Tips for selecting power electronics in solar inverters

Solar photovoltaic systems are finding more and more applications. Mobile systems in particular benefit from solar energy without spending a penny. At the same time, as the cost of conventional electricity continues to rise, solar energy is very attractive for home applications. The energy efficiency of the solar cells themselves and the solar inverters that connect the solar cells to the public grid or distributed power supply are the key to the success of this technology. Today, advanced solar inverters with a maximum output power of 5kW have a two-stage topology. Figure 1 shows the multi-group configuration of such solar inverters.

Each group is connected to its own power conditioner and then to a common DC bus. The power conditioner enables the solar cells to operate at maximum efficiency. The solar inverter generates an AC voltage that is fed into the mains. Note that the power grid shown in Figure 1 is a fictitious circuit that can be used for any inverter topology, with the addition of a mains transformer and an output filter. The transformer prevents the DC component from entering the mains.

However, there are some systems that do not use a transformer, depending on the legal background of the country where the solar inverter is sold. The purpose of countries that allow the use of no transformer is to improve system efficiency, because the transformer causes a 1-2 percentage point drop in efficiency. On the other hand, the inverter must avoid DC components, requiring currents less than 5mA. Although this is difficult to do, we have successfully achieved it in order to obtain higher efficiency. Table 1 shows the contribution of each level to system losses, system size and system cost.

It is easy to see that the transformer is the main contributor to system losses and cost. However, the transformer is mandatory in many countries and is therefore not considered for loss reduction. The output filter attenuates the current ripple generated by the output inverter stage, and the size and cost of the filter are inversely proportional to the inverter switching frequency. The higher the switching frequency, the smaller and cheaper the filter. However, this relationship is a trade-off with the relationship between switching frequency and switching losses in hard switching conditions - the higher the switching frequency, the greater the losses and therefore the lower the efficiency. Switching frequencies from 16kHz to 20kHz can meet the requirements of solar inverters due to lower audio noise and higher efficiency. Therefore, the power circuit needs further research.

The following article compares the advantages of several semiconductor technologies applicable to these two levels.

Power Semiconductors for DC/AC Boost Converters

The DC/DC converter is operated at a switching frequency of 100kHz or more. The converter operates in continuous mode, which means that the current in the boost inductor produces a continuous waveform under rated conditions. When the transistor is turned off, the transistor can charge the inductor when the diode acts as a freewheeling diode. This means that when the transistor is turned on again, the diode can be actively turned off. The following figure shows the typical reverse recovery characteristics of a commonly used silicon diode (black and red curves in Figure 2).

The reverse recovery characteristics of silicon diodes will cause high losses in both the boost transistor and the corresponding diode. However, silicon carbide diodes do not have this problem (as shown by the blue curve in Figure 2). Only due to the capacitive nature, a diode instantaneous negative current is generated, which is caused by the junction capacitance charge of the diode. Silicon carbide diodes can greatly reduce the turn-on loss of transistors and the turn-off loss of diodes, and can also reduce electromagnetic interference because the waveform is very smooth and there is no oscillation.

Many processes have been reported in the past to avoid the loss caused by the reverse recovery characteristics of the diode, such as zero-current switching of zero-voltage switching, etc. All of these will greatly increase the number of components and the complexity of the system, and often result in reduced stability. It is particularly worth mentioning that by using silicon carbide Schottky diodes even in the hard switching state, the same efficiency as soft switching can be achieved with minimal components.

High switching frequency also requires high-performance boost transistors. The introduction of super junction transistors (such as CoolMOS) brings hope for further reducing the on-resistance RDS(on) per unit area of ​​MOSFET, as shown in Figure 3.

It is easy to see that compared with the standard process, the unit area RDS (on) is about 4 to 5 times lower than CoolMOS. This means that in a standard package, CoolMOS can achieve the lowest absolute on-resistance value. This will result in the lowest conduction loss and the highest efficiency. The unit area RDS (on) of the CoolMOS process shows better linearity. When the voltage is 600V, the advantage of CoolMOS is obvious, and if the voltage is higher, its advantage will increase. Currently, the highest voltage level is 800V.

Several studies have shown that using silicon carbide diodes and superjunction MOSFETs such as CoolMOS is superior to solutions using standard MOSFET and diode processes (as shown in Figure 4).

Power semiconductors for inverters

The output inverter connects the DC bus to the grid. Usually, the switching frequency is not as high as that of the DC/DC converter. The output converter must handle the sum of the currents produced by all the group converters. The insulated gate bipolar transistor (IGBT) is an ideal device for use in this inverter. Figure 5 shows two cross sections of the IGBT process.

Both processes use wafer thinning to reduce conduction losses and switching losses caused by excessive substrate thickness. Standard and TrenchStop processes are non-epitaxial IGBT processes that do not use transistor growth processes, as such processes are very expensive because the blocking voltage is determined by the thickness of the grown crystal.

In the off state, the standard NPT cell forms a triangular electric field inside the semiconductor. All blocking voltage is blocked by the n-region of the substrate.

The absorber (depending on its thickness) allows the electric field to drop to 0 before entering the collector region. The thickness of the 600V chip is 120mm and the thickness of the 1200V chip is 170mm. The saturation voltage has a positive temperature coefficient, which simplifies parallel use.

The TrenchStop process is a combination of advanced trench gate and field stop concepts that can further reduce conduction losses. The Trench gate process provides a higher channel width, which reduces the channel resistance. The ndoped field stop layer performs only one task: suppressing the electric field with a very low off-state voltage value. This creates the conditions for designing the electric field to be almost horizontal in the n-substrate layer. This means that the resistance of the material is very low, and therefore the voltage drop during the conduction process is very low. The advantages of the electric field stop layer can be brought into play by further reducing the thickness of the chip, thereby achieving all the advantages mentioned above. Parallel connection is also possible using the TrenchStop process.

Table 2 shows a comparison of IGBTs with blocking voltages of 600V and 1200V. The power rating of the transistors used was kept constant for all three processes. This means that the current of a device with a voltage of 600V is twice that of a device with a voltage of 1200V. In other words, a 50A/600V device is equivalent to two 25A/1200V devices.

As can be seen from the table above, the 600V TrenchStop process can reduce switching and conduction losses by 50% compared to 1200V devices. Therefore, it is important for the entire system to use the excellent performance of the 600V process as much as possible. The 1200V TrenchStop process is further optimized for low conduction losses. Therefore, whether the Fast process or the TrenchStop product family has better performance depends on the switching frequency.

IGBTs also usually require a freewheeling diode to enable freewheeling, which is a special optimized version of the EmCon process. It is optimized for the 15kHz switching frequency of the 600V series devices. In the past, it was believed that the freewheeling diode must have a very low forward voltage to achieve the lowest total losses. Other optimizations can be made according to the application requirements to achieve lower total losses in the diode and IGBT. This means that in applications with IGBTs and diodes at frequencies of about 16kHz, a higher forward voltage drop is more appropriate to achieve low switching losses.

This is illustrated in Figure 6 (600V series). The left bar shows the losses of a TrenchStop IGBT and an EmCon diode in the EmCon3 process. The right bar shows the losses of a TrenchStop IGBT and a diode optimized for low conduction losses (called the Emcon2 process). The same diode in the right bar is used with an IGBT in Infineon's Fast process (600V). The yellow and orange parts of the bar graph represent the conduction and switching losses of the IGBT, respectively. The dark blue and light blue parts represent the conduction and switching losses of the diode, respectively.

It is easy to see that at a switching frequency of 16kHz, a load angle cosine of 0.7 and rated current, the Emcon3 diode produces higher losses during turn-on (dark blue), but gives better switching performance. So the diode itself is already a good choice in this regard. In addition, it also reduces the switching losses of the IGBT during turn-on. The considerations in Part 2 above also apply here. Using the optimized EmCon diode can reduce the losses by about 1W, which is an advantage. Note that when the load angle is close to 1, the switching losses will become the dominant loss because the diode is only turned on during the output inverter dead time.

in conclusion

Power semiconductors need to have different characteristics to achieve the highest efficiency in solar inverter applications. New processes, such as silicon carbide semiconductor diodes or TrenchStop IGBTs, are helping to achieve this goal. Of course, to achieve this goal, not only the individual devices must be optimized, but also the way these devices work together. This will achieve minimum losses and maximum efficiency, which are the two most important indicators for solar inverters.