1
\$\begingroup\$

Lately I've been really interested in the concept of fault tolerance. As it relates to power supplies, I'm curious how you'd maintain power to downstream components in the event that one or more of the components (e.g. a regulator) in one of the modules dies.

Let's assume the design as two modules or paths, taking 12V and converting it to 5V. What is the best practice for designing a redundant / fault-tolerance supply? How is it best to detect a failure (an indicator could be dramatic over voltage or under voltage on the output) and switch paths?

I'm a super visual person so any block diagrams or sketches would help me understand a lot! Thank you!

\$\endgroup\$
1
  • \$\begingroup\$ You might not need to detect most failures. Each module can independently try to keep the output at the correct voltage. If one stops supplying current the other will work harder automatically. Load will not be balanced evenly with this approach. The modules will have to be designed to work this way. \$\endgroup\$
    – user253751
    Apr 7 at 11:42

1 Answer 1

1
\$\begingroup\$

Redundancy and fault tolerance are huge fields. Some stuff can be made quite simple but it can also get quite complicated. For example preventing any "single point of failure" requires a lot of analysis of a whole system. Getting an understanding what parts of your system need redundancy (and how much) requires an FMEA and a deep understanding of the requirements, risks and the internals of each component.

So you want fault tolerance in your power supply? So let's ask a few questions:

  1. what kind of failures can happen?
  2. what effect do those failures have?
  3. how can that failure be contained?
  4. are there failures that are worse than others?
  5. Against how many simultanous failures do we have to be save?

So in an FMEA we'd go to each component in the design and ask how it could fail and what the consequences would be if it failed that way. We'll look at the power supply as a black box for now. So the power supply has 2 input lines and 2 output lines (GND is probably common but we don't know that for sure right now).

Let's start at the input: What could go wrong there?

  1. It could go open (not draw any current anymore)
  2. It could go short curcuit (on the input)
  3. It could create high frequency oscillations

There are probably more cases we might have to look at but those are the basic ones... We have to look what happens in which case and what influence this might have to a redundant supply if that is on the same power rail (which could be a single point of failure). Than we have to do the same with the output. What possible failures?

  1. Go open
  2. Short curcuit
  3. Overvoltage
  4. High frequency oscillations
  5. Undervoltage

And if you've answered all of those questions you can start looking into what you need and what solution might be right for you. Only now we might start talking about technical solutions.

With two power supplies (or 2 sources of everything that are connected) you have to consider how one influences the other in normal operation. So you have to make sure that the supplies do not work against each other. There are different ways to do this.

  1. Have only one supply connected at any time and have some smart stuff that switches over to the other supply in case of failure
  2. have several power supplies share the load normally. If one fails, the other(s) have to take over.
  3. random load distribution (whichever supply supplies the highest voltage "wins")

3 is easiest and can basically done with 2 diodes (or you could even rely on the output diode almost every power supply has but than make sure that it can handle the feedback).

With all solutions you probably want a crowbar cicuit in each of your supplies to prevent overvoltages...

Oh yeah, don't forget that redundancy will increase the complexity of your design and might lead to failures that would not have happened otherwise.

And make sure that you are not creating new single point of failures with your redundancy circuits.

You might also look at the discussions about such issues that have already happened here: Redundant DC-DC converter (parallel)

\$\endgroup\$

Not the answer you're looking for? Browse other questions tagged or ask your own question.