Measuring Oscillator Frequency: Counter Timebase, Loading, Drift, and PPM

Start with the Frequency Acceptance Condition

The Oscillator Design Fundamentals Guide covers networks, startup, and tolerance budgeting. This guide is a measurement protocol. State nominal frequency, permitted offset at each temperature and age, supply and load, warm-up interval, expected output amplitude, and measurement averaging time. A 1-second counter result, a 10-second average, and a production temperature sweep answer different questions. Distinguish initial accuracy from frequency change with temperature, long-term aging, short-term stability, jitter, and phase noise.

Do Not Confuse Resolution with Accuracy

A direct counter opens a gate controlled by its reference and counts input edges. An asynchronous gate boundary may contribute about one count of uncertainty. At a 1-second gate that is about ±1 Hz; at 10 MHz it corresponds to about ±0.1 ppm for this ideal quantization term. A 10-second gate reduces that term to about ±0.1 Hz, but does not remove a 2 ppm timebase offset or a drifting oscillator. Some instruments use reciprocal methods with different resolution behavior; follow the instrument manual, not the direct-count formula, when reporting a traceable result. Verify reference calibration and external-reference lock where used, and record the instrument serial number and reference state.

Worked comparison: nominal 10,000,000 Hz and a stabilized reading of 10,000,250 Hz imply +25 ppm. If the counter reference has a bounded ±2 ppm accuracy at the test conditions, claiming exactly +25.0 ppm for the device is unjustified without accounting for that reference contribution. The Oscillator Frequency Tolerance & PPM Calculator converts the measured offset and builds a stated budget; it does not certify the counter.

Avoid Measuring the Wrong Oscillator

Probe the buffered output or a specified test node where possible. A passive scope probe adds capacitance, and a frequency-counter input has finite impedance; either can alter a crystal load network, high-Q LC tank, or RC timing node. Measure input loading or estimate its size before connecting. Use an appropriate active probe or buffer if the node is high impedance, and compare readings with and without the probe when practical. Keep the output termination consistent. At low amplitude or slow edges, trigger threshold and noise can produce missed or extra counts. Check waveform shape and threshold with a scope before trusting the numeric readout. Oscillation that stops when probed is a loading problem, not evidence of a zero-frequency design.

Sweep Temperature, Supply, Load, and Tuning Deliberately

Hold the device at each temperature until its relevant internal or board temperature has stabilized. Record frequency, supply, output load, control voltage, warm-up time, and reference-clock status together. A temperature chamber air setpoint is not necessarily the crystal or board temperature. Separate reversible temperature behavior from aging by repeating a reference condition after a sweep. If testing a VCO, measure several tuning points in both upward and downward directions, then use local differences to estimate gain. For example, 100 MHz at 1.0 V and 110 MHz at 1.5 V give an average 20 MHz/V over that interval; this is not proof that the slope remains 20 MHz/V over the full tuning range. Use the VCO Frequency & Gain Calculator to document the selected points and expected local span, then inspect actual curvature and hysteresis.

Keep an Uncertainty Ledger

List counter timebase accuracy and calibration age, gate/count resolution, trigger sensitivity, probe pulling, output termination, temperature sensor placement, warm-up stability, supply measurement, and repeatability. Classify each item as known offset, bounded systematic term, or random variation. A calibration can remove a measured offset at one condition but does not erase future drift. For a conservative pass/fail decision, compare the whole bounded range with the specification; if the range straddles the limit, improve the measurement or apply a documented guard band rather than calling the part good or bad from displayed digits alone. This Guide does not calculate phase noise or Allan deviation; use specialist methods when those are the actual requirements.

• Check the counter timebase specification at the actual calibration, warm-up, and temperature conditions.
• Capture the raw frequency series, not only its rounded average.
• Record probe and termination settings and repeat the test after removing the probe where feasible.
• Run supply and temperature corners after the nominal operating state is stable.
• For VCOs, label each local gain interval and the tuning direction.
• Do not add signed deterministic offsets as if they were independent random noise.

Common Mistakes

• Trusting display digits without checking counter timebase accuracy.
• Assuming a longer gate removes oscillator or reference drift.
• Connecting a high-capacitance probe directly to a crystal or LC tank.
• Using an unstable trigger threshold on a noisy low-amplitude output.
• Comparing readings taken at different loads or supply voltages.
• Comparing chamber air temperature with a datasheet crystal-temperature limit.
• Calling one VCO endpoint slope the gain across the entire tuning range.
• Reporting absolute ppm without preserving the sign of the measured error.
• Combining systematic and random error terms without an uncertainty model.
• Confusing mean frequency accuracy with jitter, phase noise, or short-term stability.

Measurement Sources

NIST Frequency Measurement addresses accuracy, stability, instrumentation, and reporting terminology. The NIST Time and Frequency Users Manual describes counter timebase and gate behavior. Instrument-specific specifications remain authoritative for an actual test setup.