Re: gEDA-user: Power nets in hierarchy

[sdb@cloud9.net (Stuart Brorson)]

From your drawing, it looks like Global is a subclass of Design, and
it points to a bunch of net instances.  (I don't quite recognize your
schema drawing formalism -- it looks similar to UML, but it's not.
Therefore, I am just guessing.)  Anyway, I imagine that your program's
job is to read in a .sch file and create this structure.  In this
case, I guess it makes sense.  OTOH, why wouldn't you just make
"Global" an attribute of Net (i.e. a part of the Net structure)?  On
the third hand, I guess that if you want to find all global nets
easily, without searching for them, yours is the way to do it. 

As for the big picture, I am still confused where your tool will be
used.  Is it fired up from within gschem?  Or is it stand-alone?

Stuart



> Stuart Brorson wrote:
> 
> >Why not just define an attribute "GLOBAL" which you could attach to a
> >net?  A global net would then have two attributes: NETNAME and GLOBAL,
> >for example:
> >
> >NETNAME=VCC
> >GLOBAL=1
> >
> >Then the netlister would know directly what type of net it was
> >handling, instead of having to figure it out from the attached
> >schematic symbol.  In any event, I often just use named nets for power
> >instead of attaching a power symbol.  Finally, there are other common
> >nets which aren't power nets but are also globals, such as RESET.
> >Nets like RESET don't always have an accepted device symbol.
> >
> >Stuart
> >
> I like the GLOBAL concept.  How does this class diagram (schema) look? 
>  I've added a Global class on Design to represent global signals.  I 
> think this will lead to easy manipulation of these things.
> 
> For example, after flattening, I might still have a bunch of small nets 
> that are all attached to a common global.  A global net flatting routing 
> could replace them all with a single net.  Another routine could be used 
> to thread the global signals through the hierarchy in case a netlist 
> format doesn't support globals.
> 
> I think POWER device types could also be useful.  It's just another way 
> to do the same thing, but in a way that feels familiar and looks good in 
> a schematic.
> 
> Bill