[SOLVED] phase Noise/Jitter for VCO and ILO

Dear colleagues,

Your support is kindly needed as I have designed an Injection-locked ring oscillator and I needed to measure the jitter value. I made pnoise simulation and got the phase noise value before and after injection from the output graphs. However, jitter simulation seems a bit tricky to me. According to phase noise/jitter concept, I should integrate the phase noise graph to get the jitter in seconds. But, I can't determine the integration limits I should use. I made some search and some say the integration should stop at the oscillation frequency of the vco and same say double the freq. Others say it should stop at the value of the offset frequency!

Also, I performed the pnoise simulation and chose noise type "jitter" to get the cycle-jitter and cycle-to-cycle jitter and there is a very big difference in the values before and after injection. Are these two methods interchangeable (integration and pnoise jitter)? Do both measure the same jitter?

Thanks,

It really depends on your application.

In applications requiring you to specify the accumulated jitter, you need to integrate from a low frequency. For example, in SerDes, a jitter tolerance mask is specified which is a high pass filter you apply on your phase noise to get the jitter. And you can clearly understand that you can then integrate the phase noise from a frequency a decade lower than the cut-off frequency of the mask. Note that the phase noise values at close offsets are not valid because the simulator does some approximations to calculate the phase noise spectrum.

Now about the end limit. If you stop at the VCO frequency, it means that you will not observe any jitter due to a phase noise component faster than your VCO clock which makes sense as you are only sensitive to the edges. But you cannot do this if you care about the positive edge to the negative edge variation. You need to integrate it to 2xVCO frequency.

Integration of phase noise and calculation of cycle to jitter through the ADE form are both equally valid from my observations.