3
\$\begingroup\$

I am designing a DC buck power supply for some 5V and 12V electronics, and I was wondering about generally accepted good practices when it comes to sizing overcurrent protection. Searching around, I found a lot of information for larger and higher voltage systems, but very little for very low voltage (<50V) electronics. Take this example block diagram: A Buck regulator feeds three devices: a fan, a heater, and a computer. There is an eFuse between the regulator and each device.

A Buck regulator LMR14050QDPRRQ1 feeds three devices: a fan, a heater, and a computer. Each device is protected by an eFuse TPS259472LRPWR. The regulator can supply a max of 5 amps, and each eFuse is configured to clamp the current to each device at 1A, 1A, and 3A, respectively.

Suppose the eFuse current limits for the fan and heater are the real-world measured or spec sheet max current draws for those devices. The limit for the computer is equal to its thermal design power divided by the supply voltage. The regulator's current limit is not configurable. Both the eFuses and regulator will clamp current to the max value in the case of overdraw.

So, my questions are:

  1. Is there a good rule of thumb for setting current limits for simple resistive loads like the fan and heater? Spec sheet/measured max +20%?
  2. What about more complex devices like the computer? Presumably, the computer has spiky current draws when it is working, so just its TDP is probably not a good measure for setting limits, and I'm not sure if trying to measure max current spikes would be either, because we can't predict all of its operating modes.
  3. Is it possible to design in priority for more critical devices? In this example, increasing the current limits on the eFuses could theoretically overdraw the regulator. Suppose we want to give priority to the computer because it could fail if not supplied sufficient power. Would this affect how we set the current limits on the other devices (e.g., spec sheet +5% instead of +20%)?
\$\endgroup\$
15
  • 1
    \$\begingroup\$ Is there an extenuating reason for the extra protection? 12V 5A sounds low enough it might not matter. See for example: cui.com/blog/overview-of-limited-power-source-lps-requirements \$\endgroup\$ Jan 25, 2023 at 5:22
  • 1
    \$\begingroup\$ Also mind that fans aren't very resistive/linear, indeed they may draw quite large inrush current to get moving, and can draw large ripple currents in normal operation. Heaters likewise may have lower resistance when cold; incandescent lamps are probably the most dramatic example of this, having a hot/cold ratio more than 10. So, depends on the heater material and its temp range in use. And most active DC loads (like the computer) will have lots of filter capacitors in them, which need inrush current to start up. \$\endgroup\$ Jan 25, 2023 at 5:26
  • 2
    \$\begingroup\$ What are the fuses trying to ensure or do or protect? This question may sound simplistic but it isn't. \$\endgroup\$
    – Andy aka
    Jan 25, 2023 at 10:53
  • 2
    \$\begingroup\$ At this moment I see no reason to have fuses at all. I would use 3 current limiting buck regulators but, that's just me. \$\endgroup\$
    – Andy aka
    Jan 25, 2023 at 19:08
  • 2
    \$\begingroup\$ @mkeith Or conversely, if it's fanning the computer, or power supply, or etc. Nice detail. \$\endgroup\$ Jan 27, 2023 at 4:55

1 Answer 1

2
\$\begingroup\$

I don't think there is a rule of thumb. If this is for production of a consumer product, here is one approach.

You can calculate what you think the worst case current consumption is, and then build out 5 or 10 units and do 4 corner testing (or maybe more corners).

Four corner testing means in this case means high and low voltage permuted with high and low temperature. So that is hot and high, hot and low, cold and high, and finally, cold and low.

For this test, you will need a worst case firmware on the processor. The processor current draw will vary depending on what the processor is doing. You need a firmware specifically designed to create the worst case conditions for current draw on your processor. You may not know what those conditions are until you experiment a bit and try different things that your product will have to do in real life (while measuring the current).

The test must somehow detect with certainty if the efuse trips. You will have to leave the test running for a long time. Let's say at least 24 hours.

Here is a second approach. For each e-fuse, lower the limit until it trips readily in normal operation. Then double it. Cross your fingers that it will be OK.

You can also combine the two approaches by running your 4 corner testing after doubling the limits.

\$\endgroup\$
1
  • 1
    \$\begingroup\$ Thanks for spelling it out. My plan right now is to disable the eFuse OCP (or just use a MOSFET switch instead) for the sensitive devices like the computer, and arrange the devices so that there are no important devices sharing a regulator with them. For the rest, corner testing will be straightforward and sufficient to spec the current limits. \$\endgroup\$
    – Rick
    Jan 28, 2023 at 1:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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