Mail Archives: geda-user/2015/07/16/04:06:38
This message is in MIME format. The first part should be readable text,
while the remaining parts are likely unreadable without MIME-aware tools.
--8323329-1648758483-1437030885=:3383
Content-Type: TEXT/PLAIN; CHARSET=ISO-8859-7; format=flowed
Content-Transfer-Encoding: 8BIT
Content-ID: <alpine DOT DEB DOT 2 DOT 11 DOT 1507160914501 DOT 3383 AT nimbus>
On Wed, 15 Jul 2015, Chris Smith (space DOT dandy AT icloud DOT com) [via geda-user AT delorie DOT com] wrote:
> On 15 Jul 2015, at 21:37, John Doty <jpd AT noqsi DOT com> wrote:
>> And this is exactly the problem I see. A proper general mechanism would
>> allow one to associate *properties* with net segments, not just trace
>> widths.
>
> Where¢s the problem that you see? That ¡proper general mechanism¢ was
> exactly what was being discussed: the ability to split a net into
> segments, each of which could have generic attributes which could in
> turn be passed on to other tools. Trace widths was just a real-world
> example of where this might be useful.
I currently see the problem of over-generalizing a feature on basis of a
single application in a way that isn't helpful in other situations. So
IMHO, the first question which should be asked here is: what kind of
real-world situations are there where net attributes may be helpful?
I can currently come up with the following:
- managing voltage drop, oscillation, and heat by
- keeping trace resistance low and
- keeping common trace segments short, and
- improving signal quality in differential pairs by minimizing signal
run time differences
I'm not an electronics engineer, though, so please correct me if I got
this wrong.
The first is actually an attribute of type "from point/pin A to
point/pin B", the second of type "splitting a net into connected subnets",
and the third of type "A1 to B1 relates as A2 to B2".
For the second problem, there could be an elegant solution: define a
special attribute which marks a symbol as a "net attribute directive",
which is similar to "graphical=1" but makes a symbol treated as a net
segment by default. A subnet-aware gnetlist backend could then
differentiate between the subnets separated by the symbol. This doesn't
apply to the first problem, though, as it couldn't tell which of the pins
on either side require the thick trace and which don't.
Are there other situations where net attributes may be useful?
Roland
--8323329-1648758483-1437030885=:3383--
- Raw text -