6
\$\begingroup\$

Here are 3 photos of the board I am trying to build. The photos are the top, the bottom and the bottom with the analog portion (Buffer+ADC) inside a red box. It is an ohmmeter. Most SMD parts are on the bottom. The large through hole part is the 7-segment LCD on the top with the uC underneath it on the bottom. Before I start routing, I was wondering if I can still do this in 2-layer or would it be too difficult to try and route. Also, I wondering if I should isolate the analog somehow, like a split ground plane. Any criticism is welcome. Thanks.

topside

bottom

bottom analog in red

\$\endgroup\$
  • \$\begingroup\$ This all depends on what speed the signals are and what your circuit and accuracy specs are. You have a lot of room and relatively few connections. I'd only make it 2 layer, because it's cheaper. \$\endgroup\$ – Myforwik Sep 14 '12 at 9:31
6
\$\begingroup\$

The way to handle the analog part is not to split the ground plane (this increases its impedance, which you want to avoid) but to isolate the components and sensitive signal traces so that the analog signals (and their return currents in the ground plane) do not cross other signal paths.

That said, if you're worried about signal integrity or noise (IMHO, that's "yes" for test equipment like an ohmmeter--nothing worse than an instrument that lies to you), go with a 4-layer board. It'll be both easier to route and have better signal quality (Note that with 2-layer, the ground return path is via traces, so if you make the return current go all over the place, your ground will bounce, and there will be no way to repair it.) The most expensive board is the one that doesn't work.

\$\endgroup\$
  • 1
    \$\begingroup\$ +1. See "partitioning and layout of a mixed signal pcb" by Henry W. Ott for more details on the "Do not split the ground plane. ... Use routing discipline." approach. \$\endgroup\$ – davidcary Jun 22 '13 at 21:23
  • \$\begingroup\$ Yep. At work, if I see a split ground plane, it had better be that the two (or more) grounds are actually separate nets, i.e. the split is only bridged by isolators, capacitors, or transformers. Also, Ott's book is worth the money. Covers almost everything. (I think Bogatin, in his Signal Integrity Simplified book, has some good techniques on power decoupling that Ott elides in order to focus on other solutions.) \$\endgroup\$ – Mike DeSimone Jun 23 '13 at 1:36
  • \$\begingroup\$ While Henry's approach makes a lot of sense for high-speed systems it's not so appropriate for precision DC. \$\endgroup\$ – Peter Green May 27 '16 at 2:31
  • \$\begingroup\$ @PeterGreen: Please elaborate what you mean by "precision DC" and what other approaches are better for it. \$\endgroup\$ – Mike DeSimone May 31 '16 at 1:36
  • \$\begingroup\$ The key thing to remember is that the behaviour of electric currents is frequency dependent. At low frequencies current flow is dominated by resistance at high frequencies it's dominated by (mutual) inductance. Slots are a perfectly valid technique to control the path of low frequency return currents and hence keep unwanted low frequency currents away from the analog portion of the circuit. \$\endgroup\$ – Peter Green May 31 '16 at 1:48
1
\$\begingroup\$

I give a vote for a four layer board too. Boards with SMT and with components on both sides are a perfect invitation for a 4-layer construction. The power and GND planes have benefits in so many ways and these days the cost of boards such as this are no longer a killer.

\$\endgroup\$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.