The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Modeling >> Behavioral Models >> modelling nfet and pfet behavioural model verilogA
https://designers-guide.org/forum/YaBB.pl?num=1184796475

Message started by nishanis on Jul 18^th, 2007, 3:07pm

Title: modelling nfet and pfet behavioural model verilogA
Post by nishanis on Jul 18^th, 2007, 3:07pm

Hello,
I am trying to write a verilog-a behavioral model for pfet and nfet, i.e. using them as switches.
analog begin
if (analysis("static")) begin
Rch = Roff ;
end else
begin
if (V(G) >=Vth) Rch = Ron ;
else Rch = Roff ;
end;
I(S, D) <+ V(S, D) / transition(Rch,0,tr,tf);
end
endmodule
The model works fine. But when i use both pfet and nfet models to build an inverter i just get 0 at output. I think i need voltage as output. SO i tried
V(S, D) <+ I(S, D) * transition(Rch,0,tr,tf);
But this gives segmentation fault in spectre.
Can anybody tell some changes in the model where i can model voltages?
Also, is there a standard model for nfet or pfet to operate them as switches which can be used to build logic gates.
Waiting for reply.
:)

Title: Re: modelling nfet and pfet behavioural model veri
Post by Ken Kundert on Jul 18^th, 2007, 7:33pm

Try mos0, a built-in primitive in Spectre.

Title: Re: modelling nfet and pfet behavioural model veri
Post by Ken Kundert on Jul 18^th, 2007, 7:35pm

Alternatively, you can use the logic expression capability in Vlog-A to build true behavioral models of the gates.

-Ken

Title: Re: modelling nfet and pfet behavioural model veri
Post by Geoffrey_Coram on Jul 19^th, 2007, 10:25am

nishanis wrote on Jul 18^th, 2007, 3:07pm:

V(S, D) <+ I(S, D) * transition(Rch,0,tr,tf);
But this gives segmentation fault in spectre.

Segmentation fault? Cool. Are you running a current version of the simulator so you can open a support request?

When you used the curent formulation, did you remember to invert the sense of the gate voltage? I'd also have used V(G,S) rather than just V(G).

Title: Re: modelling nfet and pfet behavioural model veri
Post by nishanis on Jul 20^th, 2007, 10:38am

Hello Ken,

Thanks for replying.
Can you explain about the logic expression capability in Vlog-a. are you referring to model writer to get the behavioral model?
Thanks

Title: Re: modelling nfet and pfet behavioural model veri
Post by nishanis on Jul 20^th, 2007, 10:40am

Hello Geoffrey_Coram,

I tried your solution, its still giving the same segmentation fault.
I tried doing this also, but same error
analog begin

if (analysis("static")) begin
Rch = Roff ;
end else

begin
if (V(G,S) < Vth) Rch = Ron ;
else Rch = Roff ;
end

a = transition(Rch,0,tr,tf);
V(S,D) <+ I(S,D) * a;

end

endmodule

Hope to get further help from you.
thanks
:)

Title: Re: modelling nfet and pfet behavioural model veri
Post by Ken Kundert on Jul 20^th, 2007, 12:56pm

Go to the Verilog-AMS page and download the basic inverter. This is the starting point. Add more inputs and calculate the output using simple logical expressions (&&, ||, !, etc.).

-Ken

Title: Re: modelling nfet and pfet behavioural model veri
Post by nishanis on Jul 20^th, 2007, 2:34pm

Hello ken,

I have done one set of simulation using those codes to which you are referring. But what i am trying to achieve is i have two models (pfet and nfet) so if anybody designs gates like, AOI222, or AO or any other, i wouldn't have to model another code. spectre would use those fet models and would act as switch and implement the functioning.
I got away with segmentation fault, using code below. But at output i am getting 1.8v straight line. its not switching when vth drops. Its always at 1.8v when i/p is at 1.8v and 0 v when i/p as it 0v.

analog begin
@(cross (V(control)- vth, 0,
slack, control.potential.abstol)) begin

if (V(control) >= vth )
rout = roff;
else
rout = ron;

end
a = transition(rout,0,100p,100p);

V(vin,vout) <+ I(vin,vout) * a;
end
endmodule

Title: Re: modelling nfet and pfet behavioural model veri
Post by sheldon on Jul 25^th, 2007, 9:31pm

Nishanis,

In this case it might be easier to use the sw_no from the bmslib
to create the n-channel and p-channel device models. The model
formulation of the sw_no does not use the "if" function. This approach
works well for switching regulators[synchronous buck converters].
One point to consider is the on resistance of the switch, it needs
to be relatively high to minimize shoot-through current.

Best Regards,

Sheldon