Timing Watches

Why time watches, anyway?

All watches promise (either explicitly or implicitly) to tell you the time to within a reasonable degree of accuracy. It's why they exist in the first place. The abundance of cheap, reasonably accurate and unadjustable quartz watches has helped us to forget that, just a few decades ago, pretty much anyone who had a watch had an interest in keeping an eye on its time-keeping performance and on getting it serviced and regulated as need be.

Those of us who grew up with quartz watches may well know nothing of the maintenance requirements of mechanical watches. Our watches are so reliable and so cheap and the batteries last so long, that by the time the battery needs replacing we've probably either lost the watch, worn it out or moved on in fashion sense. In all my years of owning watches I have had a battery changed only twice, and both of those times were with second hand watches that had flat batteries when they reached me.

But then there's 'High Accuracy Quartz'. A watch that costs a fair bit of money. And also one that you can either regulate yourself or send back to get regulated by the manufacturer; both of which approaches would require you to have reliably and accurately determined the performance of your watch over a reasonable period of time. Of course, there will always be the annoying odd watch that cannot be regulated. The Bulova Precisionist and Accutron II, for example. You might still want to track the accuracy of these watches if for no other reason than to contribute real world performance data to the knowledge base of the collector community.

How should a watch be timed?

Every fraction of every second of every day for several years, ideally. The more data, the merrier. But realistically it's not going to be more often than once a day. And if you're doing that, then you are suggesting that you are timing it once every 24 hours. But if you time the watch at 07.00 on Monday and then again at 14.30 on Tuesday, then you have actually timed it across 31.5 hours. It might be 07.35, the next day, and 08.00 on Thursday. You might not get around to timing it until 15.00, on Friday. If this is the case, then there's very little point in taking daily timings. Your day-to-day graphs are already likely to be influenced by the errors in your button-pushing, and adding in another variable just renders such daily timings all but useless.

Of course, if you have a very consistent timing method and time your watches at a very consistent time, then daily timings are splendid. You can get a fairly accurate SPY projection within just a few days. The data I have gathered via my current stopwatch method suggests I should not rely on any SPY projections that my Excel spreadsheet may throw up for at least two weeks.

It is worth making clear that the only real measure of an SPY value comes at the conclusion of one full year. But if you haven't been paying attention up until that point, then the limited international warranty on your top-of-the-range Citizen will have expired before you have a chance to send it in for regulation. And of course there's also the adventure of the rise-and-fall of the graph throughout the year to be enjoyed. A modern, thermocompensated watch should maintain a fairly flat graph throughout the seasons. The ability to keep good time in spite of changing climatic conditions is, after all, one of the defining features of a HAQ. A non-HAQ, 'ordinary' quartz watch may end a year-long test with a very impressive final SPY value, but if it were subjected to daily measurement throughout the year, you'd expect to see steep rises during the cold, winter months, and equally steep falls during the summer. How such a watch averages out at year's end is pretty much luck-of-the-draw. You spend more money on a HAQ, partly because it promises to eliminate the 'luck' element and deliver dependable high accuracy. Comparing figures on the WUS HAQ forum, and seeing how HAQs perform as the year progresses, is a constant source of entertainment. I like to root for a good HAQ and get a kick out of seeing data that seems to confirm my prejudiced stance. I also quietly harbour disdain for HAQ pretenders and get hot under the collar whenever data emerge that seem to show them performing well.

What is the point of spending HKD24,000 on a Grand Seiko if a HKD2,000 Bulova Precisionist can deliver the same results? The Grand Seiko is an astonishing piece of engineering and craftsmanship. It's as 'high end' as any quartz watch could hope to be. Its movement is substantial, robust, precise, powerful, ingenious, sealed from dust, beautifully decorated, finely adjusted and assembled by highly-trained watchmakers in Japan. The Precisionist is cheap, flimsy and mostly made of plastic. From the presence of a redundant battery compartment, the base seems to have been designed for a different movement entirely. The movement is fragile, insubstantial, exposed and comprised of cheap, low-quality components. It is poorly packaged and exhibits the sort of rough finishing that you would expect from a low-cost movement assembled largely by machines in a factory in China. And yet it promises the same 10 SPY performance as the Grand Seiko! Timing, testing and monitoring these watches is crucial to exposing either the extent to which Grand Seiko's efforts are unnecessary to the delivery of high accuracy performance, or the extent to which Bulova's press statements differ from the reality of their product.

Real world vs lab.

Remember those mechanical watches that need frequent servicing and regulation? The wearer may observe a particular loss or gain, but the watchmaker determines it more accurately by popping it on a machine. Wouldn't it be nice if a machine could tell you the SPY of your HAQ without your having to time it all year long? If only it were that simple!

I acquired a highly precise timing machine for just this purpose, but I quickly discovered that I had made a serious and basic mistake. These timing machines can tell you only how your watch happens to be performing at a particular moment in time. The daily rate of a mechanical watch might be several seconds, but the daily rate of a HAQ would be absolutely tiny. My AQ1000-58B is supposed to gain no more than 5 seconds per year. So, let's imagine it's not performing to spec.. Let's say it's running at 7 SPY. If I could determine that, then I could send it back for regulation. But what does 7 SPY look like on a daily rate? 7 / 365 = 0.019. 5 SPY would be 0.013. So, you're going to need at least 3 decimal places on your daily rate in order to project an SPY value. And if your watch is running under 4 SPY, then you might like 4 decimal places. The thing is, you can't get three or four decimal places reliably from a short test. You just can't. And most timing machines will resolve a daily rate to just one decimal place, with much more expensive machines giving you two. A phenominally expensive machine might offer you three, but what's the problem? What's stopping a machine with a GPS-disciplined OXCO from giving you more decimal places? Well, you are looking for the tiniest variation in the watch over a period of just a few minutes. You're assuming that you will get a steady and consistent result at the most ridiculously small level. There are 86,400 seconds in a day (and 31,536,000 in a year). So, you want to know if your watch's daily rate is somewhere around 0.001 and 0.019? And you want to get that answer from one inhibition period? Even the longest known inhibition period - 960 seconds - covers just one ninetieth of a day. To get three decimal places for the day from a measurement of just a few minutes would demand atomic level accuracy and a whole shed load of decimal places for the timing interval. But even if you can delivery that accuracy of measurement from the timing machine, the watch's movement just isn't in the same ball park of stability. It makes fairly massive (compared to an atomic standard) variations from one tick to the next. In theory the performance over an entire inhibition period should show you the actual performance of the watch, but the inherent instability of quartz watches (even the best of them) renders this impossible. It's not just the individual ticks that vary in length, but the overall length of each inhibition period also varies slightly. You could get a daily rate to one decimal place fairly confidently. Two decimal places would be 'ball park'. Good enough for a non-HAQ, but not enough to tell you whether your watch is a 4 or an 8 SPY performer. Several hours of testing would be needed in order to pin down a 3 decimal place daily rate with any degree of confidence.

And it still won't tell you how your watch will end the year! How will it cope with a blistering hot summer? How will it cope with sub-zero nights in the dead of winter? This 'instantaneous' rate (if something derived over 4+ hours of testing can be called that) does not reflect the real world performance of a HAQ. The same is actually true of mechanical timing machines, too. The way an owner wears his watch can result in that timepiece's accuracy being markedly different from the watch-fixer's projections. So, is it worth using timing machines? Well, yes, if you want certain data. With a Seiko 9F, you know that each turn of the regulator results in an 8 SPY change. With the older Twin Quartz watches, you're shooting in the dark. Having an instantaneous rate would let you see the effect that your fiddling with the trimmer was having. You may not be able to trust the rate that it gives you, but you can at least see whether you are twiddling the trimmer the right way and whether you have overshot from gain to loss (or vice versa).

I have also added another dimesion - a heater / chiller plate. This lets me test a movement's performance at a fairly wide range of temperatures. A HAQ should, of course, return a very similar and fairly flat sort of result at all (reasonable) temperatures. It may, however, be instructive to see how a TC HAQ movement copes with temperature change in terms of its inhibition correction, and to compare the curve (or lack thereof) of a TC watch with those of HF and non-HAQ watches. And where an inhibition period is unknown (for example, in an older watch), a second-by-second analysis of the movement should reveal the length of that period.

While all this is interesting and instructive, it does not, of course, reflect the real world performance of a watch. It's a more expensive set-up than a stopwatch and rather than being quicker, it is actually an incredibly slow process. It is not, therefore, something that I do every day. It also begs for augmentation. Early results have shown strangely flat results for the Citizen Exceed, lending weight to the suggestion mooted on the WUS HAQ forum some years ago that Citizen's movements may make changes to the XO's frequency rather than (or in addition to) digital count adjustment. To test this, though, I would need a machine that can measure the frequency of the XO. Such machines don't actually tell you the frequency of the XO, but compare it to their internal timebase and present the difference as a daily (or monthly) rate for the watch. To tell if the frequency of the XO is changing, therefore, you simply have to see whether the daily rate is changing as the seconds tick by. I have recently managed to acquire an old Witschi machine that can perform this sort of test on both 32 kHz and 4 MHz watches, so as soon as I get a bit of free time I shall give it a try. The internal timebase is unlikely to have retained its accuracy after all these years, but if I actually needed it to give me an accurate result then I could simply compare the rate that it gives me for a given watch against that watch's known performance (as determined by the stopwatch method), and apply the difference to any other readings that the machine might give me.

Back in the real world, we're left wondering how to keep track of a reasonably large collection of watches when there simply aren't enough spare hours in the day to time them all. It may be possible to video several watches at once, next to a time reference, and then determine their off-set through video playback. This would mean that just one timing event is required for ten or twenty watches, rather than each watch demanding ten or fifteen stopwatch results. It should also be considerably more accurate. It would, however, require yet another investment of money and time in acquiring, setting up and configuring the necessary equipment. I'm just not ready to go there, right now.