With SprintTimer I claim that the accuracy is almost as good as a high end professional fully automatic timing equipment. This article presents some measurement I made to test this claim. Since the the most accurate method to measure a race is to to start the clock with a sound and the finish with Photo Finish, I have focused on that set up, but also tested some other variations.
The set up
To be able to consistently measure an event that starts with a sound and moves past a finish line I had to build a special device. It consist of an Arduino micro-controller, a piezo module to generate sound, a photo gate and a record player. On the record player I placed a bar that passed in front of the SprintTimer camera each lap. The photo gate registered the pass and sent a pulse to the micro-controller. On the first pass it played a short beep that was used to start the clock in SprintTimer. Six laps (around 10 s) later the finish was recorded by SprintTimer and the time was registered by the photo gate and the microcontroller. The two times were compared and normalized to 10 s.
So when the test-device was run the beep from the piezo triggered the start in SprintTimer and the camera caught the passage of the bar. The Photo finish was scrolled to the point were the bar passed and the time was noted and compared to the microcontroller.
Five different devices were tested running 10 test each.
One can say that “the newer device, the better”. On modern iPhones/iPads with a 120 fps frame rate or more, the accuracy is well within +- 0.01 s. And in most cases it is within 0.02 s. In addition 10 tests were run with Start Sender on an iPad Air detecting the sound and sending the time to the iPhone 6. Another 10 tests were conducted with the Man/Mic setting on the iPhone 6.
The average latency in Start Sender was 6 ms when running the tests and you can see that the accuracy on a good network is close to that when recording the the sound directly. The Man/Mic shows a larger spread, but is still acceptable for many races and much better than a manual start.
Finally two series of test with manual start were run. Here the accuracy is not determined by the app, but by the reaction times of the user. Still I thought it could be interesting to test if there was any difference between starting press or release of the start button.
The times were on average off by 0.1 s, which is quite normal for manual timing. The spread (at least for me) was a bit smaller when releasing the button.