designer’s notebook
Taming Electromagnetic Interference on a PCB
What’s all this noise? Look beyond the wires and connectors to think inside the box.
You can’t always hear it, feel it or see it, but every active electronic device radiates some kind of energy as it operates. For the most part, that’s the point. We want to hear the music; we want to feel the air conditioning or see the light. Those are the good things.

Meanwhile, we don’t want side effects: static on the radio, compressor noise from the AC or that annoying 60Hz hum from the light fixture bleeding over to my new bass amp. It’s these things we try our best to design out of the products we build.

We can adjust the tuner on the radio, and we can install the central AC unit away from the windows. I ordered a noise suppressor and plugged it into a socket where there’s no dimmer switches or high current motors plugged in. Then the amplifier and pedal board power cords were routed into the special apparatus, and I no longer get a wave of white noise when this MacBook Pro searches for a WiFi signal.

Even if the emissions cannot be perceived with our five senses, they can be detrimental to the performance of the circuit or spill out onto other nearby electronics. We use metal for the device housings for multiple reasons. Beyond its heat-shedding property, we want to keep the noise inside the box while preventing other noise from getting in. Those are the two central tenets of coexistence.

If you want to chain a bunch of gain-stages together, it helps to start with clean power and a common ground path. The fix for noise issues is almost always additive. The new schematic you get just before the prototype goes for FCC certification usually has a new ferrite bead, maybe a diode here and there and some retuned bypass capacitor values.

Blackboard with a new surge protector that also filters EMI and RFI
FIGURE 1. My Blackboard with a new surge protector that also filters EMI and RFI.
Noise Suppressor Circuit Visual Graph
FIGURE 2. These components are the typical building blocks of a noise suppressor circuit. (Source: Audio Karma)
Generally speaking, a few factors cause signal degradation. One of the main factors is long traces. The farther a signal travels, the more likely it will end up near a bad neighbor. In this case, “bad” can mean noisy or susceptible to picking up noise. I’m tempted to open up that big, white plastic wall-wart for this laptop to see if it has any shielding.

Getting the dirt out of the wall juice is the first line of defense. At the PCB level, the second line of defense is to shield the most vulnerable victim traces from exposure to the outside world. EMI shielding is carried out around the noise makers, as well as the sensitive signals (FIGURE 2). One of the best ways to reduce coupling between those two types of elements is simply adding space.

Placement and orientation of components and printed antennas go a long way toward mitigating noise issues. However, the luxury of a generous PCB is a rarity. No matter the amount of electronics, printed circuit boards cost money, so less is more. Doing more with less is our existential challenge. A lot of designs have an external oscillator, so let’s take a look at that.

The crystal (FIGURE 3) is placed close to the IC, but we could achieve shorter traces by rotating the circuit 90° clockwise. Note the route keep-out that creates a moat that separates the outer layer ground plane from the pins of the crystal. Putting a ground via on the edge of the shape defeats the purpose. Ideally, there are one or more XO return pins for a direct ground path back to the SOC.

A mechanical outline for a PCB typically comes from a mechanical engineer with little to no electrical training. They might give us one part of the board where tall components are allowed. Wouldn’t you know it, the tall components include the crystal. All that thing does is sit there making noise, so something has to be done with the long link back to the mothership.

Internal routing is more effective than running an aggressor on the outer layers. That notion is predicated on the fact there will be an unbroken ground plane both above and below the trace. Taking it a step further, a dedicated shape on the signal layer that encloses the trace is a good policy.

A crystal and its passive parts
FIGURE 3. A crystal and its passive parts.
Guard bands around noisy traces
FIGURE 4. Guard bands around noisy traces. The surrounding low-speed signals benefit from the isolation.
In FIGURE 4, the ground net is highlighted in green. The cyan and magenta represent layers 3 and 5. The via labels indicate the span of the microvias. We want to tie the relevant ground planes together, but not necessarily any other ground layers. If we were to flood ground on the routing layers, the full plane would not engulf these guard-bands. You might have noticed several signal vias without an adjacent ground via. The signals on both layers see the ground plane on layer 4, so the ground via is a feel-good measure.

Note: I used a 2-4 microvia stack for layer three and a 4-6 microvia stack to shield the layer five routes. Finally, going beyond a single return via for the transition from layer three to layer five, there is a cluster of 2-6 vias around the 3-5 signal vias. The idea is to form a coaxial Faraday cage in the z-axis. In cases where there isn’t this amount of space, do your best to isolate the noisy party.

We hear a lot about length-matching, phase tuning and taking care of the impedance of high-speed connections. We know vias are a discontinuity to avoid if possible. Routing over a split or void in the reference plane is a no-no. All these impedance and timing rules are also necessary to keep EMI to a minimum.

We are also drilled to keep inductive loops small when we place and route bypass caps. We are told to use multiple vias for power and ground planes. Alternating power and ground planes transforms the PCB into a bunch of free capacitors. These signal integrity and power integrity rules are there for the circuit performance reasons listed above but also to quell emission of spurious radiation. Solid design practices help bust EMI and RFI problems.

Reducing crosstalk reduces power consumption. Making noise takes energy, so suppressing the noise saves power. I mentioned noise suppression is usually an additive process. Each of those resistive, capacitive and inductive elements has its own parasitic effects. It’s like adding cream and sugar to make the coffee less bitter. At a certain point, we lose the flavor of the coffee and need a stronger brew.

An elegant layout that accounts for potential energy transfer does not require as much cream and sugar on top of the essential circuits. The ground net plays a huge role. It’s the most interesting net on the board. Vias are the one element that prevents the ground net from taking on a charge from the other nets.

Staking down the edges of the ground planes, including the voids, makes for a PCB that runs quieter and provides a substantial path for thermal dissipation. Once you’ve digested all the basic rules around EMI, you’ll see PCBs in a different way. We can immediately spot it when an artist tries to draw imaginary circuits for a logo.

The average person sees the circles and lines and does not know the difference. A trained designer spots the flaws in a heartbeat. That goes beyond the artistic rendering and down into the nuances of memory routing or even chaining a bunch of guitar effects together to create a sound I hear inside my head. I’m looking beyond the wires and connectors to think inside the box. That’s how a PCB designer sees the world.

John Burkhert Jr. headshot
John Burkhert Jr.
is a career PCB designer experienced in military, telecom, consumer hardware and, lately, the automotive industry. Originally, he was an RF specialist but is compelled to flip the bit now and then to fill the need for high-speed digital design. He enjoys playing bass and racing bikes when he’s not writing about or performing PCB layout. His column is produced by Cadence Design Systems and runs monthly.