Re: gEDA-user: moving slotting to pcb?

[John Doty <jpd@xxxxxxxxx>]

On Jun 28, 2010, at 1:22 PM, Armin Faltl wrote:

> 
> 
> John Doty wrote:
>> On Jun 26, 2010, at 11:18 AM, DJ Delorie wrote:
>> 
>>  
>>> A random thought occurred to me today - why does gschem do slotting at
>>> all?  Why does it care about footprints and packages?
>>>    
>> 
>> There's no reason for it to. There's not even any reason to select part values. Putting these things into source schematics is a barrier to reuse.
>>  
> For me it is a barrier to reuse, to have to figure out a 2nd time, what values went into a tested circuit.

You're not thinking of reuse at the same level I am. Consider a Sallen-Key low pass filter. Why should we have to draw that more than once? Somebody could post a symbol along with a source schematic on gedasymbols, and then we could all use it. But each application requires different component values, a different op amp, etc. So a file like:

R1	value	51.1k
U1	device	OP90
...

would customize the circuit for your particular application. No need to figure out anything the second time (if the requirements are exactly the same): the file has all the info you need.

>> Another issue is that hierarchy plays poorly with slotting: it should be possible to split a slotted component across hierarchical blocks.
>>  
> I'm pretty sure you opt for opening and closing brackets in different source files as well.
> But you may have your uses for this request in super-complex VLSI symbols.
>>> Would it make
>>> more sense, from a design flow perspective, to just send the symbolic
>>> information to pcb and let pcb assign footprints and pinouts?
>>>    
>> 
>> No.
>> 
>> 1. The gschem->pcb is only one of the many flows gEDA supports. Please, please, please DJ, do not forget this.
>> 
>> 2. For documentation, most will want the pin numbers, refdeses, values, etc. to match the board as built.
>> 
>>  
> yes
>>> That way, gschem does all the symbolic stuff, and pcb does all the
>>> physical stuff.  It would, of course, mean major changes to pcb to
>>> handle "elements without footprints yet" and stuff, as well as mapping
>>> multiple refdes's to single elements.  Probably make power pin
>>> management more complex too, unless we came up with a new way to
>>> manage "hidden" pins.
>>>    
>> 
>> No reason to change pcb. What we need is a schematic processor to sit between gschem and gnetlist. Expand hierarchy, assign slots and values, turn the source schematics into project schematics for both netlisting and documentation. This is a clerical task, not a graphical task, so the extra input could just be a table of some sort.
>>  
> If it gets implemented like this, i.e. I have to feed a table to assign values in a schematic,
> I might become a hermit ;-) or go on searching for a new tool chain. For me, the schematic
> is the place where I see the logic of a circuit most clearly and know, why a component is
> there and what it means. It's the natural place to define it's spec therefore.
> With lots of identical components running these things through processors and tables
> may make sense, but if you have a bunch of amplifiers, voltage deviders, filters etc.
> where no single value resembles anything else, this is totally nuts.

Well, of course, you can enter attributes into the schematic now, and in some cases that makes sense. In other cases it wastes time. So it should be allowed, but not required.

> 
> I'd say pcb does the mechanical and ev. thermal stuff, the other physics is handled with
> side-notes on paper or simulators and for me input to the schematics.
> 
> Armin
> 
> 
> _______________________________________________
> geda-user mailing list
> geda-user@xxxxxxxxxxxxxx
> http://www.seul.org/cgi-bin/mailman/listinfo/geda-user
> 

John Doty              Noqsi Aerospace, Ltd.
http://www.noqsi.com/
jpd@xxxxxxxxx




_______________________________________________
geda-user mailing list
geda-user@xxxxxxxxxxxxxx
http://www.seul.org/cgi-bin/mailman/listinfo/geda-user