Quantifying the Difference: Predictive Energy Balancing Controls for Switched-Mode Power Converters
To some, Predictive Energy Balancing (PEB) controls for Switched-Mode Power Supplies appear as a radical, untested departure from familiar practice. To others, PEB is just another loop compensation scheme, by another name. In fact, it is neither. To examine that assertion, take a minute to review the underlying principle.
To some, Predictive Energy Balancing (PEB) controls for Switched-Mode Power Supplies appear as a radical, untested departure from familiar practice. To others, PEB is just another loop compensation scheme, by another name. In fact, it is neither. To examine that assertion, take a minute to review the underlying principle.
A PEB controller is based on energy rather than voltage or current, as in a typical controller. Two energy terms, Supply and Demand, are used to provide the control. The Demand term, assessed at the output filter, is the difference between the energy at the regulation point and the instantaneous energy, that is:
(Vregulation² * C/2) – (Vinstantaneous² * C/2) where C is the output filter capacitance
The Supply term is the instantaneous energy stored in the switched inductor:
I² * L/2 where L is the switched inductance
The inductor is charged until Supply equals Demand, which is the instant of energy balance. At the precise time that these two terms are equal, the inductive charging stops, and the inductive Supply energy is transferred to the output filter capacitor. After that transfer, the output voltage will be at the regulation point. This effectively removes the output filter delay from the feedback loop, enabling single-cycle response. The gain of the loop is proportional to the ratio of L to C, and is constant for a given design. The constant gain assures inherent stability without compensation.
The PEB control algorithm is not specific to any one topology of power converter. The algorithm can improve the performance of any of the common topologies, and can regulate some forms that might be imprudent to attempt without PEB. Figure 1 shows a buck/boost topology configured as a bidirectional, unipolar power converter.
Figure 1: PEB power converter block diagram with logic and switch decoding for up to 4 power switches
Four power switches act in diagonal pairs. Closing the upper left and lower right switches energizes the inductor from the power source. That stored energy is transferred to the output capacitor through the opposite pair of diodes when all switches are open, or, synchronous rectification can be performed using the opposite pair of switches, completing the energy transfer. The topology is symmetrical, allowing power to be moved in either direction. The most significant difference between this PEB converter and a conventional buck/boost converter is the squaring of terms for expressing Supply and Demand in units of energy. While this might not seem to promise a significant difference in performance, the screen shot below clearly shows the benefit.
Figure 2: PEB power converter following a 5 kHz reference input while being clocked at 200 kHz
The reference input, in yellow, is a 5 kHz square wave of about 3 volts pk-pk. The PEB converter is scaled to output twice the reference input voltage, so the output, in purple, is a 6 volt pk-pk square wave. The PEB clock runs at 200 kHz, therefore a cycle of PEB operation is one minor scale division on the horizontal axis. Two cycles of latency on the rising edge are apparent because the severe discontinuity requires more than one cycle to energize to the balance point. The green trace is the current in the lower right switch as measured across a 100 mOhm sense resistor. Positive excursions on the green trace show energizing from Power In. Negative excursions are transfers returning energy to the Power In filter capacitor. These reverse transfers do not require prediction. The inductor is energized from the output until the output voltage is reduced to the regulation point.
The voltage regulator shown in Figure 2 has an output filter of 2.2 μf, and is supporting a 100 ohm fixed load. The Yokogawa DLM2054 oscilloscope is set to a bandwidth of 100 MHz on all three active channels.
In the next example, Picotest performed measurements on a similar PEB based converter. The converter is shown in Figure 3, mounted in a test stand with a high speed current injector and a 1-port PDN probe. Two tests are performed on this test board. In the first test a dynamic load step is applied to the regulator, via the high speed current injector while the 1-port PDN probe monitors the response.
Figure 3: The PEB test board is shown mounted in a test stand with a 1-port PDN probe monitoring the output and a J2111A high speed current injector, connected to the output, to provide the current modulation.
The dynamic load is set to step from 10 mA to 30 mA in 25 ns at a repetition rate of 5 kHz and the results are shown in Figure 4. The output voltage response to the dynamic load step is shown in the upper trace and the current step is shown in the lower trace. The oscilloscope traces are both averaged and the voltage response is bandwidth limited to 200 MHz to get a clearer view of the leading edge ring.
Figure 4: The upper trace shows the voltage response to a 10 mA to 30 mA fast edge dynamic load change at a 5 kHz repetition rate. The high speed current injector produces approximately 25 ns edge speeds. Note the slight ring at the leading edge.
The voltage regulator is then assessed for stability. Since the PEB controller is not a linear circuit, the assessment cannot be performed using a traditional Bode plot. The stability is assessed using the Picotest non-invasive stability method, which is based on an output impedance measurement. The output impedance is measured using the same 1-port PDN probe with an OMICRON Lab Bode 100 VNA in an S11 reflection based measurement. The measurement result is shown in Figure 5. The non-invasive phase margin is shown as 61 degrees, with a slight peaking in the output impedance. This is consistent with the very slight ring at the leading edge. Note that this peaking is dependent both on the operating current and the dynamic current signal amplitude.
Figure 5: The output impedance plot shows a relatively flat impedance profile, ideal for Automatic Voltage Positioning (AVP) and most Power Distribution Network (PDN) applications. The Picotest non-invasive stability measurement indicates a 61 degree stability margin from this measurement.
Changing from an AC to a DC reference input, changing the amplitude and/or frequency of an AC reference input, changing the Power In voltage, or adding a dynamic load, do not significantly alter the loop dynamics. The converter clock can be faster or slower or adaptive. The gain is calculated from the C/L ratio, so you know what to expect with other L and C values. There is no guesswork, and no compensation.
Under test, the PEB control system provides nearly single cycle response with stable performance. The high speed step load response and the Picotest non-invasive stability assessment verified these claims. The resulting flat impedance profile makes the PEB control system ideal for PDN and AVP applications. The PEB controller can be applied at much higher clock rates and at much higher power levels. The PEB controller combined with higher speed switches, such as eGaN devices could potentially provide an ideal power conversion solution.
We hope you would now agree that PEB is not just a disguised compensation scheme, and that it has been shown to provide excellent dynamic performance and stability. The PEB power converter controls described here are covered by 5 issued patents. They are available for license from CogniPower, LLC.
