Redundant Power Supply

I used this type of topology when I worked for a communication company many years back. The most common mistake I found was people plugging both power supplies into the same AC power outlet. Now a single circuit breaker can trip both of the supplies. A nuanced aspect of redundant supplies that was constantly debated was the heat sinking of the diodes. Some preferred both diodes attached to the same heat sink while others preferred separate heat sinks. The point of the debate was the voltage drop across the fully conducting diode would always be less due to the junction temperature being higher. Some engineers preferred separate heat sinks to maximize the difference in junction temperatures. Some engineers preferred a common heat sink or dual diode package to minimize the difference in junction temperatures.

Our application was to have each power supply to be large enough to carry two sets of loads. If one failed, the other would pick up seamlessly. It didn't take me long to realize that if both power supplies were set to the same voltage then both could supply current. That was not the intent.

If one power supply was a slightly higher voltage, it always did all the work. So as labeled, the redundant power supply wasn't always redundant unless the back up voltage was slightly lower. Considering the application, I had nine power supplies set to the same voltage and connected to nine redundant modules. The nine redundant modules fed nine separate loads. In fact, they were connected in a closed loop.

To prevent unintended load shifting, I added one more diode to each "B" input. This caused the "B" input to be at least 0.6 Volts lower than the setting of all of the power supplies. As a result, the primary feed went to the load unless it failed for some reason.

And yes, if a breaker tripped, the game was over. Otherwise it has been very successful since I realized that the additional external diode was absolutely the way to go. This was really important since I was creating a circular redundant feed with a common return. I currently have 15 loops in use, each loop consisting of nine power supplies of 24 VDC capable of supplying 20 amps each.

I checked Wikipedia and found no entry for redundant power supply as described above. If anyone is interested, I will insert a drawing of the circuit in the form of a circular connection. I just thought someone might find this idea useful.

Your supplies should be sized so that just one supply is capable of driving the load. You should not worry about current loading since there is nothing wrong with them sharing the load. However, you'll find in practice that even when you use a dual diode pack only one supply will drive the load at a time because of the self heating of a diode causes a voltage drop across the driving diode. Remember the Shockley diode equation includes the thermal voltage of the junction.

This was the point of the debates if separate heat sinks or a common heat sink was preferred. As the diode voltage drop starts, more current goes through that diode. More self heating occurs and a smaller voltage drop happens. Eventually a thermal equilibrium happens and one supply is driving the load.

We do something very similar in our avionics system. However, a short in the load, will kill both power sources, so we use series thermal trip fuses in series with each diode. This is needed as we have many modules that use the A/B power input, and we don't want a loss of all functions with one failure. We also have a "load share" feature between supply A and supply B that adjusts the voltages slightly, as keeping the supplies at half power each will improve their long term reliability (or so we think).

ignator

The relays change contacts when both the A and B power supplies are live. If one of the supplies drops out, then the relay contact could be used to give an alarm signal for that supply. However, should both supplies drop out, then the only way an alarm signal could be processed would be to have a third supply to power the alarm system. Where would it end <rhetorical question - NNTR>?

We know where it ends. It ends after the sun becomes a red star that swallows the Earth.

Can't we use normally close contact of the relay for alarm. When supply is O.K. relay coil will be energized and normally closde contact will become open. It will indicate that supply is O.K. In case of failure of supply, relay coil will be de-energized and normally close contact will become close to give alarm. Off course relay contact must be powered by its own power supply.

Oh, good. There is agreement.