Re: [ng-spice] Mixed signal simulation.


To ng-spice@ieee.ing.uniroma1.it
From Kev <kev@v-ms.com>
Date Fri, 13 Aug 1999 10:23:55 +0100
Delivered-To mailing list ng-spice@ieee.ing.uniroma1.it
Mailing-List contact ng-spice-help@ieee.ing.uniroma1.it; run by ezmlm
Organization V2000 Project [Mixed Signal Simulation]
References <Pine.LNX.3.96.990813070124.14420A-100000@ieee.ing.uniroma1.it >
Reply-To ng-spice@ieee.ing.uniroma1.it
Sender v2k@v-ms.com

Paolo Nenzi wrote:
> 
> On Thu, 12 Aug 1999, James Swonger wrote:
> 
> > relaxation, analog behavioral and SPICE. No process-to-
> > process sync/RPC to slow it down, and each section of the
> > circuit is supposed to be (maybe with some hand-fiddling)
> > sent to and solved by its appropriate simulator engine.
> 
> Ok, but how do you connect a digitial simulator whit an analog one (I am
> new to this field, so be patient): the timestep control of the analog 
>section
> can interefere with the digital part: I mean how we can avoid to loose a
> state change in a digital signal because the timestep is too large ?
> There should be surely a solution to this problem, otherwise mixed mode
> simulation will be not very useful.

['Analog behavioral' is implemented on top of Spice, and
'relaxation' is similar to Spice (except you don't solve
a matrix to work out where you go next).]

The algorithm for solving analog/Spice type simulation
is actually event-driven, but uses 'real' time - e.g.
starting at time 'T' you solve the equations etc.to get
a solution at time 'T@a' (known as the "acceptance time"),
then step to that time. If you mix in other simulation
which has events between T & T@a that affect the analog
simulation (say 'T@e'), you bring the acceptance time
forward to T@e (T@a := T@e) and recalculate the solution.
If a process requires an analog value (e.g. a signal voltage)
at T+e then it can be calculated by interpolating between
the T & T+a values. The key element is not actually
performing 'acceptance' until all processes in the
simulation have reached T+a (which is where the old
RPC simulators usually went wrong).

> Can you give some good refrence on the subject ?

Unfortunately not. However here is link to some old
Verilog-AMS proposals that discusses some of the issues:

  http://www.dnai.com/~orpheus/lrm/lrm.htm
  http://www.dnai.com/~orpheus/lrm/

Regards,
Kev.

-- 
http://www.v-ms.com
mailto:admin@v-ms.com   Mixed Signal Simulation


Partial thread listing: