I have been watching this thread and I need you to clarify this for me.

To simplify things lets look at the problem in one dimension instead of two. It can be easily modeled with a variable resistor. The voltage of on the wiper in respect to ground indicates where the wiper is on the resistor.

The problem is "in respect to ground". A game piece has no ground. Even if it did how are you going to get a signal back to whatever it is that will read it?

How large is the play area ?

__________________
Please post questions to the forums. PM's are for personal communication.

BCHS/3v0's Tutorials
Junebug USB PIC programmer kit., USB Bit Whacker,
The 15 Minute Printed Circuit Board! (+drill time)

 

Jason your assumptions don't make sense. Think of a resistor (as 3VO suggested) as a one dimentional device a number from 0 to infinity. If there was a simple practical solution I'm sure it would have been mentioned by now.

__________________
Bill
Smart Kits build Smart People
http://www.blueroomelectronics.com/

 

In the first idea that you mentioned, how were you thinking that the contacts on the edge of the sheet could connect with the game pieces? Please tell me if you envisioned more than one method.

In the second idea that you mentioned , would you explain the model that you had in mind in more detail. I don’t understand how the electricity could flow both up through the game piece and around the game piece at the same time – because I have been considering a game design in which the electricity has to flow through each game piece to complete a circuit.

In the third idea that you mentioned, where exactly on the game board where you thinking that it would be best to place the resistors.

What is capacitive touch?

By an array, do you mean the same thing as the grid that blueroomelectronics mentioned in the Tuesday, December 9, 2008 04:51 PM post?

If I’m correct, a Gaussian distribution is continuous. However, if I understand what you mean by array correctly, then such a design would not produce an electric output – indicative of the location of a game piece, that could be measured as having a continuous change. What type of circuit would allow me to work with a measurement of change in electricity that is continuous - and could therefore allow me to measure the location of game pieces continuously?

 

The CCD camera as stated before mounted above the playfield has my vote.

__________________
Bill
Smart Kits build Smart People
http://www.blueroomelectronics.com/

 

Right now I’m working with a model in which the game pieces are a little cumbersome because wires that are used to communicate the pieces’ properties – and that could also be used to indicate the pieces’ positions instead, connect to each game piece and run across the top of the game board to terminals. I intend to complete the circuit using a schematic similar to the one attached – which I still need to review. The signal may be varied according to an amount of resistance that I am hoping will vary somewhat consistently with respect to a distance between the game piece and each source of electricity on the game board labeled in the attachment – if and when I find the right material to make the game board out of. The play area is about two feet by two feet.

Attached Thumbnails

 

 

I may not be sure of what a resistor is. Would you explain it to me?

 

Mine too. But then I would have to connect the camera to a pattern recognition system in order to locate the positions of the pieces, and this is a little too advanced for me now.

 

I think it's pretty obvious now we had two different thoughts about how to go about things, so I'll start with yours since in theory it would work, save for some practical limitations.

Are you saying you want to turn one piece (and only one piece at any one time) into an electrode and send a current through it and measure the current arriving each of the 3 contacts?

I can see how that might work in theory, but in practicality you are going to need wires coming off of every playing piece. YOu still have to line the edge of the board with contacts so this method is no simpler than a grid method. It is probably more complicated since you now need to monitor 3 contacts with 3 very sensitive current monitors, as well as driving the current with an accurate driver (as opposed to monitoring 2 contacts with simple go/no go comparators.)

Also remember you can only inject current through one piece at any one time, or else you won't be able to differentiate between the positions of the various pieces.

You also need (fancy) math to map out the circles as well as to find the intersection. You also need math to account for how materials and measurements are never perfect so you will never get a single intersection between the 3 circles. It is much more likely that you will get 3 intersections between each pair of the three circles. These 3 points would form a triangle, somewhere within would be your playing piece, and you would need math to approximate where the best location in that area is (probably the center of the triangle, but that still takes some fairly fancy math to do).

BTW there are two ways to do that, you could use a current source and inject a known current into the board through the playing piece and measure how much current is exiting the board through each of the 3 contacts and compare that to the original amount injected. Or you could use a voltage source and measure the current exiting the board through each of the 3 contacts and then calculate the effective resistance between the playing piece and each contact. Both have the problem of non-linearity because...(see next answer)

Hypothetically, if you implenented your design you shouldn't bother with look-up tables. If your board was infinite in size you could use a single "simple" current distribution model. But it isn't. Your board is finite sized and the current will distribute differently if the current is injected into the board near the edge than if it is injected near the center. The result is that the effective resistance (if a voltage source is used to inject current into the board through the playing piece) or current at the edge contact (if a current source is used to inject current into the board through the playing piece) that is measured is non-linear with respect to the distance between contact and playing piece.

So you would have to use many different simple models for each area of the board and evaluate them all every time you get a reading or you would have to use one big complicated model. And then you would still have to follow up on the circle intersection math. Either way, this stuff is best left to MATLAB and to do it in real time requires lots of resources and time. It's much easier to map out a grid of points on the board and then place inject current into the board at every point and record the readings at the edge contacts, then put it into a look-up table. If the matrix is fine enough, then you can probably ignore the difference in current distribution effects between two adjacent points and just use interpolation to figure out where a piece is with more resolution than the matrix of test points (ie. if it's sitting between points rather than on a point). Still a lot of math and memory resources.

It's what is used on touch screens or laptop touchpads (or other things like that).
==================

Now if you wanted still want to talk about how my method worked (and I clearly explained why it does not). It's concept in one dimension is just if you have a long strip of resistive material (ie. a 1D board so to speak) and place a smaller resistive piece of material on it (ie. the playing piece) this would almost be like connecting two resistors in parallel or in series and would thus modify the resistance seen between the two contacts of the larger resistor. In 1D you can only tell if a playing piece is there or not, in 2D you would be able to tell where a piece was on the resistor.

But this is where I start with the problem of current distributing evenly throughout the smaller resistor due to surface contact and volume because of it's distance away from the sight-line between the two contacts of the larger resistor. THen there is the parallel resistance dominance effect made between the small resistor and the path of large resistor that it is in contact with (which acts like a parallel resistance connection), as well as the series resistance dominance effect (which is formed by the aformentioned parallel connection effectively being in series with the remainder of the larger resistor that is not in contact with the small resistor).

THe reason I though this was the only method was because I figured you wanted simple playing pieces with no wires or power sources in them. Either way, this method is not workable with more than one piece, and is barely workable in theory with just one piece.

 

Is it? Your original scheme had playing pieces that required engineering. What if you just made playing pieces with a small battery inside an an IR LED? And then used a camera with a filter? You end up with a pixel array with bright spots and dark spots. "Just" scan through it and record all the bright spots. Maybe some filtering too to turn the round fuzzy bright spots into a single point, but you get the idea.

As far as image processing goes, it's very lax in requirements. You don't need to compare the current frame to a previous frame so you don't need to hold entire images in memory. It's also sequential scrolling through the pixels checking each one at a time for bright/dark, and after you're done with the bit you never look at it again. In the end you just have a as many numbers as their are playing pieces, each indicating the position of the bright spots.

THe only caveat is all this implies one pixel is bright per piece and some filtering beforehand is needed to get around that which might involve holding more than one pixel's worth of data in memory which could end up being a problem. One way around this is to just skip X pixels every row and skip every Y columns of pixels. You aren't filtering so much as just reducing the resolution of the image. It would be more sensitive to "false" bright spots though, but you could always filter the recorded bright spots over time (which takes much less memory and processing) rather than filtering the image pixels spatially.

Instead of building a complicated board, you are just have a bunch of circuits that have an LED, resistor, battery, and regulator (the playing piece) which is dead easy. And the rest is connecting a camera and microcontroller with the bulk of your work being in firmware. You'd want to find the "right" camera. One that is low resolution that you can put a IR filter onto and has a simple output. The best option is the CMU Cam 3 or the POB Bot camera (you might not even need to use your own MCU for image processing since these have processors on them that you can write firmware for. They also tend to have extra memory on them so you can hold maybe an image in memory to work with rather than just pixel by pixel).

 

What type of 1D board were you talking about? Do you know where I can get such a long strip of resistive material?... I understand what you mean by saying that the math associated with some of the designs that I have mentioned so far could get very complicated, but I could also calibrate the game board according to electrical readings associated with locations on the game board after these readings were taken.

 

Were you recommending something with a surface that had specific resistive properties? I couldn’t find anything like this on the web.

 

1D was just conceptual to help explain things.

 

Thanks for your help simplifying the pattern recognition system in your earlier post. Programming such a system sounds very exciting and is something that I would like to do someday. However, I would like to keep this game as compact, lightweight, and portable as possible. It is going to be enough to have to carry around a laptop – without a camera, to make this game operational.

You said earlier that you didn’t think that two dimensional slabs with useful resistive properties were readily available. Do you know where I can find any one dimensional long strips of resistive materials?

 

If I go with the grid option, can anyone think of any problems with setting the resistance of each parallel wire in each direction with multiple wires running off of multiple locations on the coil of one of the two large variable resistors that I would have to use with this design? If you have any new ideas, please don’t hesitate to post them - because I am not conclusive about using the grid design.

 

Also, can anyone think of any three dimensional models of the flow of electricity in heterogeneous conductors?