camberay

Accuracy

How we test accuracy.

A figure on a spec sheet is only as good as the test behind it. Ours is a plain one: one car, read again and again — by the same phone and the same person, then by other phones and other people — and the scatter of every angle of every wheel taken apart into where it comes from.

01 — The words

Four words that are easy to mix up.

RESOLUTION

What the screen shows

The smallest step a reading is shown in. The app shows angles to two decimals of a degree. That says how fine the display is, not whether the next read agrees.

REPEATABILITY

Same car, phone and person

How far one read sits from the next when nothing changes but the repeat. It is the scatter you meet when you measure your own car again with your own phone.

REPRODUCIBILITY

Other phones, other people

How far the reads of other phones and other people sit apart on the same car. It catches what one phone's calibration or one person's habits add.

ACCURACY

Close to the true angle

Reads that agree with each other, and an average that sits on the car's true angle. The second part is called trueness, and it is found against a reference instrument.

How they differ

Resolution belongs to the display. Repeatability and reproducibility belong to the whole measurement: kit, phone, person, car and floor. A tool can show two decimals and still not repeat to the first — which is why every accuracy figure on our home page carries a note, and why this page exists.

02 — The method

One car. The kit placed once. The car read again and again.

Set-up

Once, for the whole study

The floor targets go down once, the way the field guide places them, and stay where they are until the last read. Each phone is calibrated in the app, the way a customer calibrates it.

A read

A full measurement

Each read is made in the app, step by step, as a customer makes it: the steering wheel straight and held, the floor walk, each wheel spun on the jack, the car down and its tyres let go, then the car-down photos the angles come from.

Between

The car moves, then settles

Between two reads the car is rolled back 20–30 cm and forward again, pushed by hand, or it drives a short loop out of the bay and back in, on a path kept clear of the targets. After a loop the steering wheel is centred and held again.

Then the car settles — all four wheels on the floor, the tyres let go, the brake on — and only then is it read again.

Again

Other phones, other people

The same car is read the same way with other phones, each with its own calibration, and by other people, each following the app on their own.

The study's set-up, seen from above A car in the middle of a garage bay, a wheel plate on each wheel, floor tiles around it and anchors scattered between them. A dashed outline shows the car rolled back a little; a double arrow beside it shows it rolled back and forward again. A dashed path leaves the bay through the front, goes round outside it and comes back in from behind: the short loop. The steering wheel is drawn held by two straps. Numbered marks 1 to 4 match the list beside the drawing. THE BAY, SEEN FROM ABOVE FRONT 1 2 3 4 1 m
  1. 1Floor tiles and anchors: placed once, never moved.
  2. 2Rolled back 20–30 cm, then forward again.
  3. 3Or a short loop: out of the bay and back in.
  4. 4The steering wheel: straight and held for every read.

03 — The grid

Every phone with every person, several times.

The reads fill a grid: phones across, people down, and in each cell the reads that one person made with one phone.

The same number in every cell

Every phone is used by every person, and every cell gets the same number of reads, so no phone and no person weighs more than another.

A mixed order

The reads are made in a mixed order, not cell after cell, so that a slow change in the car through the day — warmer tyres, a suspension still settling — does not land on one phone or one person.

The study's grid of phones, people and reads A grid with phones 1 to 3 across the top and persons A to C down the side. Each of the nine cells holds four rings, one ring for each read of the car by that person with that phone. The cell of phone 1 and person A is tinted: one cell gives repeatability. A dashed frame round all nine cells: the spread between cells gives reproducibility. PHONES × PEOPLE × READS PHONE 1 PHONE 2 PHONE 3 PERSON A PERSON B PERSON C
  • One ring: one read of the whole car.
  • One cell: one phone with one person. The spread of its reads is repeatability.
  • The frame: every cell. How far the cells sit apart is reproducibility. Three phones and three people are drawn.

04 — One read

Each read is a column of numbers.

One read gives camber and toe for each of the four wheels, the total toe of each axle and the thrust angle.

Each followed on its own

Every one of those numbers is followed through the whole grid on its own: the front left wheel's camber across every read, its toe across every read, and so on for each wheel and each angle.

Refusals count

A wheel the app refuses — re-measure instead of a number — is counted as a refusal, not left out.

What one read gives A car seen from above with its four wheels labelled FL, FR, RL and RR. Each wheel gives two numbers, camber and toe; the front wheels are drawn toed in, against dashed straight-ahead lines. Below: the read also gives the total toe of the front and of the rear axle, and the thrust angle. ONE READ GIVES FRONT FL camber · toe FR camber · toe RL camber · toe RR camber · toe + total toe front and rear + the thrust angle angles drawn larger than life

05 — What is reported

For every wheel and angle: where it sits, how far it scatters, and why.

Mean
The average of all reads: where the angle sits.
Spread
The standard deviation of the reads: how far a typical read sits from the mean.
Range
The largest read minus the smallest: the widest disagreement seen.
Repeatability
The spread inside one cell — one phone, one person — pooled over all the cells.
Reproducibility
The spread that changing the phone or the person adds: how far the cells' means sit apart.
Refusals
How many reads gave the wheel re-measure instead of a number.
Repeatability and reproducibility as two spreads A schematic with no values. Four rows, one for each phone and person: phone 1 with person A, phone 1 with person B, phone 2 with person A, phone 2 with person B. In each row, four rings are the reads of one angle of one wheel, close together around a tick at the row's mean; the rows' means sit a little apart. A dashed line marks the mean of all reads. Brackets show the spread inside one row (repeatability), the spread of the rows' means (reproducibility) and the range of all reads. ONE ANGLE OF ONE WHEEL, EVERY READ MEAN PHONE 1 person A PHONE 1 person B PHONE 2 person A PHONE 2 person B repeatability reproducibility range angle
A schematic, not results: the rings show the idea and carry no value. Each row is one phone with one person; each ring is one read.

The idea, in plain words

Workshops and factories test a gauge this way. The scatter of a measurement is split in two: the part that stays when the same tool in the same hands repeats it, and the part that other tools and other hands add. A small first part with a large second one points at a phone or a habit, not at the method. Both small means the phones and the people in the grid give the car the same numbers.

06 — What one car can tell you

It measures agreement. Trueness is a separate test.

The study says how well a read repeats on one car and one floor, and how much that changes from phone to phone and person to person. It does not put the car on an alignment machine: that comparison is trueness, and it is a different test.

How a measurement is made, step by step: the field guide. The figures on the home page and where they come from: note 1.