autorouting setup

normally they do not need to be changed. if your board will not route it

is likely due to issues of the placement and routing grids beings too

coarse.

the parameters indicate the what "cost" each of the routing entities

uses, the router will try to minimize the total cost of the board. you

will note that there are many passes with some different emphasis that

try to give you a nice board. this is characteristic of most "simulated

annealing" type optimizers, so perhaps EAGLE uses this as its algorithm.

in general, if your board routes 100% or reasonably close to it, you

would not need to alter the costs. rather spend your precious time

manually "cleaning up" any awkward results from the autorouter.

most problems i've seen with autorouting are the result of the

aforementioned grid or layer setup problems.

I´m really like Eagle software.  I believe that is a great software, very

easy to use and good enough for almost all problem involving PCB design.

My boards routes 100%.  It´s not my problem.

But I thing that trace alignment, trace distribuition and a reduced numbers

of vias are important too.  Why I would like to have better controll of

routing process.

I did several experiments, having routed some PCBs many times, changing some

parameters.  I realized that some rules are very important to get good

resoults:

1) Fine-pitch components should be aligned with the grid  (all pads);

2) Routing grid should be (at least) equal the small pitch of fine-ptich

components.  If the pitch is not a exact routing grid multiple (ex.:

pitch=0.50mm and routing grid = 0.25mm), we need a very small routing grid

to route same PCB.

3) DRC clearance should be small than pitch minus trace width (its a

physic´s problem, space is need for trace gets pad).  At least for the

clearance beetwen wires should be in according to this roule.

This few rules (some of them are obvious) and many others not listed could

be writen in a small tutorial to help people to get best resoults with

Eagle, without wasting time.

But, even using roules like those above, I´m sure that we can get more from

Eagle software.  If there are several paramaters available in routing

options, why don´t use them in our benefit ?  Why don´t clarify users about

meaning of this paramenters ?  For example, what does "costs" mean ?

Many people use Eagle because software´s qualities.  And I´d like to see

more and more people using Eagle.

Mainly, I´d like to see Eagle being used in the professional world, without

any distinction from others called "professionals softwares".  I´m sure

Eagle software has resources to make this real.  But Eagle people must teach

us to use all resources of Eagle software cans offer.

Your routing grid should be about 5 to 10 times smaller than lead pitch.

and yes, of course it must include the lead grid or the traces can not

route!

Clearances in DRC are should be determined by the limits of

manufacturing processes and are used to make sure the routed board can

be reliably manufactured, you can set them to whatever limits your board

house can manufacture. Setting them to get a good routing is wrong.

If you want less vias, you can try increasing the "cost" of vias

relative to other entities. The router evaluates which routing is best

by minimizing the "cost" of the entire board, so if you make vias 10x

more costly, it will make eliminating them 10x more important in the

routing process. This will also make routing 10x more difficult for the

computer and perhaps take 10x or even 100x longer.

If you have critical trace alignments, like a hi speed differential

pair, I will tell you that the EAGLE autorouter will never be able to

handle that, do it manually. If you want two traces to be close or far

away, well some software will do this, not EAGLE, do it manually. if you

want your board to have nice patterns on it, that usually means more

you should realize that no autorouter produces work comparable to a

manual routing. the parameters that work for one board will likely not

work so well on the next one. also the various passes of optimization

are usually more important than tweaking the values, lowering the via

count may have a side effect of awkward trace placement for example. the

object is a good balanced outcome for all parameters.

generally, i would start with power and other critical nets and manually

route them first. then i would let the autorouter finish the board if it

is complex.

It is true that EAGLE documentation for the autorouter is weak, but it

is also true that anything critical is normally handled manually, so

there is little interest in tweaking the autorouter, mainly people want

it to get 100% routing. However, if you like to experiment with it and

you think you have improved parameters, I'm sure some will like to try

your tweaked parameters on their board, so please share them with us!