EMI testing for IoT transceivers

A few oscilloscope measurements can give insight into sources of electromagnetic interference (EMI) that garble the operation of wireless IoT devices.

Examine a typical wireless device with an internet of things connection, and you’ll likely find multiple RF receivers squeezed together in close proximity onto a circuit board. Cramped quarters characterize modern IoT electronics, so sensitive RF receivers often sit nearby other circuitry that can potentially interfere with their operation. This reality has put a premium on being able to detect and correct interference problems on embedded systems.

Specialized instrumentation is available for testing electronic systems against EMI and radiated emissions standards. But oscilloscopes are generally the go-to instrument for isolating problem areas while IoT gear is in the development phase. A few tips can help speed the debugging process when scopes are involved.

Many of the interference problems associated with IoT wireless devices manifest as noise that prevents the IoT device from passing total isotropic sensitivity (TIS) tests dictated by cellular carriers. TIS is the minimum power level of a cellular signal that the IoT device can reliably receive via its cellular antennas in all directions. For low-throughput cellular technologies such as LTE-CAT-M1 and NB-IOT, transmit and receive functions share one antenna. A cellular product with multiple antennas can undergo TIS testing for each antenna separately or for both of them simultaneously. For a cellular IoT device, TIS is a measure of how far away the device can be from a cell tower before the cell signal is too weak to receive. Thus, TIS is a receive/downlink test.

(Note that there is also a transmit/uplink test called total radiated power (TRP) that adds up all the power the device transmits at a particular channel of a cellular transmit band. The better the TRP, the farther from a cell tower the device’s signal can be received. But TRP performance is generally unaffected by EMI the IoT device generates.)

IoT devices operating in the U.S. must have TRP and TIS figures that exceed minimum levels set by cellular carriers in the U.S. Otherwise, the IoT device can’t be certified for use on any cellular carrier whose minimum requirements it doesn’t meet.

Cellular TIS is measured on a single channel within a cellular downlink band. Three receive channels are generally tested per band. The exception is LTE Band 13 used in the U.S. primarily by Verizon, which is only tested at one receive channel centered at 751 MHz. The minimum allowable TIS for certification is generally around -99 dBm.

The problem is that high-frequency switching or digital signals can interact with nearby conductors, including those residing within the PCB, to produce electromagnetic radiation. If that radiation includes frequencies (spectral content) in the cellular receive bands, the radiation can drown out low-level cellular signals in those bands, thus compromising TIS test results and the IoT device’s ability to receive cellular signals from cell towers. Often, EMI is said to “desense” receivers when it is high enough to have an impact on performance. Having both time and frequency displays in an oscilloscope (Figure 1) helps you troubleshoot EMI by correlating an event with the frequency of interference.

EMI design reviews of IoT devices tend to focus on bus noise, clock signals, and particularly dc-dc converters as the main culprits contributing to noise energy. Noise sources couple into nearby circuits via four typical paths: radiation from cables, through common impedance coupling where a noisy source shares the same current return path as a noise-sensitive circuit; and through capacitive and inductive coupling.

Capacitive coupling arises where parallel conductors essentially form the plates of a capacitor. The classic example on a circuit board is the capacitance between a heat sink and a PCB return path or trace. The capacitance allows signals with a high dv/dt to pass. Similarly, inductive coupling arises between parallel wires that induce current in nearby conductors. The inductance allows signals with a high di/dt to pass. See Figure 2.

The types of signals that get inductively or capacitively coupled include clock signals. Clocks tend to generate narrow-band EMI consisting of spikes located at the clock frequency harmonics.

In contrast, dc-dc converter noise tends to be broadband and manifests itself as a rise in the noise floor of the receiver. The broadband noise is not caused by the converter’s fundamental switching frequency, which generally is in the 3 MHz range or below. Rather, noise tends to arise because of the fast rise time of the converter switches.

Rise times can be on the order of a nanosecond. Worse, there can be ringing on the switching waveform as shown in Figure 2. Given that the ringing happens after the rise or fall time of the converter switch, the ringing waveform’s rise and fall times have a more severe impact on EMI. There are peaks in the emissions spectra corresponding to the ringing frequency and its harmonics. All in all, fast power switching and ringing waveforms from today’s converters can generate EMI well above 1.5 GHz.

PCB layout practices can be another source of EMI. Experts says power and signal-return PCB layers should be spaced no further than 3 mils apart, a practice frequently ignored in four-layer boards. Another suggestion: Put the ground return plane on the outer PCB layers so they can be stitched around the edge of the board to facilitate the construction of a Faraday cage. For similar reasons, PCB layouts of dc-dc converters should minimize the board area encompassed by conductor loops (Figure 3) where there is a high di/dt.

Diagnosing problems

One of the main tools for discerning radiated emissions is a near-field H probe (Figure 4). Oscilloscope makers as well as other manufacturers make these devices which often come in kits for different frequency ranges, typically spanning from about 10 kHz to 3 GHz. They are basically inductive loops. Holding the loop near the radiator allows it to pick up radiated noise by induction. The waveform captured by inductive probes have the same rise time, ringing, pulse period, and other qualities as waveforms captured directly with a conventional scope probe. The only difference is the amplitude of the captured signal.

One advantage to using a near-field probe is that there’s no need for direct access to a circuit probe point. Additionally, near-field probes avoid the possibility of shorting a power circuit with a scope probe when power components are densely packed. The disadvantage of the near-field probe is that it may only get close to a radiating source but may not be able to precisely locate it.

A way of getting a fix on the radiating power supply component is to disconnect the supply and temporarily substitute a battery, noting whether or not the noise disappeared (Figure 5). In the same vein, converters often contain a main power inductor that can be disconnected, which disables the switching action.

Also helpful is a facility found on many higher-end scopes is time and frequency correlation (Figure 6). This capability allows the scope to trigger on an event and simultaneously generate both the voltage/time display as well as the frequency display so that frequency components corresponding to specific parts of a zoomed-in waveform time display can be examined.

Another helpful component is the RF current probe (Figure 7), useful for detecting harmonic currents on cables. These resemble ordinary clamp-on current probes but handle upper

frequencies in the 200 MHz to 500 MHz range and can detect harmonic currents low enough to be measured in microamps. These probes are basically current transformers where the cable under test is the primary and the probe is the secondary, loaded by the scope 50-ohm input.

RF current probes on cables are for measuring conducted rather than radiated emissions. They are useful for ensuring devices with cables pass the FCC Class B radiated emission test which permits 6 to 8 µV radiated fields at most.