Correlating Oscillator Supply Noise With Clock Timing Modulation

A clock can meet its average frequency requirement while supply noise modulates the timing of its edges. The relevant question is how the oscillator responds to disturbance over frequency, not simply whether the rail’s average voltage is correct. Measure the suspected coupling before changing the power architecture.

Use the oscillator’s specified sensitivity

The SiTime discussion of supply noise rejection describes power supply induced jitter and device dependent sensitivity. A quoted sensitivity belongs to specified conditions; it should not be assumed constant across every disturbance frequency or oscillator family.

Record supply range, bypass arrangement, output format, and load. Distinguish static frequency change with DC supply from modulation caused by an AC disturbance. Both can matter, but they are different measurements.

Look for a frequency relationship

Measure the supply near the oscillator using an appropriate low-noise connection. Compare periodic rail components with clock sidebands or a timing error spectrum. A converter switching frequency appearing in both records provides a hypothesis, not proof by itself.

Check whether the disturbance could instead couple directly into the clock trace, receiver threshold, or measurement instrument. Shared frequency alone does not identify the physical path.

Use a controlled comparison

Change the suspected source or supply filtering in a reversible, characterized way while keeping clock loading and measurement geometry stable. If a supply perturbation is deliberately injected, keep the resulting voltage within the oscillator’s specified operating conditions.

Compare disturbance amplitude at the oscillator pins rather than only at the injection point. The power network can attenuate or amplify particular frequencies, so equal generator settings do not guarantee equal local disturbance.

Evaluate the system consequence

Measure the clock metric relevant to the receiver and application. A discrete modulation component can matter differently from broadband random jitter. Keep integration range and any spur treatment documented when comparing results.

If a quieter supply or additional filter is selected, verify startup, voltage drop, and behavior during output activity. A filter that reduces one spectral component should not introduce an unreviewed supply transient or resonance.

Save supply and clock spectra, operating conditions, controlled changes, and the resulting timing improvement. A convincing diagnosis connects a measured local disturbance to a repeatable clock effect. It avoids assuming that every clock problem needs a dedicated regulator or that a small ripple number measured elsewhere on the board guarantees acceptable timing at the receiving circuit. Repeat the comparison with other clock outputs active if their switching changes the local supply disturbance.