Hi, I was not sure whether to ask this in #xschem or #ngspice, is there something equivalent to transient noise in ngspice/Xschem or xyce/Xschem for simulating oscillators ? So far I have used

.ic

statements to set nodes to either

gnd

or

vdd

to start an oscillation. But I am getting some weird results in a cross coupled oscillator I am trying to design. Do the SKY130 models include noise by default i.e. I shouldn't have to use

.ic

statements or do I have to manually add a noise voltage source in the circuit ? In this design

DIFF_OUT_N

seems to have a "stable" oscillation whereas

DIFF_OUT_P

continues to grow exponentially when I set the former to 1.8V. From what I can tell the design satisfies Barkhausen's criteria since the product of gain of both the stages is greater than or equal to one.

The best way to initialize a LC VCO is with a current pulse, not an initial voltage. I guess that series resistance could be better for the simulator to model the inductor quality factor. You should model the coupling between the inductors also. As far as I know, it's common to use two coils sharing the same area.

@Luis Henrique Rodovalho I have not studied about how to control oscillators yet. I have just read some basics from Prof Razavi's book. Could you please tell me, where/which node should I push/pull the current pulse to/from?

You should insert the current pulse source between the two outputs. Try to use a better modeled voltage source, with a series resistance. This should prevent the output signal amplitude to increase indefinitely, as it limits the supply current.

If you design an oscillator it will do its best to stay stable. If you design a compensated feedback amplifier it will always oscillate. /s

As the saying goes "my amplifiers will but my oscillators won't"😀