4
\$\begingroup\$

I designed a step down converter using WEBENCH. It takes 8-24V DC input and outputs 5V. The basic design works fine, but I learned from this question about why it died when I connected it to a 24V power supply that was already turned on.

Step down converter schematic

I did some experiments with the dead board. Here's an oscilloscope trace showing the max voltage when the 24V PSU is powered up and the input capacitors charge:

Oscilloscope trace showing significant ringing and 75V max voltage

A peak of 75V, even for a few microseconds, is far above the 26V absolute maximum rating of the TPS51396A.

For comparison, here is the trace if the circuit is powered up with the cable connected (note different X and Y scale).

Oscilloscope trace showing a steady rise to working voltage of 24V

The test setup for these traces is shown below:

Test setup showing an oscilloscope, power supply and the device under test

Instead of a probe I used crocodile clips - positive across the incoming supply and GND at the output (which is connected to the inner ground plane). The switch connects the output of the power supply to the input of the converter.

The application note referenced in this answer is useful. I avoided aluminium electrolytic capacitors because they are typically less robust (the PCB is not in a case), but could add one to a future revision of the board.

A TVS diode seems like a sensible choice but I am struggling to find a suitable option because the working voltage of 24V is very close to the 26V absolute maximum rating. A similar problem is described in this question, but in that case there is 12V between the working voltage and absolute maximum.

For future revisions I could add a general purpose diode for reverse polarity protection, which would give slightly wider voltage range due to the diode drop.

Is it possible to find a TVS that will clamp these excess voltages, or should I be looking for a different solution?

\$\endgroup\$
13
  • 5
    \$\begingroup\$ 2V margin is way too small. Operating continuously near absolute max is a serious reliability hit, even if you never exceed the rating. \$\endgroup\$
    – John Doty
    Sep 17, 2023 at 16:21
  • 1
    \$\begingroup\$ Please show your measurement setup for detecting 75 V peak. \$\endgroup\$
    – winny
    Sep 17, 2023 at 16:23
  • 1
    \$\begingroup\$ Yes.You want some margin. \$\endgroup\$
    – John Doty
    Sep 17, 2023 at 18:47
  • 3
    \$\begingroup\$ You need a proper 10x probe with ground spring to measure it. Overshoot is much less than you have measured. But the elephant in the room is lack of voltage margin anyway. \$\endgroup\$
    – winny
    Sep 17, 2023 at 18:53
  • 2
    \$\begingroup\$ Related: technique for using oscilloscope probe ground rings for low inductance measurement - electronics.stackexchange.com/questions/136123/… \$\endgroup\$
    – MarkU
    Sep 17, 2023 at 18:58

3 Answers 3

7
\$\begingroup\$

Is it possible to find a TVS that will clamp these excess voltages, or should I be looking for a different solution?

Sure, Zener diodes with Zener voltages in the 25 V to 26 V range exist and aren't expensive: https://www.digikey.de/short/fp404m81

But.

  1. A diode's Zener voltage heavily depends on temperature. (All components are thermosensors, some are also useful for other things… this is extremely true for diodes in high reverse bias.)
  2. It's questionable whether the diode will breakthrough first, or the semiconductor junction that becomes conductive inside your regulator IC. In the end, both are starting to conduct for the same reason.

So:

It would be wise to limit the inrush current. A simple series resistor would do – of course at the expense of wasting power. You could, however, be smart about things, like putting an P-channel MOSFET in parallel with it, and only pulling the gate low as long as the supply voltage is sufficiently below 26 V. You'd waste a bit additional power as long as you're operating "dangerously close", but my guess is that these are the times where you don't have to care about power efficiency as much, anyways.

\$\endgroup\$
1
  • \$\begingroup\$ No huge concerns about power efficiency - the step down is required to lower the total current in the wires leading to the devices. This TI paper on inrush current looks helpful: ti.com/lit/an/slva670a/slva670a.pdf?ts=1694956747966 \$\endgroup\$
    – David
    Sep 17, 2023 at 18:47
2
\$\begingroup\$

The only thing that can squeeze into that margin is a surge stopping IC: https://www.analog.com/media/en/technical-documentation/product-selector-card/Surge-Stoppers.pdf

This should work, but probably isn’t great system design. If you can respecify the converter you can go back to the cheaper and simpler passive solutions already mentioned (TVS).

\$\endgroup\$
1
  • \$\begingroup\$ Thanks. I can likely respecify the input voltage and decrease to 15V, giving 11V of margin to the absolute maximum. This is at the expense of additional cabling to accommodate the maximum current. \$\endgroup\$
    – David
    Sep 18, 2023 at 19:12
2
\$\begingroup\$

If hot plugging is a major concern, you could look at using an N-FET, sense resistor and hot-swap controller. This will also give you to inrush limiting, overload protection and reset control. There are lots of parts available, but the LTC4368 is a good starting point.

Alternatively, add a large high ESR electrolytic on the input. This will tend to damp the ringing when power is applied.

\$\endgroup\$

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.