Hi Sheldon,

ALthough I have not actually performed a set of simulations to assess the ISF of a VCO, the basic steps include (as you may already know ?):

1. At a specific time in a transient response simulation of a VCO, inject a current or voltage impulse of some amplitude into a node of a VCO. The node should be a node from which one can impact the frequency or phase as would a noise voltage.

2. Post-process the resulting transient data response of the VCO and determine the resulting amplitude of the phase step. Normalize the output phase to the magnitude of the input impulse and the maximum "charge" at the node in which you injected the impulse. This is an attempt to make the impulse sensitivity function indepedent of the amplitude of the injected impulse.

3. Record the time at which the input impulse was applied and normalize it to the period of the VCO (2pi).

4. Slide the time at which the injected impulse is applied through the 0 to 2pi time interval of the VCO oscillation period.

5. Plot or create a tabular listing of the resulting output phase step (normalized as discussed in [2]) as a function of the time (relative to 2pi of the VCO period) at which it was injected.

Page 4 of Thomas H. Lee and Ali Hajimiri's paper "Oscillator Phase Noise: A Tutorial" provides a description of the ISF and an example or a ring based and LC based VCO. Pages 2 and 3 of their paper "Jitter and Phase Noise in Ring Oscillators" from JSSC, JUne 1999 also discusses the ISF.

I am about to try this as I recently was given an assignment to compare the merits of two oscillator topologies and have not used this as a figure or merit previously.

Once again, sorry if I am re-stating what you know already!

Shawn