Hi all,

At this time I am doing intensive simulations on the VCXO, so I have several questions about transient simulations.
As I am doing simulations on a crystal oscillator, I found slow start-up and convergence time problems.

1 - I need to speed up the simulations, without giving up of conservative mode.
2 - As PSS gives me always the error that needs transient data, I would like to do one simulation only, and re-use the data for the following PPS+PNOISE simulations.
3 - If I have to change, for eg.,  a parameter in the oscillator, is there a way to avoid the re-simulation process ?
4 - Is it possible to export the netlist to Verilog-AMS and run the transient always from Verilog-AMS ? Would I gain in speed ? (I know nothing about Verilog-AMS, so my question may be completely wrong, I am sorry for that).  
5 - Why the phase noise calculated trough the phase-noise method, is different from the phase noise calculate from the Jitter analysis ?
7 - I have seen that the jitter behavior in weak inversion is much worse than in strong inversion. If I go to moderate inversion, can I decrease the Phase-Noise and Lower the power (of course this is a trade-off).
6 - Last question that I put many times: How to simulate phase noise for autonomous circuits that is: what to choose PM or FM noise ?

I know these are very different questions, and should be spread along the various topics.

Best regards,
Pedro Cardoso

Hi,

I thought before posting. Maybe I wasn't clear enough in my questions.

1 - In the transient simulation you have liberal, moderate and conservative. We know that conservative simulation takes longer. The goal here is to set some parameters, in order to decrease simulation time maintaining conservative transient simulation.

2 -  At this point (close to tape-out) the oscillator cannot have an imposed initial consition, it should start by itself. For this you need a transient to do a PSS, otherwise you don't have an inital condition for your PSS (please refer to the Spectre RF manual - Chapter 6).
Thus only one transient simulation should be done in order to save lots of time.

3 - By Jitter in weak inversion I mean Phase-noise/jitter in subthreshold region. Despite the MOSFET behaving "like a BJT" in weak inversion, once both use the same exponential law, except for a factor of "n" in the MOS equation (the square law is no longer valid in this region). As the flicker noise is up-converted that will affect phase-noise/jitter quite a lot. So the designer should minize 1/f noise.
As the BJT has much better behavior regarding 1/f noise than MOS, I thought of using the MOS in weak inversion to have a "similar" behavior as a BJT. But the phenomena that cause flicker noise in BJTs and in MOS are different, and as SNR decreases significantly, the weak inversion is not the way to go.
So the aim of the question was, if I would get a better trade-off between power and phase-noise in moderate inversion, than in strong or weak. If the phase-noise performance difference is not significantly different, I shold bias the circuit in Modrate Inversion, once I don't need a large bandwith.

Best regards,
Pedro

Hi,

Thanks a lot. I will give it a try. I was really using Shooting mode.

Best regards,
Pedro