Cadence transient PLL simulation results vary with different accuracy settings

I am simulating a circuit level PLL. I obtained the component parameters from system level simulation and everything should work. I came across this problem that if I run the transient simulation and set the accuracy (errpreset) to be

1) errpreset=liberal ==> system locks and the result is good
2) errpreset=moderate ==> system locks and the result is good
3) errpreset=conservative ==> unreasonable result and system won't lock

Sometimes if I run case 3) for a long time, the result is getting correct again. For example, if I run for 0.5us with conservative accuracy, the system won't lock (with moderate or liberal, it will lock at around 0.2us). However, if I run for 3us with conservative accuracy, the result is totally different: it locks at 0.2us and behaves exactly the same with the rest two (moderate or liberal).

Has anyone experienced the same problem before? Any suggestions will be highly appreciated.

Faye

S. here the suggested procedure how to use errpreset:

Simulators use many iterative calculations to arrive at a solution for each time instance. How long it takes to do each iteration is affected by the accuracy limits of the calculations. Generally the tighter the accuracy limits the longer the simulation takes, and the better the accuracy of the simulation. But sometimes a tight accuracy limit can lead to problems with converging to the final solution and/or give flaky results. That's why you may need to experiment with the accuracy values and know what the reasonable solution is for the circuit being simulated.

crutschow said:

Simulators use many iterative calculations to arrive at a solution for each time instance. How long it takes to do each iteration is affected by the accuracy limits of the calculations. Generally the tighter the accuracy limits the longer the simulation takes, and the better the accuracy of the simulation. But sometimes a tight accuracy limit can lead to problems with converging to the final solution and/or give flaky results. That's why you may need to experiment with the accuracy values and know what the reasonable solution is for the circuit being simulated.

Click to expand...

Thank you for your explanation.. so you mean sometimes a tight accuracy may result in a wrong result due to the reason that the simulation doesn't converge, even in the transient simulation? Have you ever seen any examples on that? My problem here is from transient simulation, with moderate accuracy, the PLL locks but with conservative accuracy, the PLL won't lock (see the attached figures).

Thank you again for your help.

Faye