Two pressure loops in the road, a light beam across the lane, and a barrier that cannot open without a payment. Between three and half past, lane four counts axles in pairs while the beam says the lane is empty.
I am the technician for the tolling system on an estuary crossing. Not the bridge. The bridge is civil engineering and somebody else's life's work. I look after the twelve metres of each lane where the money happens.
The reason I can write this down with numbers is that a toll lane is instrumented to a standard nothing else on a road is, and the reason for that is audit. Every vehicle that crosses is a transaction, every transaction is a class, and the class comes from a measurement, not from a person's opinion. So there is a lot of equipment and it all has to agree.
In each lane:
Two inductive loops cut into the tarmac, which sense metal and tell you something is there and roughly how long it is.
An axle counter, which on our lanes is a treadle: a steel plate across the lane that closes a contact under the weight of each axle. It counts axles, because axles are how a vehicle is classed and charged. Two axles is a car, more than two costs more.
A light curtain, which is a beam across the lane at about a foot off the ground with a receiver the other side. Broken beam means something is in the lane.
A camera, a barrier and a payment terminal.
Nothing moves and nothing is charged unless those agree. If the treadle says four axles and the loops say the thing is four metres long, the lane stops and a supervisor comes out, because that combination is somebody trying to get a lorry through on a car's ticket.
Lane four is the outside northbound lane, the one nearest the walkway, and it is the quietest.
Since the system was replaced in 2016 it has been logging axle events with nothing else.
The treadle closes twice, about four tenths of a second apart, which is the timing of a car's two axles at somewhere around fifty kilometres an hour. The loops do not trigger. The light curtain is not broken. The camera takes nothing because nothing asked it to. The barrier does not move and no money is owed.
Two axles, at speed, through a lane that every other instrument says is empty.
It happens between 03:00 and 03:30. Not every night. Eleven to twenty times a month, and I have eight years of it.
The interval between the two closures is the thing I keep going back to. It is not random. Across eleven hundred and some events it sits between 0.38 and 0.44 seconds, and that tightness means something: it is one wheelbase at one speed, over and over. Work it backwards at the speeds that lane sees and it is a wheelbase somewhere around two metres seven.
That is a small car. A supermini. It is also, and I only worked this out because I grew up with one, almost exactly the wheelbase of a Mark 2 Escort.
I am not putting any weight on that and I wish I had not noticed it.
What it is not, in the order I eliminated them, and some of this took years.
A faulty treadle. The treadle has been replaced three times, twice by me and once by the manufacturer under warranty, and the loom has been replaced, and the input card has been replaced, and the whole lane controller has been swapped with lane three's as a straight A B test.
It followed the lane, not the hardware. Lane three, running lane four's controller, was clean. Lane four, running lane three's, carried on counting.
Electrical interference. This was the serious candidate and it was investigated properly, not by me. The contractor put a scope on the treadle input for six weeks. An interference event looks like a spike with no mechanical signature. What the scope recorded on those nights was the contact closing and opening the way it does under a weight, with the bounce and the settle you get from a steel plate being stood on.
Something actually there that the other sensors miss. This is the one I would believe if the loops were the only other sensor, because a loop needs metal and a loop can miss a small low car on a bad day. But the light curtain is a beam of light and a bicycle breaks it, a fox breaks it, and a carrier bag blowing through at knee height breaks it. It is the most easily broken thing in the lane and it has never once been broken by this.
And the walkway. There is a pedestrian and cycle walkway on that side behind a barrier, and a cyclist on the treadle would be two axles about a metre apart, not two metres seven, and would break the beam.
The part I have not put in a report.
In 2021 we had the deck resurfaced on that carriageway and lane four was closed for nine nights. Closed properly: cones, a crash barrier across the entrance, men and a planer in it, lit up like a stadium.
I pulled the logs afterwards because I was curious whether the planer had been setting it off.
On seven of those nine nights, between three and half past, the treadle in lane four recorded two closures, 0.41 seconds apart.
There were men standing in that lane. I have spoken to the foreman, carefully, years later, in a way I thought was casual and probably was not, and asked whether anything odd had happened on that job.
He said it was the worst nine nights of his working life and he would not do it again, and when I asked what he meant he said that you get so you do not want to be the one nearest the water, and then he changed the subject, and I did not push it, because I did not want the answer enough to make a man say it out loud.
What I have done, and it is the same thing everybody in this story does.
There is a rule in the lane controller's configuration that suppresses an axle event on lane four when no loop and no curtain event accompanies it within two seconds. I wrote it in 2018. It is documented as phantom axle suppression and it is the correct fix for an axle event with no vehicle, and it has been reviewed and signed off and nobody has ever queried it.
It means the lane no longer alarms, the night supervisor no longer gets called out, and the audit figures are clean, which they always were, because nothing was ever charged.
The events still go into the raw log. I have not suppressed that and I am not going to.
I am due to retire in four years and I have already decided I am not going to tell whoever replaces me what that rule is for. I will show them how it works, because they will need to, and if they ask why it is there I will say what the document says, which is that the treadle on that lane is sensitive.
And if they are any good they will do what I did, which is read the raw log one quiet night and find about fifteen lines a month saying that something the size of a small old car came up the outside lane at three in the morning and crossed the bridge without owing anybody a penny.