Implementation of digital control design scheme for isolated bidirectional power converter

Infineon / Mitsubishi / Fuji / Semikron / Eupec / IXYS

Implementation of digital control design scheme for isolated bidirectional power converter

Posted Date: 2024-01-18

This article explores the implementation of isolated bidirectional DC-DC power transfer by adjusting a dedicated digital controller to support reverse power transfer (RPT) in addition to the standard forward power transfer (FPT) function. System modeling, circuit design, and simulation are introduced, and theoretical concepts are verified experimentally. Applications show that the conversion efficiency is always higher than 94% in both energy transfer directions.

Modular battery energy storage systems (ESS) contribute to the effective utilization of renewable electricity and are therefore a key technology in building a green energy ecosystem. The application of echelon battery ESS is becoming more and more widespread. In this submarket, it is expected that up to 80% of discarded batteries will be used in ESS to find a new life in fixed grid services, thereby extending the battery life from 5 to 15 years. It is expected that these systems will add 1 TWh of capacity to the grid by 2030. This emerging application is bound to become even more important in the energy market in the near future.

A typical implementation is to stack different battery modules and transfer their energy through a power converter to a centralized AC or DC bus (which then distributes the energy to the load in some form). The challenge with such a system is that each module has a different chemical composition, capacity and aging profile. In traditional modular topologies, the weakest module affects the total available capacity of the entire stack (Figure 1).

Figure 1. Challenges of modular ESS

To address this limitation, in the architecture shown in Figure 2, the energy in the stack is transferred to a common intermediate DC bus through individual DC-DC converters for each battery module. This energy is then passed through the main power converter to support a centralized medium voltage (MV) AC or DC bus. The voltage and power levels in Figure 2 were chosen based on typical data for ESSs on the market: 48 V battery modules, 400 V (DC) intermediate DC bus, above 20 kW (high power) main power converters and up to 1500 V Centralized busbar.

Figure 2. Battery-based modular ESS

In Figure 2, the ground reference of each module in the battery stack is different, so it is necessary to implement a separate DC-DC converter for each battery module through isolation. In addition, in order to support hybrid systems such as cascading battery ESS, each converter must also be able to transmit power in both directions. In this way, independent charging, discharging and charge balancing of each module can be easily achieved. Therefore, the core module of the application discussed in this article is the DC-DC converter, which is both isolated and bidirectional.

The following will explain how a power conversion-specific digital controller (usually built only for unidirectional power transfer) can be adapted to support bidirectional operation, so that the controller can be a good alternative to achieve the required safely and reliably. type of DC-DC converter.

Dedicated digital controllers for power conversion applications

For the control of switching devices in high-power DC-DC converters (greater than 1 kW), digital control is the current industry standard, and it is usually based on a microcontroller unit (MCU). Nonetheless, as various industrial applications place greater emphasis on functional safety (FS), the use of dedicated digital controllers may be more advantageous. From a system design perspective, simpler functional safety certification simplifies the design process, thereby reducing overall development time and achieving faster benefits, and is therefore particularly advantageous in modular implementations.

Some reasons why dedicated digital controllers are preferred over MCUs are outlined below.

Microcontrollers rely on software, contain a large number of states, and are considered unstable. Therefore, before the formulation of the IEC 61508 standard, microcontrollers were not allowed to be used in safety systems. A lot of the "functional safety" work for MCUs is in the software development phase.

In addition to the software, the MCU itself must also be certified.

Although the dedicated digital controller (as a configurable device) is still data-driven, its configuration process uses a limited variable language (LVL) instead of the fully variable language (FVL) unique to MCUs.

As a sequential digital machine, the functionality of a dedicated digital controller can be fully verified through testing, which is generally not possible with software in an MCU. Therefore, when using a dedicated controller, the device integrates core safety features.

The added security features in an MCU implementation may require considerable additional hardware compared to integrated security features in a dedicated controller. When using Failure Mode, Effects and Diagnostic Analysis (FMEDA), additional hardware often adds system-level complexity.

When using a dedicated controller, additional security (if required) can be obtained via an external MCU (usually provided at the system level).

Analog Devices' ADP1055 is a digital controller designed for isolated DC-DC high power conversion, providing a range of features to improve efficiency and safety. These features include: programmable overcurrent protection (OCP), overvoltage protection (OVP), undervoltage protection (UVLO), and overtemperature protection (OTP). Like many off-the-shelf equivalents on the market, this controller is designed for one-way power transfer, known as FPT. In order to achieve bidirectional operation, applications using this controller must be adapted to also work under RPT. The next section will explore an aspect that is important for both FPT and RPT modes, namely the efficiency of the target DC-DC converter, which must be understood before the tuning process can begin.

Achieve efficient energy conversion

Among various isolated bidirectional DC power transfer technologies, the architecture in Figure 3a has become one of the most commonly used architectures commercially due to its simplicity of implementation.

Figure 3. Power conversion topology simulation: (a) model and (b) efficiency in standard operation

This topology can be viewed either as a voltage-fed full-bridge to center-tapped synchronous rectifier in FPT or as a current-fed push-pull converter to full-bridge synchronous rectifier in RPT. To illustrate common application challenges, a typical use case is shown with a primary (DC bus) of 400 V (DC) and a secondary (battery module) of 48 V (DC) at a power level greater than 1 kW. Use LTspice to simulate the operation of a typical wide bandgap (WBG) power device with a switching frequency of 100 kHz. The parameters used in the simulation are shown in Table 1.

Table 1. Simulation study parameters

The results in Figure 3b show that efficiency drops rapidly at higher power levels when using conventional hard-switching (HS) PWM. This is even more highlighted when comparing RPT to FTP. To improve operation, we identified two main loss mechanisms that can be reduced by corresponding switching techniques explained below.

Soft switching: Figure 4a shows that in this low leakage inductance design, when using conventional PWM, the primary switches MA and MB do not turn off quickly during the passive to active switching transition. This condition creates higher switching losses throughout the system. In this case, using phase-shifted (PS) PWM (also known as zero-voltage switching (ZVS) or soft switching) helps reduce the drain-source voltage to zero during these transitions. We can do this by providing appropriate load-related dead time so that the drain-to-source capacitance of the switch can be fully discharged. The results of applying the phase shift are shown in Figure 4b.

Figure 4. Primary switch passive to active conversion: (a) HS PWM, (b) PS PWM

Active clamping: Figure 5a shows that during the turn-off period of the secondary switches MR1 ​​and MR2, a large spike and ringing is observed on their drain-source voltage. These transient events can compromise the integrity of the switch, waste energy, and cause electromagnetic interference (EMI). Implementing a digitally controlled active clamp using an additional switch (such as the MCLAMP in Figure 3) is a better alternative to mitigate the negative effects of this spike. This can further improve the efficiency of the architecture. The results of applying some form of active clamping are shown in Figure 5b.

Figure 5. Primary switch passive to active conversion: (a) HS PWM, (b) PS PWM

After implementing these strategies, the converter efficiency in RPT mode increased from less than 80% to over 90% at 5 kW. These simulation studies also predict similar efficiencies for FPT and RPT, as shown in Figure 3b.

To implement these switching functions, the ADP1055 provides 6 programmable PWM outputs for switching timing and 2 GPIOs that can be configured as active clamp sinks. Both functions can be easily implemented programmatically in a user-friendly GUI. For the benefits of these and other features of this digital controller, see the ADP1055-EVALZ User Guide, which considers standard FPT applications.

After identifying the mechanisms to achieve a feasible level of efficiency (which applies to both FPT and RPT modes for this application), we next explore how to adjust to RPT.

Adapt to reverse power transfer

To demonstrate the operation of the studied application under RPT, we created a low-voltage (LV) experimental setup for proof-of-concept. This device is based on the hardware from the ADP1055-EVALZ user guide and was originally designed for the standard case of 48 VDC to 12 VDC/240 W FPT, using the ADP1055 as the master controller with a switching frequency of fSW = 125 kHz. In order to adapt to RPT operation, appropriate modifications to the hardware and software are required. Figure 6 (top) shows the hardware portion of the signal chain for this task, with the following highlights:

Figure 6. Signal chain utilizes dedicated digital controller to accommodate RPT

Two matched isolated half-bridge gate drivers, ADuM3223, are used to turn on and off the four primary switches. The precision timing characteristics of these drivers (54 ns maximum propagation delay for isolator and driver) accurately reflect the control signal into the PWM.

The isolated power supply unit from the ADP1055-EVALZ user guide was rewired and supplemented with an auxiliary precision LDO (ADP1720) to accommodate the two ground references in the system and power all the different ICs in the application.

In the measurement section, the current measurement terminals on the shunt resistor are swapped so that the output current of the transformer secondary of the entire converter is measured in the correct direction on terminals CS2+ and CS2- of the controller.

Finally, the ADuM4195 isolated amplifier is used to safely and accurately measure the DC bus voltage. In RPT mode, the DC bus voltage is the output variable, while in FPT mode, the battery side voltage is the controlled output.

The ADuM4195-based measurement solution is an important addition to the control loop hardware. In addition to a safe 5 kV isolation voltage from the high-voltage primary side to the low-voltage control side, a wide input range of up to 4.3 V, and a reference voltage with approximately 0.5% accuracy, the ADuM4195 has a minimum bandwidth of up to 200 kHz. It enables faster loop operation compared to typical shunt regulator and optocoupler solutions, providing better transient response, which is critical for applications operating at 125 kHz switching frequency. Figure 7 shows the final experimental setup, with the added hardware in Figure 6 implemented in the ADuM4195-based measurement daughter card that was added to the original evaluation board in the ADP1055-EVALZ user guide.

Figure 7. Experimental setup for RPT proof of concept

Figure 6 (bottom) also describes the software configuration to adapt to RPT. We took a deep dive into digital control systems. The results are summarized in the description block of the process, as follows:

Correct steady-state response is achieved by changing the PWM settings so that the duty cycle changes are proportional to the secondary inductor charging. This is based on the architecture's boost-type operation in RPT mode.

We use the LCL output filter designed in the ADP1055-EVALZ user guide to determine the device's transfer function Gp(s) in the Laplace domain through AC small signal equivalent circuit technology. Unlike FPT, the response of the device under RPT is that of a second-order system with a right-hand zero (RHZ), which is the typical response of a boost converter under CCM. Note that this type of system is inherently unstable and requires a reduction in the bandwidth of the error amplifier.

Using MATLAB System Identification Toolbox, the feedback measurement Gm(s) was modeled based on the frequency response of an ADuM4195 used as an isolated follower (Figure 8). The dominant pole was confirmed to be around 200 kHz, ensuring fast response above the target bandwidth of the control system (around 10% of the observable dual frequency at 250 kHz).

Figure 8. Frequency response of ADuM4195

We chose to add a pole to the controller's standard digital compensator to reduce the overall control system bandwidth, which is necessary in such a non-minimum phase boost converter device. Therefore, we use the digital controller from Equation 1 (see the ADP1055 User Guide for constant definitions).

To keep the analysis in the Laplace domain, we create a continuous-time model Gc(s) of Gc(z) based on numerical control theory.Therefore, a computational delay (× z-1) is first added, and the final representation in continuous time is achieved by: utilizing (a) the Tustin approximation

and (b) Padé approximates discrete PWM (DPWM) delay (Tsa/2=1/4fsw) such that:

Finally, in order to design a stable response, we studied the open-loop transfer function Gol(s) = Gp(s) Gm(s) Gc(s) using MATLAB Control System Designer as a regular continuous-time control loop.

It can be observed that if the same control constants as FPT are used, the response under RPT will be unstable. Therefore, correctly designing the final values ​​of the constants in Gc(s) is crucial to ensure reliable operation. Once a stable open-loop transfer function is achieved by design, the controller transforms back to the digital domain. Figure 9 (left) shows the frequency response Gc(z) of the designed digital filter, which can be easily configured graphically using the GUI of the ADP1055 in Figure 9 (right).

Figure 9. Digital filter response configured on the ADP1055

We also configured the efficiency-enhancing features studied in the previous section (PS PWM with adaptive dead time and active clamping). It was experimentally found that in order to achieve proper ZVS in the active-to-passive conversion of RPT, it is necessary to modify the dead time in the PWM sequence. Specifically, we modify the turn-on time point of the secondary switch to occur before each active-to-passive transition interval to allow current reversal.

Testing showed that the modifications adapted to the RPT were successful, obtaining a 48 V primary output from a 12 V secondary input. Output voltage regulation is excellent for both load and input voltage changes, with relative standard deviation (RSTDEV) of 0.1% and 0.02%, respectively, as shown in Figure 10a. Figure 10b and Figure 10c show the conversion efficiency and step response to a 50% load change, respectively. In both cases, efficiency levels in RPT mode are similar to FPT mode, with a peak efficiency of 94% in the mid-power range. The step response parameters (overshoot and settling time) are (1%; 1.5 ms) in RPT mode and (2%; 800 μs) in FPT mode. We observe that lower overshoot, with slightly slower settling time, results in a stable transient response. These results demonstrate that the design process of adapting digital controllers to support bidirectional power transfer is efficient and successful.

Figure 10. (a) Output voltage regulation, (b) efficiency, and (c) 50% load step response obtained in RPT mode

in conclusion

For safe and reliable applications in the energy market, dedicated digital controllers for power conversion are a good alternative. This is because digital controllers help simplify functional safety certification compared to microcontrollers, resulting in faster system-level design time and faster time to revenue. These devices are typically built for unidirectional power transfer, and this article explores how they can be modified to support bidirectional operation. The application of isolated bidirectional DC-DC converters in battery-based ESS is demonstrated through theoretical models, simulations and experimental studies. The results verified the feasibility of the application, with similar performance achieved for energy transfer in both directions.

Review Editor: Liu Qing

#Implementation #digital #control #design #scheme #isolated #bidirectional #power #converter