Occasionally, the need arises for a variable shunt regulator capable of high-voltage operation. At high voltages, multiple transistors in series overcome safe-operating area (SOA) limitations much more effectively than parallel configurations. However, with linear operation, ensuring equal power dissipation between the devices at all times can be a problem. If one device has slightly higher gain, it will minimize the voltage across itself. This leaves the other transistors with more than their share of the load, potentially exceeding their SOA boundaries.
The circuit shown is optimized for high-voltage adjustable shunt regulation and includes a self-balancing feature to help equalize the voltage drops across the devices used (see the figure). Furthermore, the circuit’s voltage range may be extended by cascading additional stages if required.
Each stage of the regulator circuit consists of a high-voltage optocoupler (U1), driver transistor (Q1), and pass transistor (Q2) connected in an extended Darlington configuration. The control voltage (Vctrl) is applied to U1 via resistor R1 and adjusts the current drawn by Q2. Resistors R2 and R3 add stability by swamping the leakage current. Resistors R4, R10, and R16 force the voltage drops to be equal across all stages when control voltage Vctrl is zero and the power devices aren’t conducting.
Identical resistors R5, R11, and R17 are series-connected across Vin. Therefore, the voltage at the junction of R5 and R11—and the voltage at Q3’s base—is 2/3 of Vin. If the circuit is balanced properly, the voltage at the emitter of Q2 also will be 2./3 of Vin. The emitter voltage of Q3 is the same as the base voltage of Q2, which will be nominally (2/3 Vin) + 0.7 V. Hence, Q3 is just barely biased on when the circuit is balanced properly. If Q2’s emitter voltage rises relative to the reference, Q3’s base-emitter voltage also rises. The transistor shunts a larger portion of Q1’s base drive, restoring the circuit’s balance. If Q2’s emitter voltage drops relative to the reference voltage, the net effect is to reduce the voltage across all of the other stages. This increases the bias on their balancing devices and again maintains equal collector-emitter drops.
On all stages except for the last one, the balancing circuit monitors the differential between the emitter of the power device (plus 0.7 V as explained before) and the reference voltage from the divider chain. This isn’t possible with the last stage, since its emitter is connected directly to the low side of Vin. An isolated device for sensing the differential is therefore required. To balance the final stage, optocoupler U4 is used instead of a transistor. U4 monitors the differential between the collector of Q8 and a reference voltage of 1/3 Vin. If the collector voltage of Q8 falls below the reference voltage, U4 begins coming on. This performs an identical function to Q3, shunting base current away from driver transistor Q7 and restoring proper balance.
A four-stage version of this circuit was bench-tested using a 160-V dc input at varying currents up to approximately 2.0 A. The voltage drops across the power devices remained balanced to within 5% of the desired 40 V dc at all times. Obviously, at these power levels, a very generous heat sink (on the order of 1°C/W per device) is required.