Can you route better than an autorouter?

[GamerAndds]

I have never tried auto routing before, but it sounds like a good choice (looking infront of me at 5 breadboards filled with components)  . I noticed with your pictures howerver you use a much bigger trace for say your VCC. How hard is this to change after, or do you do this before?

[jolshefsky]

Here's a comparison. For the autorouted board, I had to route ground net manually for some reason (probably because I put in a ground polygon before feeding it to the autorouter), but that's mostly just via to ground plane. Otherwise, the router routed the whole board.

Well, I think your placement of parts is a bit wonky to start with. I consider myself quite a novice, but I tend to group like parts together. An example is C11 which I guess is a filter cap for a voltage divider of some sort ... maybe a temperature sensor? Anyway, I'd want to put it near the resistors its associated with and pretty that up first. By orienting them right, you can get some short traces to bind them all together and then have a couple traces that look like the inputs and outputs they are in the schematic.

On the other hand, the autorouter decided to run that trace right between the pins on the clock crystal capacitors. That looks like a recipe for trouble, with a 20MHz square wave capacitively-coupled so close to what seems to be a signal trace.

Consider also how you handled the high-current traces. I think the VCC routing under U6, U7, and U8 looks pretty good, but the wonky line around the outside of the connector doesn't. My own technique is to make copper pours of each power supply input (ground, ±V or just +V as necessary), often splitting them under a row of chips so the power pins are on the right side pour.

Overall, to put it rather bluntly, your hand-routing technique isn't very good, so rather than saying the autorouter beads hand routing, I'd say your hand routing technique is worse than this autorouter. There are boards that have some really elegant routing solutions that an autorouter would just not bother to do.

In the end, for small-signal, low-frequency boards, the prettiness of the underlying routing is not too relevant as you're not going to have issues with crosstalk and the like. As such, any old autorouter, as long as it connects all the nets, will do fine. But once you add a few high-current, high-impedance, or high-frequency traces, knowing how to route things elegantly and getting good at the TETRIS game of component placement will help.

As I say, I'm no wizard at this, but here's an example of a controller board I made for an electric conversion of a riding lawnmower. Just the top layer from OSHPark's render ... the point is the pours provide lots of connections at low impedance, and the traces are clear and straightforward. The components are laid out so functions are grouped. (The odd connectors at the bottom allow either screw-terminals or Anderson PowerPole connectors to be installed in the board.)

[kerrsmith]

I've tried running Freerouter standalone (launched from DOS prompt), and launched from KiCad if you follow the instructions here http://amichalec.net/2015/10/kicad-upgrade/. Both methods work for me.

This method worked for me, I had to install Java first as I did not initially have this but Freerouter worked straight from KiCad once the .jar file was in KiCad's bin directory.

I have been using it for the last few days (currently working on laying out a Bitx20 on multiple boards so I can test different designs for VFO, BFO, crystal filter etc) and it has certainly been a great help in getting an initial board layout routed.  Once I have the components where I want them I have been using the autorouter and then tweaking the results to my liking.

It really helped when I spent ages routing out part of the circuit and then realised I needed and extra power connector on the other side of the board - it was too much to expect that I would be able to just move a few tracks to make space, the more I adjusted it the more needed adjusting until I was practically starting from the beginning again.  I cleared all my tracks, added the power connector and then ran the autorouter and within a second or two had a fully re-routed circuit.

Personally I do not mind using autorouters (the default KiCad one does a pretty good job) as sometimes I want to get my circuits up and running fast - once I know they are roughly working I can then go back change them by hand. I prefer spending my time testing them rather than on routing out a complex layout by hand.

If there are certain parts of a circuit I really want routed in a specific way I just add a 'keep out' area, autoroute the rest and then manually route the remaining tracks.

[TheDirty]

Well, I think your placement of parts is a bit wonky to start with...

Just a reminder that the OP is from 2010 and the original poster hasn't been on the board in a few months.  These necro posts are pretty confusing.

[exe]

I'm no professional and I can't claim I tried every autorouter, but.... I've never heard an autorouter to defeat humans. Now I'm using diptrace and though manual routing takes me day (literally) the result is waaaaaaaay better than I ever seen from its autorouter. And that's besides other things like I know what traces must be kept short.  Or how to place components. I believe I tried almost any free software (that anyhow runs on linux) with the same result -- autorouter is horrible.

Also, if I can't route efficiently I may consider do changes to the schematics. Like swapping opamps in two- and four-amp packages. Or choosing a capacitor of a different size.

So, if you spent just a few minutes for routing I'm not surprised the autorouter did the job better "Just" spend a whole day or two and your layout will shine.

[suicidaleggroll]

The biggest difference I see in the OP's results is simply due to the fact that the autorouter used piddly little traces everywhere, while the OP used some big fat traces, presumably where there was decent current flow.  You can't fit a 100 mil trace between the pads of an 0805, but it's no problem with a 10 mil trace.  That makes the autorouter's job a lot easier (and would likely cause the circuit to malfunction, assuming it really did need those large traces to begin with).

[timofonic]

Bump!


Any new stress test to autorouters? What about TopoR and others?

[Twistx77]

Hi,

I have used many autorouters and they are good for what they are good... not too much anyway. If you have a simple direct connections between components it might do a well job but if you have anything a bit complicated it will not do a good job for many reasons:

1) It will go to the limits imposed to it, and that means that even if it has more space it will go to the limit and therefor is more risky for production.

2) It can't move components or change pins. A good PCB designer will move components when routing to get a better route or will go to the schematic and swap pins when possible just to get a better routing, an autorouter does not do that.

3) It doesn't have (at least in most case) into account the type of signal that is running, high speed, analog, power, etc. When you are routing a realtively complex board you have to always take into account, EMI, noise, current through the track, length, etc.

Those are the main and sufficient reasons for me to not use an autorouter unless you have a really good one were you specify many, many things to make it do it somewhat what you really want.

[timofonic]

Aren't there autorouters smart enough to surpass those issues? I can't understand it!

[Karel]

Aren't there autorouters smart enough to surpass those issues? I can't understand it!

The problem is, such smart autorouters are so complex to setup & takes so much time to specify the requirements for the traces,
that you can manually route the board in less time.