The system components are:
The sensor is contained in a box with a vertical slit opening on one side. It stands on a tripod. The height is 205 mm. In operation the box is placed on the ground with the slit facing the circle at a 21.5 meter radius (for the case of F2A, with 17.69 m lines). At this distance the box will stay clear of starting and landing models, but note that it will still need protection from dollies.
The sensor employs a phototransistor for detecting the passing of a model by optical means. It is mounted on a printed circuit board that also has some important signal conditioning circuitry.
The unit should be aligned so that it stands horizontally, pointing towards the pylon.
When a model enters the field of view, it obscures a small part of the sky, and the phototransistor current drops momentarily by a tiny fraction. The circuitry detects this drop and outputs a signal pulse, and blinks a red light on the rear panel.
From the model dimensions, its range of flying heights, and the sensor field of view as above, the photocurrent drop can be estimated. For an F2A model it is around 0.5 %, and rather independent of the flying height: Low: Near sensor, but wing seen from oblique angle. High: Far from sensor, but wing seen at nearly right angle. To detect a current drop as small as fractions of a percent, some special techniques have to be employed to adapt to the lighting conditions and exclude light variations from other sources than the models. Think about this: The sensor must be able to detect a light change of only 0.5 %. At the same time, the brightness of the sky could vary by a factor of ten! In all a factor of 2000!
There has to be a lower limit of the field of view, as objects near the ground must not disturb. The pilot, mainly (!!) but also background trees, circle fence and so on. This leads to models passing lower than around 0.6 meters (2 ft) being missed.
Two more obvious limitations are that no direct sunlight must fall onto the phototransistor, and that the sky in view must be unobstructed. No tall trees or buildings in the rear! To block stray light, the box inside surface is covered by black velvet-like material. A minimum angle of 10° between the sun and the sensor direction is recommended. Also, the sensor should not be placed in the shade from the sun by trees, flags or other objects, as this could produce fast variations in stray light.
The phototransistor current connects to an amplifier, designed to give an
output proportional to the percentage change rather than the current
itself. This way the operation becomes independent of the sky brightness.
The second stage of the electronics is an amplifier with a bandpass filter, that blocks slow variations, as for instance from clouds, as well as fast variations, from internal noise.
The third stage is a threshold circuit, which outputs a pulse when the output from the filter stage indicates a sudden drop of light larger than a certain percentage.
By these three stages, the sensor is matched to objects with the typical speeds and sizes of C/L models.
The last stage is a buffer capable of driving a long cable that connects the sensor to a PC.
The unit is powered by a small 9 V battery. An alkaline battery will give over 100 hours of operation. There is also a battery sensor, sending a warning if the battery runs low.
The F2A rules have been amended to detail the use of electronic timing in official capacity.
The program can handle not just F2A but also other categories, which can be defined by the user. There are half a dozen other categories prefefined in the software package, and it is relatively simple to add a category.
The program is able to correct for a missed lap (due to underflying) inside the laps 0 to 9. It also handles sorting out of stray pulses due to, for instance, birds, butterflies or (heaven forbid) R/C models passing.
To measure the timing errors of the sensor itself, an even more accurate
system must be at hand, and none is. Instead, some reasoning could be
done: For F2A models flying at 80 m/s with a fuselage length of 0.4 m, it
takes 4.8 ms from it enters the field of view until it is entirely
visible. Somewhere in this interval the sensor output pulse is produced,
meaning that the difference in time between the actual entry and signal
output can hardly become bigger than this. As the speed is determined by
the difference of two measurements, most timing errors will cancel. Any
net timing error is due to differing conditions for the start and stop
registrations, which could be due to flying height differences and
lighting changes. It is a fair assumption that the timing error is
well below 1 ms.
At the 2004 World Championships, where two independent systems were used, the time difference was typically less than four units in the fourth decimal, that is 400 microseconds.
Most of the mentioned errors do not accumulate lap by lap, but have the same value for the total of 9 laps as for one lap. The exception is the PC crystal controlled timer error, and this contributes an error well below 1 ms in 12 seconds, especially if calibrated.
Knowing that the separations of the stopwatch times of good manual timekeepers can be 0.03 seconds, and that sometimes they are much more, we can project this system to be 20 - 100 times more accurate!
The indicated timekeeper uncertainty of 0.03 s corresponds to 0.75 km/h at 300 km/h. As the separation of places at championship events is typically less than that, it is obvious that manual timekeeping is inadequate.
For contest use, there must be a few extra rules agreed upon, using the Sporting Code provision for "local rules". The helpers must not walk in front of the sensor, and so on. The handling of missed laps when they happen for lap 0 or 9 must also be defined. Either use the manual backup, or calculate the speed for laps 1-9 or 0-8, or, more radically, simply cancel the flight. After all, the manual timekeepers will also have a visibility problem if the model passes the timing marker at a height below 0.6 m! (Note that the rules prescribe cancellation of the flight if flying below 1 m for a whole lap.)
First test was on May 1st, with sensor alone, no PC connected, just a light emitting diode to indicate the output signal. It worked right away for F2C models. First test on F2A models with a PC connected was two weeks later, and the second in the Limfjords World Cup competition in Aalborg, Denmark, June 9-10. Then it was tested at the World Championships in Landres, where lots of flight data was recorded.
A printed circuit board was designed already then, and drawings for a box. I gave this to a few sheet metal companies, and got a decent offer from one leading to an order. Then nothing happened, the company was too busy with bigger jobs! Only after actually ordering from two more sources, after ~6 more requests, I finally got the boxes made, in the spring of 2004!
I'm indebted to Pete Soule for providing the design idea to get rid of the lens and the array of phototransistors I had in mind, and use just a single phototransistor and a slit.