I can confirm that, autorouting is completely broken. I can give you a command-line log, since I'm running on Linux. It seems like something breaks after a short while.

Yep, I have mint also, it makes no difference nothing works on any platform. Its nothing to do with the board or any of the detail on it because I have created a test board with two pins to be routed and it draws a bent line about 1/5th of the way and then borks. So, its totally buggered.

We all know autorouters are not that good and in all fairness to EasyEDA they do adise not to use theirs, check out the very last sentence of this 'tutorial' page... _ [https://docs.easyeda.com/en/PCB/Route/index.html](https://docs.easyeda.com/en/PCB/Route/index.html) _ Admittedly it has said that for some time now. _ Regards.

@mrtom528 As I said, I'm not here to discuss whether we should auto route or not.  The issue is that what routed previously now doesn't.  So no matter how bad it was is somehow worse that the bad that it already was! I think the problem is super fundamental because I think its borking at the very first part of the attempt.

I have a dedicated layer for signals and about 50 pins to connect, none of which carry sensitive signals. This is a boring task. All my power, ground, clocks, resets, fast buses, differential pairs, etc... are now done. This is exactly what autorouters are for. You can do a terrible job, my slow GPIOs won't care. Just route A to B, save me an afternoon. I ended up wiring it all up by hand, over a day. What a waste.

Yes, in the 1970 and 80s when they designed things like the 6502 or z80 people had to craft the roms and transistors and metal layers by hand using making giant masks and they obviously made mistakes.  I have a T-Shirt that says never trust the auto router! By the early 90's that was completely impossible.  30 years later and the majority of the routing inside digital chips has NEVER ever been seen or reviewed by any human because the complexity is such that it would take hundreds of years to draft a modern IC with billions of transistors.  Instead the routing has been refined over the years and then we rely on net connectivity and track length calculations to know it is correct.    I'll admit that algorithms are devilishly complex, but you have no choice but to trust the auto router.  I agree with you that for the hobbyist we just want a board where the nets are connected like the schematic.  Don't get me wrong I like a good game of Tron Light Cycle like anyone else and I really like watching the auto router try and rip up.

ok, something has now changed.  It routes.  And I can't get it to not route now.  I threw something fairly complex at it and its done as good a job as any auto router I have used. I tried to find anything wrong with the settings and changed them but they all seem to work now.  I'm not sure if something changed or if it its me. I am still using the router I downloaded last week so I know that hasn't changed but the web app may have done. I also made my board outline huge until it routed and then shrank it down.  Not sure if that has anything to do with it.  I also changed my components to 0805 minimum because by eyes are too old for 0603. 

If you changed from 0603 to 0805 without changing your Design Rules then the increased gap between the pads may have been all that was needed to get tracks to route between them and so allow efficient autorouting.

Perhaps.  Although I'd have expected it to come back saying 0% or some low figure and fail.  But it doesn't it did nothing at all, presumably stuck in a loop of indecisiveness. I'm happy with the router shown above.  I don't like routing between pads on resistors etc but at 0805 you can get away with that.  Its a powerful thing when you work with a machine to do something rather than arguing about who is better.  It doesn't get board and will do the tedious bits, I come along and change a couple of bits until I'm happy and get to spend more time with the kids.  Happy days.

I meant bored but board works here I suppose.

In experienced hands, an autorouter can be very useful. My concern about autorouters is that they are not great respecters of design rules for EMC/SI/PI so in layouts where that matters it may be faster and safer overall to manually route a board. Couple inexperience with designs including SMPS or fast edged signals and you can end up with an unsalvageable EMC/SI/PI train wreck. With manual routing there is an opportunity for the inexperienced to learn the rules which all but very high end autorouters may be cyblissfully unaware of. :)

@andyfierman Indeed but the majority are hooking up Arduino projects that were on breadboards which are as bad as its gets.  And why shouldn't they? When you get into VHF ranges its really indistinguishable from witchcraft.  We know the theory about layout and track routing but even then we find really efficient antenna designs no one knew about; all oscillators amplify and amplifiers oscillating.  I don't think any of that should discourage anyone from using an auto router.  What I do think is more of an issue is track resistance and length - but they will find that out themselves.

"When you get into VHF ranges its really indistinguishable from witchcraft." Arthur C Clarke said that "Any technology, sufficiently advanced, is indistinguishable from magic". Design for EMC/SI/PI is quite advanced but, thanks to the likes of Dr Howard Johnson, Eric Bogatin, Rick Hartley and Doug Smith, is well understood and documented so, like Nostalgia, it's not the magic that it used to appear to be. :)

I used to do EMC product testing.  The more you think you know, the more you realise you know nothing at all.  Things like Panel Link cables really do boggle my mind - to everyone else its just a cable.  To me, and I fear this means I'm getting old, I cannot get my head around how data can travel with integrity down the cable - especially where the plugs are mated.  It used to require very specialist connectors to get to a fraction of that speed.  There was always a barrier do to effect of a or b, and its often predicted that these would limit things, but here we are 20 years later and speeds just keep doubling.  Anyway off-topic. And I added some to my PCB design and now its busted again.  So its sensitive to something.  I will try to work out what by removing the bits one bit at a time.  I have a fpc 39w 0.3mm connection I felt for sure would be the problem but its not.

Can you either make your project public or add me to your team so we can have a look?

@andyfierman sure [https://easyeda.com/james_4080/esp32-generic-2](https://easyeda.com/james_4080/esp32-generic-2)<br> <br> Here is a copy of the relevant bits.  The problem is u5.  You delete it everything is fine.  I have only just worked out that u5 is the culprit so it may be something really obvious I haven't checked yet.  So save without u5 and create pcb just move outline to cover it and routes fine.  With U5 it goes nuts.  If you delete u5 just from the PCB it still goes nuts I don't really know what it does with the netlists if you just randomly delete stuff, but I tried that to find the culprit and it only seemed to work the long way around. As I said, I wouldn't auto route this design, but I got into a state where nothing would route.

@james_4080 funnily enough I'm using a cat4238 not the one on the schematic anyway. Its pin compatible with slightly different voltage reference (I hope).  I perhaps could swap it on the PCB, same footprint although its thin I think its the same footprint.

The auto router is completely broken. In the past it was working. Now it cannot route any board to any extent. I ended up spending hours and routing everything manually. I wish I preserved older versions of the working software somewhere. I did not ever expect that this company will make it useless. I am not sure why they did it. Does any one have older versions of the software that worked fine? It would be so awesome if you shared it with others.