Understanding source and return current flow on PCBs

Question

I'm looking for advice and tips related to PCB board layout and signal+ground trace/track routing.

I'm designing a 2-layer analog-only PCB that has several I/O signals leading to and from other boards and several analog amplifier ICs. I have some questions about how careful and extensive I need to be when routing ground tracks to carry the return current, and how I can best arrange these on the board to reduce noise and hum (EMF) pickup. I'm using star grounding and no ground plane on the board.

From what I have generally come to understand, you want to run signal and return tracks close to each other. To me this seems the same logic used when running interconnect cables off the PCB, e.g. between boards or equipment. On the PCB, I'm unclear about what happens with signal currents, especially when ICs are involved. For instance, I try to run a ground return next to the signal track from the board edge where the I/O connector is located back to the star ground system, but it's not always possible to run this signal+return pair all the way up to where the signal trace connects to a pin on the IC. Should this be of much concern for an analog (audio) circuit? Likewise, is it good practice to try to keep signal+return together for tracks that go between ICs on the same board? If so, how do I figure out where the return current is flowing? For instance, does the current "return" to the ground pin of the first IC, or does it flow back to the star ground system?

Does it make any sense to run a "spur" ground track, meaning one that runs along the signal track but then stops where the paired signal track reaches an IC pin, e.g. that end is left disconnected. The other end would be connected to the ground system. I can't convince myself that this would be effective as a "return" and current would flow, since one end is disconnected. Are dead-end tracks ever used, e.g. for shielding?

I have not been worried about this in the past, and I have designed boards that seem to have acceptably low EMF pickup, etc. but I have always been curious about on-board routing issues like this. I'm completely self-taught so any words of advice would be very much appreciated. I have read several PCB design guideline type documents that I could find on the web, but none seem to address this particular issue - low noise analog signal routing. Maybe its just not all that important?

Answer 1

The best advice I can give is to use a 4L board, so you can have a solid ground plane. That in itself will make things a lot easier, as you will have the return current running pretty much right under the trace carrying the forward current.

The cost difference is really not that big if what you are doing matters to you and is not crazy high volume (10.000+). You may save board space and routing time as well, which helps offset the small added cost.

Answer 2

From what I have generally come to understand, you want to run signal and return tracks close to each other.

This applies when the signals have frequency content that you wish not to couple into other areas of the board, or vice versa. Sometimes such coupling may be irrelevant, othertimes it's very relevant. So yes, the general advice might be good, but it's the reasons for it that are important, less than blindly following the recommendation.

In order to visualize how what you do affects the board behavior, you need to think of all PCB traces as having inductances, resistances, and having trace-to-trace capacitances. You can even extract this information from the PCB - even if using very crude approximations - and add it to the Spice model of your circuit.

You also need to think of the circuits as antennas that couple into other circuits. A closed circuit is a loop antenna, and you can approximately model its coupling into other "interesting" circuits by adding coupled inductors to your Spice model. When you run a signal close to its return track, you decrease the loop area of the circuit, and thus make it a worse antenna. It will radiate less, and it will be less susceptible to radiation.

You can also try to imagine the interfering signal sources. It's usually wrong to call any unwanted signal noise, since noise is by definition random. Interference is most definitely non-random and is closely related to other things that happen in and around your circuit - it's a deterministic phenomenon, for the most part.

One way to develop intuition about it with experimentation. Set up some soldered breadboard circuits, and see how susceptible they are to various interference sources.