Average speed cameras do not measure speed. They photograph a plate at one end and again at the other and divide the distance by the time, and the clocks are locked to satellite time because the calculation has to stand up in court. Section 4 gives us a journey time of two minutes thirty four.
I am a systems engineer for a company that installs and maintains average speed enforcement through roadworks.
Why the system is built the way it is.
Instantaneous speed cameras measure a speed and can be argued with. Average speed systems do not measure speed at all. They read a plate at the entry camera and again at the exit camera, both timestamped from GPS, and divide a surveyed distance by the difference.
The distance is surveyed and certified. The clocks are locked to satellite time and drift is measured. The whole system is type approved, and the reason for every part of that is that a person may stand up in a magistrates' court and say the machine is wrong, and the machine has to be able to show that it is not.
The plate reading is a photograph, kept, with the read confidence on it.
Section 4 is a fifty limit through six point one miles of works on a stretch I will not name.
About twice a month it records a journey time of two minutes and thirty four seconds.
Six point one miles in two minutes thirty four is a hundred and forty two miles an hour.
At a hundred and forty two through a coned contraflow you would be dead, and if you were not, you would be in the incident log, because there is a traffic camera every four hundred metres and a control room watching.
There is nothing in the incident log. There never has been.
The plate.
It reads the same every time and it is the same at both ends, with a read confidence in the nineties, which is a clean read on a clean plate in good light.
It is not a valid registration. The format is right for a British plate but the age identifier is a two letter code that has never been issued, and the DVLA return is no record found.
So the system is not misreading a real vehicle: it is reading, twice, six miles apart, a plate that does not exist.
What has been checked, and my company has spent a great deal on it because a system that produces impossible readings is a system a defence solicitor can attack.
The clocks. Both cameras hold GPS lock and log their drift, and the drift over a month is measured in microseconds.
The distance. Re-surveyed in 2023 by an independent surveyor.
The cameras. Both replaced. The entry camera in 2022 and the exit in 2023, different serial numbers, the second pair from a different production batch.
The images. This is what settles what kind of problem it is.
An enforcement read keeps the photograph. Both photographs exist for every one of these.
They are photographs of the carriageway with a plate in the read window and a vehicle shape behind it, at night, in the infrared illumination the cameras use, which renders a plate bright white and everything else in grey.
You cannot tell much about a car from an average speed camera image and you are not meant to. What you can tell is that there is a vehicle there, and there is.
What happens to the record.
Nothing. It cannot become a prosecution because there is no keeper to send a notice to, and it is flagged by our own validation as an invalid plate and dropped before it ever reaches the police.
That validation is why nobody outside my company has ever heard of this. The system is doing exactly what it should: it caught something impossible and it did not send anybody a fine.
I have written it up twice as a defect. The second report, in 2023, went to our technical director, who is a good engineer, and his conclusion after a fortnight was that the reads are genuine reads of a genuine object and that no fault in the equipment could produce them.
His recommendation was that we continue to monitor.
What I do.
I maintain the section, I download the logs, and the invalid plate reads go in a folder.
The one thing I have done that is mine is that I stopped doing the exit camera at night.
Maintenance on a live section is done under a lane closure with an impact protection vehicle behind you, and it is done at night because that is when the traffic is thinnest, and I have worked hundreds of nights on gantries and cabinets on live carriageways and it does not frighten me.
The exit camera on section 4 is on a pole in the verge and the cabinet is at the bottom of it, and you sit on the crash barrier with the laptop on your knees.
In September 2023 I was there at about two in the morning with the closure up and the lane empty, and the camera fired.
They fire on a vehicle. It is a flash you feel more than see, in the infrared, and there was nothing on that carriageway in either direction for about a mile, and I was looking straight up the road when it went.
I checked the log in the van afterwards. Entry read at 01:58:41, exit read at 02:01:15, plate as usual, two minutes thirty four.