Yesterday, Simon Stappleton from Procam North and I, and Dr Bernie Harper from Liverpool University took two Sony XD cams off the shelf to take a look specifcally at motion issues as I had read some stuff on line about 3d blur and movement and how things can get tricky when shooting movement. I wanted to find out for myself what happened under different circumstances. During my 3d training I have come across loads of people in the film and TV bizz who are full of advice and are quite happy to tell you how it should be done... But then when I asked the questions, I found out that although everyone is ready with advice, NOT MANY have actually shot 3d!(I guess thats just a bit like life...)We put the cameras onto the prototype Procam SbS Plate, a special plate that Jonathan Bolton at Procam North had engineered so that we could mount ANY camera Side by Side for trials. When you are training it's hard to get access to large mirror rigs, apart from the obvious reasons of cost, ie you need a team of Nasa trained scientists to program it, align it, and 'fly it' for you. Pretty much all of that technology is very expensive to hire, and because 3d is the latest 'in thing' particularly in London and the South east, access is very limited.
Training - is limited and expensive, but if you have the money then Pinewood and Shepperton run 3d training courses. Actually, on the subject of training I was very fortunate in being able to secure an 'Innovation Grant' from the RDA in our region, and that was used to get me superb access to some of the keenest and most intelligent minds on the planet who have been living and breathing 3d for 10 years or more. Now at last the technology is catching up...For our motion test shoot we shot a variety of settings - interlaced, progressive, different frame rates. We filmed loads of cars driving by and I spent hours in Noise Industries Dashwood Stereo 3d Cinema Toolbox looking at disparities and basically trying to understand what worked, what looked good and what looked bad. Whether or not to track convergence on the passing cars or leave convergence set at an optimal distance. I will talk about convergence in a separate post because it is another of those 'sensitive'' areas in the 3d community, and the idea of this blog is not to cause stress and anger, but to raise the issues for discussion...
In terms of motion - Point One: I would always shoot progressively for motion. Full stop. end. new para.
Yes, you can play with frame rates and electronic shutter speeds - but you must shoot progressively because of how the frames with movement on have motion artefacts when shooting interlaced. You may have seen it yourself in 2d TV - that weird jagged edge where the scan lines have captured slightly different moments in time so a waving hand breaks up across scan lines. If you combine stereo shots with this effect you end up with a mess that the brain can not fuse, so no 3d. Always shoot motion progressively.
Point Two: Your cameras should be synced or 'gen locked' so that each frame is written and captured at the same time - that is to say the beginning of each frame must start at that 1/50th or 1/25th of the second (or 1/60th) and scan each line, both camera's perfectly in sync. Line 1 right eye camera, scans at the same velocity and the identical moment in time as Line 1- left eye camera and so on through the full frame scan.
So Progressive and genlocked / synced. We did try some early days tests with 2 HVR-A1's that were not 'genlockable' but were synced through a dual lanc within 100ms and on the whole the results were not bad. Though we weren't filming anything with much movement - just some stuff around 'sets' for a tv drama, kitchens, living room, that kind of well lit set up in a TV studio. We used a digislate to mark top and bottom of each take to see how the sync help up over time and it was reasonable and useable.
Definitions:Motion artefacts - weird bits of video that are created when the technolgy has to 'predict' what the movement in the frame is - or how the technology sometimes fails to capture the action properly.
Temporal Occlusion - when the differences between the two shots are caused by time, by the the way the 2 cameras have recorded or 'captured' everso slightly different moments in time, and so they don't work in 3d.
My final thoughts - you can shoot on interlaced formats - I use 2 little HD JVC's shooting AVCHDs or MPEG.TS for 'Pre Viz' and ' recce' . They are not locked, not synced and interlaced!
But they show me enough to see what works and what doesn't work...
In terms of motion - Point One: I would always shoot progressively for motion. Full stop. end. new para.
Yes, you can play with frame rates and electronic shutter speeds - but you must shoot progressively because of how the frames with movement on have motion artefacts when shooting interlaced. You may have seen it yourself in 2d TV - that weird jagged edge where the scan lines have captured slightly different moments in time so a waving hand breaks up across scan lines. If you combine stereo shots with this effect you end up with a mess that the brain can not fuse, so no 3d. Always shoot motion progressively.
Point Two: Your cameras should be synced or 'gen locked' so that each frame is written and captured at the same time - that is to say the beginning of each frame must start at that 1/50th or 1/25th of the second (or 1/60th) and scan each line, both camera's perfectly in sync. Line 1 right eye camera, scans at the same velocity and the identical moment in time as Line 1- left eye camera and so on through the full frame scan.
So Progressive and genlocked / synced. We did try some early days tests with 2 HVR-A1's that were not 'genlockable' but were synced through a dual lanc within 100ms and on the whole the results were not bad. Though we weren't filming anything with much movement - just some stuff around 'sets' for a tv drama, kitchens, living room, that kind of well lit set up in a TV studio. We used a digislate to mark top and bottom of each take to see how the sync help up over time and it was reasonable and useable.
Definitions:Motion artefacts - weird bits of video that are created when the technolgy has to 'predict' what the movement in the frame is - or how the technology sometimes fails to capture the action properly.
Temporal Occlusion - when the differences between the two shots are caused by time, by the the way the 2 cameras have recorded or 'captured' everso slightly different moments in time, and so they don't work in 3d.
My final thoughts - you can shoot on interlaced formats - I use 2 little HD JVC's shooting AVCHDs or MPEG.TS for 'Pre Viz' and ' recce' . They are not locked, not synced and interlaced!
But they show me enough to see what works and what doesn't work...
No comments:
Post a Comment