I’ve always seen comments by people who don’t like autorouters and won’t use them. I’m wondering why that is. If the AR gets the board done, what’s the issue?
I can understand having to clean up a weird route now and then, but I don’t understand what’s gained by doing manually what can be done automatically. Comments?
Mike
Good autorouters are expensive. Most professional PCB designers don’t use them, because they can do a better job manually and their customers don’t mind paying for the work involved. I use the (excellent) Pulsonix router occasionally, mainly when I have lots of connector pins with non-critical signals.
Leon
leon_heller:
Good autorouters are expensive. Most professional PCB designers don’t use them, because they can do a better job manually and their customers don’t mind paying for the work involved. I use the (excellent) Pulsonix router occasionally, mainly when I have lots of connector pins with non-critical signals.
Leon
That raises a few questions. If most professional PCB designers don't use them, why do they exist in the first place? I imagine it's no small task to create a good one.
And what is meant by “a better job”? If all the nets are routed, what does “better” mean?
Mike
Because lots of people actually do use them, although the results aren’t necessarily very good. It’s just the professionals (people who spend all day designing PCBs for clients) who don’t use them.
A lot depends on component placement and selecting the correct parameters for the autorouter. A good autorouter will route the board to 100% completion, or close to it, and require very little work to be done subsequently. Even when people do use autorouters, critical tracks like supplies and grounds, and clocks, are routed manually.
The type of board is also important, no-one would think of autorouting an analogue circuit.
Leon
The one problem I see with using an autorouter is the fact that it has no way of knowing what the traces on the board are designed to do … or how much current may be flowing in a particular pathway. Some traces need to be routed in a particular way to avoid having one signal inducing noise onto another signal while other traces need to be heavier to accommodate particular current demands.
I’ve used it on occasion, but ripping up and re-tracing the errors usually doesn’t save much time. After routing a few boards by hand you start to see how you can optimize the layout so that manual routing isn’t all that hard to do.
Critical nets should be routed manually.
Leon
leon_heller:
Critical nets should be routed manually.
Exactly. I have no problem using the Eagle autorouter. I just make sure to route anything critical by hand, such as supplies, RF and anything else crucial.
Once those are done, I SAVE the board as is, save a new file and let the AR rip. If I don’t like what it did, I delete that newest board, go back to my manually routed version and start again.
An absolute key thing that I have found is making sure parts are grouped together properly. If I have an ST202 with four charge caps, I put those caps as close as possible to the ST202 and even make sure the airwires line up. Criss-crossing just more work for the AR and will deffinently lead to more vias than needed.
In Eagle, the “ratsnest” tool is your friend. Once the parts are in place, click the ratsnest and get the airwires shortened up to their shortest possible runs. This reduces work on the AR also. I don’t want the autorouter to go through it’s as*hole to get to it’s elbow.