gEDA-dev: Clarity or Sanity?

[Dan McMahill]

al davis wrote:

> The schematic should not have anything that is simulator 
> specific.  In particular, I really object to embedding 
> simulator commands in the schematic.  Even the single 
> attribute "value" where you put the simulator's text for the 
> value, is too simulator specific.

I mostly agree.  I really dislike the idea of putting model includes and 
analysis commands or other simulator setup things in a schematic.  The 
only place I don't completely agree is where you have an element which 
is really only supported in a particular simulator but that would 
typically be in a test bench schematic anyway.

> If you must put something that indicates things like "transient 
> analysis", have symbols like oscilloscope, DC voltmeter, 
> ammeter, etc.  Then let the netlister decide how to convert to 
> simulator commands.  These symbols are primarily for beginning 
> users, who might be making the transition from something like 
> Multi-sim.

I always envisioned something like the pcb mode (which I never quite 
finished) but for simulators where you'd get an extra menu.  The menu 
would be specific to the particular simulator and would be fully defined 
in a single scheme file.  I had most of whats needed in place and I 
think I had a path towards making it easy to support many different 
simulators and easily keep up with simulator capabilities but I didn't 
get around to coding it up.

For those familiar with the export dialogs in PCB, there is no exporter 
specific gui code.  If you add a new export HID, the export dialog for 
it builds itself automatically.  The mechanism is each export HID has a 
set of attributes and the export dialog box code knows how to build a 
dialog for editing an arbitrary set of attributes.  I think this concept 
could easily be applied to gschem and I started down that path.  The 
benefit is gschem actually doesn't care about any simulator details and 
you can update the simulator-specific menu by simply replacing a single 
scheme file instead of having to rebuilt gschem.

> I always either run the simulator interactively, or do multiple 
> runs in a script.  Having a box in the schematic where you put 
> simulator commands really only gets in the way.  Also, often I 
> change the circuit through the simulator based on the results.

In my mind the schematic should produce the connectivity part of the 
netlist and all simulator analysis commands should come from somewhere 
else.  I agree that having to change schematics to do a different sim is 
not the right way.

> One feature I would really like is a way to back-annotate those 
> changes, a way to update the schematic other than manually 
> changing the values.

Here is where a more powerful libgeda with a scripting interface for 
database manipulation can be really useful.  In a tool like cadence you 
can do all sorts of wierd (and all too often really useful) manipulation 
of the schematic or layout database via code.


-Dan