I'm simulating a 1GHz signal with transient simulator. Even when I set the tolerance to "1e-9, ..." which is 3 orders higher than strict, and the time step to "0.1 fsec", I still see observable difference between "Trapezoidal" and "Gear". With looser settings, the difference is larger. Though the difference does not affect my conclusion seriously, I still hope to know which is more accurate as I cannot always set so strict tolerance and time step.