What Is Electromagnetic Compatibility?

Fifteen years in this business and EMC still bites me in the ass at least twice a year.

Last spring I got called in on a project for a company making delivery vans. Their BMS kept throwing cell overvoltage faults. Not real overvoltage - the cells were fine. The monitoring IC was reading garbage. They had been chasing this for three months. Replaced the AFE chips twice. Rewired the sense harness. Nothing worked.

Took me two days to find it. Their motor controller sat eighteen inches from the battery pack. No shielding between them. The controller was switching at 8 kHz and every edge coupled straight into the cell voltage sense lines. The wires were acting like antennas. Twenty millivolts of induced noise on a measurement that needs millivolt accuracy. The BMS saw voltage spikes that did not exist and shut everything down.

That is EMC. Electromagnetic compatibility. Your stuff has to work around other stuff without either one screwing up the other.

EMC breaks into two pieces and you have to deal with both.

Emissions means the noise your device puts out. Every switching circuit radiates. Every wire with changing current creates a magnetic field. A DC-DC converter running at 200 kHz sprays RF energy all over the place. If that energy is strong enough it jams the radio in the cab or corrupts CAN bus messages or makes the ABS controller think the wheels are locking up.

Susceptibility is the other side. How much crap can your device take before it stops working right. A BMS has to read cell voltages accurately while sitting next to an inverter that is hammering the DC bus with 400 amp pulses at 10 kHz. The sensing circuits need to ignore all that and still measure the actual cell voltage.

Most engineers I meet think about emissions because that is what the regulatory tests focus on. They forget susceptibility until the product fails in the field. Then everyone panics.

Old lead-acid batteries just sat there. You bolted them in and connected two cables. Maybe you had a shunt for current measurement. That was it.

Lithium packs have electronics everywhere. The BMS monitors every cell. It talks to temperature sensors scattered around the pack. It communicates with the vehicle over CAN or some other bus. It controls contactors. It calculates state of charge and state of health using algorithms that depend on accurate measurements.

Accurate is the key word. A cell voltage measurement that drifts by 50 millivolts throws off your SOC calculation. Do that across a hundred cells and your range estimate is useless. Drift it by 100 millivolts and you might overcharge a cell or miss an undervoltage condition.

The sensing wires are the weak point. They run from each cell tap to the BMS board. In a big pack that means wires running two meters or more through an environment full of electromagnetic noise. Every wire picks up interference. The longer the wire the worse the problem.

I worked on a bus battery pack where the sense harness ran right alongside the main DC bus bars. The bars carried 300 amps during acceleration. The magnetic field from that current induced millivolt-level signals in the sense wires. The BMS thought the cells were bouncing up and down. It went into protection mode and killed the bus on the highway.

Electromagnetic Compatibility

The failure modes are ugly.

False overvoltage trips are common. Noise adds to the real cell voltage and the BMS thinks the cell is at 4.3V when it is actually at 4.1V. Protection kicks in and the system shuts down. The operator sees a fault code that makes no sense because the cells are fine.

False undervoltage is worse. The BMS thinks a cell is lower than it really is. It keeps discharging past the safe limit. Real damage happens.

Communication errors mess things up differently. A battery pack might have six or eight monitoring boards daisy-chained on an isoSPI bus. That bus runs at a few megahertz. EMI corrupts a packet and suddenly the main controller has bad data for sixteen cells. Does it shut down? Does it use the last good reading? Does it interpolate? Every option has problems.

Temperature measurement gets corrupted too. NTC thermistors produce tiny signals. A few millivolts of noise looks like a twenty degree temperature swing. The BMS might turn on cooling that is not needed or miss a thermal runaway that is starting.

Regulatory standards exist but they are a mess of overlapping requirements.

FCC Part 15 covers emissions in North America. It sets limits on how much RF energy your product can spray into the air. The limits depend on whether you are selling to consumers or industrial users. Industrial gets more slack.

For vehicles you also deal with CISPR 25 and CISPR 12. These are international standards that most countries adopt with minor tweaks. They specify test methods and limits for conducted and radiated emissions from vehicles and vehicle components.

The European CE mark requires compliance with the EMC Directive. That usually means meeting EN 55032 for emissions and EN 55035 for immunity. For automotive stuff you reference EN 50498 instead.

Then every car OEM has their own specs layered on top. Ford has their thing. GM has another. VW has a whole book. These are usually stricter than the regulatory minimums and include immunity tests that the regulations barely touch.

Getting through all this testing costs money and time. A full EMC qualification for an automotive BMS runs $30,000 to $50,000 and takes three to four weeks of chamber time. If you fail you go back and fix things and test again. I have seen programs burn six months and $200,000 on EMC testing before they got a pass.

Electromagnetic Compatibility

Shielding is the blunt instrument. Put a metal box around the sensitive stuff and most interference cannot get in. The box has to be continuous though. Seams leak. Holes for connectors leak. Ventilation slots are basically windows for RF energy.

Real shielding means conductive gaskets at every seam. It means filtered connectors or feedthrough capacitors at every wire entry point. It means no slots longer than one tenth of a wavelength at your highest frequency of concern. That last one trips people up. If you care about 1 GHz interference your slots need to be less than 30 millimeters. Good luck getting cooling air through that.

Filtering handles conducted interference. You put inductors and capacitors in the right places to block high frequency noise while passing the signals or power you actually want. A differential mode filter handles noise that travels down one wire and back the other. A common mode choke handles noise that travels the same direction on both wires.

Filter design is tricky. The impedance of the source and load matters. A filter that tests great on the bench might do nothing in the actual system because the impedances are different. I have seen engineers proudly show me their filter schematics and then wonder why the noise measurements did not change. They designed for a 50 ohm system and plugged it into something that looked like 5 ohms at the frequencies they cared about.

Grounding strategy is where the real voodoo lives. The battery pack has high voltage and low voltage sections. The HV side includes the cells and main contactors and precharge circuit. The LV side has the BMS and CAN interface and sometimes isolated DC-DC converters. These two ground systems should connect at exactly one point. More connections create loops. Loops pick up interference.

Finding the right single point is an art. It depends on where the current flows and where the sensitive measurements happen and where the noise sources live. Every layout is different. I wish I could give you a formula but there is not one.

Inside the BMS the PCB layout makes or breaks EMC performance.

Keep high speed digital stuff away from analog measurement circuits. The processor and CAN transceiver and any switching regulators spray noise. The cell voltage inputs and temperature measurement circuits are sensitive. Physical distance helps. Ground plane breaks between sections help more.

Route differential pairs together and keep them short. The sense lines from the cell taps should enter the board as close as possible to the AFE chip. Longer traces pick up more interference.

Watch your return paths. Current has to flow in a loop. The signal goes out on one trace and comes back on another. If the return path is not obvious the current finds its own way and that way might be a big loop that radiates like an antenna.

Pour copper. Big ground planes reduce impedance and give return currents a low inductance path. Stitching vias tie the planes together and reduce resonances.

A few years back I consulted for a startup making battery packs for forklifts. Their BMS worked great on the bench. In the forklift it went crazy. Voltage readings jumped around. The SOC calculation wandered all over.

The forklift had a big DC motor with a chopper drive. Every time the chopper switched the current in the main cables changed by a few hundred amps in a few microseconds. The magnetic field from that dI/dt coupled into everything.

Their fix attempt was to add ferrite beads on the sense lines. Did nothing. The interference was coupling magnetically into the loops formed by the sense wires. Ferrites block conducted noise not magnetic field coupling.

We ended up running the sense harness through flexible conduit and routing it away from the main cables. Then we added a twisted pair from each cell tap instead of single wires. Twisting reduces the loop area that the magnetic field can couple into. Finally we slowed down the sampling on the AFE chip so it integrated over more cycles of the interference. The noise averaged out.

Total fix cost about three dollars per pack in extra wire and conduit. Would have cost nothing if they had thought about it during the original design.

Electromagnetic Compatibility

If you are buying battery packs or BMS boards ask about EMC before you buy.

What standards did they test to? Ask for the actual test reports not just a certificate. The reports show the test setup and the margin to the limit. A product that barely passed is one design change away from failing.

What electromagnetic environment did they design for? A BMS meant for a golf cart may not survive in a transit bus with three traction inverters.

Did they test the complete pack or just the board? The housing and wiring and connectors change the EMC behavior. Testing a bare board tells you almost nothing about the finished product.

What kind of filtering and shielding is built in? If the answer is "we rely on the system integrator to handle that" you have a problem. You are now responsible for EMC compliance and you may not have the skills or budget to do it right.

Wireless BMS is coming. Instead of running sense wires to every cell you put a little wireless module at each cell group and it transmits the data back to the main controller. Less wiring. Easier to service.

But now your measurement data is flying through the air in a radio signal. That signal has to compete with all the electromagnetic noise from the power electronics. The wireless bands are already crowded. I have seen prototypes where the wireless BMS lost communication during hard acceleration because the inverter noise jammed the receiver.

Higher voltages make everything harder. The industry is moving to 800V and beyond for faster charging and smaller cables. More voltage means more switching energy and more EMI. The same techniques still work but you have to execute them better.

Integration keeps increasing. The BMS and the onboard charger and the DC-DC converter all crammed into one box. Less weight and cost but more opportunities for interference between functions. The analog cell sensing has to survive in the same enclosure as a 10 kW switching charger.

EMC is not going away. It gets worse every year as electronics get faster and denser and we ask them to work in harsher environments. The battery pack is the heart of the electric vehicle and keeping it accurate and reliable despite the electromagnetic storm around it is what EMC work is about.