Raspberry Pi and Arduino RFI: Separate Differential from Common Mode
Raspberry Pi and Arduino RFI: Separate Differential from Common Mode
Single-board computers and microcontroller projects make excellent station controllers, digital-mode hosts, iGates, hotspots, antenna switches and remote receivers. Their switching circuits and cables can also become part of an RF-noise system. The useful diagnosis identifies where the disturbance begins, how it travels and where the receiver converts it into interference.
RF.Guru working definition: Common-mode current is the non-cancelling phasor-sum current in a specified set of conductors, evaluated at a defined cross-section and using a declared current-direction convention. In the intended differential transmission-line mode, the outgoing and return currents are equal and opposite, so their phasor sum is zero. When they do not cancel, the remaining current must close through another reference or return path—such as the outside of a coax shield, a mast, equipment chassis, station wiring, nearby structures, earth, the operator, or distributed coupling through the environment.
This broader working definition is especially useful in practical antenna systems. On transmit, non-cancelling current on the outside of the coax can make the feedline and connected structures part of the radiating antenna system unless that path is intentional, clearly defined and properly controlled—for example by providing the required return path and placing a suitable common-mode choke at the correct boundary.
The familiar symptoms are a comb of lines, a band of hash, a spur that moves with CPU load or a noise floor that rises when one cable is attached. Those clues matter, but they do not yet identify differential mode. A power converter can drive noise between its rails, capacitively drive both rails against the environment, radiate directly from a switching loop and overload the receiver at the same time.
My integration rule: do not blame the board and do not prescribe a ferrite from the waterfall alone. Treat the board, power supply, enclosure, every cable, the receiving antenna and the receiver as one installed EMC system. Separate source, coupling path and victim before changing hardware.
Differential and Common Mode Describe Current Paths
On a two-conductor DC supply, differential-mode noise is the unwanted voltage between the positive rail and its intended return. The corresponding noise current travels out on one conductor and back on the other. Loop area, pair spacing, source impedance, load impedance and any line-to-line filter determine how it propagates.
Common-mode noise is the shared movement of the conductors relative to a declared enclosure, protective earth, station reference or surrounding environment. Common-mode current travels in the same reference direction on the conductors and returns through cable shields, equipment cases, bonding, other cables or stray capacitance.
The two modes can coexist, and imbalance can convert energy between them. Unequal cable routing, connector geometry, shield termination, capacitance to an enclosure and I/O circuits can convert a differential disturbance into common-mode cable current. But differential noise is not always the original source: a switch-mode supply can also inject common mode directly through transformer or heatsink capacitance.
For rail voltages V+ and V− measured to the same declared reference:
Vdm = V+ − V−
Vcm = (V+ + V−) / 2
State the sign convention, probe impedance and reference plane. Another standard or instrument may use a different normalisation.
The Board Is Only One Part of the Source
Fast digital edges, processor clocks, memory buses, PWM outputs, switch-mode regulators and peripheral controllers all contain spectral energy beyond their fundamental rate. Ringing and resonances depend on package, PCB, decoupling, return geometry and the connected load. A clock at one frequency can therefore produce a comb, but the spacing and amplitude can change with firmware, processor state, display activity, storage, networking and power management.
Power supplies deserve equal suspicion. A board can be quiet from a laboratory source and noisy from one USB adapter, DC-DC module or long power lead. It can also reverse the story: the board’s load transients create differential current that the supply and cable convert into common mode. A battery comparison can isolate the external supply as a variable, but it does not prove the remaining board-and-cable system is emission-free.
Raspberry Pi and Arduino are families, not one circuit. Different models, carrier boards, shields, HATs, clones and power architectures have different clocks, interfaces and return paths. Treat the exact hardware, firmware, cables and enclosure as the equipment under test.
A Waterfall Pattern Is a Clue, Not a Mode Measurement
Evenly spaced lines often point to periodic activity, while broad hash often accompanies fast edges, burst operation or several sources. Neither visual pattern says whether the receiver obtained the disturbance through a differential supply pair, common-mode cable current, direct radiation or its own overload products.
A portable receiver or near-field probe can localise a switching loop, cable, connector or enclosure seam. Probe distance, orientation and loading strongly affect the result, so this is comparative diagnosis rather than calibrated field strength. Likewise, a station SDR waterfall depends on antenna, feed-line current, attenuation, preamplifier, AGC, bandwidth and display scaling.
Conducted and radiated paths can feed each other. Differential current in a large cable loop creates a field. Common-mode current on a long cable can make that cable an effective antenna. A radiated field can induce current on another cable, which then conducts the disturbance into the receiver.
Measure the Mode Without Creating a New Hazard
- Clamp around the complete pair. Ideal differential currents cancel in a characterised current probe, so the remaining net current indicates common mode or another unbalanced return within the probe bandwidth.
- Clamp around one conductor. The result contains working current plus differential- and common-mode RF components. It does not separate the modes by itself.
- Probe across an extra-low-voltage rail. This targets differential voltage only when the probe connection, bandwidth and reference are controlled. Keep the loop at the probe tip small.
- Probe conductor to enclosure or reference. This contributes to common-mode analysis only when both conductors are measured against the same declared reference with controlled probe impedances.
- Scan in the near field. Repeat probe height and orientation. Use it to find relative hot spots, not to claim a compliance level or far-field radiation pattern.
Oscilloscope safety matters. A bench oscilloscope’s ground clip is commonly bonded to protective earth. Connecting it to a floating or mains-referenced circuit can create a short circuit, electric shock, fire or equipment damage. Do not probe the primary side of a power supply, defeat protective earth or assume a battery-powered laptop makes an instrument isolation-rated. Hazardous and floating-power measurements require correctly rated differential or isolated equipment and a qualified operator.
CISPR 16-1-2 specifies characteristics for current and voltage probes used in conducted-disturbance work, while CISPR 16-2-1 and CISPR 16-2-3 define conducted and radiated measurement methods. A shack investigation is not automatically a compliance test, but those standards show why probe transfer, detector, bandwidth, cable layout and reference plane belong in the record.
Receiver Overload Can Imitate a Noisy Computer
A digital device can emit a real disturbance while the receiver exaggerates it. Strong broadcast, amateur or digital signals can compress a front end, cause blocking, create intermodulation products or spread reciprocal-mixing noise around a carrier. An LNA ahead of an already overloaded receiver makes that worse.
Repeat the observation with known attenuation and, where appropriate, a suitable preselector ahead of the stage that overloads. Keep the receiver bandwidth, AGC and display scale fixed. If spurs collapse by more than the applied attenuation or disappear with preselection, part of the apparent noise was probably created inside the receive chain.
That does not absolve the digital source. It separates source emission from victim susceptibility so each can be improved at the correct point.
Choose Mitigation by Mechanism
| Measured mechanism | Candidate control | Evidence to keep |
|---|---|---|
| Differential ripple or edge energy on a DC pair | Manufacturer-approved line-to-line filtering, smaller loop area, correct decoupling or slower permitted edges | Rail waveform/spectrum, source and load state, voltage drop, startup and temperature |
| Net current on a complete cable or pair | Common-mode impedance at the measured boundary, improved shield/enclosure termination or route control | Current before/after, complex choke impedance, cable configuration and frequency |
| Direct field from board or enclosure | Source-layout correction, permitted clock/edge control, continuous enclosure and controlled penetrations, or greater separation | Repeatable near-field map and fixed-distance receive comparison |
| Receiver-generated spurs or raised floor | Attenuation, preselection, gain redistribution or a more linear receive stage | Level sweep, blocker frequencies, bandwidth, gain/AGC state and wanted-signal SNR |
A common-mode ferrite around an entire power or data cable can be a useful test when the cable and connector tolerate it. It is not a direct differential filter. Ferrite or series impedance added to individual conductors carries the working current and can introduce voltage drop, saturation, heating, timing error and unequal impedance.
Line-to-line LC filtering can reduce differential energy, but a generic value is not a recipe. The filter interacts with the supply, cable and board input impedance and can create resonance, startup failure, undervoltage, control-loop instability or connector heating. Verify current, inrush, ripple, damping, voltage margin and the board’s permitted supply range.
High-speed USB, HDMI and Ethernet links depend on controlled impedance, pair balance, timing and protocol requirements. Random series components, shield cuts or unsuitable isolators can damage signal integrity or create a worse common-mode discontinuity. Check the exact data rate, power negotiation, shield path, voltage/current capacity and equipment documentation. A data isolator is not automatically a safety isolation barrier.
Routing, Shielding and Bonding Must Work Together
Keep each outgoing conductor close to its intended return where the interface specification permits it. That reduces differential loop area. It does not by itself suppress common mode, which needs a controlled reference and return path.
A metal enclosure can reduce direct radiation only if seams, apertures and cable penetrations preserve the intended shielding boundary. A long shield pigtail adds inductance; a low-inductance enclosure termination is usually preferable where the interface and safety design support it. Whether a shield is bonded at one end, both ends or through a frequency-selective network depends on the interface, frequency, fault protection and system architecture—not a universal audio-ground rule.
Protective earth is a safety function, not an optional RFI experiment. Never lift it to make a waterfall look cleaner. Solve RF current with approved filtering, bonding, routing and isolation that preserves the equipment safeguards.
Use Firmware Controls Only When the Hardware Supports Them
Disabling an unused display or interface can remove one activity state from the noise map. Reducing a GPIO drive strength or slew rate can reduce high-frequency edge energy when setup/hold time, logic levels, load and signal integrity remain valid. Spread-spectrum clocking can redistribute spectral energy rather than eliminate it, so a lower peak is not automatically a lower integrated disturbance.
Use documented controls for the exact board and interface. Raspberry Pi’s Compute Module 5 documentation, for example, recommends the lowest drive strength and slew rate consistent with the application to minimise EMC issues. That is a design principle tied to that hardware and its timing requirements, not permission to apply undocumented settings across every Pi, Arduino or clone.
A Repeatable Shack Test
- Freeze the receiving setup. Record frequency, mode, equivalent noise bandwidth, attenuation, preamplifier, RF gain, AGC, antenna, feed line and display scale.
- Record the quiet baseline. Capture wanted signal and adjacent-channel noise with the digital project disconnected or in a declared idle state.
- Exercise one function. Compare CPU load, PWM, USB transfer, Ethernet, display, storage or radio activity one at a time.
- Separate supply from load. Where the manufacturer permits it, compare the approved supply with a suitable current-limited battery source on extra-low-voltage equipment. Do not bypass required earth or isolation.
- Trace cables and ports. Use repeatable near-field and clamp-current observations on power, USB, HDMI, Ethernet, GPIO, audio and enclosure bonds.
- Test receiver headroom. Repeat with known attenuation or preselection and keep the wanted-signal SNR in the record.
- Make one safe change. Change only routing, separation, an approved cable, documented firmware state or a correctly specified filter boundary.
- Restore A/B/A. Return to the first state and repeat so drift, propagation and background loads do not become the result.
The record should include the exact board and revision, add-on hardware, firmware, workload, supply, every connected cable, enclosure, antenna distance, receiver state, probe position and frequency span. “The Raspberry Pi is noisy” is not a reproducible result; “this complete configuration produces this change through this path” is.
Compliance and a Quiet Station Are Different Questions
CISPR 32 establishes emission requirements and repeatable test procedures for multimedia equipment. A product’s conformity applies to the declared model and test configuration. The finished ham project may add a carrier board, display, power supply, unshielded wiring and long cables that were not part of that configuration.
Conversely, hearing a spur on a sensitive nearby receiver does not by itself prove regulatory non-compliance. It proves an installed compatibility problem. Raspberry Pi’s product-information portal publishes model-specific approvals and test records, which is exactly how conformity evidence should be treated: by product, configuration and jurisdiction rather than by brand folklore.
IEC 62368-1 treats information and communication equipment safety through classified energy sources and safeguards. EMC modifications must not defeat those safeguards. If mitigation requires changing a mains supply, enclosure safety barrier, protective earth or certified cable assembly, keep the work with the manufacturer or a qualified person.
Primary Engineering Sources
- CISPR 32:2015+A1:2019 — Multimedia equipment emission requirements.
- CISPR 16-1-2:2014+A1:2017 — Coupling devices for conducted-disturbance measurements.
- CISPR 16-2-1:2014+A1:2017 — Conducted-disturbance measurement methods.
- CISPR 16-2-3:2016+A1:2019+A2:2023 — Radiated-disturbance measurement methods.
- IEC 62368-1:2023 — Safety requirements for audio/video and information and communication technology equipment.
- Recommendation ITU-R SM.575-3 — Receiver noise figure, blocking and intermodulation in strong-signal environments.
- USB-IF — USB 2.0 Specification and engineering changes.
- Raspberry Pi Compute Module 5 data sheet — interface and EMC design guidance.
- Raspberry Pi 5 Product Information Portal — model-specific approvals and compliance records.
Practical Conclusion
Raspberry Pi and Arduino projects do not need a Faraday cage by default, and a clamp-on ferrite is not a universal repair. Start with the exact operating state, identify the noisy port, separate differential current from common-mode current, and make the receiver prove it is still linear.
Then apply the smallest safe change at the mechanism you measured. A quiet supply will not repair a radiating data cable. A common-mode choke will not clean differential rail ripple by definition. An enclosure will not help if every cable bypasses its shielding boundary. Good EMC is integration, not blame.
Mini-FAQ
- Does a comb on the waterfall prove differential-mode noise? No. It shows periodic energy; differential conduction, common-mode cable current, direct radiation and receiver overload must be separated by measurement.
- Will a clamp-on ferrite remove differential rail noise? Not by definition. A ferrite around the complete cable mainly adds common-mode impedance; differential filtering needs a suitable line-to-line network or source correction.
- Can I probe the power supply with a normal oscilloscope? Only on a circuit and reference the probe is rated to measure. Never attach an earth-bonded ground clip to an unknown floating or mains-referenced node.
- Is a USB isolator a universal cure? No. Data rate, protocol, power negotiation, shield coupling, voltage/current and safety-isolation ratings must match the complete application.
- Does a compliant Raspberry Pi guarantee a quiet ham installation? No. Compliance belongs to a declared model and test configuration; the supply, cables, add-ons, enclosure, antenna and receiver define the installed result.
- What makes a mitigation result convincing? A repeatable A/B/A improvement in wanted-signal SNR with fixed receiver settings, supported by a measured change in the identified port, current mode or overload state.