Harmonic Balance Convergence Issue

Hi

I'm running hb and hbnoise to plot the phase noise of a clock divider that divides a 4.5-GHz clock by 10, but the simulation doesn't converge.

I configured the oversample factor to 8 (and tried 16) as the input is a square wave, number of harmonics = 5 (read in the user guide that as the oversample factor increases, number of harmonics could be small and give correct results). The circuit runs successfully in transient analysis and the waveform is generated after less than 1ns yet I configured the hb to run tran first for 5ns and save initial results. I configured the fundamental freq. as 4.5G and Freqdivide ratio=10, and accuracy is moderate.

I read somewhere that fundamental freq should be the divided one (4.5G/10) as it's the least freq in the circuit but I think in this version (cds 6.1.7) I should put the input freq (4.5G) as the sim already asks for the division ratio.

I should also note that sometimes a warning appears in hb log says the transistor leaves the linearized region (meaning the MOS PN junctions became forward), yet when transient analysis is chosen, the warning doesn't appear and the results are correct! Meaning it's a measurement error.

Thanks in advance