The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design Languages >> Verilog-AMS >> Compact model numerical precision
https://designers-guide.org/forum/YaBB.pl?num=1152263028

Message started by A_Programmer on Jul 7^th, 2006, 2:03am

Title: Compact model numerical precision
Post by A_Programmer on Jul 7^th, 2006, 2:03am

I have just implemented a compact model in Verilog A. But comparing with the original model implemented in Spice, there are some differences between their simulation results. Provided that I have eliminated all the logical errors. What are the possible sources of error? Can numberical precision harm me in this case? Thanks a lot.

Title: Re: Compact model numerical precision
Post by Geoffrey_Coram on Jul 7^th, 2006, 7:23am

My first guess would be that your physical constants are off.

Spice uses some old values for q (electron charge) and k (Boltzmann's constant), whereas the Verilog-AMS LRM specifies values from NIST. Unfortunately, the NIST values keep changing as those darn physicists get better at measuring these things. Colin McAndrew points out that this is a *terrible* thing for compact modeling; you want to fix a value for q so that the circuit simulator matches the device parameter extraction tool.

-Geoffrey

Title: Re: Compact model numerical precision
Post by A_Programmer on Jul 7^th, 2006, 11:04pm

Thanks a lot Geoffrey. Do you mean that I should add:

`define q 1.6e-19

in my verilog code?

Title: Re: Compact model numerical precision
Post by Geoffrey_Coram on Jul 10^th, 2006, 4:10am

A_Programmer wrote on Jul 7^th, 2006, 11:04pm:

Thanks a lot Geoffrey. Do you mean that I should add:

`define q 1.6e-19

in my verilog code?

Yikes! No, you should have a few digits more. Right now, the AMS LRM says
`define P_Q 1.602176462e-19

whereas BSIM3 has hard-coded
#define Charge_q 1.60219e-19

You can try that value, or you can try to find the actual value used in Spice, if you have the source code.