I have run a transient simulation on a chopping instrumentation amplifier. With the default settings, and accuracy defaults set as conservative, everything is ok: input is a sine wave and output is an amplified sine wave.
However, when I tuned gmin from 1e-12 to 1e-24, the transient output showed oscillation with the chopping frequency. The output.log said it tried "homotopu = gmin" "homotopy = source" "homptopy = dptran" to get the initial conditions. Then which settings I should trust? A smaller gmin is always more accurate and reflects the real situation?
Attached is part of the output.log file. Please help！Thanks！

Circuit inventory:
             nodes 146
           bsim3v3 217  
         bsource_1 14    
         capacitor 16    
           isource 3    
           vsource 30    


Notice from spectre.
   77 warnings suppressed.


Time for parsing: CPU = 6.999 ms, elapsed = 9.06682 ms.
Time accumulated: CPU = 313.952 ms, elapsed = 308.797 ms.
Peak resident memory used = 32.3 Mbytes.

Entering remote command mode using MPSC service (spectre, ipi, v0.0, spectre32_8384_94, ).

Warning from spectre.
   WARNING (SPECTRE-16707): Only tran supports psfxl format, result of other analyses will be in psfbin format.


************************************************
Transient Analysis `tran': time = (0 s -> 20 ms)
************************************************
Trying `homotopy = gmin' for initial conditions.
Trying `homotopy = source' for initial conditions.
Trying `homotopy = dptran' for initial conditions..
Important parameter values:
   start = 0 s
   outputstart = 0 s
   stop = 20 ms
   step = 20 us
   maxstep = 200 us
   ic = all
   useprevic = no
   skipdc = no
   reltol = 100e-06
   abstol(V) = 1 uV
   abstol(I) = 1 pA
   temp = 27 C
   tnom = 27 C
   tempeffects = all
   errpreset = conservative
   method = gear2only
   lteratio = 10
   relref = alllocal
   cmin = 0 F
   gmin = 1e-24 S

   tran: time = 500 us       (2.5 %), step = 130.6 ns     (653 u%)
   tran: time = 1.5 ms       (7.5 %), step = 91.65 ns     (458 u%)
   tran: time = 2.5 ms      (12.5 %), step = 99.63 ns     (498 u%)
   tran: time = 3.5 ms      (17.5 %), step = 166.7 ns     (833 u%)
   tran: time = 4.5 ms      (22.5 %), step = 115.6 ns     (578 u%)
   tran: time = 5.5 ms      (27.5 %), step = 85.65 ns     (428 u%)
   tran: time = 6.5 ms      (32.5 %), step = 948.9 ps    (4.74 u%)
   tran: time = 7.5 ms      (37.5 %), step = 66.76 ns     (334 u%)
   tran: time = 8.5 ms      (42.5 %), step = 103.9 ns     (519 u%)
   tran: time = 9.5 ms      (47.5 %), step = 133.4 ns     (667 u%)
   tran: time = 10.5 ms     (52.5 %), step = 106.5 ns     (533 u%)
   tran: time = 11.5 ms     (57.5 %), step = 66.69 ns     (333 u%)
   tran: time = 12.5 ms     (62.5 %), step = 111 ns       (555 u%)
   tran: time = 13.5 ms     (67.5 %), step = 1.243 ns    (6.21 u%)
   tran: time = 14.5 ms     (72.5 %), step = 172.4 ns     (862 u%)
   tran: time = 15.5 ms     (77.5 %), step = 120.3 ns     (602 u%)
   tran: time = 16.5 ms     (82.5 %), step = 1.017 ns    (5.08 u%)
   tran: time = 17.5 ms     (87.5 %), step = 1.124 ns    (5.62 u%)
   tran: time = 18.5 ms     (92.5 %), step = 139.2 ns     (696 u%)
   tran: time = 19.5 ms     (97.5 %), step = 132.1 ns     (660 u%)
Number of accepted tran steps =             72140
Initial condition solution time: CPU = 1.36779 s, elapsed = 1.36807 s.
Intrinsic tran analysis time:    CPU = 112.155 s, elapsed = 112.182 s.
Total time required for tran analysis `tran': CPU = 113.529 s (1m  53.5s), elapsed = 113.561 s (1m  53.6s).
Time accumulated: CPU = 115.155 s (1m  55.2s), elapsed = 117.525 s (1m  57.5s).
Peak resident memory used = 249 Mbytes.

finalTimeOP: writing operating point information to rawfile.

******************
DC Analysis `dcOp'
******************
Important parameter values:
   reltol = 1e-03
   abstol(V) = 1 uV
   abstol(I) = 1 pA
   temp = 27 C
   tnom = 27 C
   tempeffects = all
   gmindc = 1e-24 S
Trying `homotopy = gmin'.
Convergence achieved in 65 iterations.
Total time required for dc analysis `dcOp': CPU = 73.989 ms, elapsed = 78.5849 ms.
Time accumulated: CPU = 115.254 s (1m  55.3s), elapsed = 117.635 s (1m  57.6s).
Peak resident memory used = 249 Mbytes.

dcOpInfo: writing operating point information to rawfile.
modelParameter: writing model parameter values to rawfile.
element: writing instance parameter values to rawfile.
outputParameter: writing output parameter values to rawfile.
designParamVals: writing netlist parameters to rawfile.
primitives: writing primitives to rawfile.
subckts: writing subcircuits to rawfile.

BOE,
Thanks for the reply. I wonder if it is difficult for the simulator to converge, is it possible that the simulation completes but the result is not trustworthy?

Designer's are crazy.

I am convinced that a large fraction of the trouble designers have with simulators are self inflicted. They often run with absurdly tight settings and suffer through needlessly slow simulations and wonder why simulators are not robust. Oddly, they often don't even try default tolerances. They just assume the worst and crank down the tolerances.

You are already running with conservative settings. And gmin by default is already 1TΩ. What on earth made you think that it needed to be 10¹² times tighter? And if you were going to tighten it that much, why didn't you just make it zero? And why, when you got wacky results did you not think, well when you do something crazy you have to expect crazy results?

Having said that, it does not seem like making gmin that small should cause the problem you are seeing. But it is hard to know because you have given very little useful information. It would have been better if you actually showed us the oscillation. That certainly would have been more useful than a seemingly normal output summary.

My recommendation is for you to relax. Accept a good result when you get it and stop looking for trouble.

-Ken