Designer’s Guide Community

adesign

Community Member

Offline

Posts: 73

Mismatch simulation in Spectre
Mar 08^th, 2007, 10:59pm

Dear all,

I'm simulating a frequency divider circuit and the criticallity of design is the output duty-cycle. While simulating with Spectre without any pMOS/nMOS mismatch at cross-corners I'm getting a duty-cycle of 49% to 51%. However, at silicon the duty cycle obtained is ~43%.

Now, I want to introduce the mismatch in pMOS/nMOS devices.

Could anyone let me know how to introduce mismatch in these devices and then do the transient simulation?

All the help is appreciated.

Best Regards,

Back to top

IP Logged

mohta

Junior Member

Offline

Posts: 13

Re: Mismatch simulation in Spectre
Reply #1 - Mar 13^th, 2007, 11:46am

You could modify the netlist by hand to create the mismatch in the devices you want and re-run the simulation.

Alternatively, you can copy the schematic to a different view (say schematic_mismatch), modify the devices to be mismatched appropriately, and then use hierarchy editor to run the simulation using the same test bench by pointing to schematic_mismatch instead of the schematic view for that cell.

In both the above cases, you can simulate at most a few cases since everything is being done by hand, and it gets cumbersome if your design is large.

If your foundry provides mismatch models, then you also may be able to use the Monte Carlo tool (click on Tools --> Monte Carlo...) in ADE. The procedure for how to modify the model file to include mismatch parameters using a statistics block is available in a white paper on www.cadence.com/whitepapers

Thanks,
Setu

Back to top

IP Logged

achim.graupner

Community Member

Offline

Posts: 51

Re: Mismatch simulation in Spectre
Reply #2 - Mar 13^th, 2007, 10:46pm

Hi there,

if your design is senitive to mismatch and you want to compare simulations and silicon measurements you really need the mismatch parameters of your devices.
In order to carry out Setus recommendation of modifying the schematic you need to mismatch parameters. And if you got those parameters I think it is much less work to set up mismatch models. However, statistical models are state of the art, so first ask you PDK vendor.

Regards,
Achim

Back to top

Achim Graupner
ZMD AG, Dresden, Silicon Saxony, Germany

IP Logged

adesign Community Member Offline Posts: 73	Re: Mismatch simulation in Spectre Reply #3 - Mar 13^th, 2007, 11:35pm Thanks Achim and Setu for insight.
Back to top	IP Logged