I agree, the autorouter is a good tool for determining overall 
complexity of a design ... if it is having major problems, one is likely 
to have problems doing it by hand as well.

Marvellous to watch it run... now that is a good tough CS algorithm 
problem to tackle, kind of amazing.

Earlier on in my career I used the autorouter a lot, and set up class 
attributes, etc.  But as I and projects got more sophisticated it became 
increasingly useless.  Boards with multiple different grounds, traces 
from one section to another that can't go over various ground planes, 
differential pairs that must stay together, etc. etc.

Everything is done by hand now, unfortunately.  Takes up a lot of my 
time... but its hard to say if a far more sophisticated autorouter would 
be much of a time saver, balancing the time it would take to tag signals 
and board areas with all the constraints and rules it would need to do a 
usable job vs. just doing it myself.

Anyway, FWIW, my process now is:
1. move parts into functional groups, by looking at a printout of the 
schematic and typing "move <whatever>" and dropping it, then checking 
that part off on the printout.  Repeat for everything.

2. for each functional group, now that the ratsnest is more sane, do the 
placement.  This takes longer to do as I get better at it, ironically, 
but in the end routing is easy - its clear what has to go where, and 
very few vias wind up being needed (usually).

3. run the autorouter to see what it does with the placement I did. 
maybe change the placements a little, etc.

4. rip up all autorouter work and now move functional groups around to 
make it tight and have connectors on the board edges where they need to 
be, etc.

5. run autorouter again, see if it seems impossibly tight, and decide 
4-layer vs. 2-layer if it is still a question.

6. rip it all up - or most of it anyway - and route route route by hand...

J


David P Harris wrote:
> Martin Klingensmith wrote:
> 
>> William Chops Westfield wrote:
>>
>>  
>>
>>> Beyond the routing rules derived from "class" attributes?
>>>
>>> I find Eagle autoroutes to be somewhat useful starting points
>>> for a more careful hand-route...
>>>
>>> BillW
>>>   
>>
>>
>> That's what I do as well. The ratsnest is very hard to visualize IMHO
>> because it connects net pins to the nearest neighbor. I run the
>> auto-router and determine which parts need to be moved. I usually rip it
>> up and then run the routes by hand anyway.
>>
>>  
>>
> Yes, and watching the autorouter working can give you an idea of where 
> the problem areas are --- if it keeps retrying a certain set of traces 
> for a component, then maybe you need more room around that particular 
> component, etc.
> David
> 
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist