Measuring clock quartz frequency accurately

[Axk]

I've an old quartz clock and it has (presumably) an adjustment capacitor to adjust the frequency.

Is there a way to measure the frequency from the quartz accurately enough for good adjustment?

I've tried measuring the outputs with a 10M oscilloscope probe but it appears to completely throw off the low power oscillator to the point that the clock stops, let alone measuring it accurately.

[helius]

You need a probe with low input capacitance in order not to disturb the crystal.
An active FET probe should do the business.

[Axk]

Considering it is only 32KHz, I suppose I can use a FET input opamp and measure the output with a passive probe?

[edpalmer42]

Another trick you can try is to clip a resistor onto the end of your scope probe and then probe the circuit with the open end of the resistor.  The resistor isolates the circuit from the capacitance of the probe.  The value will depend on the signal amplitude.  Try 1M and see what happens.

Ed

[golden_labels]

Run the clock for a day and compare to a some reference. Adjust. Repeat until obtaining the target accuracy. You may start with hour-long tests if the error is already very high — but with a test lasting only a hour you can’t get better than around 0.05%.

Measuring frequency/period accurately with a scope alone might be pointless. If you would manage to measure it at 0.1%, that would still mean… 1.5 minutes of error per day.

1 hour test: error ≈ 48s / day, 24 minutes / month
1 day test: error ≈ 2s / day, 1 minute / month
1 week test: error ≈ 300ms / day, 9s / month
Since this is already in the range of single PPMs, I guess that going below that makes little sense: the error from temperature and vibrations will probably take over.

[David Hess]

Calibration can be done using an acoustical pickup and reciprocal frequency counter.  This is common for mechanical clocks.

[Axk]

What if I feed the 32KHz signal from the clock's crystal into channel 1 of a scope and a more accurate 32KHz oscillator into channel 2?

Will it be possible to compare the 2 down to say 5 ppm based on drift? Or will the scope (Rigol DS1054Z) have a greater matching error between the channels?

[ogden]

Will it be possible to compare the 2 down to say 5 ppm based on drift? Or will the scope (Rigol DS1054Z) have a greater matching error between the channels?

Yes. Info about how to do it:

https://youtu.be/KmZwAjmzHx8?t=1122

[David Hess]

While probably not practical for what you are doing, a current clamp can also be used for a non-intrusive measurement.

https://www.analog.com/en/technical-articles/sub-ua-rms-current-measurement-for-quarts-crystals.html

[Axk]

It appears the oscillator chip in the clock does have a buffered 32KHz output.
It measured 32.76750 KHz with the DS1054Z hardware frequency counter.
I've adjusted the capacitor and now it measures exactly 32.7680 KHz.
Considering some forum members measured the accuracy of the hardware counter in the scope to be within 5ppm, the clock is now hopefully a bit more accurate.

May try ordering a more accurate temperature controlled oscillator and comparing with that, if it makes any sense.

Though I suppose considering the crystals various drifts, it is as accurate as it gets now.

[ArthurDent]

Years ago I faced the same problem with a watch I had with a trimmer cap on the time base and it was consistently drifting in one direction. What I did was connect my scope ground to the metal watch case and trigger the scope from a synthesizer set to 32,768hz. I cranked the scope input channel up to maximum sensitivity and held the probe tip close enough to the back of the open watch case until I saw (along with lots of noise) the crystal frequency drifting either right or left.

I adjusted the trimmer cap to get the drift to a minimum then, just to make sure, placed a piece of aluminum foil over the back with a small hole in it to stick the probe tip through. I needed to double check the drift to make sure that when I replaced the watch back cover that it didn't cause the crystal frequency to shift from added cover capacitance. This worked well and the watch was much more accurate. If your watch is on your wrist then it kind of has a temperature stabilized environment for most of the time.

[edpalmer42]

One thing to keep in mind about watches.  I've noticed that quartz watches and clocks tend to run fast rather than slow.  I thought about that and decided that it made sense.  If your watch is fast, you'll be early to your meeting or appointment rather than late.  So you might want to nudge your watch a bit fast, but still within the typical spec of 30 sec./month.  If my math is correct, that means setting it to more than 32.768 KHz, but less than 32.76838 KHz.

Ed

[JBeale]

I've noticed the same thing, every (digital) wristwatch I've ever had ran slightly fast, usually about 1/2 to 1/3 second per day. I believe people tend to think a slow watch is broken, but a watch going the same amount fast is still ok.

[magic]

Considering it is only 32KHz, I suppose I can use a FET input opamp and measure the output with a passive probe?

Very few FET opamps have less than 1pF of input capacitance and even that might be significant compared to the 10pF or whatever normally there.
Quartz clocks are also sensitive to temperature. Just because you measure it now doesn't mean it will be the same at night or next week.
I wouldn't necessarily assume that the quartz in some random cheap scope is much better than the quartz in a wall clock.

I second the recommendation to just run it for a full day or a few and compare with a good reference (GPS, radio, NTP).
I once fixed the PSU of a mains-powered clock and noticed that it has some dependence on power supply, easily visible after a few days. So I tuned it by running it for a week at a time and adjusting the PSU.

[Zero999]

With digital clocks/watches, I believe the calibration is done to some degree in software. The oscillator is built so it's slightly fast and at the time of manufacture, then the exact frequency is tested and a calibration factor is programmed into the IC. If it's found to be two seconds per day too fast, then 20 seconds of every day are made to be 0.9s, so it the correct timing is achieved. It's also true there's probably some bias towards too fast, than too slow.

[Axk]

Ordered what is claimed to be a 0.1ppm TCXO from Aliexpress.

https://www.aliexpress.com/item/Fast-Free-Ship-Optional-100kHz-200kHz-400kHz-500kHz-800kHz-32-768kHz-20kHz-0-1ppm-tcxo-high/32717118453.html

Will see if I compare the clocks's osicllator to this one on my scope and see the drift between the two.

[jmaja]

I've noticed the same thing, every (digital) wristwatch I've ever had ran slightly fast, usually about 1/2 to 1/3 second per day. I believe people tend to think a slow watch is broken, but a watch going the same amount fast is still ok.

I would hate to have such an inaccurate watch. The worst digital watch I have had gained about one minute in half a year (before it needed adjustment for daylight saving time) and the best only a few seconds. The first one I got was in 1979 and the worst one was from the 90's.

[Zero999]

I've noticed the same thing, every (digital) wristwatch I've ever had ran slightly fast, usually about 1/2 to 1/3 second per day. I believe people tend to think a slow watch is broken, but a watch going the same amount fast is still ok.

I would hate to have such an inaccurate watch. The worst digital watch I have had gained about one minute in half a year (before it needed adjustment for daylight saving time) and the best only a few seconds. The first one I got was in 1979 and the worst one was from the 90's.

A minute over six months, is roughly a third of a second per day.

[jmaja]

I've noticed the same thing, every (digital) wristwatch I've ever had ran slightly fast, usually about 1/2 to 1/3 second per day. I believe people tend to think a slow watch is broken, but a watch going the same amount fast is still ok.

I would hate to have such an inaccurate watch. The worst digital watch I have had gained about one minute in half a year (before it needed adjustment for daylight saving time) and the best only a few seconds. The first one I got was in 1979 and the worst one was from the 90's.

A minute over six months, is roughly a third of a second per day.

Yes it is. I only have had one watch that inaccurate and I didn't like it due to it. My current one (Polar RCX5 heart rate monitor) may gain 10-20 s in half a year and I have had more accurate ones. It can be adjusted to daylight saving time without adjusting seconds. I still set it 2-3 times a year. Especially during summer it is nice to have it close to GPS time, since I do sail racing and starts are always at GPS time. The inaccurate one was also made by Polar.