Abstract: High-power output and distributed power supply are the directions of power supply technology development, making parallel current sharing technology a research hotspot. Therefore, this paper systematically analyzes and summarizes the principles and main current sharing methods of current parallel power supply technology. Keywords: parallel current sharing control, output droop characteristic module 1 Introduction The widespread application of numerous electronic devices, especially computers, communications, and space stations, necessitates the construction of high-capacity, safe, reliable, and uninterrupted power supply systems. If a single power supply is used, the converter will inevitably handle enormous power and experience high electrical stress, making it difficult to select power devices, increase switching frequency, and improve power density. Furthermore, a single power supply failure will lead to the collapse of the entire system. Using multiple power supply modules in parallel to provide high-power output is a direction for power supply technology development. In parallel systems, each module handles a smaller power output, solving the problems encountered with single power supplies. Since the 1980s, distributed power supply has become a new research hotspot in power electronics. Compared to traditional centralized power supply, distributed power supplies utilize multiple medium- and small-power power modules connected in parallel to build modular high-power power supply systems. Spatially, each module is close to the load, resulting in high power quality. By changing the number of parallel modules, different power loads can be accommodated, offering flexible design. Each module experiences less electrical stress, and the switching frequency can reach the megahertz level, thereby increasing the system's power density. High-power output and distributed power supplies have led to the rapid development of parallel power module technology. However, direct parallel connection between module outputs is generally not permitted. Current sharing technology must be used to ensure that each module shares an equal load current. Otherwise, some parallel modules will operate under light loads, while others will operate under heavy loads or even overloads. Modules with lower output voltages will not only fail to supply power to the load but will also become the load for modules with higher output voltages, leading to uneven thermal stress distribution and easy damage. The basic requirements for a power supply system with multiple modules operating in parallel are [2]: First, the output voltage should remain stable when the input voltage or load changes; second, the output current of each module should be controlled to achieve an average distribution of load current and a good dynamic response of current sharing. To improve the reliability of the system, the parallel system should have the following characteristics: achieve redundancy. When any module fails, the remaining modules continue to provide sufficient power and the entire power supply system will not collapse; achieve hot-swap, and the power supply system is truly uninterrupted; the current sharing scheme does not require an external current sharing control unit; use a common low-bandwidth current sharing bus to connect each module unit. 2 Parallel characteristics and general principle of current sharing Figure 1 shows the equivalent circuit and its external characteristic curve when two modules are operating in parallel. If the parameters of the two modules are exactly the same, that is, V1max=V2max, R1=R2, the two external characteristic curves coincide and the load current is evenly distributed. If one of the modules has a higher voltage reference value and a lower output resistance (smaller external characteristic slope), as shown in Module 1 in Figure 1, then this module will bear most of the load current. As the load increases, Module 1 will operate in a full-load or overload current-limiting state, affecting the system reliability. [align=center] (a) Parallel equivalent circuit (b) Output external characteristic Figure 1 Schematic diagram of parallel current sharing of two modules[/align] It can be seen that in a parallel power supply system, each module distributes the load current according to the external characteristic curve. The difference in external characteristics is the root cause of the difficulty in equalizing the current. The quality of current sharing performance is measured by current sharing accuracy. Current sharing accuracy is defined as: CSerror=ΔIomax / (Io/N) where N is the number of parallel modules, Io is the load current, and ΔIomax is the difference between the maximum and minimum current. Under normal circumstances, the output resistance of each parallel module is a constant value. The imbalance of output current is mainly caused by the unequal output voltage of each module. The essence of current sharing is to adjust the output voltage of each module through the current sharing control circuit, thereby adjusting the output current to achieve the purpose of equalizing the current. A typical switching power supply is a voltage-controlled closed-loop system. The basic idea of ​​current sharing is to sample the output current signal of each module and introduce the signal into the control loop to participate in adjusting the output voltage. By selecting different current signal injection points, the system reference voltage, feedback voltage error, or feedback current error can be directly adjusted to form a variety of current sharing schemes to meet different steady-state performance and dynamic response. 3 Current sharing methods According to whether there are interconnecting lines between modules in the parallel power supply system to transmit current sharing signals, all current sharing methods can be classified into two categories: droop method and active current sharing method. In the droop method, only the output wires are connected between modules; in the active current sharing method, in addition to connecting the output wires, the current sharing bus is used to connect the modules together. 3.1 Droop method [4] The droop method (also called the slope method, output impedance method) is the simplest current sharing method. Its essence is to use the current feedback signal of the module or the direct output series resistance to change the output resistance of the module unit, so that the slope of the external characteristic tends to be consistent, and the current sharing is achieved. As shown in Figure 1(b), the current sharing accuracy of the droop method depends on the voltage reference value of each module, the average slope of the external characteristic curve, and the degree of difference in the external characteristics of each module. By selecting different current feedback signal injection points, the feedback voltage value or reference voltage of the control loop can be corrected. Figure 2(a) shows the external characteristic curve corresponding to the method of changing the voltage reference value by adjusting the reference voltage. It can be seen that the smaller the difference in the voltage reference value, the better the current sharing effect. Figure 2(b) shows the external characteristic curve corresponding to the method of changing the slope by adjusting the feedback voltage value. The steeper the external characteristic slope, the better the current sharing effect. [align=center] Figure 2 (a) Adjusting the reference voltage Figure 2 (b) Adjusting the feedback voltage[/align] The commonly used droop method current sharing control block diagram is shown in Figure 3. Vi is the output signal of the current amplifier, proportional to the module's output current Ki. Vf is the voltage feedback signal. Obviously, V- = Kv × Vo + Ki × Io. When the current of a module increases, Vi rises and Ve falls. Through feedback, the output voltage of that module decreases accordingly, meaning the external characteristic tilts downward, approaching the external characteristics of other modules. This increases the current of other modules, achieving approximate current sharing. The voltage error amplifier E/A has a large DC gain Ko. Assuming Ko→∞, Vo = Vref /Kv - IoKi /Kv = Vomax - IoKi /Kv. Changing the parameters of the voltage loop and current loop can obtain the desired external characteristics. [align=center] Figure 3. Diagram of droop method current sharing control[/align] In addition, connecting a certain resistor in series between the module output and the load is also a droop method to adjust the output resistance. The disadvantage is that the series resistor consumes additional power. A more economical method is to connect a thermistor in series, whose resistance changes with the heat energy consumed in the resistor, also achieving approximate current sharing. Furthermore, Buck, Boost, Buck-Boost converters, and series resonant converters in discontinuous current mode inherently possess a certain droop rate in their external characteristics. These converters can be directly connected in parallel to achieve natural current sharing. The characteristics of the droop method can be summarized as follows: there are no interconnecting communication lines between modules; it is essentially open-loop control, resulting in poor current sharing at low currents, which improves as the load increases; for regulated power sources, a smaller external characteristic slope is desirable, but the droop method achieves current sharing at the cost of reduced voltage regulation. This method can be applied in situations where the current sharing accuracy is greater than or equal to 10%; it is difficult to achieve current sharing for parallel modules with different rated power. 3.2 Active Current Sharing Method The active current sharing method is a major category of current sharing methods. Its characteristic is that it uses interconnecting communication lines to connect all parallel modules to provide a common current reference signal. Generally, parallel converters use current-type control, i.e., dual-loop control with an inner current loop and an outer voltage loop. The power stage and the inner current loop are taken as the basic units of the converter below. The control structure and bus connection method are designed outside the basic unit to form various active current sharing methods, such as the master-slave method, the average current method, and the maximum current method. The control structure refers to how the current sharing loop and the voltage loop are configured. Figure 4 shows three control structures for active current sharing: voltage loop external adjustment, voltage loop internal adjustment, and dual-loop adjustment. In external adjustment, the current sharing loop is superimposed from outside the voltage loop (Figure 4a). The current sharing bus bandwidth is low and it is not sensitive to noise, but the current sharing control response is relatively slow due to the limitation of the low bandwidth voltage loop. In internal adjustment, the current sharing loop is superimposed from inside the voltage loop (Figure 4b). The current sharing loop can be well integrated with the current loop. The whole structure is simple. The current sharing signal is injected from inside the loop. Its bandwidth is not limited by the voltage loop. The response speed is fast. The voltage of the current sharing bus is obtained from the voltage adjustment amplifier, but it is prone to noise. In dual-loop adjustment, the current sharing loop and the voltage loop act on the basic unit in parallel (Figure 4c). [align=center]Figure 4 Three Control Structures[/align] The current sharing bus connection method refers to how a common current reference signal is obtained from all modules, indicating the master-slave relationship between modules. Figure 5 shows three types of current sharing bus connections: autonomous configuration, average configuration, and designated configuration. In autonomous configuration (Figure 5a), each module and the bus are connected via diodes. Only the diode corresponding to the module with the maximum current can conduct, and the current sharing bus represents the maximum current signal. In average configuration (Figure 5b), each module and the bus are connected via resistors with completely identical parameters, and the current sharing bus represents the average current. In designated configuration (Figure 5c), only the manually designated module is directly connected to the current sharing bus, becoming the master module. [align=center]Figure 5 Three Current Sharing Bus Connection Methods[/align] 3.2.1 Maximum Current Method (Democratic Current Sharing Method, Automatic Current Sharing Method) Figure 6 shows the control block diagram of the maximum current method. Comparing Figure 4 and Figure 5, it can be seen that the maximum current sharing technology is composed of external adjustment and bus autonomous configuration. It does not change the internal structure of the basic module unit. It only needs to superimpose a current sharing ring outside the voltage ring, and each module is connected to a current sharing bus CSB. [align=center]Figure 6 Maximum Current Method[/align] Because of the unidirectional nature of diodes, only the module with the largest current can be connected to the current sharing bus. This module is the master module. The rest are slave modules. The difference between their respective current feedback and the voltage of the current sharing bus is compared. The error amplifier output is used to compensate the reference voltage to achieve current sharing. The characteristics are: (1) Only one unit participates in the adjustment work at a time in this current sharing method. The main module always exists and is random. It is the most commonly used method to achieve redundancy; (2) There is always a forward voltage drop in the diode, so the current sharing of the main module will have an error; (3) Current sharing is a process from the module current to the current of the main module. The identities of the main and slave modules in the system are constantly alternating, and the output current of each module has low-frequency oscillation. The current sharing control chips UC3902 and UC3907 developed by Unitrode IC are based on the idea of ​​automatic current sharing of maximum current, which simplifies the design and debugging of parallel power supply systems and has been widely used. Reference [2] points out that the current sharing error of UC3902 reaches 2% when fully loaded and about 15% when 20% load. 3.2.2 Average current method The combination of the external adjustment structure and the average configuration of the bus forms the average current sharing method. That is, the diode in Figure 6 is replaced by a resistor R. If all resistor R parameters are completely consistent, the voltage of the current sharing bus reflects the average value of the current of all modules. When Ua = Ucsb, it indicates that current sharing has been achieved. If the current distribution is uneven, a voltage appears across resistor R. This voltage outputs an error voltage through an error amplifier, thereby correcting the reference voltage to achieve current sharing. The average current method is a patented technology that can achieve precise current sharing. Its disadvantage is that when the current sharing bus is short-circuited or a module is not working, the bus voltage drops, causing each module's voltage to decrease, even reaching the lower limit, leading to a fault. The solution is to automatically disconnect the faulty module from the current sharing bus. 3.2.3 Master-Slave Current Sharing Method In a parallel power supply system, one module is designated as the master module, directly connected to the current sharing bus, while the others are slave modules, obtaining current sharing signals from the bus. Figure 7 shows the master-slave current sharing method using a voltage loop adjustment structure. The master module operates in voltage source mode, and the error voltage amplifier of the slave modules is connected as a follower, operating in current source mode. Because the system adjusts under a uniform error voltage, the module's output current is proportional to the error voltage, so regardless of changes in the load current, the current of each module is always equal. [align=center] Figure 7 Master-Slave Current Sharing Method[/align] This current sharing method has high accuracy, simple control structure, and complex inter-module wiring. The disadvantage is that once the master module fails, the entire system will be completely paralyzed, and the broadband voltage loop is prone to noise interference. The wiring between the master and slave modules should be as short as possible during use. 3.2.4 Other Current Sharing Methods Based on three control structures and three bus connection methods, other current sharing methods can be designed. Figure 8 shows the literature on the current sharing method combining dual-loop adjustment and average configuration. This control method reduces the mutual influence between the voltage loop and the current sharing loop, is flexible in design, and is a compromise solution that weighs the advantages and disadvantages of external and internal loop adjustment. In addition, the thermal stress automatic current sharing method realizes current sharing according to the temperature of each module, so that the module with high temperature reduces the output current and the module with low temperature increases the current. The external controller method is to add an external current sharing controller, compare the current signals of each module, and compensate the corresponding feedback signals accordingly to balance the current. This method requires an additional controller and has more wiring [1]. [align=center]Figure 8. Current Sharing Method with Dual-Loop Parallel Adjustment[/align] 4. Summary Due to the need for high-power loads and the development of modular power supply systems, parallel connection technology for modular power supplies is particularly important in order to achieve a completely stable and reliable redundant power supply system. However, the external characteristics of each module are inconsistent, and the load current they share is also uneven, greatly reducing the reliability of modules that bear more current. Therefore, parallel operation systems must introduce effective load distribution control strategies to ensure the uniform distribution of electrical and thermal stress among modules. This is the key to realizing high-performance modular high-power power supply systems. This article introduces the general principles of current sharing technology and discusses various current sharing technologies and their advantages and disadvantages in detail. While continuously improving current sharing accuracy and dynamic response speed, current sharing control technology will develop towards increasing the number of parallel units and the parallel connection of modules with different capacities. With the gradual digitization of control systems and the development of microprocessors, applications such as single-chip microcomputers or DSPs to complete the detection, calculation, and control of power supply systems can better employ complex control strategies to achieve current sharing redundancy, fault detection, hot-swappable maintenance, and intelligent module management. References: 1. Cai Xuansan. Current sharing technology for switching power supplies. Proceedings of the 11th Annual Conference on Power Supply; 250-253. 2. Mark Jordan, “UC3907 load share IC simplifies parallel power supply design”, Unitronde application note U-129, 1993-1994. 3. Zhang Zhansong. Principles and Design of Switching Power Supplies. Electronic Industry Press, 2001. 4. Brian T. Irving and Milan M. Jovanovic, “Analysis, design, and performance evaluation of droop current-sharing method” APEC'2000, vol.1 pp. 235-241. 5. Chang-Shiarn Lin and Chern-Lin Chen, “Single-wire current-share parallelling of current-mode-controlled DC power supply” Trans. Industrial Electronics, Vol.47, No.4 pp.780-786. August 2000.