Combining Clock Source and Buffer Jitter Without Double Counting

A clock buffer’s additive jitter is its contribution to the signal, not necessarily the total jitter measured at its output. Confusing those quantities can either overstate or understate the clock budget. Identify what each specification includes before combining values from an oscillator and a distribution device.

Match the measurement definitions

The Skyworks AN766 application note describes combining source and additive buffer jitter under an uncorrelated random noise assumption. It also explains that input slew rate and operating conditions affect the buffer’s measured contribution.

Use the same jitter metric, frequency, and integration range for the values being combined. A period jitter value cannot simply replace an integrated phase jitter value because both happen to be expressed in picoseconds.

Apply the model explicitly

For independent random RMS contributions, take the square root of the sum of their squares. A hypothetical source with 100 fs RMS jitter and a buffer adding 75 fs RMS produces 125 fs RMS under that model.

Do not add the source again if the quoted output measurement already includes it. Conversely, do not treat a small additive specification as the complete output performance when the applied source is much noisier.

Examine the assumptions

Shared supply modulation or another common disturbance can correlate timing errors. Discrete spurs also deserve separate attention. The simple random noise combination should not conceal deterministic components or imply that every contribution is independent.

Check the input waveform at the buffer. A source with acceptable logic amplitude but slow threshold crossing may not reproduce the conditions used for the additive jitter specification. Keep the output format and loading consistent with the intended circuit.

Verify the complete clock path

Measure the source and final output using a setup whose noise floor is suitable for the expected difference. Estimating a very small additive contribution by subtracting two nearly equal noisy measurements can be unreliable.

Include the actual supply arrangement, fanout activity, and representative receiving load. Compare the result with the receiver’s relevant jitter requirement rather than only the most favorable number in a component table.

Save the metric, integration range, source contribution, additive contributions, correlation assumptions, and measured output. A useful budget makes its arithmetic and limitations visible. It does not turn a buffer’s isolated performance into a claim about the entire clock tree, nor does it count the same source noise repeatedly as the signal passes through several documented measurement points. Keep measured total jitter and calculated additive jitter clearly labeled in the same report.