kicad-developers team mailing list archive

[Andy Peters <devel@xxxxxxxxx>]

> On Jun 29, 2015, at 4:56 PM, Simon Richter <Simon.Richter@xxxxxxxxxx> wrote:
> 
> Hi,
> 
> On 30.06.2015 01:20, Chris Pavlina wrote:
> 
>> I'd happily implement the net tagging, though it'd require a bit of 
>> brainstorming from everyone on how those tags should be stored and 
>> shared with pcbnew.
> 
> I think it belongs in the netlist.
> 
> Ideally, I'd like this as attributes on the component pins, and then
> either inherited directly or at least verified by the ERC.

One argument against making this a pin attribute rather than a net attribute: FPGA pins can be single-ended or differential depending on the particular design. 

If your design flow says, “make a custom FPGA symbol for each design,” making a pin attribute indicating differential pairs is a great idea (as is making the pins input or output or bidirectional or tristate, instead of just making them all bidirectional in a “standard” symbol). (And this is great if you want to import pin assignments into the symbol from a constraint file.) (But I digress.)

But most people don’t go through that effort, so the attribute should be on the net.

> On one hand I like the idea of attaching pseudo components to the nets
> to add attributes, but some users might think of them as visual clutter.

We are an Altium shop at the day job, and we use differential nets all the time. I do think that the little red differential pair attribute indicator is clutter. It doesn’t indicate _which_ nets are the pairs; that’s handled by the net names with the _p _n suffix, so I think that the attribute indicator is redundant.

Now that I look at Tom’s example, where the attribute indicates everything you care about the net (including its partner), that’s not so bad! Assuming it can be hidden. And it makes the need to have a specific suffix or other magic netname go away. This user votes for Tom’s suggestion!

-