[SOLVED] Post layout Monte-carlo simulation with TSMC 180

Hi everyone,

Has anyone had experience with TSMC 180 layout extraction?

I am trying to run a mismatch/process monte-carlo simulation, using TSMC180 design kit for my extractec_view from layout. The problem is when I run the simulation on adexl only process variation works and the mismatch does not.

I saw in the netlist of the extracted_view that nch_mac (mismatch) cells are extracted as nch (regular) cells.
I extracted my layout using Assura QRC files.

So how could I extract the layout including mismatch devivices like nch_mac?

Has anyone had this problem before? Any help or suggestions would be appreciated

Afonso

Sounds like a typical Cadence issue. However I don't know Assura, but if it generates .dspf file, or actually any other netlist file maybe you can try with a text editor to find every "nch" and replace with "nch_mac". Probably even Notepad can do this.

You can just edit the netlist after extraction and replace the devices.

Why do you call "nch_mac" a mismatched model?

From what I know, nch_mac is a macromodel (i.e. a wrapper around a transistor model, that captures some of the parasitics).

In old nodes, like 180nm, they used straight SPICE models (like BSIM3 or BSIM4) and not macromodels.

Why do you expect to see nch_mac for this process?

Of course, you can replace model names by editing text file (DSPF) - but if these models are not present in SPICE library - you will get errors form your simulations.

- - - Updated - - -

QRC - Cadence now names it Quantus, or Quantus QRC - is extraction tool different from Assura.

- - - Updated - - -

Unfortunately, existing flows in IC design field do not allow to simulate mismatch caused by layout parasitics (which can dominate in advanced technology nodes).

I remember we used _mac transistors on TSMC 180 when we wanted to run Monte Carlo, because only _mac transistors had mismatch flag.
I didn't know the _mac relates to macromodel, but as I know mismatch simulates threshold voltage differences too, which is not a parasitic related parameter, so I am not sure it is fully true above that macromodel means parasitics around the device.

I have tried that and it has worked fine.

Thank you!