Volt-second Sensing Control Analysis in a DB-DTFC with the Motor Drive Analyzer

Introduction

Deadbeat-direct torque and flux control (DB-DTFC) has potential as an alternative to field-oriented controls (FOC) or direct torque control (DTC) due to its ability to achieve fast torque dynamics, manipulate losses without compromising torque, and utilize full DC bus voltage at voltage-limited operation. DB-DTFC selects the correct Volt-second input each switching period to produce the correct torque at the end of each switching period. A Volt-second sensing scheme has been developed for DB-DTFC that uses a Volt-second quantum pulse train to sense the motor terminal Volt-seconds over each switching period. Evaluating the Volt-second sensing accuracy requires a precise measurement of Volt-seconds as the reference.

The Critical Nature of Volt-second Accuracy

Power electronic inverters inherently source Volt-seconds. Thus, the Volt-seconds sourced by the inverter may serve as the key state in bridging power electronics to electric motors. In this regard, accurate delivery of Volt-seconds is very beneficial. Practically speaking, Volt-second error arises from inverter non-line arities or poor DC bus voltage measurements. Recently, researchers developed a volt-second sensing scheme yielding discrete pulses of Volt-second quantum [1]. Debugging, evaluating, and quantifying the Volt-second error in such control systems requires a measurement system that can acquire and analyze three-phase voltage and current and other control and drive signals during individual power-semiconductor device switching periods. The Teledyne LeCroy MDA810 Motor Drive Analyzer has these capabilities.

Figure 1 shows a variety of acquired signals and calculated waveforms used to understand the volt-second control performance of the system. In this example, the control system receives a torque command, and the control system generates a torque estimate. The control system uses the torque estimate to generate a volt-second command vector. Based on that vector, the control system calculates the next inverter switching state for all of the power-semiconductor devices. The advantage of the DB-DTFC control system is that it can respond instantaneously (within a single switching period) to rapid torque changes.

The following describes the various acquired signals (all located in the leftmost grids):

• Channel 1, or C1 (yellow trace, second grid from top left) is one of the three-phase line-line drive output voltages, and was acquired with a Teledyne LeCroy HVD3106 HV differential probe.
• C2 (magenta trace, top left) is the torque-command output signal in the control system (acquired with a passive voltage probe). When this signal goes high, the MDA810 triggers and acquires the signal.
• C3 (blue trace, top left grid and overlaid on C2) is the acquired torque-estimate signal as calculated (and output) by the control system. We acquired this signal with a passive voltage probe. Note that the torque-estimate signal lags the torque-command signal by one switching period.
• C4 (green, bottom left grid) is a digital output pulse from the controller GPIO that marks the beginning of the power-semiconductor device switching period. Although we used a Teledyne LeCroy HVD3106 HV differential probe for this acquisition, a passive probe would suffice. We have overlaid the DrvOutSync signal atop C4, and the transparent color-coded overlays indicate the switching periods (there are 306 periods in this acquisition).
• C6 (purple trace, third grid from top left) and C7 (red trace, same grid and overlaid on C6) are the command line-to-line voltage (generated by the DB-DTFC control system) and corresponding sensed line-to-line voltage (using the volt-second sensing technique developed in [1]), respectively. The control system’s digital-to-analog converter (DAC) generates both of these signals. These two signals should be the same, assuming ideal power conversion and accurate volt-second measurement.

The following describes the calculated waveforms (all located in the rightmost grids):

• Vdc(Vrs:Ir) (blue trace) is calculated from the per-cycle Vdc value for each switching period. It plots Vdc over time and is time-correlated to all of the acquired waveforms. Vdc(Vrs:lr) appears to the right of C2 (torque command) and C3 (torque estimate) voltages (top right grid). In this case, Vdc is the mean voltage value of the output line-to-line voltage for each switching period. We used Vdc rather than Vrms to show the positive or negative sign of the mean value; Vrms always returns a positive value. Ideally, this calculated waveform is the same as C7.
• The MDA810 calculates F3 (also a blue trace, to the right of C1) by subtracting Vdc(Vrs:lr) from C7 (sensed voltage). This reflects the error between the measured (sensed) voltage and the actual signed voltage waveform (as calculated by the MDA810) and is a measure of the volt-second sensing accuracy. Note that the voltage scale of F3 is vertically zoomed by 10x (it is 10V/div, whereas the source waveforms are 100V/div). Note that the measured voltage error increases at the point of torque change.
• The MDA810 calculates F4 (light green trace) by subtracting Vdc(Vrs:lr) from C6 (command voltage), which represents the difference between the command voltage and the actual signed voltage waveform (as calculated by the MDA810). The difference comprises the volt-second error that exists in the motor drive. Note that the voltage scale of F4 is vertically zoomed by 10x (it is 10V/div, whereas the source waveforms are 100V/div). Here again, the measured voltage error increases at the point of torque change.

In our endeavor, we sought to analyze the control system performance during the torque transition event, to measure the time between the torque command and torque estimate, and to confirm that this is occurring within one switching period (as shown in DrvOutSync Z signal). All of the following appear in Figure 2:

• The waveforms in the Full Acquisition Time tab are the original waveforms as previously described.
• In the top-right grid are zooms of the torque command and estimate (Z2 magenta trace and Z3 blue trace) with the switching period Sync signal (DrvOutS) overlaid on top of them.
• In the middle-right grid are zooms of the command and sensed voltages.
• In the bottom-right grid are vertical and horizontal zooms of the voltage errors.

Note that the volt-second sensing method for the DB-DTFC control system has very low measurement error at the torque transition event, and that the torque estimate occurs with a delay of one switching period.

Additional analysis could be performed to compare results in different steady-state and transient torque and speed conditions and different machine operating speeds. Although we used a short acquisition time in this example, the MDA810 supports up to 100 seconds of acquisition and analysis at the chosen sample rate of 2.5 MS/s (much higher sample rates are supported). Therefore, one may use the MDA810 to analyze one long acquisition with complex dynamic speed, torque, and load behaviors.

Conclusion

The Teledyne LeCroy MDA810 Motor Drive Analyzer provides unique capabilities to analyze power quantities during very high-frequency events (e.g., a power-semiconductor switching period), show the dynamic changes of these values over time, and correlate those changes to other measured power and control system signals.