The Auto-Router Challenge - will it work?

[zapta]

PCB auto routers are often dismissed here as 'waste of time' and 'non useful' so I decided to have a real world experiment using the Eagle auto router.

The experiment uses the circuit of a motorized paste injector controller I am working on (a hobby project). It includes a 16Mhz clock, inductive loads and an analog input (schematic here https://github.com/zapta/misc/raw/master/paste_injector/eagle/paste_injector_schematic.pdf )

What do you think? Will it work?

The rules are:

1. No manual editing or traces, not even post routing minor cleanups.
2. General auto router settings are ok (e.g. the cost of a bottom layer trace) but no per net (e.g. network classes) or other circuit specific settings.
3. Using a casual schematic design. Minimal bypass caps, using internal weak pullups,  no separation between digital and analog ground, etc.
4. PCB manufactured by OSHPark.
5. Components are soldered by hand.
6. Board rework is OK as long as it is not routing related (e.g. due to footprint or schematic error).


Success criteria:
1. The circuit works in the intended use cases (hobby use in room temperature, standard 5V usb charger and cable).
2. No noticeable issues, glitches, etc.

The auto routing:


Design files and schematic:
https://github.com/zapta/misc/tree/master/paste_injector/eagle

[Tac Eht Xilef]

What do you think? Will it work?

Probably; it's a pretty simple use-case on a relatively spacious board.

2. General auto router settings are ok (e.g. the cost of a bottom layer trace) but no per net (e.g. network classes) or other circuit specific settings.

Why not? Separate classes, particularly for power nets, is basic good schematic practice. Any failure due to e.g. insufficient track width or clearance isn't the autorouter's fault if you don't tell it about the requirements.

3. Using a casual schematic design. Minimal bypass caps, using internal weak pullups,  no separation between digital and analog ground, etc.

Again, why not? Those are circuit issues entirely in the hands of the designer; nothing to do with the autorouter...

[AlfBaz]

...
What do you think? Will it work?
...

It doesn't matter does it? People willing to pay for it is all that counts

[zapta]

Why not? Separate classes, particularly for power nets, is basic good schematic practice. Any failure due to e.g. insufficient track width or clearance isn't the autorouter's fault if you don't tell it about the requirements.

The less circuit specific settings, the more automatic it is.

It doesn't matter does it? People willing to pay for it is all that counts

It's a hobby. Money is only going out.  ;-)

[ez24]

what kind of paste - toothpaste ?  I see there is a clock so is this an alarm clock that injects toothpaste into someone's mouth to wait them up?

[zapta]

what kind of paste - toothpaste ?  I see there is a clock so is this an alarm clock that injects toothpaste into someone's mouth to wait them up?

Solder paste. It drives a 28BYJ-48 stepper motor.

https://www.eevblog.com/forum/reviews/solder-pasteflux-manual-syringe-applicator/msg747923/#msg747923

[DTJ]

It's a fairly simple board and lots of real estate as someone else said - it'll be ok. Manual routing would save a few vias. Maybe give the auto router some track width rules for the motor drive tracks.

On the ULN200x the auto-router has routed pins 5 - 8 together. I find this sometimes results in solder bridging between those pins. Not a functional issue but the production people can spot them as a fault/bridge during inspection.

I did a very similar job years ago and needed a small cap to gnd on the speed pot wiper to filter out the lumps.


Just out of interest - what does the vertical line connected to ground at the back of the fwd/rev switches mean. Good to see I'm not the only one that scatters "TO DO:" comments all over schematics (and layouts).
 


Be sure to upload pics of the finished & operating result !

[zapta]

Just out of interest - what does the vertical line connected to ground at the back of the fwd/rev switches mean. Good to see I'm not the only one that scatters "TO DO:" comments all over schematics (and layouts).

It's the ESD ground shield at the top of the button. Less chance of zapping something by touching the button.

As for analog noise, I can handle it in firmware if needed. It has a shared ground with the motor driver and I didn't bother adding a cap on the aref  pin. Wanted to challenge the auto router approach regarding noise.

[tonyarkles]

Why not? Separate classes, particularly for power nets, is basic good schematic practice. Any failure due to e.g. insufficient track width or clearance isn't the autorouter's fault if you don't tell it about the requirements.

The less circuit specific settings, the more automatic it is.

Interesting. I set net classes even when I'm manually routing. I think of it more as a specification than a setting.

[smbaker]

What do you think? Will it work?

Admittedly, I'm an amateur when it comes to PCB design, but I use the Eagle autorouter a lot and have pretty good luck with it producing 'functional' but not necessarily pretty or optimal boards.

The biggest problem I have with it is when you start to increase density of components and it starts to pick weird routes. I'll look at it and say 'gee, if I just rotated that resistor, it could route it much better'. I move that resistor, ripup, and re-autoroute, and it routes it in some far worse nonsensical way. It's just a dumb computer, it doesn't have intuition and it doesn't understand when I'm trying to "help" it.

I've watched some of Dave's videos and wish I had the skill and experience to do manual routing like he does. 

[zapta]

Got the board from OSHPark, assembled it and it works like charm.

All the world auto routers got vindicated!  They are useful for some practical circuits.

[homebrew]

Of course it did work - well sort of. What would you expect? As bad as these auto-routers are, they connect the correct pins ...
But that's all you get...

Every board layout falls with the placement of the components. This one isn't a very good example ...
The board would have had some benefit from a bit more bypass caps near the atmel.

It's a bit like the guys using 7805s without bypass caps. Yes it might work under some conditions. But why doing it like that if you can do far better?