Increase sampling time for motor control

Inside Machines: Pulse width modulation (PWM) using opposite voltage vector can extend time for analog dc (ADC) sampling in motor control applications.

By Jiri Ryba, PhD March 18, 2014

Modulation can increase active sampling time by opposite voltage vector insertion. In motor control applications, the proposed method increases time suitable for sampling of analog values.

While examples here show modulation for brushless dc (BLDC) electric motors, modulation can be used in other applications. Options for back electromotive force (BEMF) sensing depend on modulation techniques, and there advantages and disadvantages. Measurements examples are provided on a BLDC motor, and an implementation of a GTM (general timer module) for SPC57xPWT microcontrollers is provided.

Pulse width modulation, sampling

Analog values typically are sampled in specific time, synchronized with PWM to exclude sampling during ringing. Ringing is a transient state after the power switches to commutation. Depending on the application and type of sampled values, the sampling could be done during off-time or during on-time. The time required for ADC sampling limits duty-cycle range, which could limit motor speed. During motor freewheeling, less BEMF may occur.

The described modulation can insert an opposite voltage vector to ensure the minimal required on-time overvoltage range. The proposed method does not affect unipolar modulation at high speed, and reduces EMC compared to bipolar modulation. The modulation also is suitable for dual motor control, where current sourced from the capacitor could be suppressed by opposite current from the second motor with shifted PWM. 

1. Pulse width modulation with active time extension

The PWM (pulse width modulation) is designed to ensure sufficient active time for BEMF sampling in BLDC motor control applications. The equal distribution of switching and conductive losses over all transistors also is considered.

The BEMF is sampled in the unpowered phase where the other two phases can be supplied (sampling in active time / on-time) or connected to same terminal (sampling in off-time). The sampling in off-time is typically more sensitive, but the half of BEMF period (typically negative) is cut by bridge diodes. When the BEMF is sampled during active time, the motor node is in half of supply voltage. The trend of BEMF can be observed in both polarities.

For low duty-cycle, the active time can be shorter than the time required for BEMF sampling for unipolar modulation. The minimum required time is given by ADC sample/conversion time, number of consecutive measurements, and time of transient rippling after switches commutation. The rippling time depends on motor construction and power stage design. Any design asymmetry and parasitic impedance should be minimized to reduce the rippling time. Nevertheless, the minimum voltage will be limited for unipolar modulation.

This limitation does not occur for bipolar modulation, where the second phase is switched complementary to the first phase. In this case the average voltage is zero for 50% duty-cycle. The main disadvantage of bipolar modulation is high current rippling and higher switching losses.

The proposed PWM extends the active time in unipolar modulation to the required width (see Figure 1). The pulse width extension is compensated by the opposite voltage vector. Inserting the opposite voltage vector doubles the switching frequency. In this case the switching losses are doubled. Because the opposite vector insertion is needed only for low speed where the conductive losses are typically low, the maximum power dissipation of transistors is not increased. The need for the active time extension should not occur even in the application, where the high current is expected also during low speed. The required voltage is higher in this case due to compensation of voltage drop on winding resistance. The alternate PWM generation could be used to exclude a short time between the transistor switches (see alternative phase A and B in Figure 1). This could be required by transistors for switching losses diversion. The disadvantage of this modulation is more complex generation mainly in transient between generation with compensation pulse and without compensation pulse. There are two possibilities for solving the transient state.

The first possibility is to continue with phase shifting. In this case the negative pulse naturally comes to positive pulse. This could continue until both positive pulses are same. This method is natural for the selected modulation, but in the specified range allows ADC sampling only every second period. In this case the PWM frequency is the same even when the negative pulse is generated. In reality the behavior is different; compare to the modulation in Figure 1. In the first step, every second pulse decreases the pulse width. In the second step, this pulses change polarity.

The second possibility is to reduce frequency to half as soon as the negative pulse reaches zero width. In this case the motor voltage UAB will be exactly same as in Figure 1, but the handling of the modulation is more complex. The PWM frequency also is doubled by inserting the opposite voltage vector, as Figure 1 shows. 

2. Generation of PWM by GTM

The modulation was tested in a simple BLDC motor control application. The application is designed on a PowerPC based microcontroller, while the PWM is generated by the GTM. The PWM could be generated on a timer output module (TOM) channel or an ATOM [ARU (advanced routing unit) connected TOM) channel. Both modules have similar functionality. The main difference is that the ATOM is connected to ARU so the input data could be automatically loaded from other modules, such as a multi-channel sequencer (MCS) or parameter storage module (PSM). On the other hand, the TOM channels allow connection to the sensor pattern evaluation (SPE) for automatic commutation, but this feature is not used in this application as the current version of the GTM does not support selecting four different duty-cycles at same time.

The PWM update is executed every period. The PWM requires different settings for the odd and even periods. The compare registers are updated according to the required duty-cycle based the on odd or even period. To reduce CPU load, the update could be processed by MCS or PSM.

The reload of the register occurs in the middle of the active time.

Figure 2 lists compare values for when an opposite vector is not entered and for when an opposite vector is entered.

The formulas are valid for the voltage vector where the first phase (A) is positive, second phase (B) is negative, and third phase (C) is off. The other vectors are generated in same way, but the phases are changed accordingly. See Table 1. The input is required duty-cycle. 

Table: Phase State

2.1 Compare value calculation when the opposite vector is not entered

Modulation in phase A is following:

  • Top transistor first event: TFPOFF = Duty/2 – DT;
  • Bottom transistor first event: TFPON = Duty/2 + DT;
  • Top transistor second event: TRPON = Period – TFPOFF;
  • Bottom transistor second event: TRPOFF = Period – TFPON;

Phase B is connected to negative terminal (bottom transistor is on)

Phase C is disconnected (both transistors are off). 

2.2 Compare value calculation when the opposite vector is entered

Modulation in phase A is the same as in the previous case, but the duty is replaced by the minimal duty-cycle to ensure required active time (PWMMIN).

  • Top transistor first event: TFPOFF = PWMMIN /2 – DT;
  • Bottom transistor first event: TFPON = PWMMIN /2 + DT;
  • Top transistor second event: TRPON = Period – TFPOFF;
  • Bottom transistor second event: TRPOFF = Period – TFPON;

Modulation in phase B is following:

  • Top transistor first event: TFNON = Period/2 – (PWMMIN – Duty)/4 + DT;
  • Bottom transistor first event: TFNOFF = Period/2 – (PWMMIN – Duty)/4 – DT;
  • Top transistor second event: TRNOFF = Period – TFNON;
  • Bottom transistor second event: TRNON = Period – TFNOFF;

Phase C is disconnected (both transistors are off). 

Measured results, opposite vector

The described modulation was tested with a 12 V Pittman motor 3441S001-R3. Figure 3 shows the control signal of transistors, phase-to-phase voltage, and motor current. Inserting the opposite vector doubles the frequency of the first harmonic.

Insertion of the opposite vector ensures the required active time to measure needed analog values. The modulation was successfully tested on the PowerPC based microcontroller. Modulation slightly increases current rippling compared to unipolar modulation, but the current rippling and losses are less than in standard bipolar modulation.

– Jiri Ryba, PhD, is field applications engineer, STMicroelectronics. Edited by Mark T. Hoske, content manager, CFE Media, Control Engineering, mhoske@cfemedia.com.

Key concepts

  • Pulse width modulation (PWM) using opposite voltage vector can extend time for analog dc (ADC) sampling in motor control applications.
  • Insertion of the opposite vector ensures the required active time to measure needed analog values.
  • Modulation slightly increases current rippling compared to unipolar modulation, but the current rippling and losses are less than in standard bipolar modulation. 

ONLINE extra

www.st.com 

References

[1] V.Viswanathan, Dr.S.Jeevananthan, "A Novel Current Controlled Space Vector Modulation based Control Scheme for Reducing Torque Ripple in Brushless DC Drives," International Journal of Computer Applications (0975 – 8887) Volume 28- No.2, August 2011

[2] Dae-Kyong Kim, Kwang-Woon Lee, and Byung-Il Kwon, "Commutation Torque Ripple Reduction in a Position Sensorless Brushless DC Motor Drive," IEEE Transactions on Power Electronics, Vol. 21, No. 6, November 2006

[3] ST Microelectronics, "RM RM0334 Reference manualSPC574Kxx – 32-bit Power Architecture® based MCU for automotive applications" Rev 2 October 2012

[4] Bosch GTM (Generic Timer Module) Specification

See also other motor and drive articles below.