[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Gnucap-devel] Re: VerilogAMS

From: Jonathan David
Subject: [Gnucap-devel] Re: VerilogAMS
Date: Fri, 8 Sep 2006 00:17:52 -0700 (PDT)

the last one.. 

>> 2. What is different about the cadence AMSDesigner is that
>> the digital and analog are VERY closely integrated..  you can
>> use both digital and analog constructs in the same module..
>> Pretty sweet.. I'd like the same capability in a free tool,
>> so I'm willing to help work on it..

>I would like to know more about the user experience.  What is 
>the command set like? etc...

Which command set? 
Verilog(AMS) files for design description
compiled into libraries ( that supports multiple views per cell)
Spectre .scs files for models
a set of configuration files for 
which connect rules to use, pdk model sections to include, and 
and amsControl.scs for the analog side simulation controls.. 
(reltol, abstol, integ method etc.. ) 

Since this is all integrated with ncsim, use justs adds "-ams" to the command 
ncvlog -ams file.vams to compile a cell (MANY more options)
ncelab -ams <configuration> to elaborate (link) the design and make a snapshot
ncsim -ams <snapshot> to start the simuation, giving the user a command prompt 
for simulation commands
ncsim>run 20ns 
ncsim>probe "/tb/dut/out_p" (add signal to probelist)

you can also specify a tcl file to source at the begining of the simulation for 
all the run /probe commands

- of course cadence provides guified interfaces as well.. 

What I'd want to do differently? 
1. move the model library format from spice to verilog-A 
except for the minimum required spice primitives (R, L, C, V, I, E, G, )
all models (all diode and above models) I'd use compiled verilog-A compact 
and a Verilog-A syntax to specify the model parameters and the equivalent of 
"sections" (and where needed subcircuits) as used in spice models.
2. I don't really care about "interpreted" vs "compiled" .. an Interpreter 
might give you the ablilty to modify the network on the fly.. 
but you tend to sacrifice speed.. and probably a lot of it. 
My main goal for this is to have something I can use to start experimenting on 
a measurement language. so speed would not be that important... I'm not sure 
the flexibility buys me any advantage though. 

reply via email to

[Prev in Thread] Current Thread [Next in Thread]