MP6543A – Distorted negative phase-current measurement with low-side current sensing

Hello,

I am developing a 3-phase BLDC motor controller using an RP2350 MCU and an MP6543A motor driver.

I am measuring the three motor phase currents using the MP6543A low-side current-sense outputs (SOA, SOB and SOC), connected to the RP2350 ADC.

For each current-sense output, I use:

  • 22 kΩ from SOx to VCC

  • 22 kΩ from SOx to GND

  • 100 pF capacitor on the ADC/current-sense signal

The motor current is relatively low, below approximately 0.5 A.

For this test, the motor phases are driven using triangular excitation signals, phase-shifted between the three phases.

The ADC is sampled at 500 kS/s. For each phase, I only retain the ADC samples acquired while the corresponding low-side MOSFET is enabled, and I use the median of these samples as the measured phase-current value for that PWM interval.

The attached plot shows the resulting 12-bit ADC readings for the three current-sense channels.

The zero-current level is approximately 2020 ADC counts.

The positive-current portions look smooth and consistent with the expected waveform. However, when the current becomes negative, the measured waveform is strongly distorted. On phases 1 and 2, the measured value even periodically drops all the way to zero.

This behavior is very repeatable and synchronized with the phase excitation.

I would like to understand whether this behavior is expected from the MP6543A low-side current-sense architecture.

In particular:

  1. Is the SOx output a valid representation of the phase current in both directions whenever the corresponding low-side MOSFET is enabled?

  2. During negative phase current / current recirculation, can the current flow through a path that is not sensed by SOx even though the low-side gate input is active?

  3. Is there a specific part of the low-side conduction interval in which SOA/SOB/SOC should be sampled? For example, should samples close to the switching edges be excluded?

  4. Is taking the median of the ADC samples acquired during the low-side ON interval a valid method for reconstructing the phase current, or is a more specific sampling strategy required?

  5. Could the automatic synchronous rectification of the MP6543A explain the strong difference between the positive and negative portions of the measured current?

I would also like to know whether the 22 kΩ / 22 kΩ bias network and 100 pF filtering are appropriate for this application.

Any recommendation regarding ADC synchronization with the MP6543A switching states would be very helpful.

Thank you.