different current through two series resistors

In general getting femto-amp accuracy from the simulator is unlikely, unless you've gone out of your way. The default value of iabstol is 1e-12 (i.e. a pA), so you would not expect accuracy of currents significantly less than that. If you really care about currents that small, you could always tighten iabstol.

Also check if other things are connected to this node by name (check the resulting netlist). Also check the presistor to see if it's a schematic or just a primitive instance in the netlist - maybe there's a path to ground inside the subckt.

Andrew.

Thank you

I changed the iabstol to tighter values but there was no change in current difference.

No, presistor is a simple resistor.

I just tried it with a simple circuit and got sensible results (even without tightening tolerances). I still can't see why you care about fA currents, but if this is a problem for you I suggest you contact Cadence Customer Support with the design data so that an Application Engineer can have a look. What you've shown isn't enough to understand what the problem (if any) is. Here's what I did:

// spectre netlist to try to show small resistors
r1 (n2 n1) resistor r=0.05
r2 (n1 0) resistor r=3.2
v1 (n2 0) vsource dc=7.5f

dc dc
oppoint info what=oppoint where=screen

And here's what it output:

Instance: r1
Model: resistor
Primitive: resistor
          1 : V(n2) = 7.5 fV
          2 : V(n1) = 7.38462 fV
          v = 115.385 aV
          i = 2.30769 fA
        res = 50 mOhm
        pwr = 266.272e-33 W

Instance: r2
Model: resistor
Primitive: resistor
          1 : V(n1) = 7.38462 fV
          2 : val(0) = 0
          v = 7.38462 fV
          i = 2.30769 fA
        res = 3.2 Ohm
        pwr = 17.0414e-30 W

 Couldn't this just be the current from Gmin? Set Gmin to a low number to check.

Rob

These resistors are a part of post-layout extracted circuit.

So, few current leakages are accumulated in other nodes and lead to a few hundred fA.

I tried with gmin=1e-20 too.

I doubt it would be gmin, since gmin resistors are only added across specific non-linear device junctions (e.g. Source-Drain junctions on a MOSFET), plus nodes with no path to ground. So it's unlikely there will be any gmin resistors near these resistors.

Andrew.