Consider it Murphy's Law: You buy two 15V supplies or one ±15V dual supply to power some op amps. Then, when the last QA inspector or first customer flips the switch, the power supply hangs up. This hang-up occurs because the positive supply yanks its output to positive. The negative supply reads this output as a reverse polarity before it has a chance to start up. So, the supply's negative regulator gives up. The output sits there just above 0V, draining positive current into its output clamp diode, and you have no negative supply diode. Some supplies have this problem, and some don't.

Let's assume that you've connected a start-up diode from the 5V power bus to the positive supply, which causes the positive supply to start up. Fig 1 presents a simple solution using an enhancement-mode n-channel MOSFET to ensure that the negative supply also starts up. In this circuit, the 5V supply through diode D₁ forces the positive supply to start as before. However, if the 15V supply is a little slow starting up, Qstays off, thereby decoupling the load until the supply's output voltage is sufficient to turn it on. At this point, the supply can't hang up because its output voltage is in the operating range and the supply has started.

Q₁ soon turns on and behaves like a short circuit (as long as you've picked a MOSFET with low enough on-resistance at 15V). Most NMOS types are quite happy with 15V on the gate and 30V on the drain. During start-up, Q₁ may have to handle full power, but only for an instant. For the rest of the time, the MOSFET is either all the way off (no current) or on (no voltage) and dissipates negligible power. A small-footprint NMOS device with no heat sink can handle the power dissipation. Naturally, if you decide to handle output short circuits, you have to plan for more power in Q₁ . If there's no 5V supply to make life easy, another NMOS in the negative leg (or PMOS in the positive leg) does the same job for the positive power supply. (DI #1639)