wierd convergence problem with cadence spectre transient simulation

[congcongaihuihui]

Hi,
I am simulating a power amplifier(PA) with a bias ciucuit in the cadence spectre transient simulation, a separate simulation of PA circuit and a bias circuit waveforms are good, but when putting the PA and bandgap reference bias circuit together with no direct connetions, I find the output of bias biginning to oscillate with a swing of several millivolts.
Later, I found that the transient simulation in the cadence, when placing the low frequency circuits such as bias circuit, LPF circuit next to the high-frequency sinusoidal signal source with both ends vacant, the signal of low-frequency circuit will begin to oscillate.
****I mean this high frequency source has no connection anywhere with the low frequency circuit ,yet the transient simulation fails? please help me.
And is it the problem of Cadence software or of the circuit design?

 

[erikl]

... when placing the low frequency circuits ... next to the high-frequency sinusoidal signal source with both ends vacant,
the signal of low-frequency circuit will begin to oscillate.

placing ... next to suggests you're talking of postLayout analysis?
Then it's probably cross-talk via parasitic capacitances or via parasitic inductive coupling.

 

[congcongaihuihui]

yeah，postlayout analysis. In cadence spectre,transient simulation.
I mean that the high frequency source is just placed away from the bias circuit, with two ends floating,having no connection with the bias circuit.So,I can not understand how cross-talk hanppen？

 

[erikl]

Re: cross-talk due to parasitic capacitive coupling

Check the postLayout netlist. You'll probably find extracted parasitic caps between the corresponding nodes.