A roller brake tester spins each wheel on a pair of rollers and measures the force the brake puts back into them, and it is calibrated with a lever and a certified weight. With a car's nearside rear jacked clear of the rollers, spinning in the air, the tester records a braking effort on that wheel, and it is the right effort for that car.
I am the MOT tester who runs the brake lane, which is a pit with rollers in it.
How the roller tester works.
A car drives onto the rollers one axle at a time. The rollers spin the wheels, you press the brake, and load cells under each roller measure how hard the wheel resists, which is the braking effort. Left and right are compared for balance, front and rear are added for the total, and the number has to be a proportion of the car's weight.
The cells are calibrated with a lever arm and a certified mass. It is a direct measurement. Nothing is inferred.
A wheel off the rollers puts nothing on its cell, because nothing is pushing on it.
One car, a small hatchback that comes in every year, reads a braking effort on the nearside rear with that wheel off the rollers.
I found it by accident. The car had a slow puncture that side and I had it jacked to look at the tyre while the rear axle was still over the rollers, and I ran the rollers without thinking, and the screen showed both rears braking.
The nearside rear was in the air. I could see it. The roller under it was spinning against nothing.
The cell read one point nine kilonewtons, which is what that wheel reads when it is on the rollers and the brake is on.
What has been checked.
The load cells. Calibrated, then the nearside rear cell swapped with the offside rear. The reading moved with the wheel, not with the cell.
The rollers. Nothing was in contact with them. I have had a colleague lie in the pit with a lamp while I ran it with the wheel in the air, and he watched the roller spin free with a clear gap to the tyre, and the screen read one point nine.
The car. Any other car, jacked the same way, reads zero on the lifted wheel, as it should. This one reads its own brake.
The brake. With the wheel in the air you can turn it by hand and feel the brake bind when the pedal is pressed, which it does. What you cannot do is push on a roller through a gap of air, and the cell says something is.
The direction.
A load cell reads force in a direction. Braking effort is the wheel dragging the roller backwards, and the tester reports it as positive.
The lifted wheel's reading is positive, and it rises when the pedal is pressed and falls when it is released, on a pedal I am pressing with the car in the air, on a wheel that is turning freely and touching nothing.
I have had the tester's manufacturer look at the data log. Their engineer said the cell recorded a genuine load in the braking direction and asked what was on the roller.
What is done about it.
The car passes its brake test every year, on the rollers, with all four wheels down, with readings that are good for its age.
I have not recorded the other reading anywhere official, because there is nowhere for it. The certificate is for a test done correctly and the test done correctly is fine.
What I do.
I run the brake lane.
The one thing I have changed is that I do not jack that car over the rollers, and if it needs lifting it goes on the other ramp.
The reason is the second time, a year later, when I did it again on purpose to see whether it was still there, and it was, and I left the rollers running with the wheel in the air and the pedal off, and went round to look at the wheel.
It was turning. The rollers were spinning, the wheel was clear of them, and it was turning at the rollers' speed, driven by nothing, and when I put my hand on the tyre to stop it, it kept turning against my palm with about the force of a roller, and the cell under it, my colleague told me from the screen, read the force of my hand.