whose figured out how to successfully use DSLR primes? Is the only other option Zeiss, Red Pro primes or other Cinema glass?

Thanks,
Phil

At the moment we are fixing to build systems around various “primes”.
The problem with off the shelf lenses is the helicoid(s) do not affect a truly linear translation. If you imagine the axis of the lens as being Z then as you change focus the principal point of the lens drifts in x and y and also pitches and yaws slightly; and of course matching the correct focusing distance between left and right lenses can be tricky.

Without giving too much away here’s what we have come up with, bearing in mind that this is for high precision stereo metrology.

1. Rather than rely on the lens helicoids, we lock the lens off and instead build a translation mechanism for the lens [precision ball screws and crossed roller bearings]. You need to have enough flange clearance so that a small bellows mechanism can be introduced between the lens and sensor. The bellows can be recessed to come forward of the camera body and still achieve infinity focus. This basically eliminates the X Y drift.


2. Use a linear encoder or micrometer +/- 5 micron to measure the degree of translation of the lens.


3. Use an autocollimator to “fix” back focal distance to a known measurable location. Also use the same set up to determine the principal point of the lens as focused onto a particular X,Y location on the sensor. Clever design of the various assemblies mitigates issues of thermal expansion but thermal calibration may be necessary.


4. Need extremely rigid and stable mounting system between camera body and lenses.


5. Focusing needs to be calibrated using encoder or micrometer drive for left and right lenses.


6. Extra setting jigs can be used to make sure the lense is absolutely perpendicular to the sensor, (autocollimators, and rotating jigs, to correct wobble).


7. Radial distortion and magnification are inextricably linked, so to get any meaningful matched magnification you have to correct and calibrate for distortion.

For our work we like where possible to use symmetric Biogon type lenses as the radial distortion can be very much lower and also yield much more accurate calibration. Retro-focal lenses tend to have a distortion between 1 and 2 1/2%. This radial distortion also creates a very characteristic Y parallax pattern across the field of view If any kind of horizontal image translation is applied for correct viewing of the stereo imagery, (so it’s always good to correct for distortion).

So correcting normal photographic primes for coherent and matched focusing for high quality stereo is a complete hassle unless you have a VERY good reason to do so. We haven’t calibrated enough systems to say definitively what the variations of raw magnification vs focal distance are for a given chunk of glass as compared to a random or consecutive sample of lenses from a production line.

Zeiss claim very high precisions and repeatability for their master primes.


Just like what OptiTek said, calibrate for magnification rather than focal distance setting. Here it might be possible that you could rig up a small vernier scale between the focusing ring and the lens body. This should give you much more precise focusing. Also you could use a sprung mechanism to apply pressure to remove back lash from the helicoids, i.e the translation slop between rotating the focusing ring clock wise and then anti clock wise (you might grind your helicoids a bit). By shooting targets and grids in a controlled set up you then generate your own scale markings on the focusing ring for a given magnification, these will be different to what is marked on the lens and be different from left and right lenses. If you are prepared to go through many iterations then it might be possible that you converge on a useful calibration and set of tables for hyper focal distance. If you take great care with this then you should be able to generate results that are better than at least half of what you see in the cinema.


What OptiTek said… hyperfocal distance and tweak in the field once its right, don't even breathe on the equipment after that! If you are a "gifted" tweaker you should get acceptible results, a lot of trial and error and sticky tape.

[Phil Bates]

Wow, Eric, thanks for the detailed post!

Also, thanks to everyone else for contributing. :)

P