|
crowd sourcing the experts |
|||||
| |
|||||
 |
|||||
(Page 1 of 2 pages for this article 1 2 >) Filed under: 3D •Cameras •Hardware •Production •Training • Hands-on with the Panasonic AG-3DA1 S3D CamcorderUsing the 3DA1 at the Createasphere / Panasonic 3D Workshop
Last weekend I attended the two-day Createasphere 3D Production Workshop Utilizing Panasonic AG-3DA1 Cameras in Burbank, and got a good overview of the issues in shooting stereo 3D (S3D) content, as well as some immensely instructive hands-on experience with the AG-3DA1 S3D camcorder.
I went as a 3D newbie; I had no previous experience shooting S3D, and wanted to learn the basics. I was concerned that the focus on the 3DA1 camcorder would detract from the generality of the course, but my fears were unfounded: general S3D theory was discussed in detail, and the comparative simplicity of the 3DA1 (fixed interaxial [IA, a.k.a. interocular, or IO]; single-camera operational convenience) allowed plenty of opportunity to get into 3D compositional trouble without getting bogged down in the mechanical hassles of aligning a more traditional S3D rig.
An AG-3DA1 feeds a 3D plasma screen over HDMI in the main classroom.
Day 1 started off with an overview of S3D theory and practice, followed by an introduction to the AG-3DA1. We then had a hands-on workshop in the late afternoon. Day 2 commenced with another hands-on workshop, continued with a review of editing and finishing options for 3D material, and concluded with a final Q&A session. I attended as a guest of instructor Bob Kertesz (whose email sig says, “DIT and Video Controller extraordinaire. High quality images for more than three decades - whether you’ve wanted them or not.©”); there were several seats left over in the 40-person limit for the course, so I wasn’t displacing a paid customer. Mr. Kertesz tag-teamed with Dave Gregory, S.O.C. (whose card says, “Your very own gray-haired LIGHTING CAMERAMAN and EDITOR”), an optical effects specialist whose filmography dates back to the original “Star Trek - The Motion Picture” (the first—and so far the only—big-name Hollywood picture I worked on, grin) and “Star Wars - Episode IV”. These two gentlemen talked through the general theory and practical issues of shooting stereo 3D for the first half of the first day.
S3D TheoryI can’t recount the entirety of what was discussed (for one thing, I’m still absorbing it all myself), but I will list some of the things I made notes on. Please excuse me for not reprising basic S3D theory here! • Without absolute pixel-by-pixel synchronization between the two views (genlock, at a minimum), motion in the scene shows up as positional disparity (since there’s a time difference between the L and R frames). If it’s horizontal motion / disparity, it reads as false stereo depth; if it’s vertical, it just hurts the eyes. That’s one reason “two Canon 5Ds nailed to a 2x4” won’t work as a stereo rig. • When you go to a 3D theater, the best seats are at the back, on the centerline. The worst are in the front row and/or off to the sides. • BSkyB (European satcaster) says “too much coming out of the box is intrusive in people’s homes.” Only 2% (of screen width) disparity is allowed. • 10% of people don’t see stereo 3D. • Converging the eyes (to see stuff close) is easy, up to a point. Divergence (seeing stuff past the convergence point; going walleyed) is OK up to about 1 degree; beyond that and “eyes pull out of sockets”—very disturbing, if not downright painful or impossible for most folks. Staying within that 1 degree divergence is generally the limiting factor in S3D depth budgeting, and it depends on screen size and how near that screen the viewer is!
Bob Kertesz demonstrates disparity using crossed sticks representing the L & R viewing axes.• “Depth budget” - the near/far limits of comfortably viewable depth in an S3D image, defined by the disparity of near and far elements. Depth budget is controlled by IA, the convergence point, and the size of the the final image is viewed at. Smaller screens allow a deeper depth than larger screens. • “The one thing that can’t be fixed [relatively easily and inexpensively] in post is depth volume.” If you shoot with too wide an IA, it can be fixed by interpolating between the L and R images; if you shoot too narrow, you have to dimensionalize (e.g., recreate one “eye”) from scratch. • Vertical disparity is horrifically disturbing, since you don’t see that in real life and your eyes don’t work that way (other bad disparities: size, rotation, distortion, color, exposure; that’s why matching lenses and matching camera settings is so critical). • Humans normally use binocular depth perception only out to 130 meters. • “The rule of 1/30”: IA should be no more than 3% of the distance to the nearest object in the scene. Some say 1/50 (2%); or even 1/100 (1%) for large screen presentations (e.g., 3% for showing on a 42” screen is OK, but 1% is more appropriate for 60 ft. cinema screens). You can violate this rule for short, quick shots, but no for long takes. Soft focus allows more violations, but has its own issues. • Corollary: close-ups are “danger, Will Robinson!” S3D prefers medium shots and MCUs over ECUs. • Horizontal edge violations (where one eye sees something the other doesn’t, because the edge of the image crops it off) are terrible; fix in post with a “floating window” to crop the excess info, but do it with half-second wipes in the affected channel, not hard cuts. • In S3D, convergence is the substitute for shallow depth of field in 2D. • Beamsplitter (over/under) rigs have more issues than side-by-side rigs (or all-on-one cameras like the 3DA1) due to the semitransparent mirror: differing reflections and polarization effects. These and other “bandits” (elements appearing to one eye, but not to the other) are especially prevalent with water waves, rain, etc. Organic beamsplitters avoid a lot of these problems, but they’re $1000 each, and they suffer from fungus growth in wet environments, as they found on out on “Pirates of the Caribbean 3D”! • Setting up a two-camera rig is time-consuming. Even a quick one like the Kernercam rig (which is built for 2/3” box cameras and matched Fujinon zooms) takes about 10 minutes to build and another minute for a “homing” procedure to calibrate the servos to the cameras and lenses—and that’s a very fast one. Calibration needs to be re-done on every lens change, so it’s preferable from a time and money standpoint to stick with calibrated zooms instead of swapping between primes from shot to shot.
The Kernercam 3D beamsplitter rig, with custom-fit cabling preinstalled.• As far as the learning process is concerned, there are four levels of 3D competence: Unconscious 3D Incompetence - you don’t know that you’re doing it wrong. Conscious 3D Incompetence - you know that you’re doing it wrong, but you don’t know how not to do it wrong. Conscious 3D Competence - you know how to do it right, but you have to think about it to avoid making horrid mistakes. Unconscious 3D Competence - you have internalized the rules, and you just shoot good 3D naturally. The AG-3DA1After lunch, the ubiquitous Jan Crittenden, Panasonic’s product manager for the 3DA1 (among other things), took us through Panasonic’s integrated S3D camcorder.
Jan Crittenden demonstrates the 3DA1 as Dave Gregory looks on.• The AG-3DA1 is “so much like a standard, 2D camcorder that your initial 3D footage will stink!” You’ll shoot 3D as if it were 2D, because the camera makes it so easy, and boy! Will you ever be sorry! (I can speak from my experience in the following 24 hours that she was saying nothing more than the absolute truth, grin.) • The zoom is 5.6:1. It started out as 10:1, but as it was optimized for 3D production the zoom range shrank.
Basic tech details for the 3DA1, courtesy Panasonic.• The 3DA1 was nine months from concept to production, half the time of a typical camera. • The camera is switchable between 50 Hz and 59.94 Hz standards. It will shoot interlaced, but don’t do it! Instead, shoot 25P or 30P, deliver as 1080i (that way there won’t be any temporal disparity between the channels regardless of presentation technology). Of course, the 3DA1 shoots 24P as well. • Convergence is controlled with a side wheel, shared with iris/gain. Settings from C00 (nearest convergence, about 2 meters / 6 feet) to C99 (shooting parallel). The EVF and LCD can show left eye (normal), right eye, or “mix” mode with both eyes overlaid; the camera is converged where the images overlay without disparity. • The 3DA1 has a fixed IA of 65mm (roughly the same as a grown human’s IO), and its most appropriate shooting range is from 3-30 meters, or 10-100 feet. • It has wired remote start/stop, focus, iris, and zoom, like other Panasonic handycams (think Bebob and Varizoom remotes), and adds remote convergence control. HD-SDI carries start/stop signals for synchronized offboard recording. • Camera has HDMI 1.4 output (needed for full-res, two-channel S3D) and dual HD-SDI outputs. HDMI and HD-SDI are mutually exclusive; need to switch HD-SDI off in the menus to get HDMI output. • “ProRes LT has more than enough headroom for AVCHD transcoding.” (I’m not sure I agree; I see differences between LT and normal when I transcode AVCHD; even differences versus HQ, though those differences are visually insignificant. But I’m splitting hairs; LT is probably fine for most needs.) • The camera always records PH mode: 24 Mbit/sec overall bitrate (long-GOP AVCHD, with 8-bit 4:2:0 sampling). Class 4 or faster SDHC cards; 180 minutes on a 32 GB card. Sandisk or Panasonic cards recommended, “otherwise, you’re on your own.” Audio is only recorded on the left-channel card, so the left card will fill up faster than the right card: don’t freak out, this is normal. • The 3DA1 was designed to take you from 2D’s “Guitar Hero Easy” (control focus, iris, and zoom) to 3D’s “Guitar Hero Medium” (control focus, iris, zoom, and convergence). You don’t have to fuss with the rest of the 3D setup as it’s controlled by the camera; the left and right channels are factory-tracked and factory-matched. • The camera has a “3D Guide” function, showing you your approximate depth budget for your current zoom and convergence settings. It has two presets: when “3D” is shown in white, it’s assuming a 77” screen; when it’s green, it’s working based on a 200” screen. • The 3DA1 uses shift lenses for convergence, not toe-in of the lens axes, so there’s no problem with keystoning. • Iris ring also controls gain, but according to Bob Kertesz, noise in 3D isn’t as apparent as in 2D, so it’s not a big deal. (I find this to be the case; looking at the camera’s output on the 25” LCD and 50+” plasma without 3D glasses, the gained-up images are pretty noisy, but once I put on the glasses, I simply don’t notice it. Interesting…). • It’s $21,000 instead of the $4,000 you’d assume (in essence, the guts of the 3DA1 are comprised of two $2,000 AG-HMC40s side-by-side) because the zoom lenses are hand-matched, and each camera is manually calibrated with servo look-up tables to ensure that the two lenses track each other precisely.
A Panasonic slide showing the 3DA1’s basic schematic.
Hands-on, Day 1We broke up into four groups of seven or eight people each; two groups went with Dave Gregory and two went with Bob Kertesz. Each group was given an AG-3DA1. Theoretically, one group with each instructor was supposed to shoot a dramatic scene in 3D, while the other shot a “making of” docco, also in 3D. Practically speaking, both groups strove to get interesting and informative shots, regardless of assignment. A couple of folks from the “drama” group were “volunteered” to be actors, so the drama group would have something to shoot (grin). These unfortunates were usually dealers or rental-house folks; it was felt, not unreasonably, that they would have ready access to 3DA1s after the workshop, so their hands-on time in class was less critical. It must be noted that the actors acquiesced to their fates with dignity and good humor. I was in the docco group with Mr Gregory (practically speaking, this was the time I shot stills; all action pix were shot that afternoon, before and between the times I got my hands on a 3DA1 or I was helping to wrangle cables or drag monitors around). Since the 3DA1 sensibly limits near convergence to 2 meters (lest the background divergence go so wide as to instantly fracture a viewer’s skull), a large stage was needed, and we dragged a desk, desk accessories, and two chairs outside to a patio so the cameras would have enough room to work. The two actors went through the motions of a job interview, while the drama crew tried a variety of shots.
Setting up the “drama” cam on the patio.The docco crew shot the drama crew and also branched out into experimentation: a telephoto shot of someone walking towards and away from the camera, odd angles, shooting through foliage, and so on.
The “docco” crew composes a shot.The drama camera was tethered via dual HD-SDI lines to a large 3D plasma in the room next to the patio, so folks could watch the camera’s output live. The docco camera’s monitor had only a short HDMI cable, so it went untethered, and the docco crew used the camera’s mix mode alone to monitor convergence. Several folks used the 3DA1 handheld, just because they could.
Handheld, eye-level.
Handheld, using the LCD.Try that with a traditional, two-camera stereo rig! When I got my hands on the camera, I tried a variety of things I was told I shouldn’t: shakycam (not that I like shakycam, mind you; I just wanted to see how much worse it looked in 3D), extremely close foreground elements; tight closeups, edge violations. After a feverish period of shooting, everyone went back into the room to watch playback. Everything we were told to avoid looked really bad in playback. Shakycam was unwatchable—even more so than normal. Edge violations by things in front of the screen plane were execrable. Tight closeups were disturbing. Converging too close, in an attempt to keep close action “inside the box” (behind the screen plane) caused too much disparity in the background, making it impossible to fuse visually. Cutting between shots with widely-varying convergence distances caused headaches. Y’know? All that stuff about depth budgets and choosing your convergence point carefully? All those rules? They aren’t fooling! I can’t speak for the other folks, but I went away from this exercise with a newfound respect for stereographers and the constraints under which they work.
(Page 1 of 2 pages for this article 1 2 >) 
You must be registered to comment. This is an effort to reduce spam. Please REGISTER HERE.
|
|
||||
|
|
||||
Litepanels Illuminates Chroma and Luma at NAB 2012Check out two on-camera lights that were featured at the show
We have more from Litepanels as Toby Sali showed us their newest on-camera lights from the floor of NAB 2012. He gives us some details about the Luma, which is a daylight only light,… |
Orad News and Products from NAB 2012Learn about what we discovered when we stopped by the Orad booth
Shaun Dail from Orad took the time to tell us about what they had to show and talk about at NAB 2012. He tells us about TD Control, a sports specific MAM system as well as their… |
Check out a Number of Hardware and Software Options from B&HEverything you need in one place
We grabbed Jerry Zorek, Manager of Business Development at B&H, to learn about what B&H was showing off at their studio booth. He shows us a Resolve system with the… |
Anton Bauer Unveils Gold Mount Solution and MATRIX Cheese Plate at NAB 2012Take a close look at both products from the show floor
At NAB 2012 Anton Bauer unveiled its QRC-CA940 Gold Mount solution for the new Canon EOS C300 digital cinema camera that is designed to streamline battery management for users.… |
To be considered for listing, contact pr (at) provideocoalition (dot) com
Copyright © 2012, HD Expo, LLC a division of Diversified Business Communications. DBA Createasphere
All rights reserved. HD EXPO, High Def EXPO, Createasphere, E-Tech, Entertainment Technology Exposition, 3D Production Workshop, VariCamp, P2 Camp, ColorCamp 101, and Lighting, Filters & Gels for HD are all trademarks of HD Expo, LLC.