Designer’s Guide Community

Tommy

Junior Member

Offline

Posts: 19
Tokyo, Japan

Time varying variable to laplace_nd()
May 31^st, 2006, 2:17am

HI,
I know this is possible with Verilog-AMS.
My question is specific to Cadence Verilog-A in IC5.1.41

Here is a bit of code that fails in Cadence Verilog-A.

analog begin
@(initial_step("static")) begin
nout[0] = V(in)*(R/(R+R_L));
nout[1] = V(in)*(R/(R+R_L))*(R_C*C);
dout[0] = 1;
dout[1] = V(in)/R;
dout[2] = L*C;
end

V(out) <+ laplace_nd(V(dr) , nout , dout);

When I pass a voltage (which is an input to the block) to laplace_nd using an array
the simulation fails. Is there a work around?

Thanks
Tommy

Back to top

IP Logged

Geoffrey_Coram

Senior Fellow

Offline

Posts: 1998
Massachusetts, USA

Re: Time varying variable to laplace_nd()
Reply #1 - May 31^st, 2006, 9:47am

According to the Verilog-AMS LRM (version 2.2), in Table 4-23, Section 4.4.15, the zeros and poles arguments to laplace_nd are supposed to be constant expressions. Thus, I am surprised that the simulator let you even compile the model.

The LRM is a little sloppy in 4.4.12.4 about a "vector of M real numbers"; I believe it is intended to be a concatenation like the example in 4.4.12.5:

V(out) <+ laplace_nd(V(in), {0,1}, {-1,0,1});

Back to top

If at first you do succeed, STOP, raise your standards, and stop wasting your time.

IP Logged

Tommy

Junior Member

Offline

Posts: 19
Tokyo, Japan

Re: Time varying variable to laplace_nd()
Reply #2 - May 31^st, 2006, 10:11pm

Geoffrey_Coram wrote on May 31^st, 2006, 9:47am:

According to the Verilog-AMS LRM (version 2.2), in Table 4-23, Section 4.4.15, the zeros and poles arguments to laplace_nd are supposed to be constant expressions. Thus, I am surprised that the simulator let you even compile the model.

The LRM is a little sloppy in 4.4.12.4 about a "vector of M real numbers"; I believe it is intended to be a concatenation like the example in 4.4.12.5:

V(out) <+ laplace_nd(V(in), {0,1}, {-1,0,1});

Mr. Coram,

Here is the exact bit of code that compiled in verilog-AMS
The coefficents are'nt in an array.

analog begin
//numerator coefficients
n0 = V(in)*(R/(R+R_L));
n1 = V(in)*(R/(R+R_L))*(R_C*C);
n0_2 = V(in)/R;
n1_2 = V(in)*C;

//denominator coefficients
d1 = (C * (R_C + ((R*R_L)/(R+R_L)))) + (L/(R+R_L));
d2 = L*C*((R+R_C)/(R+R_L));
d1_2 = (L/R) + (C*R_C);
d2_2 = L*C;

if (analysis("static")) begin
V(out) <+ V(in)*V(d);
end

V(out) <+ laplace_nd(V(d), {n0, n1}, {1, d1, d2});
I(outsense) <+ -1*laplace_nd(V(d), {n0_2, n1_2}, {1, d1_2, d2_2});
end

Back to top

IP Logged

Geoffrey_Coram

Senior Fellow

Offline

Posts: 1998
Massachusetts, USA

Re: Time varying variable to laplace_nd()
Reply #3 - Jun 1^st, 2006, 4:10am

Tommy wrote on May 31^st, 2006, 10:11pm:

Here is the exact bit of code that compiled in verilog-AMS
The coefficents are'nt in an array.

"verilog-AMS" is a language, not a simulator. What *simulator* were you using?

But the LRM says what it says; I would argue that that simulator has a bug if it allows the coefficients to be voltage-dependent, and you can't rely on it. What happens if you connect some feedback, such that V(in) depends on V(out)? Convergence in an analog simulator could be completely messed up.

That said, it looks to me that you could simply factor out V(in) from the numerator coefficients. I assume that R, C, R_L, and R_C are all parameters? In that case, I'd suggest making the coefficients parameters (localparams if your simulator supports them, which prevents someone from overriding them):
parameter real R = 50;
parameter real L = 10n;
parameter real C = 1p;
parameter real R_L = 50;
parameter real R_C = 50;
parameter real n0 = R / (R+R_L);
parameter real n1 = (R/(R+R_L))*(R_C*C);
parameter real d1 = (C * (R_C + ((R*R_L)/(R+R_L)))) + (L/(R+R_L));
parameter real d2 = L*C*((R+R_C)/(R+R_L));

and then you can write
V(out) <+ laplace_nd(V(d)* V(in), {n0, n1}, {1, d1, d2});

Back to top

If at first you do succeed, STOP, raise your standards, and stop wasting your time.

IP Logged

Tommy Junior Member Offline Posts: 19 Tokyo, Japan	Re: Time varying variable to laplace_nd() Reply #4 - Jun 1^st, 2006, 9:31pm Mr. Coram, I used silvaco smartspice which refers to Accellera Verilog-AMS std . Yes factoring out V(in) in this case works. But I was looking at a case where I cannot factor out the term. Thanks Tom
Back to top	IP Logged

Geoffrey_Coram Senior Fellow Offline Posts: 1998 Massachusetts, USA	Re: Time varying variable to laplace_nd() Reply #5 - Jun 2^nd, 2006, 4:17am Do the filter coefficients really depend on the input voltage as a function of time?
Back to top	If at first you do succeed, STOP, raise your standards, and stop wasting your time. IP Logged