TVS Diode and Schottky Diode in Parallel - Check my understanding

[pigtwo]

Hello all,

I've been stuck on a problem for a while now and I feel like I must be missing something. Customer has a wall wart supply they want to hot plug into a board. That board has a 5V regulator which keeps dying while plugging in the wall wart(connection to the board). We found there is a negative transient and a large amount of overshoot exceeding the regulators max input voltage.

We're tasked with testing a solution before design change. The answer we came to is some input protection in the form of a TVS diode to protect against the positive excursion and a schottky to protect against the negative excursion. Both diodes in parallel and reverse biased(IE cathode to positive rail, anode to gnd). Upon testing we've found that the schottky alone seems to correctly prevent negative transients. But when we add the TVS, the schottky seems to stop preventing them. The TVS always works regardless of whether there is a schottky or not.

After this testing we've removed the schottky and it appears to be undamaged(open in one direction, 0.6v in the other). We repeated this process again with the same results. Is there something I'm missing about the interactions of these diodes? The TVS diode should be activating when it's reverse voltage is exceeded(in this case around 24V) and the schottky should only activate when it's forward voltage is exceeded(forward = 0.6V, reverse = 60V). The diodes are directly in parallel right next to each other. Am I missing something? Or is something else likely happening?

Thank you!

[TERRA Operative]

Don't most TVS diodes conduct in both directions once their threshold voltage is reached? You may need to source a unidirectional TVS diode.

What about putting an appropriately rated bidirectional TVS diode across the incoming supply to absorb any over/undervoltage events followed by a forward biased schottky diode to prevent any negative voltage reaching the regulator?

[pigtwo]

Yeah I should have clarified I was using a unidirectional TVS.

I had started with just a bidirectional TVS and this one also seemed to not work but something else may have been happening. I think I'll give it a shot. For the forward biased schottky diode you reference this would be in series correct?

Thanks!

[wasedadoc]

BTW you need a lot of current through a Schottky to get 0.6 volt drop.

[TimNJ]

A TVS can be an acceptable solution, yes. Alternatively, have you considered the root cause of the spike? Does the 5V regulator have ceramic input capacitance?If you have a space constrained application that must use surface mount MLCCs (can't fit an electrolytic), you can add a parallel RC. A rule of thumb is that the capacitance in the RC be roughly the same as the existing capacitance in circuit. R on the order of 0.5ohm to 5ohm typically. I mean, all of this depends on how much damping you need to keep the spike down to an acceptable level, how much inductance in the DC cord, and so on.

https://www.analog.com/media/en/technical-documentation/application-notes/an88f.pdf

[David Hess]

The voltage spike from the leakage inductance of the wall wart transformer may be what is killing the regulator.  As TimNJ points out, the large input capacitance at the regulator input would normally protect it.

[pigtwo]

Alternatively, have you considered the root cause of the spike?

I spent pretty much the whole day in the lab trying to nail this down and I'm not sure I really do know the root cause. Just a thing to note, I'm an FPGA guy so when we get into the analog world I'm not as familiar.

Before getting into it, just to clarify I seem to have two events happening. One is a transient that happens before mating is complete. This is just a spike that goes positive and negative and eventually settles back to zero. IE not the wall wart providing power. Then there is the Vin spike as the wall wart provides power and overshoots by a lot. The overshoot problem seems to be resolved by a TVS diode. Which at least for now is sufficient. It's the positive and negative going transient that I'm really having trouble with.

The root cause seems to be some sort of ground 'equalization' between the board and the wall wart. The board(without probing) is floating normally. IE no other wall connection or connection to anything else. The wall wart has two prongs so no earth connection. If I connect the ground of the wall wart to the ground of the board I will see this large transient that goes positive and negative. Maybe about -10V to +10V. Looks pretty high frequency. Peak to peak is about 2-10ns. This is scoped on the board at Vin+ and ground. IE the board's power input from the wall wart. During this test all I'm doing is actually just touching the outside of the barrel connectors together. Both are connected to ground.

The reason I think this is some sort of ground equalization(there may be another term for this I don't know) is because if I take a wire from the earth ground from the wall and connect it to the outside of the barrel connector of the wall wart then mate the connectors(or just touch the grounds together) I don't get the transient. So it would seem the board is being grounded to earth via the scope ground clip and for some reason the ground on the wall wart is sitting at a different voltage. So when these connect there is some current. This sounds very odd to me as I would have thought the wall wart being isolated would have prevented anything like this. But I don't know.

Confusingly isolating the scope from earth still gives the transient. But the board is sitting on a grounded ESD mat so I imagine even some parasitic stuff here can get the board back to earth reference. This seems to make sense because if I use a non-isolated scope and just the wall wart connector(no board, just wires from the connector) and I mate them I get the transient. But if I isolate the scope I don't get the transient. So the board on ESD mat, isolated scope still gets transient seems to track.

Ok so finally the problem I'm at is nothing I do seems to bring this transient down. Putting diodes and/or caps right on the barrel connector output has no effect. Series diodes have no effect. Probing before and after these additions doesn't seem to matter very much. So my thought would be is that this would be more like EMI. It could be a lot higher frequency than I think and the scope is aliasing. And so maybe just passes through these components and doesn't get clamped(remember FPGA guy, I don't know if that makes sense). But then the question is if these diodes don't even activate from it how is it hurting the regulator? I would think those diodes also wouldn't activate and so wouldn't get damaged. And I don't have an answer to this. I think I'm missing something.

I put some ferrites in series and that actually did seem to help. It didn't remove the transient entirely but maybe cut it down by half. And they were random ones from the lab. So maybe proper ferrites tuned for the situation would work better. But I'm still a little skeptical this this transient is really killing the regulator. So other things I've considered is the rate at which the wall wart ramps could be a problem but I found no spec in the data sheet for this. The ramp rate is pretty quick to me. Maybe 50ns(going from memory) to go from 0v to 25v. I'll have to look into ways to slow this down, if possible. Just to rule that out.

Anyways I'm rambling now. I don't now if any of that makes sense to someone with more experience in these matters. It is all very odd to me. I'd love to put this problem to bed so any ideas are welcome. Thanks!

[pigtwo]

It's occurring to me now that I likely don't need to solve this ground equalization problem. If the whole problem arises from the board being earth referenced. Either from probing or esd mats. The actual design is in a molded housing and can't become earth referenced. We actually have a previous design that never fails and I couldn't figure out why. But all those provided to us came in that molded housing. I'll have to test to verify. And I'd still be interested in how this would be solved with out the housing solution.

[T3sl4co1l]

This is probably relevant: https://electronics.stackexchange.com/questions/704381/why-did-my-dc-dc-buck-circuit-fail

Tim

[TimNJ]

The 2-prong adapter is referenced to earth through Y-caps which bridge its primary-secondary. The oscilloscope has its own AC/DC converter and Y-caps too. Some things to keep in mind when measuring anything mains derived, even if you float the scope, etc. "Isolated" in the traditional sense of the word, but there's always going to be some amount of coupling.

So, what frequency is "high frequency"? This will give you an idea bout the types of capacitances and inductances involved. If it is a matter of the output negative becoming equalized with earth, then you could have the transient response, and resulting ringing a parallel resonant circuit of the Y-caps in parallel with the inductance of the scope probe. Expect ringing in the MHz-ish range, I think. It's been widely discussed here that the output negative of a 2-prong AC/DC adapter can float at ~50% of mains voltage. So, if you earth it, yes there will be some transient response.

The question is whether what you are seeing really creates a differential-mode voltage across the input of the regulator? Or is what you are measuring merely a common-mode measurement artifact (for lack of a better description). If you adding a variety of "differential mode" components across the input does not change what you see on the scope, then what you are seeing may not be differential mode.

Like mentioned, for the other issue, high Q ceramic capacitance causing issues is probably the root cause there, if in fact you are using MLCCs on the input, with no parallel electrolytic or RC to damp it. Check the attached simulation showing the oscillation/overshoot on the input when the supply is plugged in @ 5ms, and when the load current is applied @ 20ms. Whether you stimulate an LC with a step voltage or step current, you are likely to get oscillation on both voltage and current.