0
\$\begingroup\$

I've spent this evening blowing up TPS54231s. Even a couple of TPS54331s.

schematic

(note that C31 is currently DNP)

The input voltage is specified as 17-24 VDC, and this circuit's worked admirably well at 17 volts for a couple weeks now. This evening I've been testing it at 24 volts. With an adjustable bench supply (not a fantastic one - it's a surplus laptop power supply and a Chinese eBay adjustable buck converter), it works ok at 24 volts (to be fair, however, at startup the source supply is drooping a tiny bit, so it may not be a fantastic test of maximum voltage capability).

The trouble starts when I plug in a 24 volt surplus laptop power supply. The TPS54231s instantly die and short out the input. Now, I admit I haven't tried to characterize this input supply's startup characteristics by capturing it on my scope. I'm starting to run low on replacement chips and I'm done with that for now. But the steady-state voltage is correct under load and not - 24V. The TPS54231's absolute maximum input voltage is 30V. Could it be at all likely that the input supply is spiking beyond that - however briefly - and that's toasting these chips?

Assuming that's true, would replacing the S1M with a 24V TVS diode solve the problem?

I really don't want to reduce the input supply voltage spec.

\$\endgroup\$
  • \$\begingroup\$ What's your load current? Did you use webench to do the circuit load analysis and feedback loop adjustment? Did you use real inductor and capacitor esr values in the webench components? 10uF cap is a ceramic? If it's 35V, it'll have something like 1uF capacitance at 24V so you have to use that with webench to see how it really behaves. Same with output cap. \$\endgroup\$ – Barleyman May 24 '16 at 8:40
  • \$\begingroup\$ The 10 µF is ceramic and the 47 µF that's unpopulated is the footprint for a polymer. \$\endgroup\$ – nsayer May 24 '16 at 14:53
  • \$\begingroup\$ The load draws 2A at the beginning, but ramps downwards to closer to 750 mA after a few minutes of warm-up. But the behavior occurs both under load and under no load. And it happens so quickly that I posit that the load is insignificant. \$\endgroup\$ – nsayer May 24 '16 at 16:33
1
\$\begingroup\$

The TPS54331 is rated at a max input voltage of 30 volts and you are connecting a nominally 24V DC supply and you say the TPS54331 is dying virtually straight away.

Who can say what inductance is in series with the 24V DC output inside the power supply - can you see the internal circuits? I mention inductance because an inductor is pretty common inside power supplies both to reduce noise and reduce EMI.

So you connect the power supply and current flows through the 10 uF capacitor (C1) to charge it up. That current is flowing through the output inductor (speculation maybe) in the power supply and when the capacitor is fully charged to 24V there is still energy in that inductor which will carry on charging the capacitor to maybe twice 24 V (worst case) just like this: -

enter image description here

The above shows what can happen when 3V is applied to an LC circuit that has low damping (zeta is low). The red curve would be the resultant voltage across C1 - it is nearly double the input voltage. Now imagine what would happen when 24V is applied.

I'm not scare mongering - this has happened to me so I know it's possible.

Solution - put a 27V zener across the power terminals of the regulator.

\$\endgroup\$
  • \$\begingroup\$ Yeah - a 24 volt TVS acts kinda sorta like a 27V zener. Now that you mention it, there is a "lump" at the end of the cable from that supply. It's obviously a ferrite bead to knock down the EMI, but could it have that much impact on the output voltage? \$\endgroup\$ – nsayer May 24 '16 at 14:35
  • \$\begingroup\$ The lump may be insignificant or highly significant or somewhere in between. \$\endgroup\$ – Andy aka May 24 '16 at 14:47
  • \$\begingroup\$ Populating C31 and replacing the S1M with a 24V TVS appears to have made a difference. I've power cycled it 5 times in a row with the suspect supply and it's come up every time. It would have blown the first or second time before. \$\endgroup\$ – nsayer May 26 '16 at 23:13
1
\$\begingroup\$

In addition to the answer from Andy:

I read the datasheet for the TPS54231 and it has this very interesting note:

The device is designed to operate from an input-voltage supply range between 3.5 V and 28 V. This input supply should be well regulated.

Now this is particularly unusual for a voltage regulator (that the input should be well regulated in the first place).

That leads me to believe that the input stage is very susceptible to fast transients (and notably, line transient response is not characterised although load transient response is).

I would populate the 47μF device and, as you surmise, find a transient suppressor in place of the input diode.

Note that some transient suppressors can have as much as 5V between guaranteed stand-off and guaranteed breakdown, so a careful look around might well be necessary.

\$\endgroup\$
  • \$\begingroup\$ Most of the TVSs that I'm considering have a maximum guaranteed breakdown of less than 30V. I guess all I can do is hope? \$\endgroup\$ – nsayer May 24 '16 at 14:38
  • \$\begingroup\$ I would check the parametric tables at various manufacturers. There are some really good devices out there. \$\endgroup\$ – Peter Smith May 24 '16 at 17:06
0
\$\begingroup\$

I just want know about before you provide the input supply have verified the input supply polarity. When polarity is reversed due to the diode it get short circuit.

\$\endgroup\$
  • \$\begingroup\$ Um, yeah, I'm sure the input has the right polarity. That diode is supposed to short out the input to protect the circuit if you hook it up backwards. \$\endgroup\$ – nsayer May 24 '16 at 14:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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