System-Level Design

Once you have the power supply built and tested as a standalone circuit, you need to consider how it’s going to interact with the circuit it’s meant to power. (I call this the “load circuit” below.) The characteristics of that circuit affect the power supply, so you have to consider the two as a complete system.

If the load circuit has enough reservoir capacitance, it can still be running when the regulator shuts down. If that happens, the load circuit can reverse-bias the protection diodes around the regulator. If done repeatedly with a big enough current dump, this can kill the protection diodes, and then on the next power cycle, the regulator, too. If that’s happening to you, see the next section.

If instead the load circuit shuts down first, that pretty much ensures that the regulator won’t die during shutdown. But, that can happen for two very different reasons, which affects whether there can be a problem during startup instead.

The first possibility is that the load is relatively light, so that the power supply’s big power reservoir keeps the regulator running until after the load circuit shuts down. Such a light load will generally not cause problems on startup, either, so all is fine.

If on the other hand the load circuit has a high current drain but also a big power reservoir, it’ll behave similarly to the light-load case during shutdown. But on startup, the regulator will see the load as nearly a dead short circuit until the load reservoir charges sufficiently. If the regulator is already running when the load comes up, the full power supply voltage across that near short circuit will cause a higher charging current than if the regulator is coming up along with the load. If the regulator’s protection circuitry is inadequate, the regulator may die in the higher load situation but not when both circuits come up together.

If you’re using the power supply to run a DIY amplifier, you have a great resource available: search the forums to find reports from people with similar configurations to yours. This reduces the probability stuff above to rubble — the collective wisdom of the forums trumps generalities.

Switch Between the Load and Power Supply

If you always power the load up and down by flipping a switch between the two circuits, you eliminate the possibilities that the regulator doesn’t start up correctly, and that the load is killing the regulator during shutdown. It simply becomes a question of whether the startup load kills the regulator or not. If you don’t know in advance which way things will play out and you just want to pick one method and try it, I think this gives you better odds of success. If this switching setup works for you, the only downside is that there’s a constant power wastage whether the system is running or not, that being the regulator’s idling current. It’s only 10-20 mA, but some people care about such small losses.

If for some reason you have a problem in this configuration, moving the switch to the AC side to fix it is a bad bet if you don’t know why you’re having the problem. It may be that moving the switch is exactly the right fix, but if you’re just trying things without understanding exactly what the problem is, this would just be a desperation move, its only virtue being that it’s easy to try. The fixes in the next section are going to have a higher chance of success.

No Switches

If the power supply is cased separately from the load circuit, no switches are actually necessary. You can turn the whole system on and off by plugging and unplugging the AC cord to the power supply. If you want the power supply to come up before the load circuit, you simply don’t plug the power supply into the load circuit until you’ve plugged the power supply into the AC supply. The connectors are the switches.

Two Switches

Now, if the two circuits are in a single case and you want the power supply to come up first, it can still be useful to have a switch on the AC side as well. The AC side switch would probably be on the back of the enclosure, and normally be left on; it would only exist for those times when you want the power supply to be completely shut down, such as when going on vacation. The switch between the two circuits would probably be on the front side of the enclosure, for convenience, as it’s the one normally used to turn the load circuit on and off.

Most of the time, the regulator dies due to an incorrect switch arrangement. If you’ve already tried changing that, here are some things you can try to keep your regulator alive and happy.

Reduce the Load Capacitance

The single biggest mistake people make when building amplifiers is using too much rail capacitance. The thought process goes that if low rail capacitance is bad, then lots and lots must be good. It is, up to a point, but you can go overboard. There are two problems stemming from having too big a rail cap bank.

The first problem is, the bigger the cap bank, the more energy it takes to charge it. The energy required is also related to the charging voltage, so this problem is more severe if you have the switch between the power supply and the load. (But see the next paragraph before you decide that you shouldn’t put the switch there!) For an excellent description of the processes involved, see this page. If you work through the math on that page, you find that there is no way you can charge big rail capacitors as fast as they can theoretically be charged. Some current limiter somewhere is going to kick in early on, so the charge time will end up taking something on the order of a second. That’s plenty of time for disaster to happen down in the microelectronic world.

At turn-off, we have a different problem: now we have to dump all that energy in the rail cap banks in the load circuit and the power supply. A good part of it will go into driving the load circuit as it shuts down, but if the power supply’s reservoir cap falls in voltage faster than that of the load circuit, current will start flowing back from the load into the power supply once the power supply’s voltage drops far enough below the load. Enough current flowing backwards through the regulator will kill it, which is why the protection diode exists: it starts conducting before the regulator starts running backwards, saving the regulator. But, if the rail cap bank is big enough, the discharge current through this diode will kill it eventually. With the protection diode blown, the regulator will probably die the next time the system is powered down. You avoid this problem entirely if you put the power switch between the power supply and its load circuit, because the power supply can shut down separately from the load circuit. You still need D2 to protect against C8 running the regulator backwards during shutdown, but it’s a much smaller cap.

Bigger rail cap banks exacerbate all of these problems. If the documentation for the circuit being powered puts some limit on the recommended amount of capacitance, heed it. Using the maximum recommended capacitance vs. only a fair fraction of it usually gives only a small improvement.

Change the Regulator Type

If the regulator is dying and the switch is on the AC side of the power supply, you can try changing to a higher current version of the same regulator you’re using now. For example, use the LM338 instead of the LM317. The higher output current ability means it will have less trouble charging the load’s rail capacitance than the smaller regulator. It may also actually perform better, too, because higher output current implies lower output impedance, always a good thing in a power supply.

If the regulator doesn’t actually die, but rather fails to start correctly when you power both circuits up together, it’s probably that you’re using an LDO and the load capacitance is too high. The fix is simple: switch to a standard regulator. You may be able to fix this by moving the power switch to the DC side of the power supply, but because this goes against the datasheet’s recommendations, I can’t recommend it. Use the right part for the job.

Add a Soft-Start Circuit

If the switch is on the AC side of the power supply and the regulator is dying due to having too high a load on startup, you can add some kind of soft-start circuit to fix this. There are a number of choices.

The simplest is to put an NTC thermistor between the supply and the load. An NTC thermistor has a relatively high resistance at room temperature, so at turn-on, it limits the current charging the load’s rail capacitors. As the thermistor heats up, its resistance drops, so the power supply becomes relatively low impedance. A typical thermistor made for this purpose starts at 33 Ω at room temperature, and drops to about 0.5 Ω at 0.3 A of load current. (This is just one example: you’ll have to run the numbers for the thermistor you actually use.) The downside of this is that even half an ohm is pretty high as output impedances for power supplies go, but if it keeps the reguator working, it may be a price you’re willing to pay.

There are many more complicated circuits of this type. Google for “soft start,” “inrush current limit” and other such terms.

Use a Bigger Protection Diode

If the regulator dies, the first thing to test is whether the protection diode across it (D2) has also died. If it has, you know that it died first, and then the regulator was killed because the diode wasn’t there to protect it. If that’s happening and none of the above solutions works or is palatable, you can simply try using a bigger protection diode. You can find diodes capable of handling up to 10 A that can be crammed into this position. It’s a bit brute-force, but it can work.