Can I replace watch quartz with DS3231 32 kHz output?

Question

I'd like to know if it's possible to synchronize classical quartz clock by replacing inbuilt quartz crystal with DS3231's 32 kHz output. I need to synchronize my analog watch with RTC DS3231.

Am I supposed to simply remove quartz crystal and conenct DS3231 output? There are two wires coming from the watch IC, the crystal is connected between them. I think one is XTAL1 and the other one XTAL2, but which is which? I wonder is it even possible?

I heard it's about connecting 32 kHz output to XTAL1, so "2?" and frequency corrector are not connected then? I'm clueless.

Answer 1

Usually, yes - assuming you mean 32.768kHz when you say 32kHz, or the watch will run at the wrong rate.

First you need to know which IC pin is the input and which is the output, as Ignacio commented. A quick measurement with your oscilloscope will show a strong 32.768kHz signal at one pin on the crystal, and a weak signal at the other. The weaker signal is the input.

Now you need to connect the watch ground to the DS3231's ground, and capacitively couple the DS3231's 32.768kHz output into that input, removing the crystal of course. The correct value of capacitance is a guess, but 22-100pF is probably the range, and I would guess 33pf as a starting point. After the capacitance, you should see a signal approximately the same as the original. If it's much larger, reduce the capacitance and vice-versa. (You may have to remove the trimmer capacitance, or turn it to its minimum setting, i.e. maximise the input signal)

Oh and you still need the correct supply voltage of course - e.g. the original battery.

There should be a good clean copy of your 32.678kHz signal on the original (now unused) output pin.

Answer 2

I have measured the 32kHz output signal of four DS3231 modules with a strange result. The board of the modules are labled ZS-042 and has a DS3231M chip mounted. Instead of the high accurate 32768 Hz signal I have measured 32641Hz, 32710Hz, 32730Hz and 32743Hz. When I enable the 1PPS signal at all four modules, the period is the same with little deviation of some ppm. The measured range of the 1PPS is between 1000.009ms and 999.994ms without waiting for long time stabilization.

All frequency measurements are done with a calibrated version of a TransistorTester. The calibration was done with a trimmer (10-35pF) at the crystal using a 1PPS signal from a GPS receiver as reference.

As a result of this measurement I would say, that the 32768 Hz of the DS3231M can not be used as precise reference.

Probably this is only a problem of the cheap DS3231M!

Answer 3

The 32 kHz output of the DS3231M is not temperature compensated while it is for the non-M variant. The MEMS oscillator of the M model is highly sensitive to temperature and the frequency will drift like crazy on that unit. The temperature compensation takes as part of the timekeeping functionality, so it still meets spec in regards to the time (but not frequency) outputs. The non-M variant has temperature compensated time and frequency. See https://blog.heypete.com/2017/09/05/major-differences-between-the-ds3231-and-ds3231m-rtc-chips/

Answer 4

I was able to upgrade my bedside clock radio to use the 32 kHz from a DS3231SN.

Rather than completely replacing the xtal signal (which I tried but didn't work), I fed a tiny bit of the DS3231SN 32 kHz signal into one side of the xtal oscillator circuit.

For my case, which was an MCU running on 3.3v, I found that feeding the 5v 32 kHz signal via a 100k resistor was enough to "coax" the existing crystal circuit to sync to the DS3231.

I tried this on what I thought was the "input" side, but it didn't work until I switched to the other side (I may have had the pinout wrong). So, some experimentation may be needed, but I have been delighted with the success of my project.

See Adding a Temperature-Compensated Crystal Oscillator to a Consumer Clock Radio