Accuracy tests between Trakmate, Chronocone and EttSexEtt

[Marcus Seyffarth]

With the 'new' computer driven timings systems at hand there is no doubt that competitions could run smoother, results can get out faster and errors are less prone to affect the final standings. There wont be any mixing of lanes, typos when moving times from timer to spreadsheet and such.

One question still stand. Are they accurate?

Since no hardware specialized computer geek could give me the answer the only way to find out was to do some testing!

So I downloaded Chronocone 4.x. Bought a joypad and smashed it to pieces, borrowed Jani's TrackMate and went to the electronics store and bought a super fast relay so that I could connect all systems to the same start and finish without mechanical interference.

When all things were connected in my personal little timing lab I switched on the computer and started noting times. I used my PIII 1Ghz computer to run both systems at the same time, and the Trakmate that is specially designed for measuring times as a benchmark.

I did 50+ tests, most in the range 10-15 seconds since this is the time a skateboard heat normally takes. I also did a few at each second mark to see if the errors where linear or if they occurred at start or finish. Note that I have only used hundreds of seconds on all systems at the comparisons. So the fluctuations in the errors become larger at lower times, and this is due to truncations of the times.

What I found out was that both computer driven timing systems did lag about 0.4%. This can of course be due to my processor being slower than it should or something similar. The INTERESTING part though is that the error (given the Trakmate is exact) is linear and pretty much the same all the way. And as long as you use the same system for a whole race, for both lanes etc the times and results will be just fine. Though maybe not exact for a world record or such, but exact enough if the time is adjusted according to the used computer/processor.

What you don’t want is a system that is unpredictable, i.e. the start/finish signal is not caught straight away each time. This gives an error that can not be adjusted and the times can therefore not be trusted to the last decimal.

Unfortunately a few outliers, or unpredictable errors occurred using the Chronocone. The predicted times was 0.05-0.06 seconds different from what was noted. I’m not saying it’s all bad, I’m just pointing out that when racing gets tight and there is money at stake at first place I wouldn’t wanna lose to someone by a few hundreds or even thousands knowing the system is not 100% correct.

All test results are available at www.ettsexett.com/timing.xls

I know you should probably do longer series of tests. And being one of the creators of a system I’m probably not the right guy to bark, but I did the tests as exact I could excluding all the error sources I could come up with.

The reason for these tests was originally to find out if there is any idea to put more hours into a system that I previously didn’t know how exact it was. And since I read that the Chronocone should be used at major US events I figured I might include it in the test.

[John Gilmour]

I hope to see a merging of these systems. Hardware, features, software, and display.

The CPU likely doesn't measure out an exact second every time. But for that matter neither do many watches.

Some Physics depts have very accurate timers- they might be an even better reference. If you know a college student likely they use these for constant velocity or constant acceleration experiments in early physics in college.

Does the Trak mate use an oscillating quartz crystal or a 555 chip for base timing?

[Pat Chewning]

I will make some tests of the Chronocone accuracy against a calibrated traceable standard. We have several calibrated pulse generators, counters, and digital oscilloscopes at HP where I work.

Chronocone uses the QueryPerformanceCounter function and QueryPerformanceFrequency functions available in the computer. These functions are normally used in software development to measure the time it takes to execute computer instructions. This is commonly precise down to about 1 microsecond, so I really feel very comfortable that the system is repeatable and precise, but not necessarily accurate.

Repeatable: Get the same measurement value for several instances of identical stimulus. (Repeat "x" stimulus, always show "Y" result)
Precise: Ability to discern small variations in stimulus (Give a stimulus of X, get Y; give a stimulus of X + .0001, get Y +.0001)
Accuracy: X = Y where X is calibrated and traceable to a known standard.

I did some preliminary testing of the precision of Chronocone and satisfied myself that it is precise to .00001 seconds, but only repeatable to .0001 seconds. That is why Chronocone reports results to "only" .0001 increments.

It could be that Marcus was measuring the normal variation of his mechanical relay rather than the accuracy and precision of the timing system. When measuring events in the .001s to .0001 s range, simple mechanical variations like the bouncing of relay contacts can make a big difference. Some timing systems might trigger on the 1st bounce of the relay contact, others will wait for a more stable pulse before triggering. Most devices that rely on mechanical switch contact will have some variation of a "debounce" circuit in them. Not all "debounce" circuits are alike. So really the best way to test a timing system would be to eliminate all mechanical devices and drive the input from a known electronic switch (transistor) which should switch on and be stable in a couple of nanoseconds.

Look for test results soon.

John Gilmour: No real timing system will use a 555 timer as the time base. I'm very confident that the Trackmate uses a crystal oscillator as the timebase. (As do Ettsexett and Chronocone).

-- Pat

[John Gilmour]

John Gilmour: No real timing system will use a 555 timer as the time base. I'm very confident that the Trackmate uses a crystal oscillator as the timebase. (As do Ettsexett and Chronocone).

-- Pat

I made a system in the 1990's which used the 555 chip only to hold the switches closed long enough to register very fast trippings of the timing strips. I saw the 555 chip was used for some timing applications- but never used it- as I also figured a crystal oscillator was the way to go.

BTW should someone feel particularly industrious....in your opinion do you think it would be hard to make the software work on a Palm or Windows CE machine....(Not for use at an actual race...but for training and recreational racing)

[Wes Eastridge]

BTW should someone feel particularly industrious....in your opinion do you think it would be hard to make the software work on a Palm or Windows CE machine....(Not for use at an actual race...but for training and recreational racing)

That would kick ass! I just got used Sony PalmOS device and would love to be able to use it as a practice timer.

[Marcus Seyffarth]

I will make some tests of the Chronocone accuracy against a calibrated traceable standard. We have several calibrated pulse generators, counters, and digital oscilloscopes at HP where I work.

GREAT! That's pretty much what I wanted. I think the tests that I did where ok, but if you have better gear it's all good!

It could be that Marcus was measuring the normal variation of his mechanical relay rather than the accuracy and precision of the timing system. When measuring events in the .001s to .0001 s range, simple mechanical variations like the bouncing of relay contacts can make a big difference. Some timing systems might trigger on the 1st bounce of the relay contact, others will wait for a more stable pulse before triggering. Most devices that rely on mechanical switch contact will have some variation of a "debounce" circuit in them. Not all "debounce" circuits are alike. So really the best way to test a timing system would be to eliminate all mechanical devices and drive the input from a known electronic switch (transistor) which should switch on and be stable in a couple of nanoseconds.

I don't think that I measured the variation of the relay. The relay is standardized and have a guaranteed switchtime that is a few milliseconds (can't quite remember how many but I have this at home if needed). Second out I truncated the answers to only hundreds of seconds so even if it should be that the USB port needed a more stable contact then the COM port on the same computer (which I doubt) it should not show.

Another thing that I can't really understand is why timing equipment in slalomskateboarding should show thousands or tens of thousands of a second. Skateboarding at 20mph (approx 32km/h) you travel less than 3/8" or 9mm in a thousand. In my opinion it's not skill that decides if you win by this little. But that is perhaps not the topic here...