Measuring Crystal Frequency Without Retuning the Oscillator

A crystal oscillator can change frequency or stop oscillating when a probe is placed directly on one of its pins. Adjusting load capacitors while that probe remains connected can tune the measurement setup instead of the finished product. Establish an observation method that preserves the oscillator’s intended loading before making frequency decisions.

Use the device’s supported observation path

Microchip’s 32.768 kHz crystal oscillator application note describes observing a clock output on an I/O pin or using a suitable high-impedance crystal measurement probe. The appropriate path depends on the actual controller and oscillator. A generic scope probe should not be assumed transparent at a crystal pin.

If firmware can route a buffered clock or divided clock to an output, document that configuration. Confirm the source selection and divider rather than assuming the output represents the crystal directly. A processor may switch to another clock source after a startup failure.

Build a frequency error budget

Record the crystal specification, board conditions, supply, temperature, and any calibration state. Include the counter or scope timebase accuracy and gate interval. A very detailed frequency display does not guarantee equivalent absolute accuracy.

As an illustrative conversion, an error of 1 Hz at 32,768 Hz is approximately 30.5 parts per million. This arithmetic helps compare results with a ppm requirement, but it does not identify whether the source of error is crystal tolerance, loading, temperature, aging, or the measuring instrument.

Separate frequency from startup robustness

• Measure after a defined stabilization interval.
• Repeat startup under the relevant supply and temperature conditions.
• Confirm that the observation firmware does not change oscillator settings unintentionally.
• Record whether attaching the measurement connection changes behavior.
• Preserve results before and after any capacitor adjustment.

A frequency close to nominal on one warm board does not prove reliable startup across the product’s operating conditions. Likewise, startup success does not establish the required timekeeping accuracy. Keep these acceptance questions separate in the test plan.

When direct probing is necessary, follow the component vendor’s recommended method and record the probe’s loading and connection. Avoid correcting the board to compensate for an undocumented measurement capacitance. Release the chosen capacitor values with evidence from the intended production configuration, including the final board layout and component variants. This gives later debugging a stable reference and prevents repeated tuning whenever a different bench instrument is used.

Include any factory trim or software correction in the record. Report the raw oscillator result separately when the purpose is to assess the hardware itself.