When small interaxials just don't cut it! - Page 9

If you are going to exploit two cameras for 3D video, why not take advantage of the flexibility for audio as well? In stereo mode in both cameras you have the same set of choices - in effect you can choose between wider-spaced mic modes (two cameras left and right or two rights or two lefts) and two close-spaced mic sets that are to the left or right. In surround mode, well the possibilities are endless for choosing the best soundfield. C'mon guys, you need to go all the way here!

That's an interesting observation, and I agree. Minor differences in contrast and detail between the two camcorders also might play a small part. I believe that the stereo image that is the sum of both my eyes allows me to see more clearly than either eye by itself. A contributing factor is that one of my eyes sees the world with a bit more contrast, while the other sees things with a bit more resolution. I'm not sure why that is - maybe differences in the way light hits objects, maybe just because of some inherent physiological difference, maybe a combination - but there is a perceptible difference.

Thanks to Frank's intro shot for the test, it occurred to me that I hadn't used my Fuji W3 since I got the JVC, so I went to the St. Louis Zoo today and took a few 3D stills. The wider IA of the Fuji definitely makes 3D look more natural and pleasing to me, although I have to adjust how far I stand from my subjects. The stills from the Fuji look great. I'm going to put together some Blu-ray 3D iso's of my stuff from the last year. Thanks, Frank.

The IA's for the Fuji and the JVC should be similar when I get my stereo base extender. Unfortunately, I can't go as wide with the JVC, so some shots still won't work as well.

Don,

If I go with a dual camera rig, I think I'll do it with dual Canon's similar to Frank's. I'll probably sell my Canon HF10 and get a couple of newer Canon's that shoot at 25mbps. The HF10 is limited to only 17 or 18mbps. I'll also want tandem LANC control, like Frank's. Fortunately, I have Frank if I need answers to some of the specifics of how to put all this together.

Hey, Frank, could you describe exactly how to get Vegas to accept the Fuji W3 mpo files (or how to pair the split left/right images)? I can't get it to work.

Ha ha!
Old joke for audio techs. That doesn't work!
But for those who don't understand physics, picking up sound from separated mic sources causes echoes due to speed of sound delay factors. On stage where widely separated mics are present sound techs go to great lengths to isolate mics that are widely separated. On my tv shows I went to great lengths to to isolate mics on separated speakers even those sitting next to each other.. Besides the reverb the sound you get will also suffer phase cancellation at different frequencies that cause distortion. The best sound field is very simple, single location with directional mics near the camera for ambience, and close placed mic to speakers that are isolated to prevent pickup of other sounds. Mix the close mics with the surround mic. Achieving the isolation is the tricky part.

Joe, I would think you would have more 3D options with two TD 1's. The lanc control has little purpose unless you are trying to do what Frank is doing with a remote control live setup. Having two 3D gives you more options in shooting with traditional 2 camera productions from two POV's. This can be especially rewarding with performances where you set one cam up in full wide and then use the other to shoot your closeups. Cut between the two.

I've shooting two years ago with dual Sanyo FH1 at 1080p60 (and dual kodak zi8 at 720p60) without lanc and after syncronisation in Vegas9 (temporal) StereoMovieMaker (spatial) had not bad at all 3D

Now thinking about 3D rig based on sony's camcorder HDR-CX130 with lanc and 1080p60 (or 50p)

I must have missed something...

I thought the JVC didn't have Lanc, yet you just synced it with the Sony.

Of course, I have no intention of using Stereoscopic Player 3D and don't know what sort of limitations there is with that software. I will be editing everything with Vegas Pro and Vegas Pro has no trouble syncing multiple cameras without chase lock or Lanc but you have to know how to do that. The sync is accurate to the start of each frame. Joe may have some issues caused by his need to do some pre Vegas conversion that could cause sync problems but I doubt it. Reason is that Vegas can quantized to frames for all formats.

On another subject, I had a couple hours today free while wife was at her conference so I went to a hardware and electronics surplus store and picked up some metal and knobs to begin construction of the bench. Can you make up a chart you have found to work for IA and distance range of shooting? I believe there is one in one of my texts but haven't located it yet. The main frame I bought is aluminum channel 5 ft long and I'd like to reduce that to something I can manage for travel. 2.5 ft would be nice but not sure what range that puts the rig in. If only 1000 ft, it may not work for my project intention. I want to shoot mountain ranges and canyons 1-3 miles. If I can get that as maximum and then any distance less than that I'll be happy.

"Ha ha!
Old joke for audio techs. That doesn't work!
But for those who don't understand physics, picking up sound from separated mic sources causes echoes due to speed of sound delay factors. On stage where widely separated mics are present sound techs go to great lengths to isolate mics that are widely separated. On my tv shows I went to great lengths to to isolate mics on separated speakers even those sitting next to each other.. Besides the reverb the sound you get will also suffer phase cancellation at different frequencies that cause distortion. The best sound field is very simple, single location with directional mics near the camera for ambience, and close placed mic to speakers that are isolated to prevent pickup of other sounds. Mix the close mics with the surround mic. Achieving the isolation is the tricky part."

While my suggestion was mostly tongue in cheek, leaving aside micing stages with speakers (loud speakers or human speakers?) and focusing on actual use in the field, the following is relevant:

There are three well-established microphone stereo techniques (among others) for recording natural sounds (soundscapes, crowds, musical concerts):

1. spaced omnis

If the mics in the cameras are omnidirectional mics (non-directional), then spacing the left and right mics much farther apart with two cameras recording the sounds could give a better soundfield and stereo. Distances are typically in feet, however, not inches. Good for the Grand Canyon 3D effort.

2. ORTF

This technique uses cardioid (unidirectional) mics spaced about 17 cm apart with a specific angle for each mic (110 degrees). The stereo mics on one camcorder are spaced much closer than this, so achieving a greater spread between the mics (left and right from two cameras) could result in better sound, if the camera mics are unidirectional and are angled close to 110 degrees.

3. XY or coincident

This techniques has the mic heads exactly coincide (one point) with the mics pointing at angles in different directions. I do not think that is what is used in camcorders, but it is popular among the small, high-quality audio recorders with built-in mics (Zoom, Sony). Having two cameras won't help for this technique.

I was talking about people speaking on camera with their own mics. Over the years there have been different ways to deal with this. With people speaking, the best was developed by Sure called Intellimix. In Orchestral recordings, the mics are all up close and low gain to prevent pickup of surrounding instruments. The mix board is used to balance everything. If recording the same stage with stereo then only two mics are used and they are located center audience, commonly hung from the ceiling. The stereo rig uses two cardioid patterns aimed opposing and at 90 degrees to create a center phantom channel that is in phase.

Mark, your imaginative ways of recording stereo won't work. Been doing this too many years professionally and know how these things are done. I do not know how the internal modern mic configurations are done in consumer camcorders. Most today use two picups and DSP. I think the TD10 uses four with DSP for 5.1 but I could be wrong on that. Your methods will record sound but it will be loaded with distortion and reverb. For stereo, use directional cardioid mics, close together, aimed opposing at 90 degrees. Or, make life easy and just buy the stereo mics.
Bottom line, wide separated stereo mics to improve stereo separation is not just a myth, it is goofy dumb! But not to be too hard on this there is one special case where this works. It is when you only have one sound source, AND you can feed the two mics into a SURE Intellimix.

"For stereo, use directional cardioid mics, close together, aimed opposing at 90 degrees."

That is just one stereo technique, and it is neither the best nor used the most (actually hardly at all) by audio engineers who specialize in recording sound for a living.

Don, ORTF and spaced omnis are standard methods of recording stereo with two mics. What is hanging from the ceiling for concert recordings could be either an ORTF configuration, spaced omnis or even a "Decca Tree," which has three mics, the middle to fill the hole from spaced omnis. There is also Blumlein, mid-side. . Some companies use scores of mics for different sections as you say, but I am just talking about two-microphone stereo techniques. And placement matters a lot too. Most professionals I know prefer two-mic techniques for recording natural acoustic music, but often recording live concerts dictates closed-space multiple mics because of an obsession with quiet.

Two mics at 90 degrees with some kind of spacing (if no spacing it's likely XY, but that is rare) is not a well-respected stereo technique and I do not know any professional engineer who does that. I did not say anything about "separation" for the sound field, just the mic spacing to achieve good sound, as measured by multiple dimensions directionality, frequency response, depth, width.

Now, whether these classic stereo, two mic techniques will work with combinations of camcorders' mics, I agree we can't know without knowing their specs. But I know what I need to find out. Anyway, this is just fun to think about. One could just more easily just set up two mics and plug them in, as you say, but certainly not with an arbitrary spacing at 90 degrees! And cardiod has limited bottom frequency response compared to omnis, and often more noise at a similar price point.

Frank. You are absolutely correct that the editor doesn't know that the two videos are not framed at the same moment in time, UNLESS, you provide a reference. In professional video that would be timecode and genlock. However, neither consumer camcorder has this reference. The pro versions do provide time code that can be used to get frame accurate video. SO, absent the timecode, what is done for a reference is the use of audio that is "glued" to the video. Thus my prior references to the clapboard. The simultaneous pop sound plus the visual of the boards coming together is as old as nagra sound recording for film and the method is still a rather accurate procedure to get the two close enough. There have been references to the use of a photo strobe flash as the light is less than a frame of video in duration but this method has been shown to be less accurate than sound. In Vegas the use of sound waveforms on the two cams of a sound pulse of short duration can be quite accurate. Still, I agree, that it will never be as accurate as two genlocked cameras. The need for the two cameras to be that close may not be necessary for amateur video. The work you did looks just great to me and was quite surprisingly close. If you feel your shoot was horribly out of sync then I think you need to get your eyes examined again. We're not planning to broadcast this stuff you know.
Also, LANC is a rather mysterious signal set. I've had great success making a converter for LANC time code convert to LTC AND GENLOCK to sync broadcast cameras to a consumer camcorder's LANC output. It doesn't surprise me you have found equipment that can do the same. If I complete the task of building the twin TD10 bench and it fails sync as you suggest, I can always buy the device you are using and resolve the problem. But, your results tell me I won't need that.

Obviously, I haven't used LANC with dual cameras, but it seems to me that a huge problem would be lining up any shot that used anything other than max or minimum zoom. If you want to create a shot that doesn't zoom in or out while live (the vast majority of shots I'd want to use), isn't it very difficult without LANC to set exactly the same zoom range for both cameras? This would be a problem completely unrelatd to genlock.

Also I am no expert but have given this some thought. We want an angle with vertex on the subject in the foreground and 2.5 inches (my estimate) eye distance at the point where the zoom pretends to be. I think Frank's numbers are right, to multiply the standard ia (but maybe of 2.5) by the zoom factor. By zoom, of course I mean anything different from a normal lens emulating the eye.

If one runs a constant speed zoom, the ia could be changed also at a constant speed whenever holding down the zoom button, so the mechanics or electronics need not be complicated.

Also, I don't think the ia distance needs to be completely consistent from shot to shot, but a rig could pretty readily recalibrate itself at each extreme of zoom and ia distance, when it can't go any further no matter how much we mash that button.

One more thing. I like to keep the cameras parallel for simplicity, but if you want convergence, just build the rig to shift the cameras along an arc instead of a straight line.

I got an opportunity to "play" with bench rigs that did adjust the angle of the cameras but the guy there explaining how that worked said the arc of adjustment is not the same along the bench travel. it has to vary depending on the distance between the cameras. Since a bench would not be circular but more parabolic ( angle less as the IA distance increases), what they did is have the two cams on a servos that automatically adjusted the angle as the cameras changed distance along the bench. This was really complex when the cameras used under-thru configuration. Aside from being complex, the rig was quite expensive.

But there really is an easier way which is the keystone adjustments in the Vegas 3D stereoscopic effect. Just keep the bench straight and the cameras parallel and use Vegas to calibrate the images convergence.

Joe- I agree and that is why I feel that when using twin cameras it is quite important to use the proper tool to get things aligned properly. I know Frank believes the LANC is the best tool but having sat through the Sony class on how to do this, the process is actually quite easy with Vegas. Vegas Pro has the tool to align the two images both with automatic and manually. The adjustment monitor is your eyes and brain. Even if the shot has a change in z distance the adjustments can be corrected along a keyframe timeline. If the shooting zoom with twin cams is not changed during the shooting but the zoom locked down to initial calibrated framing each, then there should be no need for LANC to keep both zoom lenses in sync. I have no experience with dual camera control using LANC so I have no idea how accurate the zoom sync would be. I would be amazed if the zoom on each camera tracked accurately. In Franks test he had to carefully manually adjust the JVC / Sony so that the image was about the same. I think he did an amazing job of that as the two looked to be exact. Also I felt the exposure and color balance looked amazingly the same. It certainly shows that two dissimilar camcorders can be used to create a 3D image.

BTW- In an over through or under through setup the method of calibration is to set the cams for zero IA and super the cams images with transparency, calibrate the zoom on each for exact match and then the zoom is locked down. the cams are then spread to the desired IA on the bench.

Frank- OK I understand your sync comment better now. You claim you got lucky but I think the timing is actually easier than you think as long as one has a reference to calibrate the videos with. I haven't done this with 3D but I have with lip syncing for music videos and it was not that hard to do except that some of the cameras in the project were not the same so the audio / video tracking was off. The easy way here was to unlink audio and video and with quantizing to frames off, slide the audio a few milliseconds and re group the sound to the video then match up quantizing to video frames. I don't see why this will not work with 3D. But at this point it is a hypothesis and until I test it, it remains as such. If it fails and your Lanc controller system works I can always spring for that.
First things first, I need to build the bench.

Frank- I didn't say I am against adding the Lanc controller. I said I would add it if the testing proved it was an advantage. Your demonstration between the two JVC and Sony clearly showed it can work without the controller. It seems that using two Sony TD10's will be even easier to align and adjust as well as using Vegas to do the tweaking.
I do know how difficult timing the alignment of each left and right video can be without a reference, but using one or the other or both sound track plus a "clap board" sync reference, my experience has been that it is not that difficult. In my multiple camera productions, my timing sync just used LTC time code and it was very quick and easy to align using nothing but time code in DF mode. I had 4 cameras on many of my TV shows. They were all started at different times but the timecode was chase locked, yet only two using identical DVCAM recorders were in sync, another using a different DVCAM recorder was always off by exactly 3 frames and a fourth was a hard drive recorder was out by 1 frame. Once I knew the timing adjustments, I made them for ever clip and the the error correction factors always held, year after year of shooting. Now I know that you say the calibration needs to be more accurate than what I did but I will have to see if that is an absolute requirement or just your guess that it is. Additionally, I need to see what the cost of the Lanc device is, it is works like the magic you claim and not expensive, I'll just spring for it. Unlike you, I DO have a budget and need to watch the money carefully. I'm retired with little supplemental income from my business and I work with a budget. I am thankful that you are doing such a generous job of testing these things so I don't have to. I do have to be careful, however as not everything you do and every end goal I have is a match to yours. Therefore some things may not be necessary in my end use. e.g I have no use for a live windowed display on a laptop or a checkerboard display monitor. MY goal is still to shoot some very large panoramic landscapes and maybe some fireworks. I also need to take the 3D package into the public and not have it be overly bulky or draw unusual attention.


Ran across this info some here may find interesting:

http://www.technica3d.com/3D-alignment-chart/

Frank- Is this the Lanc device you are using? For the Sony, did you need to find your own cable adapters for connecting to the TD10? The Lanc port on the TD10 is not the older style stereo mini plug / jack but a D shell 11 pin I recall connector that also contains AV signals.
I read through the setup manual on this device and it seems it is not as simple to use either. A bit pricey too.

Can't make out what bird that is in the tree, A Bald Eagle maybe? Could be a crow too as the detail in the thread image is not enough to tell.

Most of the price is likely in the R&D recovery. Hope your guy can do better in price.

By the description on the web page, there are additional factors to consider they claim to resolve, such as exposure sync and shutter speed sync. I admit I hadn't thought of those issues but probably when out of sync, could be additional major detractors from a 3D stereographic rendering.

So building a 3D stereo bench for wide IA using a second TD10 could get more expensive than I thought-
Second TD10-- Budget for $1150
LANC sync-- Budget $435
2 AV/r to LANC jack adapters $50
Metal and other hardware-- $50
Total $1685