Hi All
I finally got time to do some ground tests.
I set up the camera on a sliding bar that allowed me to vary the interaxial distance (distance between the lenses) to demonstrate different amounts of 3D.
Original 17mm
65mm - human eyes
100mm
150mm
200mm
250mm
300mm
It is quite fun to see the difference. As the I.A. increases the depth between the nearest and farthest object expands. As the depth expands (hyper stereoscopic) there is also a miniaturization effect as out brain interprets the wide 'eye' separation instead as a miniature world seen with 'normal' eyes.
For my taste the original 17mm I.A. is way too small for flying but could be good inside a cockpit or in a car, boat where you are much closer to the ground plane. Human eye 65mm looks very natural but once off the ground an extra I.A of 100mm - 150mm would look great for multi rotor proximity flying. For fixed wing you could go much wider. My max was 300mm (1ft). It looks very exaggerated on the ground but I bet would look fine flying 100ft over trees etc.
The other thing to consider is alignment and the convergence point. I quickly aligned the cameras using hot glue. It is good enough for this test. Accurate camera alignment is essential for comfortable stereo. There should be no vertical misalignment in any point of the frame. Only horizontal.
Convergence is the point at which the two images over lap. The more you 'toe in' the cameras the closer the lines of sight will converge. Generally it is thought to be good practice in stereo photography to allow the scene to drop away from the frame. The frame acts like a window and we view the world beyond. I generally follow this idea. But I still like to have objects pop out if I get close.
My preference is to set convergence at a distance I can use as a warning. When flying my quadcopter I like the convergence set at about 4 ft. As I approach a gap in a tree etc I can use the frame (stereo window) to measure when I get to 4ft distance. When a branch is closer than the frame I know it is closer than 4ft.
Over all the camera performed well. The supplied extension lead did not work until I added two wires and connected up two more pairs of pins (thanks trypilot). I think this was just a mix up during shipping. I got the wrong cable.
The supplied lens is 3.6mm FOV and gives a very natural view especially when combined with 65mm I.A. But for my own preference I like to fly with a wider FOV eg 2.6mm or even 2.1mm. This creates a wide angle or fisheye view like a Gopro and is very useful for seeing objects in your peripheral view. It is easy to swap lenses and they only cost a few $ each.
On my transmitted video there were two horizontal bands of shaky video. The right eye has a band in the center of the frame and the left has a band hidden at the top of frame in the sky. Because it is different in each eye it has the effect of wobbling the depth. I think I saw the same thing in Trypilot's video which at the time I thought was caused by vibrations but as my test is on a tripod it has to be part of the camera or transmission system. It is present in both extended and original configurations so it is not caused by the extension cable. I would want to get this problem fixed to be fully happy with the camera.
I have made a video for you to view. If you have field sequential 3D compatible goggles then down load the original 1 GB video from here...
It was recorded live from the Vtx to a minidv tape. This format preserves the field interlacing. If you have a 3D monitor that uses polarized glasses it will also work. You just need to play it at original size and line up the movie to the screen polarizers so that it is left right correct.
I also created this side x side version for youtube. To do this I had to separate the video fields and use them to create the left and right eye views. As the fields are created 1/30th of a second apart from each other the left and right eye views are technically out of sync by 1/30th of a second. You can see this as a kind of underwater wobble effect if I move my arms horizontally of a car may look like it is floating as it drives past etc. This time mis sync is not noticeable when viewed with field sequential 3D goggles but is very noticeable when I view it on youtube.