The Problem Is Harder Than It Sounds
A machine taking money has to answer a question that a person answers effortlessly and unconsciously: is this a real coin, and which one is it?
It has to do so with nobody present, in a fraction of a second, for a stream of coins arriving in any order and any orientation, while distinguishing between denominations that may differ only slightly in size. It has to reject anything that is not a coin without jamming, because a jammed machine earns nothing and costs a visit. It has to keep working for years in a cold doorway with dust, damp and sticky residue getting into it. And it has to be cheap, because it is fitted to a machine selling items worth very little.
The constraints are unforgiving in a particular way. A machine that is too suspicious rejects real money, which infuriates customers and loses sales. A machine that is too trusting accepts anything, and word travels. The acceptable error rate in both directions is close to zero, and the two errors pull in opposite directions.
Every stage in the development of these devices is a response to somebody having defeated the previous stage, and the sequence is a small arms race conducted inside a slot the width of a finger.
Like our content? Follow us for more.
The First Answer Was a Series of Gates
The earliest mechanisms were entirely mechanical and worked by elimination. A coin was not identified so much as repeatedly given the opportunity to fail.
The first test was diameter. The coin passed through a slot that anything larger could not enter and fell onto a track with an opening that anything smaller would drop straight through. Two gates, and only a narrow band of diameters survived.
The second was thickness, by the same method in the other axis. The third was weight, using a pivoted arm or a cradle that would only tip at the correct load, rejecting anything too light to operate it and anything heavy enough to carry past the stop.
These are all comparisons rather than measurements, which is why they worked without any adjustment or power. They are the mechanical equivalent of a shape-sorting toy, and a correctly made coin is the only thing that fits every hole in sequence.
The arrangement had one great virtue beyond cheapness. A rejected object was never held; it simply continued falling, down a separate path, and emerged in the return cup. The reject route is gravity and nothing else, which is why it sits at the bottom and why it works even when the machine is dead.
And a Test for What the Metal Did
One further mechanical test was properly clever, and it tested a material property rather than a dimension.
A coin released onto a sloping track and allowed to roll, or to bounce, behaves according to how elastic and how dense it is. A disc of soft metal bounces poorly and loses energy; a disc of hard alloy bounces well. By arranging a short drop onto an anvil and a gap placed where a correctly behaving coin would land after bouncing, a mechanism could reject objects that were the right size and weight but the wrong material.
The same principle was used in the other direction with a magnet placed beside the coin’s path. A coin of non-magnetic alloy continued past it unaffected. A steel disc was slowed or held, and fell out of the accepted route.
That magnet is worth noticing, because it is the first appearance of the idea that eventually dominated the whole field: that the most reliable thing to test is not the coin’s shape but its substance.
And the mechanical methods still were not sufficient. The persistent weakness is that several of these tests can be satisfied simultaneously by an object which is not a coin, because diameter, thickness and mass are properties that many metal discs share. There is a well-known category of hardware that is, dimensionally, very nearly a coin, and the machines of that era had no reliable way to tell the difference.
Measuring the Metal Itself
The step that solved it uses a coil of wire carrying a rapidly alternating current, positioned beside the path the coin falls along.
When a piece of metal passes close to such a coil, small circulating currents are induced within the metal, and those currents in turn affect the coil. The coil’s electrical behaviour changes measurably as the metal goes past, and the precise way it changes depends on the metal’s properties: how well it conducts electricity, and how it responds to a magnetic field.
Those two properties are characteristic of the specific alloy. Two discs of identical size and weight made of different metals produce clearly different signatures. The machine is therefore no longer asking whether the object is the right shape. It is asking what the object is made of, and getting an answer good enough to distinguish alloys that look identical.
Using more than one coil at different frequencies gives more information still, because currents induced at a high frequency stay near the surface of the metal while lower frequencies penetrate further. A plated coin, consisting of one metal over a core of another, produces a different combination of responses from a solid coin of either, which means the machine can detect a layered structure without seeing inside anything.
And because the coin is moving, the shape of the signal over time also encodes its diameter and the speed at which it passed. A single sensor assembly delivers material composition, size and timing at once, from a coin that is still in the air.
Which Makes the Alloy a Security Feature
This is the consequence that most people have never considered. The metal a coin is made of is chosen partly so that machines can recognise it.
When a country changes the composition of a coin, every coin mechanism in the country has to be reprogrammed or replaced, because the electromagnetic signature it was looking for no longer matches what is arriving. This is a substantial and expensive operation, and it is one of the practical reasons coin specifications change rarely and with long notice.
It also works the other way. A coin can be deliberately given a composition that is awkward to reproduce cheaply, or a layered construction whose multi-frequency signature is difficult to imitate with a single homogeneous disc, precisely so that forgery becomes uneconomic. The security is in the metallurgy and it is invisible.
There is a knock-on effect that surfaces whenever currencies sit side by side. Two countries’ coins of different value but similar size and similar alloy will be confused by machines, and this has happened repeatedly and at scale. The resolution is always the same: alter a composition, or update the machines, and both are expensive.
So the design of a coin is a negotiation between a mint, a machine industry and whatever economics apply to the metal itself, and the public sees none of it.
Timing, Sequence and the Things People Tried
Two further checks exist because of behaviours the designers had to anticipate.
The first concerns anything attached to the coin. A coin that is restrained rather than allowed to fall freely passes the sensors at the wrong speed and does not travel on through the mechanism in the expected time, and modern mechanisms compare the timing of signals from sensors at different points along the path. A coin that is detected at one point and then fails to arrive at the next on schedule, or which is detected arriving and then detected leaving in the wrong direction, is treated as suspect and the transaction is not credited. Additional sensors on the return and storage paths confirm that what was accepted actually went where it was supposed to go.
The second concerns persistence. Mechanisms count rejects and anomalies, and a machine seeing a long run of unusual events can respond by refusing further input, because a legitimate customer does not produce that pattern.
It is worth being clear that this piece describes why these checks exist rather than how any of them might be circumvented. The relevant point is architectural: a modern mechanism does not make one decision about one coin. It maintains a running model of what should be happening inside itself, in sequence and in time, and anything inconsistent with that model is rejected on principle rather than on the merits.
That is a much stronger position than any single test, and it is the same shift in thinking that happened in a great many other security problems at around the same time.
The Business End: Sorting and Change
Identification is only half of it. Once a coin is accepted it has to be put somewhere, and a machine that gives change has to be able to get it back out.
Accepted coins are directed by small gates into separate tubes, one per denomination, stacked in order of arrival. Each tube feeds a dispenser at the bottom which can push out exactly one coin at a time. A machine therefore knows how much change it can give by knowing how full its tubes are, which it tracks by counting in and counting out, and which it can also sense directly.
When a tube is full, further coins of that denomination are diverted into a cash box that cannot be dispensed from. This is why a machine can be stuffed with money and still unable to give change: the dispensable part and the stored part are physically different places, and only one of them is reachable.
It is also why machines ask for exact money. The request is not a preference. It is a statement that at least one tube is empty, which the machine knows precisely and cannot do anything about until somebody visits.
And the escrow arrangement is the reason an abandoned transaction returns your money rather than keeping it. Coins inserted during a transaction are held in a temporary position rather than being committed to the tubes, so that the whole lot can be returned if the transaction fails or is cancelled. Only on completion are they released into storage. A machine that took your money and gave nothing has, usually, failed to complete that release correctly.
Why It Stopped Mattering
The coin mechanism is one of the most refined pieces of consumer engineering of the twentieth century, and it is being retired.
The reason is not that it stopped working. It is that the whole problem went away. A machine that takes a card or a phone does not need to identify a physical object at all. The question changes from what is this disc made of to does a remote system authorise this amount, and the hard mechanical and electromagnetic engineering simply becomes unnecessary. The slot, the tubes, the dispensers, the reject chute and the coil assembly all vanish together.
What is lost with them is a device that worked with no network, no power in some cases, no account and no identity, and which gave anybody with the right piece of metal in their pocket exactly the same service as anybody else. That is a real property and it was never designed in deliberately; it was simply what a coin-operated machine was.
The mechanism’s real memorial is the thing nobody notices. For about a century, a box in a doorway could reliably distinguish a real coin from a convincing disc of metal, in the dark, in the rain, with no supervision, in less time than it took the coin to fall past a sensor, and get it right so consistently that the public came to assume that recognising money was easy.
Like our content? Follow us for more.
