How honest can a phone acceleration timer be?

Timing a car from rest to a hundred kilometres per hour sounds like a stopwatch problem, and it is not. The hard parts are knowing exactly when the car started moving, knowing exactly when it crossed the target, and having enough measurements in between to place both moments precisely. A phone can do this well enough to compare one run against another with confidence, and it cannot do it well enough to argue with a magazine figure. Knowing which side of that line you are on is the difference between a useful tool and a number that flatters you.

The update rate is the limiting factor

Most phones deliver a position and velocity fix once per second. Between two fixes the receiver tells you nothing, so a target crossing that happens midway has to be interpolated from the readings on either side. Under strong acceleration a car can gain seven or eight kilometres per hour in that gap, which is why a naive implementation that simply reports the timestamp of the first fix above the target can be out by several tenths. Interpolating between the two neighbouring fixes recovers most of that, and the result is repeatable to roughly a tenth of a second on a consistent surface. Some devices can be driven faster than one hertz, and where that is available the interpolation error shrinks accordingly.

Detecting the start is harder than detecting the finish

The finish is a threshold crossing on a smoothly rising curve, which is a well-behaved problem. The start is not. A satellite receiver sitting still reports small random velocities, so a timer that triggers on any non-zero speed will start early, on noise. Waiting for a clearly non-zero speed instead starts the clock after the car has already moved. The usual compromise is to detect a sustained rise above a small threshold and then work backwards to the moment the curve leaves zero, which is an estimate rather than an observation, and it is where most of the remaining error lives. It also explains the two ways a run fails to register at all: a fix that was still settling when the car moved, so the early samples were rejected, or a launch gentle enough that the curve never left the noise cleanly. Waiting for the accuracy figure to settle and starting from a genuine standstill rather than a slow creep removes both.

Rollout, and why published figures are quicker

Manufacturers and magazines usually quote times measured with rollout, a convention borrowed from drag racing in which the first foot of movement is not counted because the timing beam is placed a short distance ahead of the start line. It sounds trivial and it is worth two to three tenths on a typical car. A satellite timer starts at true zero, so it will read slower than the published number for exactly the same run. Neither is wrong; they are answering different questions, and the comparison is only meaningful if you know which convention each figure uses.

Everything else that moves the number

Surface and temperature decide how much of the first second is wheelspin. A slight downhill is worth a great deal and is not always visible from the seat, which is why serious testing is done in both directions and averaged. Fuel level, passengers and luggage change the mass; air temperature and altitude change the engine's output; a hot gearbox shifts differently from a cold one. Two runs on the same day in the same place can differ by half a second for reasons that have nothing to do with the instrument, and that variability is the honest baseline against which any tuning claim should be judged.

What the numbers are actually good for

Comparing your own car with itself. Same road, same direction, same fuel level, same tyre pressures, several runs averaged: that is a measurement with real information in it, and it will show a change after a service, a different tyre, or a season change. Cross-car comparisons are much weaker, because you are then comparing two sets of conditions as well as two cars. The other genuinely useful figures are the rolling ones, sixty to a hundred and the like, because they start from a moving baseline and therefore skip the entire start-detection problem — which makes them the most repeatable numbers a phone can give you.

Doing it safely and legally

None of this belongs on a public road. Acceleration testing means a private road, a closed circuit, a track day or an airfield, with permission, with nothing else moving, and with the phone mounted and started before you set off rather than operated while driving. The app that hosts this guide keeps personal bests for each timer so there is never a reason to look at the screen mid-run, and it labels its demo mode clearly so you can see how the whole thing behaves without moving the car at all.