The plant is full of drives, welding sets and radio traffic. How can I be sure the pulses I see come from the motor?
Noise rejection is the core field skill in motor PD measurement. Use at least three independent discriminators before you write a diagnosis.
Measure the background first: Record the site with the motor de-energised and, for online tests, with the sensor on a healthy reference feeder. Any pulse appearing at both locations is not motor PD.
Voltage dependence: Raise voltage in steps and plot amplitude against voltage. True PD has a defined inception and extinction voltage and grows with stress; most interference is flat or erratic.
Three-phase comparison: Genuine winding discharge is normally dominant on one phase. Interference coupled through the earthing system appears on all three channels with near-identical timing - a classic common-mode signature.
Time-domain pulse shape: Discharge pulses are short (nanoseconds) and repeat at the supply frequency; thyristor and PWM switching is periodic with the carrier and can be removed by gating synchronised to the drive.
Frequency and polarity: Shift the measurement bandwidth away from the dominant interferer, and use pulse polarity - PD from a void reverses with the applied half-cycle, whereas capacitive crosstalk does not.
Cross-confirm: A signal seen by both an HFCT and an acoustic sensor, with a consistent time delay, is almost certainly internal discharge.