Research & Development A bidirectional DC/DC converter in DAB topology
Many of the advantages of DC/DC conversion, which can be found individually in different standard topologies, are integrated in the DC/DC converters with DAB technology (Dual Active Bridge). This article describes its design and control.
The vision of using electrical energy from regenerative sources as efficiently as possible requires suitable energy storage systems. This also requires suitable power electronics for integrating the storage devices into the supply network, as well as a sophisticated operating strategy. Finepower has developed a DC/DC converter based on the DAB topology that combines advantages such as bi-directionality, galvanic isolation, and good parallelizability. In the following, the article outlines the advantages of such a DAB converter and what has to be considered in its design and control.
Depending on the principle, the cell voltage of an electrochemical energy storage device varies with the state of charge. Therefore, in almost all cases an electrical converter is necessary to provide an interface with a defined voltage to the outside. In order to make this possible for both the charging and discharging directions, either two separate converters or a bidirectional converter can be used.
Two active full bridges for charging and dischargin
In systems where installation space and costs play only a minor role (e.g. container storage), it may make sense to use two separate converters that are optimized for the respective operating range. Bidirectional converters are a better solution for cost-critical applications with limited installation space. In many cases, galvanic isolation in the converter is also necessary. This is a very important aspect, especially in-home storage solutions. This is because galvanic isolation of the memory from the mains and a memory voltage of up to 60 V allow battery modules to be exchanged or commissioned on-site without the help of a specialist.
Due to the currently low feed-in tariff, an increase in own consumption through the use of a home storage system is more attractive than ever. Bidirectional galvanically isolated converters are also becoming increasingly important in the automotive sector. A topology that very well combines the two properties of bidirectional energy flow and galvanic isolation is the dual-active bridge topology (DAB for short). The Dual Active Bridge is a topology with two active semiconductor bridges. Figure 1 shows the structure of a DAB structure.
A high-frequency transformer is connected between the two full bridges. On the one hand, this determines the coarse transmission ratio between the primary and secondary sides, and on the other, it ensures galvanic isolation. The third core component is the longitudinal inductance L1. Depending on the operating frequency and rated power, it can in the best case consist exclusively of the leakage inductance of the transformer. This procedure can save costs and installation space, but with the disadvantage of increased losses. The correct design of the transformer, therefore, requires some know-how and experience.
Due to the total of eight switching elements in this topology, the number of different control methods is very high and there are several degrees of freedom that can be used for optimization. Depending on the field of application of the converter, however, it is not always necessary to use all degrees of freedom for the control, whereby the complexity of the control can be considerably reduced. For a very limited input and output voltage range, for example, the simpler phase-shift control is sufficient, where all switching elements on the primary and secondary sides are controlled with a 50% duty cycle. The two bridge branches of a full bridge have a phase shift of 180°. Only the phase shift φ between the primary and secondary full-bridge is used to control the power flow. Figure 2 shows the basic control of the semiconductor switches. With the aid of these idealized current and voltage curves, the large signal transfer function and the maximum transferable power can be calculated analytically.
As Figure 3 shows, the usable working range for a control angle φ of -90° to +90° is given. When looking at the formula for the maximum transmittable power Pmax (Fig. 7), it becomes apparent that this depends on the input and output voltages, the switching frequency fs and the longitudinal inductance L1. The input and output voltages are usually already determined by the application area. Therefore, only the frequency or the longitudinal inductance can be used to dimension the converter. If a certain nominal power is to be transmitted, this results in an upper limit for the product of L1 and fs at fixed voltages. The determination of the optimum frequency then depends on the semiconductor switches used, magnet materials and the available installation space. As with all switched converters, the size of the inductive components in the DAB is indirectly proportional to the switching frequency.
The efficiency achievable with the DAB topology lies between 96% and 99% and is primarily determined by the nominal input and output voltage as well as by the switching frequency. A further influencing factor is the desired range of the input and output voltage. This is due to the fact that at low voltages and high powers the ohmic losses represent the dominant loss component and scale these quadratically with the RMS current. In addition, when phase shift control is used in combination with a wide voltage range, there are inevitably working ranges in which an increased reactive current component is created in the transformer (Fig. 5), resulting in higher losses in the transformer winding.
The other loss components in the DAB are given by the switching losses in the semiconductor switches, the core losses in the transformer and the skin and proximity losses in the transformer winding. Fig. 4 shows the efficiency of a DAB converter developed by Finepower from 400 V to 48 V. The DAB converter is a DAB converter developed by Finepower. The nominal power of the converter is 1500 W; the MOSFETs are switched at 140 kHz.
In addition to the ohmic losses, the switching losses of the power semiconductors represent a not negligible proportion of the total losses. In order to select suitable semiconductor switches, the operating conditions in the circuit must first be analyzed. These include the maximum RMS current through the respective switches, the maximum reverse voltage, and the switching currents when the semiconductors are switched on and off. The operating frequency of the converter is also an important criterion.
In order to achieve a high power density, it is necessary to increase the switching frequency. As a result, the usable switches are limited to MOSFETs. IGBTs are no longer suitable for frequencies above 100 kHz due to their high switching losses. Now it is important to select the right MOSFET for this application.
Switching the MOSFET with Zero Voltage Switching
If one considers the switching behavior of the components when operating the DAB with a high power, it quickly becomes apparent that all MOSFETs can be switched here with Zero Voltage Switching (ZVS). In the ZVS, the energy stored in the inductive components is used to recharge the switching nodes of the MOSFETs with virtually no loss. To achieve this, a positive drain-source current must be switched off. The current driven by the inductance then recharges the switching node of the MOSFETs and commutates into the body diode of the second MOSFET of the half-bridge. If the MOSFET is now switched on, its drain-source voltage is nearly 0 V. This means that there are no significant switch-on losses, which reduces the overall loss balance.
If, however, the DAB is operated at very low power using the standard phase shift control, it is no longer possible to switch with ZVS (Fig. 5). Then, when the MOSFET is switched off, the current commutates into its own body diode and the voltage of the switching MOSFET corresponds to the input or output voltage. This means that all the energy stored in the output capacitors (Coss) of the MOSFETs must be converted into heat. In addition, the current flowing in the body diode to be disabled at the time of switching must commutate into the transistor to be switched on within the switching time. During this so-called reverse recovery of the body diode, a cross current flows over the half-bridge, which is completely converted into heat. In order to avoid extreme losses and dynamic overvoltages, the body diodes of the MOSFETs must, therefore, have a very short reverse recovery time (trr).
The body diodes of silicon-based trench MOSFETs have too long blocking delay times in the high-voltage range (>200 V), so that Si MOSFETs can only be used in the low-voltage range. In the high-voltage range, on the other hand, switches made of so-called wide-bandgap materials such as silicon carbide (SiC) or gallium nitride (GaN) must be used. The bandgap between the valence band and the conduction band here is about 3 times larger than with silicon. Due to this nature, the service life of the minority charge carriers is considerably shorter here. This means that the charge zone is cleared much faster than with a silicon body diode.
In addition to the significantly better switching behavior of the body diode, wide-bandgap semiconductors also have other advantages. Due to the larger bandgap, the structures can be reduced in size with identical reverse voltages and forward currents, thereby reducing the parasitic capacitances (Cgs, Cgd, Coss). This also reduces switching losses. In addition, the maximum operating temperature is higher, which simplifies heat dissipation from the components. Depending on the voltage range, both SiC and GaN may be the better choice.
The manufacturing process of silicon carbide semiconductors is similar to that of silicon semiconductors. Silicon carbide can be produced in a large number of polymorphic crystalline structures (polytypes) by a homo-epitaxial process and can thus be used as an independent substrate. In semiconductor physics, epitaxy means the application of a crystalline structure to a so-called substrate. If the substrate consists of the same material as the substance to be deposited, this is called homo-epitaxy, otherwise hetero-epitaxy. Gallium nitride, on the other hand, does not yet have an efficient process for producing high-quality single crystals. So far, it has mostly been applied hetero-epitaxially to foreign substrates such as silicon carbide or silicon. However, the lattice mismatch resulting from the different crystal structures of these substrates and the GaN applied leads to undesired impurities, which has a negative effect on the quality of the components.
For this reason, GaN-MOSFETs are usually built laterally, whereas SiC-MOSFETs and the silicon variant are built vertically. The vertical design is an advantage for high reverse voltages, since the electric field does not have to be guided along the surface. The maximum reverse voltage of the lateral GaN-MOSFETs is therefore limited to <1000 V. The maximum reverse voltage of the lateral GaN-MOSFETs is therefore limited to <1000 V. The maximum reverse voltage of the lateral GaN-MOSFETs is therefore limited to <1000 V. However, GaN-MOSFETs can be manufactured on existing production lines for silicon semiconductors, which enables more cost-effective production.
For reverse voltages up to 900 V, GaN-MOSFETs can be seen as a good alternative to SiC semiconductors due to their price advantage and somewhat better switching characteristics. For reverse voltages >900 V SiC has better characteristics.
The use of Wide-Band-Gap semiconductors has many technical advantages, but the available MOSFETs are much more expensive than the silicon version. Therefore, it is obvious to consider whether by exploiting the degrees of freedom when driving the converter, ZVS can be achieved permanently and thus the use of wide-bandgap semiconductors can be dispensed with. In order to enable permanent ZVS, the drain-source current must always have a positive value when the MOSFET is switched off. This must also have a certain minimum value so that the energy stored in the choke is still sufficient for recharging the output capacitances of the MOSFETs.
To meet this criterion, in addition to changing the phase angle φ between the primary and secondary bridges, it is also necessary to change the phase γbetween the two primary or secondary half-bridges. Figure 5 shows the operation of the DAB with an unfavorable input/output voltage ratio in the lower partial load range with phase-shift control. If one compares the switching currents Isw1 and Isw2 with the switching currents in Fig. 2, it becomes apparent that Isw2 now has a negative sign. This means that switching is no longer performed with ZVS at this point. In the extended control shown in Fig. 6, the phase γ is now also adapted within the secondary full bridge.
Through an optimized control of the phase angles φand γFinepower has succeeded in maintaining ZVS over the entire working range, even with a wide input and output voltage range, and is thus able to build a cost-efficient circuit based on existing Si-MOSFETs. Since semiconductor switches represent the largest cost factor in DAB alongside inductors, the conversion from full bridge to half-bridge is another possibility for cost savings.
This article was first published in German by Elektronikpraxis.