Measuring load capacitance of a 32.768kHz crystal oscillator

[jeremy]

Hi all,

I have a question which came up recently which I don't have a good answer to: suppose you had a crystal oscillator which was nominally tuned for 32.768k, but you had no idea of the correct load capacitance and there are no useful part markings or datasheets, etc. How do you measure the load capacitance of the crystal? For example, digikey part number SE2413TR-ND comes from the Epson FC-135 series of crystals, which are available in 7pF, 9pF and 12.5pF options. But how do you confirm that it is the one you expect?

The best answer I could come up with was to build a pierce oscillator circuit, then hook up a good frequency counter to the output and keep swapping values of caps until you get as close as possible to 32.768k. But since you are soldering to change capacitors, this could be quite tedious and is prone to some sort of heat damage or at least thermal drift.

I found a boat anchor which is supposed to be designed for measuring crystals, the HP E4916A. But it only goes down to 1MHz. I also thought of trying these Murata voltage controlled capacitors: https://www.murata.com/en-global/products/capacitor/variable/overview/lineup/variable but they only go down to ~16pF, which is a little bit too high.

[David Hess]

I would use a variable capacitor and then measure the capacitance of the variable capacitor.

Two or more varactors can be used in series to divide their capacitance.

[srb1954]

The best answer I could come up with was to build a pierce oscillator circuit, then hook up a good frequency counter to the output and keep swapping values of caps until you get as close as possible to 32.768k. But since you are soldering to change capacitors, this could be quite tedious and is prone to some sort of heat damage or at least thermal drift.

You should also measure the start-up performance of the oscillator as incorrect load capacitors may cause slow or erratic start-up. A very low-capacitance scope should be used on the oscillator output to avoid the scope probe loading from affecting the oscillator. Failing that a resistor, say 1k , in series with a standard scope probe will reduce the effect of the probe capacitance affecting the oscillator.

When determining the load capacitance remember that the amplifier input and output capacitance and circuit strays also make up part of the effective crystal load capacitance.   

[David Hess]

You should also measure the start-up performance of the oscillator as incorrect load capacitors may cause slow or erratic start-up. A very low-capacitance scope should be used on the oscillator output to avoid the scope probe loading from affecting the oscillator. Failing that a resistor, say 1k , in series with a standard scope probe will reduce the effect of the probe capacitance affecting the oscillator.

Buffer the oscillator output with an emitter or source follower as part of the circuit and you can ignore the probe capacitance.  There are suitable operational amplifiers available now which could be used as a follower for buffering if you need DC precision.  Look for a fast part with low common mode input capacitance.

[dietert1]

Some MCUs like MSP430 include programmable capacitors for this, so the fine tuning can be done during initialization.

Regards, Dieter

[voltsandjolts]

If it's an RTC crystal for an MCU, configure the MCU to output a frequency derived from the 32768Hz. Measure that frequency (because it's buffered) and use it to adjust load capacitance as necessary to achieve 32768Hz. This way you can account for MCU pin capacitance, stray capacitance, load capacitance and crystal offset all at the same time.

Furthermore, as dietert1 said, an MCU can do run time compensation using a HF crystal as reference and so compensate for ambient temperature changes. MCUs may have internal capacitance adjustment or some digital ppm divider adjustment.