• Page
• Discussion
• View source
• History

From Odwiki

Contents 1 MotionControl: 2 DepthOfField: 3 Maya vs. Houdini: 4 Commercial Software: 5 Books:

MotionControl:

• The accuracy of a motion control rig is often measured by it's ability to repeat movements. so, an accuracy like .5mm over each repetition at a given time has little correlation to the absolute accuracy that we are interested in from the digital side. for example, if a rig is told to move 20 meters across the stage, it probably can't be assumed that it will will move 20 meters += .5mm even if it's supposed to be accurate to .5mm translationally.

• Usually when you ask the motion control group for some transform data for the rig they output nodal information for the camera head. This data is easy to import but requires a lot from the motion control software. When it bakes together all the information that it's using to position the camera into a transform for the film plane it can be filled with errors in calculation or calibration that the software has no way to check. It not until you get this data into your 3-d system that you can see the problems and at which point it can be very hard to fix.

• The last problem is just getting a good registration of where (0,0,0) is in the data relative to your objects of interest. Getting this measured accurately is non-trivial particularly if the relative distance between the camera and the objects is small - which it usually is. Rob Bredow

• Lynx and Kuper data is easy to import - it can be exported as ascii columns with a one or more line(s) header: almost a .chan file. Tim Segulin

DepthOfField:

this can done efficiently by rendering a z-depth matte and performing the blur in composite. inferno/flame can use a dof spark to create really nice dof with good control if a z-depth matte is used. conversion of z-depth matte to luma map: sesi's izg or in cops

Film Cameras & Film Formats:

• In Houdini, if you put in a 50mm lens/focal length, you must also put in the proper aperture in mm. Page 64 and 65 of the user guide show ways of deriving the proper measurments. The five channels that go hand in hand to getting your camera locked in Houdini are; focal, aperture, resx, resy, and aspect. resx and resy would match your film plate in pixels, aspect would match the pixel aspect of the scanning device. John Willete

• And be aware that if you are working at video resolutions that you are doing the camera matching using footage that has not been optically scaled in telecine. JasonIversen

• The stretch factor for an anamorphic (or cinemascope) lens is 2:1. JasonIversen

1. . You can type the focal length in millimeters directly into the focal length parameter of the Houdini camera. However, for it to mean anything, you also have to tell Houdini the *real world dimensions*, in millimeters, of the images represented by your scans as they existed inside the camera.
2. . To do this, you could literally grab the negative, cut the part of the image that appears in the scan out with an exacto knife and measure it with a ruler. But your client would probably object.
3. . So how do you find the real world dimensions of the image? If you are working with cineon files that were scanned in a Kodak scanner, you're in luck: Kodak publishes the dot pitch of the CCD array used in their scanners. It's 166.666666 pixels per millimeter, at the native resolution of 4K. I'm guessing you have 2K Cinemascope scans, but you would calculate using the 4K native resolution. Since 4K cinemascope cineon files are 3656 X 3112, you know your scans are 3656/166.66666 X 3112/166.66666, or 21.936 X 18.672 millimeters in size. If you are not working with scans from a Kodak scanner (just because they are cineon files doesn't matter), the 166.6666 pixel per millimeter figure will be close, but not accurate. Try to find out the dot pitch of the scanner that was used, and use that figure.
4. . An additional step is required since you must compensate for the cinemascope squeeze. How big, in millimeters, would the image have been before it was intercepted by a cylinder of glass and squeezed 2:1? Fortunately, the math is easy: 21.936 x 2 = 43.873 millimeters, which is the width of your frame. If you are working with the cinemascope unsqueezed, this is the number you put in the aperture field (which corresponds to the width) of the Houdini camera.
5. . The vertical dimension will be taken care of by the aspect ratio info you give Houdini. I will assume you are working in the resolution of you scans. If they are 2K cscope cineons, that'll be 1828 x 1556, with a pixel aspect ratio of 2.
6. . There is one more nasty little gotchya. Cinemascope is a film format that leaves area on the film for an optical soundtrack. So, the image is pushed to the left on the film. The axis of the lens is still centered over the film, so the center of projection will be shifted right relative to your images. If your scans came from a Kodak scanner, you're in luck again, because you can calculate the offset based on the relative resolutions of Full Aperture (which is centered relative to the film and lens) and Cinemascope, which has the right side of the picture chopped off for the soundtrack. You would put .5*(4096-3656)/3656, or 0.060175 into the Crop/WindowX parameter of the Houdini camera. Make it negative (-0.060175) to push the center of the picture to the right. Be sure to turn off "Apply Zoom to Background" for the viewport or the shift will take your backplate along with it.

Denis Gauthier

But more on the optical centre topic...

Actually, with a camera set up for Academy, the optical center is in the Academy center, i.e. slightly to the right of the negative center. If that were not the case, a nodal pan/tilt setup would be impossible (which it pretty much is anyway with anamorphic, as the horizontal and vertical nodal points aren't necessarily coincident). Ergo, you should not need the crop parameters.

"185" is an Academy composition, and hence, needs a lens setup where the optical center is to the right of the frame center. Any compositions that have "Academy" in them require this, and that includes Cinemascope, which is shot anamorphically into the complete Academy region.. The two variants of Super-35 (common-center and common-top) are the only projected formats that need a full-aperture setup. Antoine Durr

For most 35mm footage, the lenses are centred on the Academy area. A scan representing this area should be optically centred. Anamorphic originated material is also usually centred on the academy area. The optical centre of your image becomes an issue when you have a full gate scan. Normally you need to offset the centre to compensate for the extra "soundtrack" area on the left of the image. Many productions are now shooting in Super35 which uses the extra width of the soundtrack area of the camera original negative shot spherical but the image is later optically or digitally squeezed and offset to fit the standard anamorphic academy area. Super35 cameras are modified so that their lenses are centred on the full gate aperture.

To add further to the confusion (slightly), the standard full gate aperture width is 24.89mm but a standard Cineon 2k/4k scan actually covers slightly less than this - 24.576mm - you lose a thin sliver on the left. Not all scans conform to this........ Dan Glass

Maya vs. Houdini:

Houdini express its camera apx value in millimeters. Maya express it's apertures in inches. (Actually it expressing them in "decimal-inches". ie Mayas horzFilmAp look like this "1.42"in) Do some conversions between the two and look into the USERSGUIDE for its short treatis on Houdini cameras. The other values you may need can be derived from the ones already supplied in houdini. James Peterson

• Rangi Sutton answering a post on the SESI list:

houdini Aperture * 0.03936981 = maya Horizontal Film Aperture

Or backwards: houdini Aperture = 25.400173 * maya Horizontal Film Aperture

Yep, maya focal and houdini focal are the same deal... I've checked the above by outputing rib's from each package and it's good...

Commercial Software:

• 3d equalizer - exports to Houdini
• boujou - exports to Houdini
• PFTrack (formerly known as Icarus) - exports to Houdini
• SynthEyes - exports to Houdini and costs $400
• ras track
• maya live

Books:

• american cinematographer