poslathian
LeafLabs
GNU-Go is surprisingly strong! I was just getting my butt kicked the other day via http://361points.com/computergo/ but sadly they seem to have disabled that site now.
GNU-Go over wifi to the gameboard is an awesome idea. 10 days until maker faire...hmmm.
gbulmer
Moderator
Well here's a small portable board which might save some effort:
http://www.sureelectronics.net/goods.php?id=1122
They seem to be out of stock, but likely some on ebay or somewhere.
Or you could make your own:
http://www.sureelectronics.net/goods.php?id=230
Building it in a small box, preferably somewhat scruffy and Rube Goldberg, would make being beaten at its hands (pins?) more terrible :-)
gbulmer
Moderator
Yes, pymite is a microcontroller version of Python:
http://code.google.com/p/python-on-a-chip/
From my perspective, its weakness is it needs a host PC.
I'd prefer a self-contained system, so a dumb terminal would be enough.
Golang is being designed by several very talented people (Ken Thompson of UNIX, regular expressions, UTF-8, ... fame, Rob Pike of UNIX, Plan 9, Squeak, ... fame, and Robert Griesemer of Java HotSpot fame). They are employed by Google to research and develop tools for, among other things, building servers.
gbulmer
Moderator
From the bin of things-that-only-work-in-your-head:
using 3 piezos on the underside of the gameboard, triangulate the location of where the human player's stone was placed based on the sound it makes when it hits the board.
I have never tried it, but I am told that accelerometers can work quite well for measuring propagation time through solids.
I was thinking of a portable, LED-based board, so that it could even be hung on a wall.
poslathian
LeafLabs
I thought about accelerometers here, but I figured the bandwidth was too slow to get an accurate triangulation via propagation time. For instance, if the speed of sound in the solid is 5Km/s and I have two sensors 1m apart, then differentiating between a strike next to one sensor versus the other should be detecting a timing difference of 5ms. most accelerometers have a bandwidth ~1KHz, so you could really only get a resolution of ~20cm using that method. Of course thats all pretty handwavy back of the envelope, ideally we need a resolution of less than 1cm.
poslathian
LeafLabs
I think thats the max sample rate, putting the bandwidth at 1.8 ... right? Also, its possible for an analog output part, while the posted bandwidth might be low, you may still be able to trigger off of fast events.
I was thinking just going piezo->schmidt trigger->external digital interrupt
but theres some analog work there that has screwed me in the past on a similar project. Namely, rectifying those piezos can be a pain. I tried just a 4 diode full wave rectifier and ended up with a 60Hz antenna.
gbulmer
Moderator
I was surprised when I was told about the accelerometers. I should see him next Wednesday (not tomorrow), so I can enquire further.
I would have done a similar estimate to you;
12,000 m/s (Diamond - we'd want it to look nice :-),
or
5.0e3 m/s is good
1cm resolution is 1 part in 5.0e5, so < 2.0e-6, or 2uSec, (i.e 500KHz)
Digikey sell some low-cost, reasonably high frequency (60KHz+) for under $2, but I don't know what I'd get for 500Khz.
I think there was something about sampling, maybe something like this, maybe in Art of Electronics? I think they used an op amp to amplify/ give low impedance output, then a diode to rectify/clamp (i.e. half-wave rectify).
Would make a nice user interface, clicking the stones down on a hardwood, or even plate glass, board.
I'd be tempted to experiment with a glass board, and a web cam underneath, if I had the spare cash and time.
Good luck, and keep us posted!
poslathian
LeafLabs
So the problem with half wave rectify is this:
I have experimentally verified that you cant guarantee the polarity of the wave front when it arrives at the sensor. This confuses me, but I'm not totally surprised. This means that If i tap the surface in the middle, the wave front might hit sensor A with a negative polarity, and sensor B with a positive polarity. Since the frequency of the wave from my finger tapping is pretty low, lets generously say its 1KHz, but i suspect it depends strongly on the surface material and is probably close to whatever the resonance for that material is, the distance between the positive and negative peaks is 1/2 wavelength, or 500us of a time lag. If you are trying to measure timing differences between the two sensors with a resolution of only 1-20us, you have to ensure you trigger off of the first peak that arrives, not the first positive peak that arrives.
Perhaps you could get around this with some smarter approach...such as measuring the time of arrival of the first two positive peaks at both sensors, and measuring the phase difference between them. Hmm...thats actually a pretty good idea - and half wave rectifiers are much easier/lower noise to make. Im gonna go try that!
poslathian
LeafLabs
grumble, that actually doesnt make any sense. You can say "lets measure the phase difference between the positive peaks received by both sensors" but you really cant say whether or not this phase difference is 'D' or 'D + pi' because you wont know whether it was one of those situations where one sensor was getting the mirror image of the first.
Maybe theres still something smarter....or, I suppose, you could do it with amplitude, but that requires looking at the signal with the ADC's and dealing with all the noise and.... pft, too hard.
gbulmer
Moderator
poslathian - as I've said before, I am not an electronics engineer, so I may be talking from the wrong end of my alimentary canal ...
Looking at both connections of an AC-coupled microphone, neither side connected to ground, wont one side always give the complementary signal to the other?
Assuming that is true, then each side can be independently fed to separate op amps and half-wave rectifiers, and each of those fed to two input pins (not one).
For a specific wave, one side will drive a positive signal into the input pin, and the other will be 'clamped' by the diode. So the software is just looking for the first pin of two being triggered by the microphone.
Am I missing something?
(PS - Maybe this discussion should be moved out of the ADC benchmarking thread?)
poslathian
LeafLabs
You might have something there. I always succumb to the urge to ground one of the leads to avoid dealing with fully differential (0 centered signals) amps. But since were trying to rectify the signal anyway, you dont have any need for a differential gain stage. Ill give it a try. Unless things progress miraculously in the next few days I dont think I'll have time to put it together for maker faire. The other projects were polishing up already had some sort of proof of concept of whatever the "hard part" was. But for the piezo triangulation, I havnt gotten anything working yet.
You must log in to post.
