sheldon
|
Amar,
In general, my feeling is that simulating a VCO is not a place to consider cost-performance. They are difficult circuits to analyze and are sensitive to numerical noise because the their gain, kvco, tends to be high. In addition, VCO tend to be relatively small so the penalty of simulating with high accuracy is not as high. Having said that, I would try one of the following options when using Spectre:
1) moderate 2) conservative 3) conservative w/ method=traponly 4) moderate w/ relref=alllocal or pointlocal
In general, using the trapezoidal integration method is better for oscillators, there is a detailed discussion of the topic in Ken Kundert SPECTRE book. If you are seeing differences between moderate and conservative, then options # 3 & 4 may help explain the difference.
Best Regards,
Sheldon
|