Most of documents about this article show two basic formula –

For PWM off time – Vphase = 3/2*e

For PWM on time – Vphase = 3/2*e + Vdc/2

And show us beautiful voltage waveforms with PWMed BEMF, where ON and OFF time parallel. But i have another situation with mine 8kW hub-motor. Look at this:

Typical examle, where off-time BEMF differs from on-time on Vdc/2. Here’s mine:

Second PWM cycle inverted. Green area- affected by diode, integration starts in red area, after zero crossing. Closer view:

At the beginning of the cycle is seen resonant oscillations, for my motor they lasts about 10uS, ADC samples in center of ON time and OFF time, what give us clear data, without interference.

What is source of this mis-alignment? For proper BEMF integration i should take phase voltage, subtract neutral point, and integrate the resulting value. For off-time it is right, if we take neutral point as zero, coz other two phases on the ground, resultant step-switch very accurate with installed hall sensors. But if i take raw value from on-time, subtract Vdc/2, voltage is incorrect, step-switch comes too early. This means neutral point moves with BEMF, actually it is always moves, coz of third harmonic, but this is seen only when two other phases is powered.