Speed (comparative testing)
Validating the speed on a moving vehicle whilst on the track is difficult as there is no obvious “reference” to compare against, GPS already represents the state-of-the-art with respect to high accuracy speed measurements – so for this test it is really a comparison against competitor systems rather than a true “validation” of accuracy. It is also worth noting that on the day of the test, the track was damp in places (although not wet, and certainly no standing water) which is known to cause problems to optical speed systems – we would have expected better results from the optical system on a dry track.
Direct comparison of measured speeds
The speed accuracy from all systems was good, and looking at the data on a large scale the results are overlaid almost line-on-line, the only apparent difference is the occasional dropout from the optical system, apparently caused by light water spray from the track.
On a closer scale the difference between the 5Hz and the 20Hz sample rates becomes apparent, although the accuracy of the 5Hz system is still excellent at update times.
Examining the data in detail, the slightly higher noise of the VBOX3 becomes apparent, as well as the small latency in the optical sensor output. This final graph is measured over a moderate acceleration rate of 0.6g.
Distribution of differences in measured speeds
To more fully compare the outputs of the different systems, the differences in measured speeds have been plotted as a frequency chart. The results are from an extended 20 minute run that contained some steady state data (low and high speed) as well as a great deal of transient data. Clearly the 20Hz DL2 and the VBOX are in exceptionally close agreement with a mean of 0.001kph and a standard deviation of just 0.036kph. Clearly this is only a comparison between the VBOX and the DL2, looking at this result in isolation there is no way to tell which system had the lower error, we can only summarise the speed error from both systems was very low under all the static and dynamic conditions.
The 5Hz DL2 also performed well, with a very low mean error and a standard deviation of under 0.1kph. Notably both the mean and standard deviation of the optical system were significantly higher.
Differences in measured speed time-alignment
Much is made of the latency of GPS speed measurements, and this has been a source of significant error in older GPS systems. The DL2 is a post processed system, so there is no live output of speed on the box – however it is still essential that the GPS data is correctly aligned with the other inputs (wheel speeds for example). To test the time alignment of the two systems, calculated speeds from both the VBOX and the DL2 were “cross-correlated”. The time alignment of the recorded speeds from the 20Hz DL2 and the VBOX was found to be around 4ms.
The reason for this small time discrepancy is that the VBOX3 show the average speed over the previous 10ms, so the centre point of the sample is actually 5ms before the provided data. Because the DL2’s speed solution is post processed, no such error exists. However, a time alignment error of 4ms - 5ms is small enough to be insignificant for vehicle testing applications.