From: KJ Sullivan Subject: Re: TECH: Seeking info on optical tracking Organization: Solbourne Computer, Inc., Longmont, CO USA Glen wrote: >I am interested in making use of a ccd and diodes to track a headset. >From what I have seen in this newsgroup, some sort of fancy >"defocusing" technique is used to increaase the resolution of the ccd. >Sorry if this is a frequently asked question, but does anyone have >information on this technique? Any papers, journal articles? I have a copy of a patent which I helped develop the hardware for. I will dig it out and provide you with a patent #. We used a United Technologies Trans-Impedance AMplifier instead of a CCD. A trans-impedance amplifier is a X-Y array which converts a spot of light into a proportional current across 4 sinks (+/-X, +/-Y) We were able to drive a low noise op-amp to produce a .01 to 10VDC signal. The noise was kept below the 10mv bias. We then ran this into a Data Translations 12bit A/D convertor, and translated the data into viewer space. We used visible blocking filters on the detectors and user-mounted IR LEDs. To get the signal levels we needed, we had to use 50W IR leds!! Expensive little buggers! We thought about regular light bulbs but the slew/decay rates were too slow. This was 4 years ago, I'm sure that there are better gadgets available now. Well, I designed and built the detectors, we used 24mm C-mount video lenses and 'defocused' them slightly to stablize our readings. Too focused, and the data would bounce around too much. This was working against some software used to 'toss' out eratic readings. Later refinements in the op-amp design reduced the noise and increased signal stability. I also helped design and build the LED sequencer. The sequencer was controlled by the PC serial port and was basically a 1-out-of-16 selector which drove open-collector 15VDC drivers. I connected the drivers to the LED array via phone cable and RJ-45 connectors. We were pulsing the LEDs sequentially with 15VDC for about 200ms each. By knowing which LED was on at a given moment, the geometric configuration of the LED array, we were able to display a real-time image of the user's location and orientation to within 0.1 millimeter and 1/x degree. We were using a Dell PC 286/20 w/FPU. 'C' was used to develop the software. >From this experience, I built a head-mounted setup which has LEDS arranged in a pyramidal shape. I intend for the detector to mounted overhead and communicate with the LEDs via a modified TV IR remote chip set. The standard TV IR remote chip set was a bandwidth of 100KHz, I could divide that into 4 channels, thus providing 2 audio channels for stereo input, 1 channel for voice control, 1 channel to tell which LED to turn on. The detector would use the same Trans-impedance/op-amp setup and use a 12mm red marble for the lense. A quality lense is not desired. The user would wear this headset with a fanny pack of professional Video rechargeable batteries providing the current for the IR LEDs, microcontroller, and headphones, thus free of an umbilical cord. I have modified a pair of SEGA LCD glasses and mounted them into a nice pair of SCOTT ski goggles, this arrangement would require the user to watch the stereo display on a large screen TV. In the original project, all 6 axis were important, so we used 2 detectors arranged horizontally 90 degrees from each other. This was to prevent 'losing' sight of important data points. With the single overhead mounted detector you would lose accurate 'Z' data. But then, I don't care if the user jumps or kneels. Hope this is enough to get you started, I'm the only one working on this new project now, and don't have the funds to continue or the time (gotta earn a living) to program the 8097 and 8031 controllers to get everything talking. I will be happy to answer questions if you have any. The bottom line is "I know that it will work, I did it already". The main advantage of IR detectors and LEDs obiviously is that you can do it in normal room light, although dim is better. Florescent lights give out a lot of IR, but not as much as incandescent lights. Sincerely, KJ Sullivan kj@solbourne.com -- KJ Sullivan + . . . . . . . Solbourne Computer Support + : : : : : : : kj@solbourne.com 800-447-2861 + 2.5 2.5 2.5 2.5 2.5 2.5 2.5 Cross-eyed Stereo Signature + D D D D D D D