[jakecee (OP)]

i am currently working on my final year project for diploma in electrical and electronics engineering. the project that im working on is about a coin counting jar, which means when coin is inserted into the coin slot, a 16x2 LCD screen will display the value of the coin inserted and the sum of all the coins. at the same time, if an amusement park token or counterfeit coin is inserted, the machine is able to reject it. the PIC that i have selected is 18F4550 because it has been provided in PIC workshop before. however, the method to detect counterfeit coin is still unsure. i tried to use FSR (force sensitive resistor) but it is not sensitive enough. is there any other component, like a strain gauge or load cell or other alternatives that can measure grams, to get the weight of the coin? or other methods to detect counterfeit coins? please help me out!!

[imatfaal]

i am currently working on my final year project for diploma in electrical and electronics engineering. the project that im working on is about a coin counting jar, which means when coin is inserted into the coin slot, a 16x2 LCD screen will display the value of the coin inserted and the sum of all the coins. at the same time, if an amusement park token or counterfeit coin is inserted, the machine is able to reject it. the PIC that i have selected is 18F4550 because it has been provided in PIC workshop before. however, the method to detect counterfeit coin is still unsure. i tried to use FSR (force sensitive resistor) but it is not sensitive enough. is there any other component, like a strain gauge or load cell or other alternatives that can measure grams, to get the weight of the coin? or other methods to detect counterfeit coins? please help me out!!

Hi Jake

I wouldn't have a clue about designing a circuit to identify a counterfeit coin.  I will however warn you that many of the responses you might receive to messages such as this will assume that, in fact, you want to make a counterfeit coin rather than spot one.  It is a common meme on the message boards for questions to be made about prevention of fraud when the reality is that the poster actually wants to know about circumventing methods of detecting frauds. 

Just a quick warning so that you dont feel put upon and disheartened by strange answers.

[CliffordK]

I would wonder if digital analysis technology is getting to the point where it could recognize different series of coins, even with an amount of dirt and corrosion on the coin.  And have the technology miniaturized. 

In the USA, there are both Zinc pennies and copper pennies in circulation which I would assume are slightly different in weight.  Over the last decade several new series of nickles and quarters were released. 

And, of course, there is the problem that Canadian coins closely resemble US coins.

[graham.d]

Mars Money Systems used to be an important company making such systems. There are several techniques used, often together, to detect counterfeit coins and to recognise the value of genuine ones. Here are some:

Select coins according to size
Measure the mass for a particular size (sometimes this is done via what could be described as a large mass spectrometer) - this checks the density matches with the material the coin is supposed tobe made of.
Measure the magnetic properties of the coins
Check for the validity of writing on the coin and correct serrations on the edge

There may be more. Of course the sophistication will depend on the cost of the system and whether this is necessary depends on the likelyhood of fraud - a certain level may be acceptable against the cost of the anti-fraud system. Also, the more complicated systems may start to have an unacceptable false detection rate.

[jakecee (OP)]

imatfaal:
yes i agree on the part where i'll be making a fake coin rather then spotting one. which is why i chose an amuse park token. or maybe just a cardboard or smtg.

CliffordK:
for my project, other currencies are not taken into consideration. im only fosucing on malaysian coins - 10, 20 and 50 cents.

graham.d:
yes. indeed. i have included a coin sorter to sort out the coins inserted according to their diameter using the gravitational-based coin sorter. and yes ive read about measuring the magnetic properties of the coin. however, what are the sensors needed to do the job? i couldnt seem to get much info on that. not even the sensors to detect weight. cause all i can get are mainly on load cells, which is too expensive for my budget. what do u think about a proximity sensor?

i really need to settle down with a method to differentiate real coins from the fake ones and work on it asap. my deadline is drawing near

[SeanB]

Most common is to use a series of coils, driven at various frequencies, so as to detect the physical properties of the coin. A small low frequency coil detects the bulk material by the change of resonant frequency of the coil as the coin passes. This tells you if the coin is steel, nickel, brass or copper. It is the first coil so you can then use the following coils to further refine.

The second coil has a ferrite core that is larger than the biggest coin, and gives a signal that will allow you to see how big the coin is, by the amount of energy absorbed by each coin. A third coil is driven at a high frequency, and this is high enough to only be affected by the skin effect from the coin, giving info as to the engraving on the coin, or to detect the outer material in a bimetal coin. If you add a coil normal to the coin edge it will be able to tell you if the edges are milled or smooth.

To detect mass you can use the simple method of allowing the coin to fall and have it's fall stopped by a rubber mat, then roll off the mat onto a short inclined slope before being allowed to fall freely under gravity. This then will sort the coins according to mass, as you have a constant horizontal impulse, with the mass and gravity determining the path the coin takes, so you just need to detect the place where the coin lands to give a rough mass measurement.

All this has been around for many years as the coin mechanism in pay phones. Have a look at one to see the basics, most are quite easy to dissect to see how the mechanisms work. Make sure you do it to a phone that you own if you want to do destructive work on it.

[CliffordK]

Sean,
Most interesting description,  So, a lot can be done with magnetic properties, resistance, and induction. 

It sounds like size sorting is easy.  But, could also be accomplished by using a slide slot like many washing machines, pool tables, and some video games use.

I find it very annoying when soda machines falsely reject dollar bills, as well as coins.

[jakecee (OP)]

thanks sean! based on what u said, are u talking about a solenoid?

[SeanB]

No, a wound coil with a ferrite core, arranged so the coin in passing past the pole pieces form a part of the magnetic circuit. This allows the coin to affect the coil by changing the inductance, or by absorbing energy from a LC oscillator using the coil. You get both a frequency change and an amplitude change, both of which give information as to the material in the coin. You have to either use the coils in a free running oscillator and measure both the frequency and output amplitude, easy enough with a microprocessor, or even using some op-amps and simple diode peak detectors to do simple detection. Otherwise drive the coil through a high impedance and detect the output voltage drop caused by the coin. You might also have 2 coils arranged so the coin passes through the gap between them, the output signal giving information as to the composition and loss of the coin.

The analogue methods are very old, having being in use in payphones for many years, whilst the digital methods are more common in vending machines.

[MikeS]

Older pay phones compared coin diameter, thickness and weight to known parameters by using individual coin slots and chutes.  Everything else was rejected.  They were remarkably efficient.

[CliffordK]

In Italy, they standardized the pay phone coinage to a single coin called gettone. 



The three slots would naturally exclude other coins.  I suppose you could counterfeit them, but it would be a big hassle for little benefit

Whenever I wanted to call home, I'd fill the phone up with as many gettone as it would take (a dozen or so).  Then during the call I would feed them into the phone as fast as I could, and I could generally talk for a minute or so before it would cut me off.

[jakecee (OP)]

sean, then what type of sensor should i use to measure the changes in inductance? is it advisable to change my method to the one ure saying now or should i stick to force sensitive resistor?

[SeanB]

Using a mass measure to do this is going to lead you into a world of hurt, high noise level, having large shock loads from the coin hitting the balance, settling times and such. Add to that the problems of really good non monotonic 18 bit or better ADC's and the terrible tempco of most strain gauges and i would suggest the alternative methods.

Measuring inductance change is easy, use either a clapp or harley oscillator and use a microcontroller frequency counter to do the work. You can simply measure the current in the oscillator to detect the loss in the coin. The coil is simply wound in a former with the coin path as part of the magnetic circuit. Depending on the frequency and the area of the coin path you can get info about the coin. A large pole allows you to detect different sizes, as the smaller coins of the same material absorb less energy. A small coil allows you to detect composition by loss in the circuit.

You would have your coin mechanism do basic sorting, to not accept thick coins or to drop out small coins, as well as to provide a constant speed past the various detectors.

Build up a gentle slope allowing coins to run down, make a detector or two of various sizes and see what you get out for assorted coins. This will give a guide as to selection of frequency and the magnetics. Not an arcane art, but just get the oscillators to start reliably every time, and then see what they do when a coin runs past. Try a few washers as well to see the required conditions to reject a coin.

[jakecee (OP)]

sean, yes i have thought about a gravitational-based coin sorter. in which coins of different diameters will fall into different channels made of perspex (transparent plastic). judging by the way u said just now, and since i'll be detecting 10, 20 and 50 cents, are u suggesting that i use 3 wound coils at the end of all the 3 channels so when the coin falls down it will cut the magnetic flux lines and change the inductance?

[Geezer]

Would it work if you just analysed the charcateristics of the sound the coins make when they drop on to a hard surface? You might grab a piece of the sound then do a FFT on it. The frequency spectrum might turn out to be a pretty good "fingerprint".

[RD]

Would it work if you just analysed the charcateristics of the sound the coins make when they drop on to a hard surface? You might grab a piece of the sound then do a FFT on it. The frequency spectrum might turn out to be a pretty good "fingerprint".

You might have something there ...

 [ Invalid Attachment ]
http://www.freesound.org/people/Jace/sounds/17502/


But if someone kicked the coin box (with some coins in it) it would produce a true coin noise which could fool your detector.

[CliffordK]

But if someone kicked the coin box (with some coins in it) it would produce a true coin noise which could fool your detector.

Or, had a speaker that one could reproduce the sound of the dropping coins plus the inverse of the distortions getting to the recording device.

I suppose it is a bit different, but weren't there rumors that pay phones used to send tones over the phone lines to confirm payment, and some people could reproduce the tones to indicate that a payment had been made when in fact nothing had been paid.

[RD]

... and some people could reproduce the tones to indicate that a payment had been made when in fact nothing had been paid.

That sounds like "phone phreaking" ...

Joe Engressia, a blind seven-year old boy. Engressia was skilled with perfect pitch, and discovered that whistling the fourth E above middle C (a frequency of 2600 Hz) would stop a dialed phone recording. Unaware of what he had done, Engressia called the phone company and asked why the recordings had stopped. This was the beginning of his love of exploring the telephone system.

Other early phreaks, such as "Bill from New York", began to develop a rudimentary understanding of how phone networks worked. Bill discovered that a recorder he owned could also play the tone at 2600 Hz with the same effect. John Draper discovered through his friendship with Engressia that the free whistles given out in Cap'n Crunch cereal boxes also produced a 2600 Hz tone when blown (providing his nickname, "Captain Crunch"). This allowed control of phone systems that worked on single frequency (SF) controls. One could sound a long whistle to reset the line, followed by groups of whistles (a short tone for a "1", two for a "2", etc.) to dial numbers.

http://en.wikipedia.org/wiki/Phone_phreaking

[RD]

A better spectrogram of the coin dropping noise: clearly showing three resonant frequencies.
The ones at ~15KHz and ~21KHz will be inaudible by most people, (as will the faint artifact tone at 18KHz),
Spectrogram produced by SoX (freeware).

 [ Invalid Attachment ]
http://www.freesound.org/people/Jace/sounds/17502/

[techmind]

I've not looked closely at coin-sorters, but have long-assumed that these days they rely largely on electromagnetic/inductive properties - which is largely consistent with what SeanB says. By using different frequencies, it will be possible to establish a signature for the metal/alloy and as SeanB has suggested, concievably you could generate some electronic 'measure' of the overall size too.

Turning that concept into a working practical demonstrator is not something I'd want to have to do in a very short amount of time (both the analog driving/detection of the coils and the required signal processing is probably a bit specialist). With two frequencies (probably one around 50-100Hz, and another in the high 10's kHz - either in two coils, or one coil/pair which alternates between the two frequencies) you should be able to distinguish between magnetic and non-magnetic coins relatively easily... you might even make a crude judgement of size - which might be good enough for your project.

If you were doing a real sorter for the real world, you might look at employing a dsPIC which can do much more hardcore signal processing.


Magnetic/inductive techniques basically just check the alloy and overall size/weight. I'm not aware they could check anything regarding the embossing/marking/knurlings.
The advantage of these techniques are that they don't involve delicate mechanics (which need oiling/maintanence) and they're fairly robust against abuse/vandalism/gunk/pocket-lint in-the-works.

I'm doubtful that the acoustic method mentioned will work very well in practice.