I'm continuing to make slow progress on Project Egbert. Last week, I finally got EAGLE to finish routing all the connections for the life-sized version of the board, then forked off the larger-than-life lab version, which incorporates many headers for debug purposes.
It's amazing how big old-fashioned 0.1"-pitch headers look when you drop them onto a board full of modern little SMT parts. And I though that 14-lead SOIC looked big!
Anyway, adding the debug headers caused the router to be one confused little iggle, and, in the few runs I've managed to fit in since Thursday afternoon (each run takes around 15 hours), it's persistently run into a flurry of pathological no-routes around halfway through, eventually managing around 99.5% completion, which Just Isn't Good Enough.
Well, I decided to attack the 99.4%-complete version produced by last night / this morning's run. Turns out some of these no-routes can be handled with minor ripups, which leads me to wonder why the autorouter couldn't handle them. Other need more major overhauls... so now I'm trying the tactic of ripping up several traces in the vicinity of each no-route, and then telling the autorouter to try again. It seems to be helping; at last report, it was down to three no-routes (and, of course, autorouting a few dozen connections runs a bunch faster than doing the whole board).
So, there's still hope for releasing Gerbers tomorrow. And Purchasing needs to be unleashed on the parts list (and I sure hope Purchasing is somebody else, 'cause that's a heck of a long parts list, and I've still got software to write).
Comments