Best way to match lenses for mirror rigs

[Callum Just]

Hi all,

Simple question really, what is the best practice when trying to match sets of lenses? I’ve worked on a few 3D films and have helped shoot charts etc but it never seems to be an exact science. I’ve spent days looking through UP’s and Cookes looking for matches.

What are people’s experiences? I know Mr Collister has a lot of UltraPrimes on Spiderman at the moment; it would be great to know how you did it?

 

[Phil Bates]

Hi Callum,

My answer comes from the "tight budget" viewpoint, my experience may differ from others. Also, I was purchasing for a parallel rig, not beamsplitter.

We purchased six sets of Nikon DSLR primes. We matched the best two of each and sent the rest back. Our testing was done not shooting charts, but finding an outdoor scene with contrast and detail throughout the frame. The comparison was done in After Effects, using the Difference transfer mode.

However, no two lenses match exactly, so there is always post work. In fact, to be honest, since I have to do post work anyway, sometimes I wonder why it was so important to get them so close.

Regardless, I have had to do less post work with the primes than what was required on a professional aerial shoot with a sterographer carefully matching and collimating Optimo zooms.

The post work involves adjusting position, rotation, scale and even corner pin distortion at times. Some of that is the rig's fault, not the lenses.

I have limited experience, but am getting great results with the primes.

Hope that helps,
Phil

 

[Ketch Rossi]

Phil's approach is a nice one, however Still Glass has a lot more variation and disparity variation form glass to Glass then Cine Glass do,
and Cine Glass varies from Manufacturer to Manufacturer.

When matching Cine Glass for 3D work, it is necessary to be in a Controlled environment, a Lab, and match that glass via technical methods
including Projection, Master Primes with less then 1% variation are amongst those that almost certainly you could pick an lens from any set
and get a perfect Match all the time, need only to Calibrate them on a projector an you are good to go

Cookes and Ultra Primes should be close to the MP's, I heard of only needing 3 Sets of Cookes to find a god Matching Pair.

 

[Phil Bates]

Phil's approach is a nice one, however Still Glass has a lot more variation and disparity variation form glass to Glass then Cine Glass do,
and Cine Glass varies from Manufacturer to Manufacturer.

When matching Cine Glass for 3D work, it is necessary to be in a Controlled environment, a Lab, and match that glass via technical methods
including Projection, Master Primes with less then 1% variation are amongst those that almost certainly you could pick an lens from any set
and get a perfect Match all the time, need only to Calibrate them on a projector an you are good to go

Cookes and Ultra Primes should be close to the MP's, I heard of only needing 3 Sets of Cookes to find a god Matching Pair.

Ketch, I suspected that Cine Glass would be more consistent. Still, any disparity I come across is within that 1%. So, Ketch, do you do post work to fix that last 1%? Or is it so minimal, you let it go? I fix anything I can see to within 1 pixel. Really curious about this since I am living in my own 3D rabbit hole and have't been up for air in a LONG time.

Phil

 

[Doug Horton]

There's always discrepancies between lenses it seems, no matter how minimal. Correction towards these tiny disparities is very subjective as well. What one deems as good enough, another might see as unacceptable.

I prefer primes over zooms because when aligning, it at least starts me with less variables when trying to match lenses. However, we've used our zooms SOO much that I sort of know at this point how close I can expect them to be, on a variety of R1 bodies.

Anything within 1 pixel can almost be considered "un-detectable" as far as your brain's concerned, but as a Rig Tech, obviously the better you can do your job, the happier people are.

 

[JohnF]

When matching Cine Glass for 3D work, it is necessary to be in a Controlled environment, a Lab, and match that glass via technical methods
including Projection, Master Primes with less then 1% variation are amongst those that almost certainly you could pick an lens from any set
and get a perfect Match all the time, need only to Calibrate them on a projector an you are good to go

Hi Ketch,

The lab test I understand in that using a controlled, lab based test to find the closest optically matching lens pairs.

But...

I'm trying to figure out what is meant technically by "calibrate them on a projector". What is involved in a projection test and what does one calibrate?

JohnF

 

[Peter Lyons Collister ASC]

On Spiderman we were/are lucky enough to have Dan Sasaki at Panavision match all of our Ultra Primes. He did have to disassemble some of the lenses and re build them to match.

You will find that the wider focal lengths are the hardest. Of course we had exacting specs and there is some fudge room.

Peter

 

[Stephen Strangways]

Oftentimes for matching lenses for 3D, you're not looking to match the look of the lenses - the mirror throws most of that off anyway - but you want the focusing to match mechanically between the two lenses. Some cinema lens manufacturers will have a set of slightly differently marked focusing rings, and test each lens once assembled to find the best match. Some lens may get the "A" focusing ring, some the "C" and so on. For 3D work it is beneficial to get two lenses from the same focusing ring set.

 

[Pedro Guimaraes]

Oftentimes for matching lenses for 3D, you're not looking to match the look of the lenses - the mirror throws most of that off anyway - but you want the focusing to match mechanically between the two lenses. Some cinema lens manufacturers will have a set of slightly differently marked focusing rings, and test each lens once assembled to find the best match. Some lens may get the "A" focusing ring, some the "C" and so on. For 3D work it is beneficial to get two lenses from the same focusing ring set.

this is a non-issue now with c-motion lens mapping feature.

 

[chuck colburn]

Angenieux is offering matched zoom sets that cover S35

http://www.bhphotovideo.com/c/product/753382-REG/Angenieux_309330_Optimo_3D_Ready_Package.html

But matching lenses is just the start in 3D work. The whole 3D rig itself must be set up to assure that the two cameras are aligned to each other and are seeing the same framing. This is where your optical mechanical tech comes into play. The rig alignment is more important than the focale length varience in the lenses themselves. We shot much 65mm film in 3D without being too concerned about lens matching. Of course it's much easier when shooting video as one can resize the frame easier in post then you can with film.

 

[Ketch Rossi]

Hi Ketch,

The lab test I understand in that using a controlled, lab based test to find the closest optically matching lens pairs.

But...

I'm trying to figure out what is meant technically by "calibrate them on a projector". What is involved in a projection test and what does one calibrate?

JohnF

There are several methods of Lens calibration after the Mechanical part is been taking care of, and each company
has their preferred method.

This is an Abstract of them:


By using the least-squares fitting approach, the calibration procedure for fringe projection profilometry becomes more
flexible and easier, since neither the measurement of system geometric parameters nor precise control of plane moving is
required. With consideration of camera lens distortion, we propose a modified least-squares calibration method for fringe
projection profilometry. In this method, camera lens distortion is involved in the mathematical description of the system for
least-squares fitting to reduce its influence. Both simulation and experimental results are shown to verify the validity and
ease of use of this modified calibration method.

 

[JohnF]

Cheers Ketch!

That makes sense. Do you have any idea if the lens being tested is used for the projection or tested on-camera to measure a projected pattern?

JohnF