Mail Archives: geda-user/2015/10/14/12:37:31
On Wed, Oct 14, 2015 at 09:55:03AM -0600, John Doty wrote:
> Great stuff! Is this published somewhere?
No, the rest of the scheme file is in need of serious cleanup and very
specific to conventions between the person doing the layout and myself,
in contrast with the fairly clean net writing part I showed in the
previous message.
Gabriel
>
> On Oct 14, 2015, at 2:06 AM, Gabriel Paubert (paubert AT iram DOT es) [via geda-user AT delorie DOT com] <geda-user AT delorie DOT com> wrote:
>
> > On Tue, Oct 13, 2015 at 03:09:25PM -0600, John Doty wrote:
> >>
> >> On Oct 13, 2015, at 2:39 PM, DJ Delorie <dj AT delorie DOT com> wrote:
> >>
> >>>
> >>>> Please not in libgeda or the geda-gaf core tools. That has
> >>>> tremendous potential to break things. geda-gaf is a mature toolkit
> >>>> in production use: changes to the core should not be taken
> >>>> lightly. New tools and new plug-ins are great. We have a transparent
> >>>> schematic format and powerful scripting mechanisms to give us power
> >>>> and to allow us to keep that power from interfering with each
> >>>> other's objectives.
> >>>
> >>> What about exposing attributes-on-nets to the backends?
> >>>
> >>
> >> I’ve wanted that for a long time, but I note that you can effectively attach attributes to nets using symbols. The fine primitive gnetlist:graphical-objs-in-net-with-attrib-get-attrib (the longest name in the gnetlist namespace) allows a backend to extract them. This is more versatile, as it allows the back end not only access to attributes, but allows a symbol to define relationships between nets. Consider the following, which could allow a backend to deduce that a pair of nets is a transmission line, and give it a name:
> >>
> >
> > That's what I've been using for about 3 years to generate netlists
> > with hints for the PCB designer. I have only two symbols:
> > - one with a single pin for 50Ω transmission lines,
> > - one with two pins for differential lines.
> >
> > I've attached them to this file. They are as small as I possibly could
> > make them because I found that larger ones were too distracting (and
> > also consuming too much space on schematics).
> >
> >
> > Here is a the net generating part of gnet-rinf.scm (for
> > CadStar):
> >
> >
> > ;;
> > ;; Display the individual net connections
> > ;; starting from the third if they exist.
> > ;;
> > (define rinf:write-connections
> > (lambda (netpins port)
> > (if (not (null? netpins))
> > (begin
> > (format port " ~a ~a~%"
> > (caar netpins)
> > (cadar netpins))
> > (rinf:write-connections (cdr netpins) port)))))
> >
> > ;;
> > ;; Display all pins from the net starting from the passed one
> > ;; which is the second one.
> > ;;
> > ;; This is the line with the .TER command.
> > ;;
> > (define rinf:write-net-pins
> > (lambda (netpins port)
> > (format port ".TER ~a ~a~%"
> > (caar netpins)
> > (cadar netpins))
> > (rinf:write-connections (cdr netpins) port)))
> >
> > ;;
> > ;; Write net: first line starting .ADD_TER
> > ;;
> > ;; Only write nets that have at least 2 pins: gnetlist
> > ;; produces quite a lot of single pin nets, even when
> > ;; a pin is explicitly connected to a No-connect symbol!
> > ;; There are also cases where the list of pins is empty!
> > (define rinf:write-net
> > (lambda (port netnames)
> > (if (not (null? netnames))
> > (let ((netpins (gnetlist:get-all-connections (car netnames)))
> > (routing (gnetlist:graphical-objs-in-net-with-attrib-get-attrib
> > (car netnames)
> > "device=Netlist_Directive"
> > "net_route_code"))
> > (grouping (gnetlist:graphical-objs-in-net-with-attrib-get-attrib
> > (car netnames)
> > "device=Netlist_Directive"
> > "bus_name")))
> > (cond ((> (length netpins) 1)
> > (format port ".ADD_TER ~a ~a ~s~%"
> > (caar netpins)
> > (cadar netpins)
> > (car netnames))
> > (rinf:write-net-pins (cdr netpins) port)
> > (if (not (null? routing))
> > (format port ".ATT_TRE ~a ~a net_route_code ~a~%"
> > (caar netpins) (cadar netpins) (car routing)))
> > (if (not (null? grouping))
> > (format port ".ATT_TRE ~a ~a bus_name ~a~%"
> > (caar netpins) (cadar netpins) (car grouping)))
> > (newline port)))
> > (rinf:write-net port (cdr netnames))))))
> >
> >
> >
> > ;;
> > ;; Write the net part of the gEDA format
> > ;;
> > (define rinf:nets
> > (lambda (port)
> > (let ((all-uniq-nets (gnetlist:get-all-unique-nets "dummy")))
> > (rinf:write-net port all-uniq-nets))))
> >
> > But you have to know about one caveat: when the net traverses the
> > hierarchy, you have to put these symbols on the highest level sheet
> > where the net appears in the schematics. This could be construed as
> > a bug in gnetlist.
> >
> >> So, I see no urgency to exposing attributes on nets.
> >
> > I still think it would be convenient. Not absolutely necessary, but
> > convenient. However I like the fact that the differential symbol
> > explicitly shows which nets form the differential pairs. For single
> > ended transmission lines, I feel that the 50Ω symbols, even very small,
> > clutter the schematics.
> >
> > Gabriel
> > <50Ohm-1.sym><diff-1.sym>
>
> John Doty Noqsi Aerospace, Ltd.
> http://www.noqsi.com/
> jpd AT noqsi DOT com
>
>
- Raw text -