PDA

View Full Version : Help! Novice -- Wireless stereoscopy


stereofan
Jul 02, 2006, 10:46 AM
I'm facing a large learning curve since I've just recently become interested in radio and diy electronics. I'm comfortable with most of the terminology and some of the concepts, but I need help to execute my plan and I'm unsure of the best solution: I've already priced and planned a live video feedback project using a/v cabling.. the multiplexer comes out to be the most expensive part since I need to place two genlocked cameras in stereo parallax and broadcast field-sequenced ntsc (or page-flipped vga) to get a 3d signal.
I'm trying to figure out how to minimize hardware size and power demands. In considering a wireless version, I'm wondering whether -- a) in the ntsc (or pal) version, using a timing switch or alternately hopping frequencies according to the timing of even/odd field acquisition is better in terms of stability, ease of assembly and expense; and b) in using the vga mode (for which, in sufficient quantity, I can get genlockable sensors) what is the minimum amount of hardware I can get away with? Most people tell me to use a pc, but I've gotten the idea that ICs and microcontrollers (perhaps GPUs, too) are the way to go.
The range on this project is tiny -- within 10 meters, so I'm also looking at using a specification called ZigBee which is used for Wireless Personal Area Networks. The spec on this, however, is 250 kbit/sec on the 2.4 GHz band and I don't yet know the data transmission demands for stereovideo (nor have I looked at the data transmission reduction resulting from using an MPEG-4 codec).
If I hopped frequencies, I'd want to switch the receiver up and down while holding the transmission constant on two separate bands; if I use a timing switch, I suppose I have the option of powering the circuit off/down (off probably causes too much noise, down would limit transmission radius, effectively switching the received signal -- although I'm unsure of the signal graphs and timing you would have to use) or... using a capacitor-based shunt, maybe?
I'm sure that if I undertake this wireless design, there are amateur to pro considerations that I might not take into account :confused: and that I certainly don't have enough background to spec out these considerations,:eek: although I might have enough to evaluate the options once I know how to contextualize the options and/or have the facts at my fingertips.
Thanks in advance to the DIY Electronics RC Groups forum for the time and expertise provided -- I hope this discussion will prove fruitful.

Andrew McGregor
Jul 02, 2006, 12:53 PM
ZigBee isn't going to be fast enough. 250kbit/s is theoretically enough for MPEG4, but then you have to look at the protocol overhead... won't cut it.

Do you want full frame rate in both eyes? If half rate is OK, then just use a wireless video sender, genlock the cameras together and switch in the inter-frame interval (I bet that switch module is out there somewhere). If you need full rate for some reason but would be OK with reduced resolution, use a security camera multiplexer and the same wireless sender.

If you want to go full digital, you're in for a bunch of fun. You'll want some USB MPEG4 hardware encoding cameras, a gumstix linux box (google them... cool toys), and an 802.11g (2.4GHz) or 802.11a (5.2GHz) radio. That'll get you about 20Mbps, which will handle two MPEG4 streams in just about any encoding. Then receive it with a nice fast PC. Problem with that route is it's probably more expensive and you don't get the genlock.

stereofan
Jul 02, 2006, 01:47 PM
>Do you want full frame rate in both eyes? If half rate is OK, then just use a wireless >video sender, genlock the cameras together and switch in the inter-frame interval (I >bet that switch module is out there somewhere). If you need full rate for some >reason but would be OK with reduced resolution, use a security camera multiplexer >and the same wireless sender.

A switch module may be out there somewhere, but I'm not finding it. :( The ntsc 3d is 60 hz, and --near as I can tell -- the signal is composed of alternating odd and even fields, so the resolution for each eye is already reduced even though the stream itself is full frame rate. As per Zigbee, if I was using one on each camera, and each stream is alternate fields, then the resource is only operational half the time. This is why I was thinking that there might be some way to use it?? (

Most security cam multiplexers I see are designed for multiple-io instead of 2:1 and also use p-in-p, although I have seen video source switching; but how can I be sure that it switches at the field rate. I haven't tried programmable switches yet, though.

Terry S
Jul 02, 2006, 04:35 PM
I have been thinking of this for a long time, not for 3d but just to send 2 video signals from a plane with only 1 transmitter. The project is beyond me so I wish you luck, maybe you can work it out.

Terry

jeffs555
Jul 02, 2006, 05:41 PM
http://www.cs.ubc.ca/~jennings/vid_sw/html/article.html