# Techniques for Optimizing Return Paths

First of all, what is a return path? The ground plane is the simple answer. For every wave of energy (measured in volts) that goes out from a source, an equal but oppositely charged wave tries to make it back from the far end to the source. This counter-wave completes the circuit so that more waves of energy can follow. I say “tries” as the voltage seeks the path of least resistance.

## Return on Layer Two?

That return path may or may not be the one that was designed for the circuit. A poorly designed return path will result in the return currents wandering around and escaping as electromagnetic interference. This phenomenon annoys other devices, messing with sound and picture quality among other things. The Federal Communications Commission has a threshold for these emissions. An acceptable product will be immune from the emissions of other products and will not disturb those products with its own emissions. Pretty simple to explain and often difficult to implement. Curbing EMI is one of the most important aspects of return path design.

Image Credit: Velodyne LiDAR - Return paths in a system with many PCBs can be convoluted.

In an alternating current (AC) application, the return path alternates from one of the two wires to the other. One wire has a positive voltage that swings negative while the other goes from a negative voltage to a positive voltage of the same value as the positive lead such as plus or minus 110 VAC. A three wire system, like you would see on power tools and HiFi gear, has that third lead at (or near) zero volts. That is considered “neutral” as it should not have any significant energy transfer going on. The third “rail” is there specifically to tie the chassis to the ground so that you, the user, do not become the ground path through an internal short within the equipment. Any time you are being used as a conductor, you’re probably having a bad day.

Cause and effect are in play with direct current (DC) as well. What’s missing is the reciprocal action that promotes energy transfer over great distances. What’s left is a cleaner medium to work in. Without AC noise, we can turn our power on and off in an established pattern to do useful things. Morse code is still relevant as a machine language, but the data rate makes the return path of little concern. ASCII and other binary code zip around in massive bursts on our devices and beyond. The switching speeds are somewhere between ridiculous and ludicrous. It is in this world that completing the whole circuit really matters.

## The Thin of the Plot

Our best defense against the losses through emissions is to make the current loops as short as possible. Bias circuits require bypass caps to complete the loop. It is not enough to place the capacitor near the power pin and call it a day. The orientation of the cap should be considered if there is a ground pin in the vicinity. A low-impedance trace to that pin or to a common ground via will be a good move.

Even if the layout at hand is low tech and can get away with a quick and dirty execution, it still has to share a world where other devices do their thing without interruption.

Better still for power supplies is to bridge the power pin to the ground pad under the regulator with the capacitor acting as a direct part of the path. I know that the assembly house will advise you to place all of the components in the same orientation. It is so rare for that method to lead to the optimal performance, that I’ve done a placement that way approximately zero times out of thousands of PCB Designs. The end user’s right to the best possible performance comes before the factory’s guidelines; at least to a point. This holds true especially for analog, but digital designs quite often have edge rates that require optimal placement just to get over the threshold of minimum performance.

Even if the layout at hand is low tech and can get away with a quick and dirty execution, it still has to share a world where other devices do their thing without interruption. One of the other things that I don’t buy into is the notion that something doesn’t need much care because it’s just a prototype or a test fixture. Bologna! Science projects have almost as much chance at becoming a product as the projects that were intended for eventual production. Always be prepared for a design to “go viral” and for the stakeholders to be reluctant to have you make any further changes. It has happened before and will happen again.

## Routing Whoahs!

Before you route over a gap in the plane, have a look at some of the simulation videos available online. A broken Faraday cage generates an alarming field of electromagnetic interference or EMI. The bad design work itself does not make the radiation. The power source of the equipment drives the radiation and is enabled by the leaky design.. That’s right, you pay extra in power inefficiency in order to fail FCC certification. Fix the mess before it becomes one.

Image credit: Yurly Shlepnev - He has a number of good videos available.

A light bulb doesn’t emit any photons until you flip the switch. The switching regulator is not so different. Installing it near the device that loads the regulator’s output pin will help minimize the obnoxious current flow around the area. Yes, it is noisy by nature, but it can be contained. All of the good vendors show you how to be successful using their product. They are going to be conservative and use more space that you might be able to allocate. Good (ground) fences make good neighbors in this case.

Optimal return paths help devices coexist. A little bit of space won or lost can make a lot of difference. The formula for coupling is a nonlinear function such that doubling the air-gap between two items results in a square of the isolation. Likewise, sliding two traces just a little closer to each other can dramatically increase the amount of cross-talk on those lines. This, by itself, is reason enough for those stakeholders to get squeamish when you want to “DFM” their “science project”. Risk aversion rises as we go along the product life cycle so it pays to be brave and do it right as soon as you can. From placement to fan-out through routing and tape-out, design each connection as a whole circuit. It is always more gratifying to tell a success story. Make it happen.