[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: gEDA-user: Marketing gEDA - was - Re: Professional PCB help using geda?



On Thursday 06 September 2007, Larry Doolittle wrote:
> PCB can't do the automated parasitic extraction and signal
> integrity simulations that high end software suites can.

Actually, a first cut, without crosstalk, is fairly easy to do.  
First, we need a translator, to translate from the PCB format 
to a simulation language.  Then, we need electrical models of 
the traces, vias, etc.  Then, we need to model the drivers and 
receivers, which are usually specified as IBIS models.

The easiest translation is to Verilog-A, the structural subset, 
which the gnucap snapshot now accepts.  It should be a lossless 
transformation, with a back translation also available.

The electrical models are mostly transmission lines.  In high 
end tools, they usually use a field solver to extract 
electrical parameters from physical dimensions.  If the layout 
is restricted to certain simple shapes, which it is, there are 
closed form equations that can do the extraction.  This is very 
simple to do.  It can be programmed as Verilog-A modules.

I have IBIS-3 about 90% complete. Completing it has not been a 
priority.  It could become a priority if the rest is done.  
Actually, it should be rewritten to take advantage of the new 
plugin system, and to eliminate experimental code.  IBIS is a 
very difficult language to implement correctly.  It's easy to 
implement incorrectly.

A full simulation with crosstalk is much harder.  The added 
requirements are a multi-port coupled transmission line, and a 
field solver to extract the parameters.  There is Free software 
available to do this.  It just needs to be incorporated into 
the system.

As to the likelihood of it happening....  If someone else steps 
up to do part 1, I will follow with part 2 and 3.  If someone 
is willing to pay for it, it rises to the top.



The high end software suites usually have the extraction and 
simulation as separate products.  Usually there is a multi-step 
process to transfer the data.  Since they are separate, it is 
common to mix products from different (even competing) vendors.

Most of the SI simulators are not Spice based.  Spice 
performance is poor when most of the circuit is transmission 
lines.  The SI simulators use a different algorithm that is 
much faster for this type of circuit, but does not handle 
general Spice circuits.

The gnucap algorithm handles transmission lines similarly to 
transmission line simulators, so it should run fast like a 
transmission line simulator.  It does run faster than Spice for 
this, but I have never properly benchmarked it against any 
transmission line simulator, so I don't know for sure.


_______________________________________________
geda-user mailing list
geda-user@xxxxxxxxxxxxxx
http://www.seul.org/cgi-bin/mailman/listinfo/geda-user