EMC troubleshooting for mixed-signal boards

A mixed-signal board can pass every functional test yet fail EMC because the same clock edge corrupting an ADC reading is also driving common-mode current onto an external cable.

Finding that interaction requires more than probing the noisiest-looking net. The task is to identify the source, coupling path and susceptible circuit, then prove which change has interrupted the mechanism without creating a new one elsewhere.

Why mixed-signal EMC faults are difficult to separate

Mixed-signal EMC problems rarely respect schematic boundaries. A design may have separate analogue and digital functional blocks, but electromagnetic energy follows impedance, geometry and return paths rather than net names. Digital clocks, switch-mode converters, memory interfaces and processor I/O can couple into ADC references, sensor inputs and external cables through electric fields, magnetic fields, shared impedance or radiation.

The measured symptom may also appear far from its source. Periodic ADC noise could be caused by direct capacitive coupling from a clock trace, modulation of the reference rail, ground movement at the converter, aliasing of an out-of-band signal or conducted interference entering through a sensor cable. Treating all five as an ADC problem wastes debugging time.

Start by defining the signature. Record whether the disturbance is synchronous with sampling, converter activity, a digital clock, a power-converter switching cycle or a communications transaction. Examine both the time and frequency domains. A narrow spectral line and its harmonics suggest periodic activity, while broadband elevation may indicate fast edge transitions, power-stage ringing or multiple uncorrelated sources.

Digital clocks are usually edge-rate problems

A 25 MHz clock is not merely a 25 MHz EMC source. Its harmonic content is governed largely by rise and fall time, trace discontinuities, loading and ringing. A fast output buffer can generate significant energy well beyond the clock fundamental, even where the receiving logic does not need such aggressive edges.

Probe technique matters here. A long oscilloscope probe ground lead forms an inductive loop that can exaggerate ringing or pick up nearby fields. The resulting waveform may send the investigation in the wrong direction. Use a short ground spring, an active probe or a properly selected differential probe, while respecting probe voltage and common-mode limits.

If series damping reduces ringing and emissions without compromising timing margins, the likely mechanism is transmission-line excitation rather than clock frequency alone. Place the resistor close to the driving device. Locating it at the receiver leaves the full trace exposed to the undamped edge and usually provides poorer control of radiated and coupled energy.

Do not change termination values blindly. Check the driver impedance, trace characteristics, topology and receiver thresholds. Excessive resistance can slow the edge enough to cause threshold dwell, duty-cycle distortion or timing failures.

ADC noise may enter through four different paths

Input and reference coupling

High-impedance analogue nodes are susceptible to electric-field coupling. Clock traces routed beside an ADC input, reference or amplifier feedback path can inject displacement current through small parasitic capacitances. The current may be tiny, but the impedance of the victim node can convert it into a measurable voltage.

Reducing parallel routing, increasing separation and adding grounded copper can help, provided the shielding copper has a low-inductance connection to the reference plane. A decorative copper island connected by one narrow neck may resonate or capacitively couple without providing an effective return.

Reference pins also deserve direct measurement. Noise on the reference can scale or modulate the conversion result. Decoupling components should follow the converter manufacturer's guidance, with short paths and minimal shared inductance. Adding more capacitance at random can destabilise a reference buffer or move an anti-resonance into a troublesome frequency band.

Power distribution coupling

Digital load steps produce current transients across the impedance of planes, vias, package leads and decoupling networks. If analogue and digital circuitry share part of that impedance, the transient becomes a local supply or ground disturbance.

Measure rails close to the device pins, not only at the regulator. A clean waveform at the regulator says little about the voltage presented to an ADC several centimetres away. A low-inductance probing arrangement is required, otherwise the probe loop can report the local magnetic field rather than the rail voltage.

Grounding and return-path discontinuity

Splitting a ground plane is not an automatic cure for mixed-signal EMC. A high-speed digital trace crossing a slot forces its return current around the discontinuity, increasing loop area and magnetic-field radiation. The diverted current can then pass through sensitive analogue territory or a connector bond.

A continuous reference plane with disciplined component placement often performs better than physically separated grounds. Keep noisy current loops local and stop digital routing from entering analogue regions. Where a converter manufacturer specifies a particular ground arrangement, follow the current revision of its data sheet and layout documentation rather than applying a generic split-plane rule.

Board-level grounding must also be considered alongside chassis bonding. An external cable shield terminated through a long PCB trace or pigtail has increasing inductive impedance at higher frequencies. That impedance permits shield current to develop a voltage and couple into circuit ground. A short, wide, circumferential chassis connection is normally more effective where the connector construction allows it.

Aliasing of out-of-band interference

An ADC can translate interference above the signal band into an apparently low-frequency error. This is particularly misleading because the observed spur may not match any obvious clock on the board. Varying the sample rate can reveal the mechanism: an aliased component moves predictably, whereas a genuine baseband disturbance may not.

Review the analogue anti-alias filter, converter input bandwidth and source impedance. The filter must be assessed at the interfering frequencies, not only across the wanted passband. Parasitic component behaviour and PCB layout may dominate well before the nominal component values suggest a problem.

A structured mixed-signal EMC investigation

Change one variable at a time and preserve a baseline. Before modifying hardware, record firmware version, operating mode, cable arrangement, supply, enclosure state, clock configuration, ADC settings and measurement instrument settings. Without that record, an apparent improvement may simply be a change in workload or cable coupling.

Near-field probes are useful for localisation, but their output is comparative rather than a direct compliance measurement. A magnetic-field probe is effective around high di/dt loops, power inductors, clock return paths and connector bonds. An electric-field probe is more responsive around high-impedance nodes and high dv/dt conductors. Probe orientation, height and cable routing must remain controlled if scans are to be compared.

A spectrum analyser or receiver can then correlate local activity with conducted or radiated observations. Suitable signal generators can support controlled susceptibility investigations, while the broader selection of signal generators, noise sources and function generators can help inject known disturbances during development. Injection must be current-limited and planned so that device ratings, safety constraints and unintended port overstress are respected.

For spectrum occupancy and intermittent-signal work, equipment such as the Narda SignalShark 3110 or Narda SignalShark 3310 may be relevant, subject to the required frequency coverage, dynamic range, analysis functions and measurement environment. Instrument selection should follow the signal being investigated rather than familiarity with a particular analyser.

Typical scenario

Consider an illustrative data-acquisition board containing a microcontroller, Ethernet interface, switch-mode power supply and precision ADC. The ADC output contains periodic spurs, while radiated pre-compliance measurements show corresponding clock-related emissions. Connecting a long sensor cable raises both effects.

The first decision is whether one mechanism explains both symptoms. The team could operate the ADC from a quiet internal source, disable Ethernet activity, vary processor clocking and compare results with the sensor cable connected and disconnected. Near-field scans around the clock source, converter, DC-DC power loop and connector provide localisation, while current-probe measurements on the cable can indicate whether common-mode current is involved.

If slowing an unused-fast GPIO edge reduces cable current but not ADC noise, two paths probably exist. Conversely, if improved chassis bonding at the connector reduces radiated output and converter spurs together, external cable current may have been driving ground movement near the analogue front end.

Using the wrong method can obscure that distinction. A near-field probe alone cannot establish formal radiated emissions performance. Likewise, a bench spectrum analyser connected without suitable attenuation, DC blocking or port protection can be damaged or overloaded, producing misleading intermodulation products.

Early investigation permits inexpensive changes such as clock damping, return-path correction, connector bonding or filter-footprint revisions before enclosure tooling and PCB release. EMC Hire can support this work through equipment hire, practical setup advice, pre-compliance engineering, on-site testing and access to test facilities. Formal compliance testing can then be planned against the applicable product or product-family standards, with the manufacturer retaining responsibility for the legislation, conformity route and technical file.

When to Hire EMC Equipment

Mixed-signal troubleshooting often needs several instruments for a short period: a suitable analyser or receiver, near-field probes, current probes, signal sources and accessories for safe coupling or monitoring. Purchasing the wrong bandwidth, input capability or analysis option creates a long-term asset that may not suit the next programme.

Hiring avoids unnecessary capital expenditure where demand is irregular. It also allows a development team to scale its capability during a concentrated debug window without accepting continuing storage, servicing and calibration overheads. This is particularly useful when a project needs a specialised instrument for days or weeks rather than throughout the product lifecycle.

Equipment must still be matched to the job. Frequency coverage, displayed average noise level, dynamic range, preselection, detector behaviour, input protection and available probes all affect whether the wanted signal can be distinguished from instrument artefacts. EMC Hire can help define a suitable configuration before the hire period starts.

Where measurement traceability is relevant, EMC Hire uses test equipment with calibration traceable through an appropriate ISO/IEC 17025 accredited calibration provider. Suitable traceable calibration supports repeatability, comparison between development and formal testing, and stronger recorded evidence. It does not make an informal bench setup equivalent to a standards-defined compliance test.

Common EMC Testing Mistakes to Avoid

Moving cables between measurements

Cables are often the dominant radiating structures below the frequencies where a small PCB becomes an efficient radiator. Moving a cable by a few centimetres can change coupling and common-mode current, making a modification appear effective when only the antenna geometry changed. Photograph and dimension cable positions.

Assuming every spur is generated by the board

Ambient radio services, nearby switch-mode supplies and laboratory equipment can appear in a scan. Compare EUT-on and EUT-off traces, change antenna or probe position, and verify whether the signal tracks an EUT operating mode. Poor ambient control can cause false diagnosis and wasted redesign.

Using unsuitable analyser settings

Resolution bandwidth, detector, sweep behaviour, attenuation and preamplifier state influence the displayed result. Compliance-oriented emissions work may require peak, quasi-peak or average detection and bandwidths defined by the applicable standard. One setting is not suitable for every frequency range or measurement purpose.

Breaking return paths with split planes

Routing a clock across a ground split increases loop area and can couple energy into analogue circuits or cables. The schematic may show separated domains, but the physical return current still needs a continuous low-impedance path.

Testing an unrepresentative operating mode

An idle processor, static display or inactive communications port may miss the highest-emission or most susceptible state. Exercise clocks, converters, interfaces and analogue channels in representative worst-case combinations, then record the exact firmware and configuration.

Failing to document the setup

A plot without cable positions, board revision, enclosure state, instrument settings and EUT mode is weak engineering evidence. It cannot reliably support regression testing, a technical file or later mitigation decisions.

Frequently Asked Questions (FAQs)

Should analogue and digital grounds always be split?

No. A continuous plane with controlled placement and routing often gives lower return-path impedance. Splitting may be appropriate in a manufacturer-defined architecture, but an uncontrolled slot beneath high-speed routing can increase emissions and analogue coupling.

How can I tell whether ADC noise is aliased?

Change the sample rate while keeping the suspected interference source stable. An aliased spur generally moves according to the relationship between the interference and sampling frequencies. Confirm with controlled source changes and review the analogue filter response.

Can a near-field scan predict formal radiated emissions?

Not directly. Near-field scans are strong localisation tools, but probe response, distance and coupling differ from a standards-defined radiated measurement. Use them to compare sources and modifications, then verify performance in a representative pre-compliance or formal setup.

When is shielding preferable to filtering?

Shielding is useful when field coupling dominates and a low-impedance enclosure or local shield connection can be achieved. Filtering is more effective where energy is travelling along conductors. Many faults require both, with careful treatment of the shield-to-filter boundary.

Does passing pre-compliance prove EMC compliance?

No. Pre-compliance improves confidence and can produce calibrated engineering data, but formal evidence depends on the applicable requirements, test configuration and documentation. Manufacturers should verify the latest active standards, limits, test levels, operating modes and conformity obligations.

What should be recorded during a debug session?

Record board and firmware revisions, EUT modes, supplies, cables, terminations, enclosure configuration, probe position, analyser settings and every hardware modification. This evidence makes results reproducible and supports later risk assessment, technical documentation and formal test planning.

Plan the next measurement before changing the PCB

Effective mixed-signal EMC troubleshooting comes from separating source, path and victim, then using controlled measurements to test each theory. EMC Hire can help select suitable hire equipment, arrange on-site testing, support pre-compliance investigation, discuss formal compliance testing where appropriate, or book space at the EMC Hire test facility.

For a practical review of the test problem or an equipment hire quotation, contact the engineering team on +44 (0)1462 817111 or email sales@emchire.co.uk.