Insights on Timing Recalibration

[jadney]

This is specific to my 95QGR-33, but it should apply equally to any quartz watch that has a 32,768 Hz crystal and an adjustable capacitor for calibrating the frequency.

It has previously been mentioned that just removing the back from the watch will change the oscillator frequency and so will connecting a probe to the test point on the watch, so simply setting the measured frequency to 32,768 does not result in good timing. Instead, we need to determine a change in frequency that will get us to the right place.

I had previously posted about setting the timing by trying to figure out the frequency error in ppm (parts per million.) That can work, but it's tedious. It occurred to me that an easier way would be to consider the fact that a frequency error of 1 Hz would result in a time error of 1 second after each 32,768 seconds. 32,768 equals 9.1002 hours. I wanted to find a simple way to use this to determine the frequency error over some integral number of days. It turns out that 3.05 days (~72 hours) comes to 8 "time periods" of 32,768 seconds. So here's what I did with my 95 QR that had been gaining a lot of time.

I set the watch to my computer time and then checked it again in exactly 3 days. (Same time of day, 3 x 24 hours later) If the frequency had been off by 1 Hz, the time error would have been 8 seconds. My watch was fast by 15 seconds, so the frequency was fast by about 1-7/8 or 1.875 Hz. I choose to reduce it by a bit more, 1-15/16 or 1.937 Hz to put it on the slow side, to allow for future ageing.

I opened the watch and connected my HP frequency counter via a low capacitance probe. It was difficult to get a stable reading but I managed to read 32,769.566 Hz.

Subtracting 1.937 gives 32,767.629 Hz as my goal, but I was not able to adjust it that finely. Instead I tried to shoot for something slightly lower, because normal ageing of a quartz crystal will always cause an upward drift in frequency.

After testing for an additional 3 days, this watch still appears to run fast by perhaps 1/2 second. I'm now letting it run for 30 days, which should give me a much better picture on what additional adjustment might be necessary, if any. If I'm right about being ~1/2 second fast over 3 days, that comes to 5 seconds over 30 days, which is pretty good compared to the original spec of +/- 30 seconds per month. Still, I'd like to get it a bit on the slow side, so that it will improve over the years to come.

Note that this method does not require an accurately calibrated frequency counter, but it does require one that can measure to an accuracy of 3 or more decimal places. It may be useful to me to record the frequency I end up at; this could be used in the future on other 95Qs if I use the same probe and if the counter has not aged upwards significantly.

Hope others find this useful, or at least interesting.

[Kasper]

Thx interesting read

[jadney]

Update on my Casio 95QGR33 recalibrations:

I set my watch to CST using the time shown on https://www.timeanddate.com/time/zones/cst about 6 weeks ago. Then I wore that watch normally for 30 days. To get the best accuracy possible with this method, I compared the watch time to the URL above after 30.4 days.

3.03... days is 8x 32,768 seconds, so 30.3... days is 80x 32,768 seconds. If this watch is off by 1 HZ, it would be 80 seconds off after 30.3... days.

I actually compared times after about 30.4 days, because these watches tend to run fast, and I am hoping to overcompensate, getting them to run very slightly slow. This is because all crystals tend to age upward in frequency over time. Getting back to the slow side will make them improve over the next year or so.

After 30.4 days, I found that the watch had gained 8 seconds. (Note that I had already done this exercise once before, starting with a watch that seemed to gain over a minute a month, so it was far out of calibration.) This means that, on average, this watch's crystal was then running 0.1 Hz fast.

Using an HP frequency counter that gives me 8 places after the decimal point, and a low capacitance x10 probe, I measured the watch's frequency and then adjusted it down by a little more than 0.1 Hz. While I was set up to do this, I took a different 95QR33 watch and adjusted it to the same frequency, or as close to the same frequency as I could get. Needless to say, it's VERY difficult to do these tiny adjustments; there's a lot of trial and error.

The results:

That was 2 weeks ago. Since then, the two watches have kept in lock step with one another and so far, in those 2 weeks, they are both about 2 seconds behind. That works out to falling behind by 4 seconds a month, which is MUCH better than the +/-30 seconds per month accuracy that these were originally guaranteed to keep. AND, it's likely to get better, at least for some time, until the error climbs past zero and they slowly start to run fast. It will be interesting to see how quickly this happens; I expect it to be quite slow.

[jadney]

Here's a long term update on the timing calibration I did on my 95QR-33s.

Per my previous posts, above, I set the crystal frequencies of both of these back in March, to run a tiny bit on the slow side. Now, 6 months later, I have compared the times shown on both watches with the time shown here:

https://www.timeanddate.com/time/zones/cdt

After 6 months my 95QR-33 is now behind by about 50 seconds while the 95QGR-33 is behind by only about 20 seconds. That works out to less than 9 sec/mo. for the 95QR-33 and about 3.333 sec/mo. for my 95QGR-33 If we consider that the watches when originally sold were claimed to do better than 30 seconds per month, this is excellent, and it's likely get better over time, as the crystals continue to age. I will reset both watches to the correct time today.

Crystals all tend to age upwards because they are a mechanical resonator and they slowly shed mass as sub-microscopic bits of quartz flake off. Less mass causes the crystal frequency to rise. This is why extreme cleaning is part of any crystal manufacturing process, to remove all possible traces of material that might tend to shed off in the future. No cleaning process is perfect, however, and there will always be bits of quartz that are loosened over time.

I have some photos I'd like to post, but it looks like I need permission to add attachments.