I know APS is supposed to be as accurate as spectre, but I'm having problems simulating a large block that includes an LC-tank PLL. I see the tank amplitude vary widely throughout the sim, whereas in plain spectre, it just grows exponentially to its limits as one would expect. Are there any options to APS that I might be missing? I'm running with default error presets and integration method. The tank center frequency is 10 GHz, and the simulator step size is fairly small - I see 2 - 5 ps steps most of the time.