Can anyone please help me with a very simple circuit of three capacitors in series. Attaching its netlist below:
// Library name: FloatingGate // Cell name: cap //
View name: schematic
C8 (net5 0) capacitor
C10 (net05 0)
capacitor c=1p ic=1
C11 (net5 net05)
simulatorOptions options reltol=1e-3 vabstol=1e-6 iabstol=1e-12 temp=27 \
tnom=27 scalem=1.0 scale=1.0 gmin=1e-12 rforce=1 maxnotes=5 maxwarns=5 \
digits=5 cols=80 pivrel=1e-3 sensfile="../psf/sens.output" \
tran tran stop=1 errpreset=moderate
writefinal="spectre.fc" annotate=status maxiters=5
finalTimeOP info what=oppoint where=rawfile
modelParameter info what=models
outputParameter info what=output
designParamVals info what=parameters
subckts info what=subckts where=rawfile
saveOptions options save=allpub
With my knowlegde of circuit and experience on spice simulator of Tanner, I expect net5=1v and net05=1v
However, Output in cadence the voltage at net5 and net05 start with 1v and decreased exponentialy wo zero.
The capacitor is enable to retain its charge.
This issue looks simple but it is creating issue in all my designs, which are working well in TSpice.
Please help me
Hopeful for quick reply from some expert
This is because of the gmin conductance that is added from floating nodes to ground. Nodes in reality are never entirely floating - there's always some leakage path. The issue is for a circuit simulator, that floating nodes can cause a convergence problem, because it leads to an ill-conditioned matrix (effectively there are potentially an infinite number of solutions for a floating node). To help with floating nodes and off junctions, spectre inserts a gmin conductance to ground (for floating nodes), or across the off junction. The default value of gmin is 1e-12 Siemens (i.e. 1e12 ohms) - and is explicitly stated in your netlist. With a 1pF cap, that's a 1 second time constant - hence, the decay you're seeing over a 1 second simulation. You could set gmin=0 in this case (Simulation->Options->Analog in ADE, or the simulatorOptions line in the netlist above), and you'll then get no decay, but be warned in general gmin helps avoid convergence problems (most circuit simulators do similar things; this is not unique to spectre).