Ball park coin resistance

[toster]

Yeah - I thought it was a great idea too. But it turns out to be more complicated than it sounds.

First off arduino can't act as an USB host by itself - so a webcam doesn't work out of the box.
Even if I rip the webcam apart and connect directly to the device, the arduino doesn't have enough mem to store the image.
It's impossible to analyze the data as it flows by, the arduino is too slow to even read the data real-time.
This means saving it to some other device, then loading and processing...

So yeah. Need something more powerful than an arduino for that.

[4to20Milliamps]

Yeah - I thought it was a great idea too. But it turns out to be more complicated than it sounds.

First off arduino can't act as an USB host by itself - so a webcam doesn't work out of the box.
Even if I rip the webcam apart and connect directly to the device, the arduino doesn't have enough mem to store the image.
It's impossible to analyze the data as it flows by, the arduino is too slow to even read the data real-time.
This means saving it to some other device, then loading and processing...

So yeah. Need something more powerful than an arduino for that.

I don't think the arduino is the problem:

http://www.arducam.com/tag/arduino-camera-shield/

http://nootropicdesign.com/projectlab/2011/03/20/arduino-computer-vision/

[toster]

Oh, they actually have a shield...
The arduino still can't load that much info but they've taken care of the pain of saving that all to an SD card.

"Due to the speed limitation of AVR series MCU, and all the Read and Write access to LCD GRAM is emulated by IO operation, saving a 320×240 RGB555 BMP into SD/TF card will take up 10 seconds."

So.. 10 seconds to save the image to SD. And after that I have to read it in w/ the arduino. And then process it without reading the whole thing. Just driving the file pointer around and reading it byte by byte.

I suppose it could work but that's like 30 seconds per coin or sth :/

Hum.. the other link looks promising. /me reads.

[Chet T16]

I've a counting money box here all all it does is measure the diameter of any coin put in with a sliding arm mechanism.

[toster]

Hum.. could you explain it in more detail? How do you go about measuring the diameter?

[Simon]

Like i said it will have a disk or arm with holes in it for each coin to slip into, it will then be carried to the correct sorting section

[toster]

Yeah I thought so too but he did say "measure the diameter of any coin". Hope. It's still not dead.

[Simon]

I think he meant any withing the context of his own currency,

You could have something along the lnes of a slider pot that moves based on the diameter of a coin, motor "crushes" coild between two surfaces  and when it stops the resistance is taken

[toster]

Yep. That would work. But I still don't know neither the range the resistance could be in nor the resolution I'd need. This discussion has gone a long way off the original question

[Simon]

not really, you want to reinvent the wheel and we are trying to help, at the end of the day unless you use something like the sparkfun unit you have a lot of mechanics to come up with, never mind the electronics

[toster]

As I said - I don't mind doing that and your help is highly appreciated.

[JoeyP]

As a learning experience (only), I encourage this exercise. The oxidation issue could be somewhat mitigated by using kelvin clamps. Google "4-terminal resistance measurement" for more info. The problem though, is that the resistance will be very very low, so you will need a large constant current source and a very sensitive voltmeter. Because the voltage will be so small, thermocouple effects will come into play as well. As an example, copper vs copper oxide has a Seebeck coefficient of about 1mV/degree. It will be more about the analog circuit design than the processing.

[Simon]

I've dabled down there with a current regulator and a very low shunt resistor to mitigate power dissipation, the voltages get so low that expert analogue desgin skills are required.

[JoeyP]

Agreed, which is why I encourage it. This is the sort of exercise that helps beginners grow into experts

[sleemanj]

[JVR]

Built one like this a few years ago, worked epically fine.

Other option is a arm connected to a pot, and a weak spring, as the coin goes under the arm, measure the peak resistance of the pot, and you know the diameter.

[toster]

Other option is a arm connected to a pot, and a weak spring, as the coin goes under the arm, measure the peak resistance of the pot, and you know the diameter.

O_O
Now that's what I call elegant! I must try this out. Like. Now!
My hat's off to you sir!

[c4757p]

Thanks, JVR, for pointing out the opto solution. I hoped someone would. The camera idea is silly when you can just align one edge of the coin with gravity and use a couple phototransistors to find the edge. I like the pot idea in theory, but it may be hard to find one that is loose enough that it won't just stay put and trap the coins.

[JVR]

I like the pot idea in theory, but it may be hard to find one that is loose enough that it won't just stay put and trap the coins.

Yeah, it would need a bit of work, bit it would be infinitely more configurable that the opto's.  Large slope would help, or even a modded servo to "catch" the coin and push it past the arm.

[Galaxyrise]

Yeah, it would need a bit of work, bit it would be infinitely more configurable that the opto's.  Large slope would help, or even a modded servo to "catch" the coin and push it past the arm.

There's other complexities too, like the arm will have momentum, and pushing the coin in faster will result in a higher reading.

I gravitated towards the optical solution.  I had to sleep on it, but I think I came up with how to implement the optical solution such that it can read a very large variety of heights with just a few beams:

The idea is to point one of the opto beams vertically and then time when it's broken relative to the horizontal beams.  By computing their ratio rather that dealing with absolute times, it shouldn't matter what speed the coins are inserted at.  Backspin could still matter if the horizontal window is large enough.  It'll be a bit tougher to build to get that vertical beam so it doesn't cause coins to get stuck; either it'll have to go a bit horizontally as well or you'd have to put something clear into the track for it to shine through.

[Rerouter]

if you want to avoid coins being forced in, use a small vertical shaft, say 2-3 diameters at the coin slot, that way every coin will loose most of its horizontal momentum and run your timing arrangement from the same starting speed,

[cwalex]

maybe a way to stop the problem with people pushing the coins in with too much momentum you could arrange the slots inside so that if the coin is moving too fast it will just keep going past the point where it will drop into your measurement section and come out the place where change would normally come out. Then they can have another go at putting the coin in. I'm not good at drawing so I hope my description is easy enough to understand.