Use op amps to make automatic-ORing power selector

Replace diodes with FETs to reduce power loss.

Bob Zwicker, Analog Devices, Fort Collins, CO; Edited by Martin Rowe and Fran Granville -- EDN, August 11, 2011

Many systems must select among two or more low-voltage dc-input sources, such as an ac adapter, a USB (Universal Serial Bus) port, or an onboard battery, for example. You can implement this selection using manual switches, but automatic switching is preferable. You usually want to use the highest-available input voltage to power your system. You can accomplish this task using a Schottky-diode ORing scheme (Figure 1). Unfortunately, the forward-voltage drop of a Schottky diode ranges from 300 to 600 mV. This voltage wastes power, creates heat, and decreases the voltage available to your system.


Positive current flows from the MOSFET drain in an N-channel-FET design. In a P-channel design, the current flows from the MOSFET source. The MOSFET’s drain-body diode would defeat rectifier operation if the usual current flow (for switching or amplification) were used.


Your second design task is to choose an op amp. The op amp must be able to operate with the voltages involved and to adequately drive the MOSFET’s gate voltage. The P-channel design requires a rail-to-rail I/O type. A single-supply op amp is adequate for the N-channel design. Another important consideration is the op amp’s input offset voltage, V_OS. The total ±V_OS window must be less than the maximum desired voltage drop across the MOSFET. For example, if you permit a 10-mV forward-voltage drop at full load, then the op amp should specify an offset voltage of ±5 mV or better.

Read More Design Ideas

R₁/R₂, R₁₁/R₁₂, and R₂₁/R₂₂ form the input-voltage divider, which biases the op-amp input at a level slightly below that of the input voltage that it is controlling (figures 2 and 3). This offset must exceed the op amp’s maximum offset voltage to ensure that all op-amp parts in production always turn off the MOSFET when you apply reverse voltage. In the example of the P-channel 5V design, R₁ and R₂ bias the inverting op-amp input at 99.9% of the input voltage, or 4.995V dc. In steady-state operation, the op amp servos with the conducting MOSFET to keep the other op amp’s input at the same voltage, within the tolerance of the op amp’s offset voltage. With a perfect 0V-offset op amp, light-load currents cause the MOSFET to only partially enhance, so the circuit delivers a 5-mV MOSFET forward-rectifier drop. This mild effect is the only disadvantage of R₁ and R₂’s input offset biasing. If the MOSFET resistance is too high to allow it to maintain 5 mV at full load, then the op amp fully enhances the MOSFET as its output swings to the rail, and the ORing circuit delivers the MOSFET’s fully enhanced on-resistance.


You can consider the MOSFET’s variable on-resistance as the element with which the op amp senses current. When you apply reverse voltage, the MOSFET de-enhances, the I×R voltage drop increases, and the op amp’s output ends up at the appropriate supply rail, driving off the MOSFET as hard as it can.

With light-load conditions and a given offset voltage, the op amp tries to servo the voltage on its power-output-sensing input to the voltage on its power-input-sensing input plus the offset voltage. With R₂ open-circuited, the op amp has no intentional external offset. If the op amp’s offset voltage were in the unfavorable direction, a sizable reverse-cutoff current would occur if the input-power bus were to fall to a lower potential than the output-voltage bus.

Figure 4 shows current-voltage test data for the operating region. The complete design, including intentional offset, produces the green curve. The equivalent of an unfavorable internal offset and no intentional external offset produces the blue curve. Although the green curve sacrifices some forward-voltage drop at light-load conditions, its forward voltage is always less than the full-load maximum. The intentional offset avoids any significant reverse current in the MOSFET. This design can switch at the 0A current transition, at which the leakage-current MOSFET’s drain-body diode is likely to dominate.

On the other hand, the blue curve, without intentional offset, permits significant reverse current under some circumstances. This example shows approximately 100-mA reverse current with 2-mV reverse voltage across the MOSFET before the circuit switches off the MOSFET. Both the P- and the N-channel designs have undergone testing, and the P-channel design is in production.