In conventional back-EMF zero-crossing detection techniques for BLDC with 120◦ electrical conduction mode, a zero-crossing event of the back-EMF signal occurs 30◦ electrical before each commutation instant, Fig. 4.2. The back-EMF signal of the unexcited phase is compared to a sample of the neutral point voltage. Differ- ent methods are proposed to obtain a good measurement of the back-EMF signal [Lai2008, Darba2013b, Shao2003, Shao2006]. The back-EMF zero-crossing in- stant occurs when the back-EMF voltage value equals the voltage of the star point of the stator windings. For delta connected windings and machines where the star point is not available a virtual neutral point is used. In [Lai2011] it is shown that the back-EMF detection depends upon the PWM techniques and an approach to
54 4.2 Conventional Zero-Crossing Detection Methods commutation spikes clearance delay back-EMF monitoring interval phase back-EMF phase back-EMF commutation instant [n+1] commutation instant [n] Td zero-crossing timingTz,1 zero- crossing instantof thefiltered signal Tz,1-Td (a) phase U back-EMF phasedU back-EMF commutation instant [n+1] commutation instant [n] commutation instant [n-1] actual zero- crossing instant phase V back-EMF zero-crossing timingTz,2 zero- crossing instantof thefiltered signal Tz,22 (b)
Figure 4.2. Different methods to determine commutation instants using back-EMF zero- crossing detection, (a): by using time difference between commutation and zero-crossing instant, (b): by time difference between successive zero-crossing instants
detect the zero-crossing points is explained in detail. Counting from the back- EMF zero-crossing detection, the commutation instants can be triggered 30◦after
Self-Sensing Commutation of BLDC Machines Using Back-EMF Samples 55
each zero-crossing. When the previous commutation instant [n] occurs, a timer starts to count and the time interval between this instant and the zero-crossing of the back-EMF is measured, Fig. 4.2(a). This time Tz,1is often assumed to be the
time interval during which the rotor position has changed by 30◦electrical. With this assumption, the commutation instant [n + 1] is triggered a time Tz,1after the
back-EMF zero-crossing event. In another method the time interval Tz,2 between
two successive commutations is measured and half of this time Tz,2
2 is used as an
equivalent of the 30◦ electrical position Fig. 4.2(b).
In [Iizuka1985], this has been implemented on a microprocessor where a low-pass filter is used to remove the high-frequency components in the measured back-EMF signal. These high frequency components are the results of the switching actions of the power converter in order to control the average value of the voltage applied to the machine. However, the time delay of the low-pass filter has a constant value for different speeds (as indicated in Fig. 4.2 by Td) and hence this limits the speed
range.
Different improvements have been implemented for the zero-crossing method. For example in [Kim2011] a zero-crossing detection of the line-to-line measurement of the back-EMF signal is implemented but it shows the same problems of conven- tional zero-crossing methods. In [Shao2003, Shao2006] an alternative back-EMF zero-crossing detection is proposed using the negative terminal voltage of the DC- bus instead of the neutral point of the machine. It uses the zero-crossing detection concept so adding a fixed 30 degree delay is still necessary. In [Damodharan2010] the terminal voltages measured with respect to the negative DC-bus voltage are used to determine the zero-crossing instant of the back-EMF. In [Tsotoulidis2015] zero-crossings of the filtered zero sequence voltage (ZCV) are used. This method also uses a filter to obtain a more smooth ZCV. Another method to detect the com- mutation instants of a BLDC machine using the back-EMF zero-crossing concept is the integration of the back-EMF of the unexcited phase as it results in the rotor flux and hence position as discussed in [Becerra1991]. This method is less sensit- ive towards switching noise so it does not use any filter but it has some offset errors arising from the integration action.
In the aforementioned studies [Iizuka1985, Kim2011, Shao2003, Shao2006, Damodharan2010, Tsotoulidis2015, Becerra1991] obtaining the instant of com- mutation is realised by adding a 30◦ electrical delay to the zero-crossing instants of the back-EMF and show therefore the same inherent problem.
If the motor phase windings are star connected and the star point is available for measurements, summation of the three phase voltages results in a third harmonic voltage. Integration of the third harmonic voltage results in an estimation of the third harmonic flux of the motor. Zero-crossing instants of the third harmonic flux correspond to the commutation instants [Moreira1996]. This method shows the exact instant of commutation and solves the problem of calculating exactly 30◦
56 4.2 Conventional Zero-Crossing Detection Methods
electrical, but the noise-to-signal ratio in the lower speed range is more critical compared to other methods because of the relatively low amplitude of the third harmonic component. Moreover, the access to the neutral point of the motor is re- quired which is not always the case and the drift can be an issue for the integration. In [Shen2006] integration of the third harmonic is proposed for ultra high speed self-sensing BLDC drives. The performance of the proposed method is not evalu- ated for speeds lower than 40 krpm which makes it suitable only for very specific high speed applications.
Another indirect method to detect the zero-crossing is obtained by using the con- duction state of the free-wheeling diode as proposed in [Ogasawara1991]. This method works well in a wide range of speed but requires additional and complex electronic circuits to detect the current of the free-wheeling diodes and extra isol- ated power supplies. In [Cheng-HuChen2007] the commutation signals are directly detected from the average line-to-line voltages obtained from filtered PWM wave- forms and machine model which solves the 30◦ electrical shift problem but deals with the disadvantages of the filter delay.
To conclude, using the zero-crossing instants of the back-EMF results in inaccurate estimations of the current commutation instants. This is a result of the assumption of a constant rotor speed during which the variation in rotor position is propor- tional to the corresponding time interval. However, during speed transients such as acceleration and deceleration, using Tz,1 or Tz,22 results in an inaccurate or faulty
estimation of the 30◦ electrical position. Such an error can have a great influence in the speed estimation and could result in unstable behaviour of the drive during generator or braking mode. Fig. 4.3 shows the operation condition for a BLDC ma- chine in generator mode in open-loop speed control. When a sudden speed increase occurs a lower braking torque is generated by the self-sensing generator drive due to incorrect commutation instant detection (Fig. 4.3(b)). This loss in braking torque results in a monotonous increase in the speed, Fig. 4.3(c) which on its turn causes future incorrect commutation instant detections [DeBelie2015].
In [Darba2013a], the authors proposed a self-sensing method that detects the com- mutation instants more accurately referred to as threshold technique. It uses the back-EMF voltage variation instead of time differences to detect the 30◦electrical shift as it is done in most of the conventional methods. Simulation results pub- lished in [Darba2013a] show the advantages of this method. Experimental results are published in [Darba2013b, Darba2015b]. The basic principles of this method and the results of simulations and experimental tests are explained in Section 4.3. In [Darba2014, Darba2015a] another back-EMF based self-sensing commutation technique is proposed which also provides an accurate estimate of rotor position information. In Section 4.4 this method will be explained and simulation results as well as experimental validations will be used to prove the performance of the proposed method.
Self-Sensing Commutation of BLDC Machines Using Back-EMF Samples 57 50 60 70 80 90 100 110 120 20 0 20 PositionV[rad] Back EMFV[pu] PhaseVW PhaseVV PhaseVU (a) 50 60 70 80 90 100 110 120 1 0 1 Position [rad]
Braking torque [pu]
(b) 50 60 70 80 90 100 110 120 15 10 5 0 Position [rad] Speed [pu] (c)
Figure 4.3. Simulation results, continuous increase in the rotor speed of a self-sensing BLDC generator during a sudden increase in the driving mechanical power