Resolving EMI on an RS-485 BMS Link

I reproduced the fault with the accumulator, BMS, and RS-485 wiring installed in their operating configuration. Communication remained stable before power delivery, but valid BMS messages were lost when the powertrain became active. Repeating the operating sequence produced the same behavior, establishing a clear relationship between powertrain operation and the communication failure.

The failure was identified through a consistent event chain: voltage, temperature, and current messages stopped arriving, the diagnostic logs reported lost BMS communication, and the safety system opened the AIRs because valid battery data was no longer available. This made a permanent protocol or software defect less likely and provided a repeatable test condition for the physical-layer measurements performed in Step 2.

I monitored both conductors of the RS-485 pair with a Tektronix DPO4054 oscilloscope while recording BMS diagnostic messages and repeating the powertrain operating sequence established in Step 1. This allowed the conductor voltages, communication errors, and inverter operating state to be compared during the same failure event.

The captures showed fast transient spikes recurring at approximately 25 kHz while the inverter was operating. This repetition rate matched the inverter IGBT switching frequency, and the disturbance appeared during the same events as the loss of BMS messages. The combined electrical and functional evidence identified inverter-generated EMI as the likely source of the RS-485 communication failure, rather than an independent application-layer fault.

Based on the measured transient waveform and its effect on the RS-485 communication levels, I developed a first-order RC filter to attenuate the interference. The design had to reduce the transient amplitude while preserving sufficient differential voltage, transition speed, and timing margin for reliable frame detection.

I evaluated candidate component values under the same inverter operating conditions used during diagnosis. After each iteration, I compared the conductor waveforms, BMS messages, and diagnostic errors to determine whether the interference was reduced without excessively degrading the communication signal. In parallel, I reviewed the grounding and shielding arrangement to reduce the coupling paths rather than relying on filtering alone.

I integrated the selected RC network into the accumulator communication wiring while keeping the added conductor lengths short to limit parasitic effects and additional interference pickup. The connections were soldered, mechanically secured, and visually inspected before the system was energized.

I also reviewed and adjusted the local grounding and shielding arrangement to reduce interference coupling into the communication path. After inspection, the completed installation was prepared for validation using the same powertrain operating sequence and measurement points established during diagnosis.

I repeated the original powertrain operating sequence using the same oscilloscope connection points and BMS monitoring established during diagnosis. Comparing the pre- and post-mitigation captures showed that the repetitive transient bursts were substantially attenuated while the RS-485 communication levels remained clearly distinguishable.

The improvement was then confirmed at system level throughout the tested sequence: BMS voltage, temperature, and current messages remained available, the diagnostic communication errors no longer appeared, and communication loss did not cause an unintended AIR opening.