Redundant power is a critical component in high availability systems. In the simplest solution, two power supplies can drive a load through diodes to OR their outputs together. In this way, the power supplies can either share the load or have one active and the other in standby. A field effect transistor (FET) ORing controller is a more practical solution because it avoids the diode voltage drop, power loss and heat dissipation. More creative and economical systems can be configured with the low voltage loss MOSFET. Examples of redundant power supply configurations for server applications are discussed.
Redundant power techniques for servers
High availability system power supplies are bused in OR, N+1 configurations, or a combination of the two. Usually, diodes are not used in low-voltage, high-current applications because of the forward voltage drop and resultant heat dissipation. FET ORing is the preferred technology. However, integrated and discrete design MOSFET controllers have many of their own problems.
A controller monitors differential voltage across the MOSFET, VAC in Figure 1. Based on VAC, the controller sets the MOSFET's gate voltage. The actual switch point voltages where the MOSFET turns on and off, and the method and speed of control, define the performance and stability of the controller to successfully emulate a diode.
The TPS2410 controller is designed specifically for server applications. Server loading is a low-voltage, relatively constant, high current. Reverse current into a failed power supply is unacceptable. Some examples of power supply configuration for redundancy follow. Example figures use the boxed diode symbol shown in Figure 1 as a shorthand notation for the N-channel MOSFET and Controller.
Figure 1. The "Boxed Diode" is a shorthand notation for the controller and MOSFET