The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design Languages >> Verilog-AMS >> AC Analysis of Verilog-A Module
https://designers-guide.org/forum/YaBB.pl?num=1184363901

Message started by Nano_Model on Jul 13^th, 2007, 2:58pm

Title: AC Analysis of Verilog-A Module
Post by Nano_Model on Jul 13^th, 2007, 2:58pm

Hi,

I have a complicated 4-terminal device module written in Verilog-A. It works for DC and transient analyzes. However, when I use it to run AC analysis, what I get from the output are all unexpected zeros. What's a possible reason for this?

I read previous post saying that the simulator builds an equivalent small-signal circuit based on DC information it has. How does it actually work? Like...does it build a typical hybrid-pi model, a T-model? or? Can I "probe" the small-signal capacitances calculated by the simulator in this case? Give the complexity of my module, I wonder how can the simulator build a SS circuit for it. Does it just treat my module as a black-box and probe its terminal behavior?

Thanks a lot,
Nanodel

Title: Re: AC Analysis of Verilog-A Module
Post by Ken Kundert on Jul 13^th, 2007, 4:39pm

Did you provide a small-signal stimulus?

-Ken

Title: Re: AC Analysis of Verilog-A Module
Post by Nano_Model on Jul 15^th, 2007, 1:47pm

Yes, I used (in SPICE) .ac analysis...

Title: Re: AC Analysis of Verilog-A Module
Post by Geoffrey_Coram on Jul 16^th, 2007, 4:46am

You didn't answer Ken's question. Having an ac analysis doesn't mean that there's actually an ac source (small-signal stimulus) in the circuit. If you don't have a stimulus, then you'll always get zero output.

The following is a 5V dc source *with* a magnitude=1 small-signal stimulus.
V1 1 0 5 ac 1

The simulator takes the (symbolic) derivatives of all the expressions you used in your module and uses them to determine the ac response. It needs the derivatives anyway for using Newton's method to solve the dc operating point.

Title: Re: AC Analysis of Verilog-A Module
Post by Nano_Model on Jul 16^th, 2007, 10:56am

Thanks Geoffrey. Yes, I did include an ac stimulus like what you've said. Nevertheless, I still got 0.0 as my output.

Now from your answer, I have another question. Is the simulator always able to compute the symbolic derivatives of any expression? My drain current expression is not analytical. It involves solving a self-consistent equation iteratively. So, in this case, I wonder how the simulator computes the derivatives and linearizes my module.? :-?

Title: Re: AC Analysis of Verilog-A Module
Post by Geoffrey_Coram on Jul 19^th, 2007, 4:44am

I believe the simulator is always able to compute the symbolic derivative. Even if you have an iterative solution, it's all composed of assignments:

//dx_dv = dx0_dv;
x = x0;
for (...) begin
// dx_dv = dx_dv + dupdate_dv;
x = x + update;
end

where the //commented lines are added by the compiler generating the C code from the Verilog-A.

Title: Re: AC Analysis of Verilog-A Module
Post by Nano_Model on Jul 21^st, 2007, 10:44am

Cool.

Are there any manuals/books/documents for me to understand how the simulator works? Like...how the verilogA code is converted to C...and how the simulator runs various analysis... :)

Title: Re: AC Analysis of Verilog-A Module
Post by Geoffrey_Coram on Jul 27^th, 2007, 6:36am

I don't think there are any books (nor even papers) on how Verilog-A is converted to C. There might be something about how Verilog-A equations are translated to equations and unknowns in the simulator in "The Designer's Guide to Verilog-AMS." (Chapter 3, Analog Modeling, Section 2.1, Conservative Systems)

As for how the simulator runs analyses, "The Designer's Guide to Spice & Spectre" is probably going to cover it for you.

If you click "Books" above, you can get to some links that will chip a few cents (maybe a buck or two?) back to help support this site.

For your actual problem: can you try writing an explicit function for the drain current (maybe the "initial guess" for the iteration)?

Also, check your if() statements carefully. Sometimes people things like:
if (x < -100)
Iout = 0;
else
Iout = f(x);

forgetting that Iout = 0 also means dIout/dx = 0, and this would give you a zero output in ac.