1
\$\begingroup\$

Here's the issue with eagle autorouter and me trying to connect a 62256 ram chip to a 74HC574 latch (even with clock and latch enable disconnected)

I measured the width in the trouble IC (74HC574) and its 6 mm. I went paranoid and decided to allow eagle to use 0.1mm track width and 0.1mm track spacing. I used 0.3mm spacing between pad and anything else next to it. I set the restring (of the pads) to 0.24mm and to make the pads small and circular instead of oval, I set Elongation to 0 instead of 100.

In the miscellaneous section, I unchecked "check grid", "check angle", "check font" and "check restrict".

I think I did very good so far.

So here's my math if I used 0.1mm track width and 0.1mm track space:

 0.3mm initial spacing from end of pad = 0.3mm
+0.1mm 1st track width = 0.4mm
+0.1mm 1st space = 0.5mm
+0.1mm 2nd track width = 0.6mm
+0.1mm 2nd space = 0.7mm
+0.1mm 3rd track width = 0.8mm
+0.1mm 3rd space = 0.9mm
+0.1mm 4th track width = 1mm
+0.1mm 4th space = 1.1mm
+0.1mm 5th track width = 1.2mm
+0.1mm 5th space = 1.3mm
+0.1mm 6th track width = 1.4mm
+0.1mm 6th space = 1.5mm
+0.1mm 7th track width = 1.6mm
+0.1mm 7th space = 1.7mm
+0.1mm 7th track width = 1.8mm
+0.1mm 7th space = 1.9mm
+0.1mm 8th track width = 2mm
+0.3mm space to next pad = 2.3mm

and even if I doubled the track width, I'd still have space.

I also tried changing all cost settings under the Route tab in autorouter to "0" except I left the polygon setting alone and I even set the autorouting grid to 0.2mm.

Here's part of the pcb completely unrouted

Unrouted

Here's how I hand-routed the board but it took me 10 minutes and I don't want to spend 10 minutes per complete routing attempt.

Hand-routed

This is how eagle routes it:

Autorouted

It didn't finish.

In every auto-routing attempt I made, there was not one time that eagle would start a track horizontally from an IC pin (such as what I did with my ground pin of the 74HC574), and this is even after I set the preferred track direction to horizontal.

How do I configure eagle to auto-route these tracks properly (or at least how do I configure eagle to make tracks run horizontally out of the ground and VCC pins of digital IC's)?

\$\endgroup\$
  • 1
    \$\begingroup\$ How much did you pay for the autorouter? \$\endgroup\$ – Andy aka Oct 31 '16 at 17:19
  • \$\begingroup\$ nothing but still, you'd think it at least do that much. I even tried polygon setting to zero now and still no luck \$\endgroup\$ – user116345 Oct 31 '16 at 17:40
  • \$\begingroup\$ Since the pins on those parts are onn 0.1 inch centers, it would probably be easier to route on an 0.025 inch grid, than on a metric grid. \$\endgroup\$ – Peter Bennett Oct 31 '16 at 17:57
  • 2
    \$\begingroup\$ After having tried EAGLE's autorouter several times, I decided that I will route boards manually. Do not know why but I get drive from drawing. Also, when you draw, you may realize that you need to change layout of the components on the board. In your case, with your manual routing I would say you would need some experience to do it nicer and faster (plus align chips to proper grid to have tracks straight), but for sure you have the talent :). I give you 1000% your manual routing will be more effective and efficient. \$\endgroup\$ – Anonymous Oct 31 '16 at 18:36
  • 1
    \$\begingroup\$ Manually routing a severely complex PCB is one of the most satisfying things in electronics. It also proves the need for the incomparable skill of a human brain to recognise patterns and to find alternatives rapidly. Computers cannot do this because they simply do not know what your circuit is supposed to do. You would have to spend a month telling the computer how to do it, when you could simply draw it all yourself in 2 days. That said, "Toporouter" is a pretty impressive bit of kit. \$\endgroup\$ – Wossname Nov 2 '16 at 20:48
1
\$\begingroup\$

This is why manual routing is the way to go

What you have discovered is that autorouters are extremely limited tools compared to the ability of a human brain to recognize patterns and prune unfruitful branches quickly and visually.

There are a few suggestions I can offer though:

  • Route on an imperial or soft-metric (i.e. imperial units converted to metric) grid vs a hard-metric grid. The DIP is a North American invention, and so are many other component packages -- hard metric packages have only come into play relatively recently with (T)QFPs and (T)SSOPs/MSOPs at the 0.8, 0.65, and 0.5mm pitches, as well as BGA and QFN. Furthermore, metric sizing for SMT passive packages is still not universal -- while tantalum EIA sizes are metric, ceramic caps and chip resistors are often still imperial sized.

  • Pay close attention to what your escape routes are -- it's easy to "box yourself in" when routing and wind up needing vias you otherwise wouldn't need. This is doubly important when routing ubiquitous nets like power and ground, or if you can't get to your fast toggling lines (clocks, latch enables, etal) early on in the routing process to avoid sending high-frequency, hard-driven signals noodling all over your board, emitting EMI as they go.

  • Pay attention to locality -- if two components are part of the same circuit, they should be close together on the board. This is a major EMI-abatement tool as well as excess loop areas just make your board a better antenna, which is generally not what you want. It also makes it easier to follow when troubleshooting and debugging.

  • And finally, keep scale in mind. This isn't a huge concern with DIPs, but when working with SMT packages, it's very easy to lose sight of just how small things really are and wind up with a board that is a nightmare to assemble because you have parts crammed in so tightly that you can't get your iron in to solder them without doing a contortionist act.

\$\endgroup\$
  • \$\begingroup\$ I get boxed in too often and it seems the only way out is to manually connect ground later after making the circuit board which I think is a bad idea, so either way its a lose-lose \$\endgroup\$ – user116345 Nov 3 '16 at 1:25
  • \$\begingroup\$ @Mike -- keep working at it and I'm sure you'll find a solution (maybe routing grounds earlier would help?). Going to a four-layer board helps tremendously with routing, of course, but that's not an option for everyone, which is why I didn't suggest it in my answer. \$\endgroup\$ – ThreePhaseEel Nov 3 '16 at 1:27
  • \$\begingroup\$ If I could do a four-layer board by hand, I'd be rich by now, but I'm just a poor hobbyist. I posted another thread about seeking a new chip that may help with my routing. I don't care if I have to switch chips just to get the routing to work. \$\endgroup\$ – user116345 Nov 3 '16 at 1:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy