Abstract     Abstract: This paper mainly introduces the current sharing control principle of high frequency electric power supply, and presents several current sharing control schemes, describing their respective advantages and disadvantages. Finally, the UC3902 is used to realize the current sharing control of the system. Keywords : Current sharing control ; UC3902 ;   1 Introduction   Power supply systems play a crucial role in modern power systems, primarily used in power plants and substations as vital power sources for DC mechanisms, relay protection, signals, automatic control, emergency lighting, instrumentation, and emergency loads. Their performance and quality directly impact the safe operation of the power grid and the safety of equipment. This paper analyzes commonly used current sharing methods and models two automatic master-slave current sharing methods using small-signal models, ultimately selecting the automatic master-slave current sharing method. This method enables current sharing among modular power supplies, improves the response speed of the current sharing loop, and achieves redundancy in the parallel modular power supply system. Dynamic current sharing is implemented using the UC3902 integrated current sharing control chip, and a control circuit based on automatic master-slave current sharing is designed to achieve system current sharing control. 2. Basic Principles of Current Sharing Control To ensure that each power supply equally shares the load current, it is necessary to first analyze the mechanism of uneven current distribution and then take corresponding countermeasures. The situation after multiple power supplies are connected in parallel is shown in Figure 1. All power supply outputs are connected together, so their output voltages are the same. However, the setpoint and feedback proportional coefficient of each power supply are slightly different. Considering that the offset voltage of the operational amplifiers are also different, the error signals of each power supply are also different (the subscript k indicates the k-th power supply). The voltage regulators of each power supply typically use proportional-integral (PI) regulators. When the outputs of the power supplies are first connected in parallel, some power supply error signals are positive, the voltage regulator integrates in the positive direction, the open-circuit voltage increases, and the output current increases; others are negative, the voltage regulator integrates in the reverse direction, the open-circuit voltage decreases, and the output current decreases. When the load current is less than the maximum current limit of a single power supply, the final steady-state condition is that one power supply's current is zero, the regulator works normally, while the other power supplies' currents are negative, and the regulators are in lower limit saturation. All the load current is borne by the zero-current point source, and the output current of the other power supplies is zero. Figure 1 shows the power supply system after parallel connection. When the load current is large, exceeding the maximum current limit of a single power supply, the situation becomes slightly more complex. Some power supplies have positive current, some have negative current, and at most one power supply has zero current. The output current of a positive power supply is the maximum current limit, the output current of a negative power supply is zero, and the output current of a power supply with zero current is between zero and the current limit. The sum of the output currents of all power supplies equals the load current. Based on the above analysis, the reason for the unequal output currents after power supplies are connected in parallel is that, under the condition of the same output voltage, the error signals of the voltage regulators are different, reflecting the dispersion of circuit parameters. In order to compensate for this dispersion and make the output currents of each power supply equal and all equal to zero, control measures must be taken. Therefore, the basic principle of parallel operation current sharing control is to detect the output current of the modules, determine its degree of current imbalance, and use this signal to change the given or feedback amount of the DC output voltage, so that the voltage regulator error signal is zero and the current given output of each module's voltage regulator is the same, thereby achieving the purpose of current sharing control. 3 Current Sharing Control Scheme   There are various methods to achieve current sharing, and their current sharing accuracy and principles also differ. Based on the control mechanism, they can be divided into two main categories: the descent characteristic method and the dynamic current equalization method. The dynamic current sharing method includes both structural planning and current planning. 3.1 Descent Method   The droop characteristic method, also known as the drop (tilt) method or voltage regulation method, as shown in the figure below, works by adjusting the droop of the converter's external characteristic (output impedance) to distribute current reasonably among the modules. Essentially, it utilizes the open-loop technology of the switching power supply's output impedance to achieve current output balance. When the current in a certain module increases significantly, rising and falling, the output voltage of that module decreases accordingly, meaning its external characteristic decreases, approaching the external characteristics of other modules. This causes the current in other modules to increase, achieving approximate current sharing, but the voltage regulation rate of the modules themselves deteriorates. The disadvantages of this current sharing method are obvious; it is essentially an open-loop control. Because typical converters are designed as low-output-impedance voltage sources, achieving current sharing through the droop characteristic weakens the load regulation capability. The distribution performance is better under heavy loads, and each module requires individual adjustment, making it difficult to achieve current sharing for modules with different rated power. Figure 2: Principle diagram of current sharing control using the droop characteristic method. 3.2 Dynamic Current Sharing Method   The dynamic current balancing method consists of two parts: control structure and current planning strategy. There are three basic control structures: inner loop regulation structure, outer loop regulation structure, and external controller structure. (1) In the inner loop regulation structure, each module uses a common reference voltage, feedback voltage, and regulator. The advantage of this structure is that the current balance is stable and the output voltage regulation is accurate. The disadvantage is that it reduces the modularity and fault tolerance of the system. For switching converters with an inner current loop, it is very convenient to implement this structure. Only one voltage outer loop of the converter is needed to control the current inner loop of all converters. (2) In the outer loop regulation structure, each converter uses the current planning error to adjust the voltage outer loop setpoint to achieve current balance. Its main feature is that each converter module has independent output voltage feedback. The advantage of this structure is that it has modularity and standardization suitable for industrial production, the system is easy to expand and maintain, and has good fault tolerance for the failure of a single module. The disadvantage is that the transient process may be unstable and the voltage feedback gain is limited. (3) The external controller structure uses an external controller to control all converters to achieve current balance. It compares the load current signals of each module and adjusts the feedback signals of each module to balance the current. This system achieves excellent performance but requires an external controller and interconnection between the controllers and the power supply. However, having a single controller control all modules reduces system reliability, and extensive interconnections can lead to system insecurity. However, with the development of modern distributed power system technology, this technology has shown many advantages and is increasingly used. It employs centralized control, which facilitates dynamic pulse interleaving to reduce electromagnetic interference, facilitates error monitoring, and allows for the utilization of existing monitoring systems, achieving excellent current balancing and output voltage regulation performance. Many current sharing methods can be derived from current planning. Current planning strategy is a crucial part of dynamic current balancing strategy. Its role is to coordinate the voltage regulators of each parallel converter, obtain the current balancing error of each module, and feed it back to the voltage regulator through an error amplifier. Currently, there are two main types: average current planning strategy and master-slave current planning strategy. 1) Average Current Planning Strategy The average current planning strategy obtains the reference current value of each module by weighting the sum of the currents of each module. This reference current value is compared with the feedback current to obtain the current error, which is then fed back to the voltage regulator of each module through an error amplifier. Each module adjusts its own output voltage to maintain its proportion of the total output current. Currently, the automatic current sharing average current method is widely used among relatively mature technologies. The internal current sharing principle of each module is shown in Figure 3. The voltage drop across resistor R represents the current sharing error voltage of that module. By adjusting the output through this error signal, the equivalent DC potential at the output terminal of the module is reduced, making the output current closer to the calculated average current, thus achieving the current sharing effect. Its disadvantages are obvious. Any module failure or short circuit in the current sharing bus will cause the system to be unstable, and when the output of a certain module reaches its limit, the output voltage of other modules will drop to its limit. Figure 3 Current sharing control principle of average current control method 2) Master-slave current sharing strategy The master-slave current planning strategy uses the current signal of the selected master module as the reference current value of all modules, compares it with the feedback current to obtain the current error, and adjusts the output voltage of other master and slave modules to maintain their proportion in the total output current. The disadvantages are that the failure of the master module will paralyze the entire system, there is no redundancy performance; the high-frequency band loops of each module are connected together, which brings a lot of noise interference to the system. This type of strategy mainly includes: designated master module method, rotating master module method, and automatic master module method. Figure 4 illustrates the simplified principle of autonomous current sharing. By using a unidirectional buffer circuit to control the bus voltage of the module with the largest output current, automatic selection of the master module and automatic isolation of faulty modules can be achieved, enabling system redundancy and hot-swapping. Open or short circuits on the bus do not affect the independent operation of each module, improving system reliability. Figure 4: Simplified diagram of low-difference automatic master-slave current sharing control principle. Based on the above analysis, the current sharing accuracy of the descent method is relatively low, and the average current method cannot achieve redundancy, thus the reliability of the parallel power supply system cannot be guaranteed. Adding an external controller makes the system quite complex, hindering technology transfer. Combining three-loop control with current programming in dynamic current sharing can yield many control methods with higher current sharing accuracy. In demanding systems, dynamic current sharing is increasingly widely used, especially the master-slave control method. 4. Circuit Design   Currently, the UC3902 and UC3907 current sharing control chips produced by Unitrode are ideal. This paper uses the UC3902 to design a power supply current sharing control circuit. The UC3902 load current sharing integrated controller is a monolithic current sharing control integrated circuit that can accurately distribute the total system current for multiple independent power supplies connected in parallel. The UC3902 itself is designed based on the principle of autonomous current sharing, so it integrates a current sharing compensation circuit. It only requires a small number of passive components. Its internal structure block diagram and external parameter connections are shown in Figure 5. Before setting the parameters, the rated output voltage, maximum output current, and maximum allowable voltage adjustment range must be understood. The selection of the bus voltage should comprehensively consider the sensitivity to noise, current sharing, and the number of parallel modules. Because the actual current sharing bus is only driven by the master module, each slave module represents a 10kΩ resistance on the bus, meaning that each additional module will increase the supply current of the master module by 100uA. To improve current sharing accuracy, a high-precision resistor should be used for the sensing resistor. (The current sensing amplifier's highest output voltage is 10V, which is the smaller of the two). 6V is selected. The current sensing amplifier gain is 40, and the module's maximum output current is 40A. Therefore: RSENSE = 6/(40×40) = 0.00357Ω 1) Figure 5 shows the internal structure block diagram of UC3902, which adjusts the amplifier's maximum current. Its value should be kept within a certain range, as a slightly lower value may cause the system to become sensitive to noise. Its actual current is determined by the ADJR pin voltage (2.6V) and the resistor connected to it. 2) The actual value should be selected. The selection of Ω should be minimized, with a typical value between 20-120, to ensure that the normal voltage feedback of the converter is not affected. 3) Because the current sharing loop is embedded in the existing voltage loop, interference between control loops should be avoided to ensure the stability of the voltage loop. Therefore, the crossover frequency of the current sharing loop should be at least 1/10 of the voltage loop's crossover frequency to minimize the impact of the current loop on system stability. The design of the current sharing compensation capacitor is to obtain the desired crossover frequency of the current sharing loop. 4) Where—transconductance of the current sharing loop error amplifier, typically 4.5ms;—current sharing amplification factor, a fixed value of 40 in UC3902;—gain of the voltage loop at the crossover frequency. Based on comprehensive analysis, the current sharing loop crossover frequency is set to 600 rad/sec, at which point the voltage loop gain is 40dB, calculated to be 3.06, but actually taken as 5 (two 10dB capacitors in parallel). 5) Therefore, the actual value is 51. 5 Summary   This paper analyzes a modular power supply parallel system to derive the basic principle of current sharing. This principle ensures current sharing by assuming the voltage regulator error signal is zero and that the output current of each module's voltage regulator is identical. Considering the large output current and the high voltage and high current load on components in this project, the UC3902 chip was used to achieve autonomous current sharing. This resulted in a stable and reliable current sharing system and signal acquisition and feedback circuitry, solving the reliability problem of the parallel system and ensuring the stable operation of the experimental setup. References   1. Lu Qiusheng, Zhang Yanjie. Current sharing technology in parallel power supply. Communication Power Supply Technology. 2000, (6): 12~14. 2. Xie Qinlan, Chen Hong. Current sharing technology in parallel switching power supply system. Marine Electronic Engineering. 2003, (4): 75~78.