When running ac analysis and plotting transfer function of amplifier I am getting noisy results at low frequency.
Its looks like accuracy of the simulator and when changing gmin it some times helps.
Is there a way to eliminate this noisy results?
I wouldn't expect that level of numerical noise, especially not when the gain is so high - that's rather strange. I also don't think that gmin would affect it, assuming that gmin is reasonable (near the default of 1e-12).
Please contact customer support. This is rather unusual.
We have seen this kind of problem rather often, usually on RLCK-extracted circuits, and currently have CCR 2470950 open for such a problem. In this case, plain Spectre could simulate the circuit just fine while APS had big problems.