TVS for Protecting a 38Vmax DC-DC that needs to operate at 24Vin??......

[Smokey]

It's still two in parallel.  Since this is smaller, I could add a third.
But ya, it would be nice if they had a thermal impedance plot.

[tszaboo]

TVS is not the right part for surge protection. You use it to suppress eg. ESD events, not surge events. The energy in those are orders of magnitude different. So you TVS can be anywhere at 40-50V because having 35V on a 38V device for such a short amount of time is generally OK. Plus you have capacitors shunting the high frequency inputs.
But your problem is with surge protection, that requires different circuit.

[T3sl4co1l]

TVS is not the right part for surge protection. You use it to suppress eg. ESD events, not surge events. The energy in those are orders of magnitude different. So you TVS can be anywhere at 40-50V because having 35V on a 38V device for such a short amount of time is generally OK. Plus you have capacitors shunting the high frequency inputs.
But your problem is with surge protection, that requires different circuit.

So all those TVS sold for and specifically approved for surge application,

...aren't?

Oh, or do you mean that TVS specifically?  ( https://assets.nexperia.com/documents/data-sheet/PTVS30VZ1UPA.pdf )   Well, they clearly say 150A 8/20us, which is pretty damn good I'd say, for what it is. But it ain't gonna be much beyond that, and it'll be naught but a tiny crater from load dump.

Tim

[temperance]

a TL431 the pass element could be a P-fet for practically no loss in normal operation

You didn't read what I wrote or didn't understand what I wrote.

Of topic but some insight into LDO's with PNP, P-FET regulators.

With a P-FET the pass device voltage is referenced to the input of the regulator and because of this the circuit is pretty bad at handling input transients. The transient will drive the P-FET further into conduction while the control circuit with it's limited bandwidth is taking it's time to respond. Making the control circuit faster is very difficult if not impossible. Early LDO regulators had difficulties dealing with input transients. Some newer types are much better but nowhere near the performance of a "normal" regulator.

TVS is not the right part for surge protection. You use it to suppress eg. ESD events, not surge events.

I think you are confused by the terminology used for those diodes. There are:

TVS diodes with ratings for surges only also called tranzorb diodes by Vishay.
TVS diodes suitable for ESD and to some extend surges. Example: SP0115-01ETG
TVS diodes suitable for ESD only without surge rating specifications.

[tszaboo]

TVS is not the right part for surge protection. You use it to suppress eg. ESD events, not surge events. The energy in those are orders of magnitude different. So you TVS can be anywhere at 40-50V because having 35V on a 38V device for such a short amount of time is generally OK. Plus you have capacitors shunting the high frequency inputs.
But your problem is with surge protection, that requires different circuit.

So all those TVS sold for and specifically approved for surge application,

...aren't?

Oh, or do you mean that TVS specifically?  ( https://assets.nexperia.com/documents/data-sheet/PTVS30VZ1UPA.pdf )   Well, they clearly say 150A 8/20us, which is pretty damn good I'd say, for what it is. But it ain't gonna be much beyond that, and it'll be naught but a tiny crater from load dump.

Tim

This will survive 150A at IEC 61000-4-5, which is a surge standard. The lowest level of IEC 61000-4-5 is 500Ohm with 2Ohm source resistance for power supplies, which would be 250A current. It would be appropriate for surge protection for signal lines, but OP want to protect a power supply. And that it's rated peak pulse current, honesty I don't think it would survive that repeatedly. Or IEC 61000-4-4 which IMHO are much nastier, since it's repeated pulses. And these two standard are only applicable for household stuff, automotive standards are even more nastier.

[langwadt]

a TL431 the pass element could be a P-fet for practically no loss in normal operation

You didn't read what I wrote or didn't understand what I wrote.

Of topic but some insight into LDO's with PNP, P-FET regulators.

With a P-FET the pass device voltage is referenced to the input of the regulator and because of this the circuit is pretty bad at handling input transients. The transient will drive the P-FET further into conduction while the control circuit with it's limited bandwidth is taking it's time to respond. Making the control circuit faster is very difficult if not impossible. Early LDO regulators had difficulties dealing with input transients. Some newer types are much better but nowhere near the performance of a "normal" regulator.

with pfet the idea was not to make a regulator but disconnect the load when the voltage it high

https://electrooptical.net/static/oldsite/www.analog-innovations.com/SED/OverAndReverseVoltageProtection.pdf

[temperance]

with pfet the idea was not to make a regulator but disconnect the load when the voltage it high

You wrote earlier to to make a crude linear regulator. That what you've shown now is something different and probably more than fast enough.

[T3sl4co1l]

This will survive 150A at IEC 61000-4-5, which is a surge standard. The lowest level of IEC 61000-4-5 is 500Ohm with 2Ohm source resistance for power supplies, which would be 250A current. It would be appropriate for surge protection for signal lines, but OP want to protect a power supply. And that it's rated peak pulse current, honesty I don't think it would survive that repeatedly. Or IEC 61000-4-4 which IMHO are much nastier, since it's repeated pulses. And these two standard are only applicable for household stuff, automotive standards are even more nastier.

61000-4-5 is a supporting standard; typically it is adapted for particular use.  It's used at full strength for most mains testing (2Ω differential), but reduced for common mode testing (22Ω mains-GND, or similar or higher for TNV circuits, etc.).

This isn't an unusual confusion; another example:
https://electronics.stackexchange.com/questions/682917/en61000-4-5-surge-protection-of-pcb-power-supply
No way in hell that handles mains level surge, but the specifics of EN50121 match suspiciously close to the part rating, which seems a quite reasonable explanation for the product claims.

Which is, of course, how they are able to test all those various voltage TVSs at 8/20us or 10/1000us or whatever, at specified current.  Simply adjust the transient voltage, or use an attenuator (or both), to maintain specified waveform while delivering rated peak current.  I'd be surprised if they did it by series resistance alone (as most other tests do), since they would have to specify 1.2/50us, or something inbetween, depending on rating, which would be either very unexpected, or very messy.  But that is actually a good question, what exact waveform they test with and how much of it ends up clamped by the TVS (OCV peak vs clamping ratio).

Automotive isn't really nastier, but it depends what it is.  Ye olde ISO-7637 has two nasty pulses, one around 10A and some ~ms, which I'd want to model or test, but this device may well pass; and the notorious load dump, which this absolutely will not pass.  All the other pulses are much lower level: EFT and ESD sorts of things at mere hundreds of volts, and impedances of 10s of ohms.

There are others of course, like ISO-10605 ESD, which defines a more stringent test (ESD gun up to 330pF/330Ω), more or less because a seated occupant has more capacitance to surroundings than when casually walking by an office desk.  Though the voltage can be whatever ("as defined in test plan"), presumably specified in other standards, or as agreed between supplier and customer.  (The rail standards are generally similar, with more emphasis on agreement, since rolling stock isn't exactly a consumer commodity.)

Tim

[T3sl4co1l]

with pfet the idea was not to make a regulator but disconnect the load when the voltage it high

You wrote earlier to to make a crude linear regulator. That what you've shown now is something different and probably more than fast enough.

Ah, by mentioning "P-fet" I think they implied a different circuit.  P-FETs are commonly used for fairly gross on/off limiting/switching application in automotive circuits like this.  This wasn't exactly clear, of course.

For my part, I didn't bother mentioning the other traditional protection methods (switch, pass limiter, shunt limiter), because I think OP is already well aware of them, and is concentrating on TVS for, well, unstated reasons.  Or perhaps out of poorly made assumptions, such as "it worked for 12V"; but unstated in any case.

The current level is a little high, enough that a pass limiter might not be attractive, or that the added bulk, or parts cost (especially if one contemplates a LT4363 "Surge Stopper" or the like $$$!), might not be acceptable.  Though that doesn't leave much of any choice, as the pure-TVS solution is rather bulky and expensive itself.  Perhaps the regulator must be changed after all.

Tim

[langwadt]

with pfet the idea was not to make a regulator but disconnect the load when the voltage it high

You wrote earlier to to make a crude linear regulator. That what you've shown now is something different and probably more than fast enough.

Ah, by mentioning "P-fet" I think they implied a different circuit.  P-FETs are commonly used for fairly gross on/off limiting/switching application in automotive circuits like this.  This wasn't exactly clear, of course.

For my part, I didn't bother mentioning the other traditional protection methods (switch, pass limiter, shunt limiter), because I think OP is already well aware of them, and is concentrating on TVS for, well, unstated reasons.  Or perhaps out of poorly made assumptions, such as "it worked for 12V"; but unstated in any case.

The current level is a little high, enough that a pass limiter might not be attractive, or that the added bulk, or parts cost (especially if one contemplates a LT4363 "Surge Stopper" or the like $$$!), might not be acceptable.  Though that doesn't leave much of any choice, as the pure-TVS solution is rather bulky and expensive itself.  Perhaps the regulator must be changed after all.

Tim

this is quite a bit cheaper, https://www.digikey.dk/en/products/detail/richtek-usa-inc/RT1720GF/6002895

what is the current actually? with a 3A regulator making, say, 5V from 24V the input current is only going to be 500mA

[temperance]

This will survive 150A at IEC 61000-4-5, which is a surge standard. The lowest level of IEC 61000-4-5 is 500Ohm with 2Ohm source resistance for power supplies, which would be 250A current.

That's something very different from what you wrote earlier stating TVS's are only applicable to ESD. A TVS able to survive a 150A surge pulses can be ok for some applications because the diode might not be the only component dissipating some energy.


The automotive industry uses other standards for surge testing.

https://www.vishay.com/docs/88490/tvs.pdf
https://www.tdk-electronics.tdk.com/download/185840/e2cd27e63b1ece1bf787906a9474904a/ctvs-protection.pdf

[tszaboo]

This will survive 150A at IEC 61000-4-5, which is a surge standard. The lowest level of IEC 61000-4-5 is 500Ohm with 2Ohm source resistance for power supplies, which would be 250A current.

That's something very different from what you wrote earlier stating TVS's are only applicable to ESD. A TVS able to survive a 150A surge pulses can be ok for some applications because the diode might not be the only component dissipating some energy.


The automotive industry uses other standards for surge testing.

https://www.vishay.com/docs/88490/tvs.pdf
https://www.tdk-electronics.tdk.com/download/185840/e2cd27e63b1ece1bf787906a9474904a/ctvs-protection.pdf

If it cannot be used for a power supply input that needs surge protection, but it could be used for something else, and we are talking about protecting a power supply input, then how exactly is it different from what wrote?
Have you looked at the devices that are mentioned in that vishay appnote? They are not regular TVS diodes, they clearly have some sort of MOV or other device integrated into them, and are called "TVS devices". I mean, yeah, if someone names their MOV a TVS diode, and then that makes me technically wrong, congratulations, here is one internet point for proving me wrong: 1 

[temperance]

Here is what you wrote earlier:

TVS is not the right part for surge protection. You use it to suppress eg. ESD events, not surge events.

I don't see how I offended you. I only pointed out that the term TVS doesn't automatically imply ESD protection only.

If it cannot be used for a power supply input that needs surge protection, but it could be used for something else, and we are talking about protecting a power supply input, then how exactly is it different from what wrote?

As stated earlier but not explicitly, a 150 A 8/20 µs TVS can be used on a supply line if other components on the power supply line dissipate or smear out some part of the energy, which is more often than not the case. One small is enough to reduce the peak current trough the TVS. Of course the TVS on it's own will just go up in smoke.

Have you looked at the devices that are mentioned in that vishay appnote? They are not regular TVS diodes, they clearly have some sort of MOV or other device integrated into them, and are called "TVS devices".

I don't think there are MOV's in there because MOV's deteriorate on every surge event. Those are probably zeners with very descent heat slug at least the package looks like that.

[T3sl4co1l]

If it cannot be used for a power supply input that needs surge protection, but it could be used for something else, and we are talking about protecting a power supply input, then how exactly is it different from what wrote?
Have you looked at the devices that are mentioned in that vishay appnote? They are not regular TVS diodes, they clearly have some sort of MOV or other device integrated into them, and are called "TVS devices". I mean, yeah, if someone names their MOV a TVS diode, and then that makes me technically wrong, congratulations, here is one internet point for proving me wrong: 1 

They... they give part numbers?

e.g. https://www.vishay.com/docs/88382/sm5s.pdf
It's a diode.

They don't say what it's made of, but "passivated junction" doesn't make much sense outside of planar semiconductors, and given the voltages, and as it was the style at the time, it's certainly silicon based.

At worst, it's a diode in parallel with something, but absolutely nothing else but an avalanche diode will give you a 39V Vpp at 93A while also breaking down at 28V (5mA).

They give waveforms too, and clearly the peak voltages are in line with the datasheet.

They mention "EPI PAR" (and non-) types, without defining what those mean.  And in fact PAR seems to only appear in this article, its alternative forms, and the product line itself.  I'm guessing it's "passivated anisotropic rectifier", but they don't actually show the abbreviation side-by-side.  Shoddy notation.

Now, there are plenty of boners to critique in the application note -- it is as notoriously bad as any other appnote -- but, the ways in which it is bad are more interesting than simply stating it's bad.  I mean, they're all bad, that's kinda the starting point.

...

...Aha, this is probably the patent:
https://patents.google.com/patent/US5399901
They must mean "anisotropic" as in part of the fab process, i.e. using KOH etch to make a pyramidal depression.  Which is done to expose the edge of a depletion region, forming a mesa, which is then passivated (presumably with SiO2, hence "glass passivated").  I assume because the mesa can have more even current density / freedom from defects, and perhaps guard rings and surface states weren't well enough understood at the time (but I mean, this was the 80s, HV MOS was already well under production; maybe as a largely(??) bipolar/rectifier company they didn't have much knowledge/tech on those topics, who knows?), and this was done to make cheap robust rectifiers with reasonable avalanche ratings, indeed including load dump ratings.  Probably modern Vishay parts are similar, though they could just be your average modern planar epitaxy something or other.

Oh, and epi vs. non would just be how the PN junction is formed, epitaxy or diffusion, evidently the latter having higher internal resistance or built-in potential, whatever the case is.


Anyway, I don't see where you get an MOV feel from this, or why it should be anything other than what it says it is.

Again, we can formally critique the appnote if we like, but this extreme conclusion doesn't seem supportable in any way.

As stated earlier but not explicitly, a 150 A 8/20 µs TVS can be used on a supply line if other components on the power supply line dissipate or smear out some part of the energy, which is more often than not the case. One small is enough to reduce the peak current trough the TVS. Of course the TVS on it's own will just go up in smoke.

Again, what power line?

If in present context, then OP never specified load dump parameters (peak voltage, duration, impedance), so at best all we can do is say whether it could work at all (which, I think is safe to conclude not; I'd be shocked if the  survived even 12V load dump).  Whether any particular level, who knows.

And, load dump is slow enough that reactive components do little if anything, at least at realistic values; that said, maybe someone wishes to put one of those automotive amplifier "boost" capacitors in there, a couple/fraction of a farad across the line, which would do the job, actually surprisingly well; but for production purposes, that'd be an odd pick at least.  (But an interesting one to see in a real product, hah.)

If you're still stuck on IEC 61000-4-5 surge and 150A "not existing" in that standard, then, perhaps; but, again, there are numerous applications of that standard where the peak current is significantly reduced, so it's perfectly reasonable that a small device can handle something like that.

I don't think there are MOV's in there because MOV's deteriorate on every surge event. Those are probably zeners with very descent heat slug at least the package looks like that.

MOV manufacturers give curves for infinite lifetime.  Only undersized MOVs, or excessive surge conditions, cause significant deterioration.

Tim

[Marco]

Sure avalanche diodes are not as robust as MOVs, but a load dump ain't lightning either. The standard allows a fair amount of resistance, Vishay shows them surviving ISO 7637 testing. Likely there will never be a load dump bad enough to destroy the TVS and if there are, you point to Vishay's application note and run.

To me blocking makes a lot more sense than dumping though. I'd just try SEPIC with a high voltage MOSFET and an extra input diode, it's going to be a little inefficient ... but I doubt it's completely impractical. (Vin for the controller can have a fair amount of added series resistance, so that can be easily protected with a TVS diode.)

[bdunham7]

Sure avalanche diodes are not as robust as MOVs, but a load dump ain't lightning either. The standard allows a fair amount of resistance, Vishay shows them surviving ISO 7637 testing. Likely there will never be a load dump bad enough to destroy the TVS and if there are, you point to Vishay's application note and run.

To me blocking makes a lot more sense than dumping though. I'd just try SEPIC with a high voltage MOSFET and an extra input diode, it's going to be a little inefficient ... but I doubt it's completely impractical. (Vin for the controller can have a fair amount of added series resistance, so that can be easily protected with a TVS diode.)

I see discussion of 8/20µs pulses which is not relevant to load dumps which are 100-400ms.  I think the Vishay reference to ISO 7637 testing is for nominal 12/14V systems, not 24/28V.  This app note is a good read:

www.ti.com/lit/an/snva681a/snva681a.pdf

[temperance]

MOV manufacturers give curves for infinite lifetime.  Only undersized MOVs, or excessive surge conditions, cause significant deterioration.

It seems I should have made that more clear.

Repetitive surge ratings for MOV's infinite lifetime are quite a bit lower than the usual surge pulse capability of those devices listed in data sheets. Most manufacturers don't even state infinite lifetime curves for their MOV's. I do have an old Harris application handbook (paper version) with such data. The peak power handling of those devices is extreme. But the number of allowed pulses is very low at the extremes. (I understood from literature that you actually burning tunnels trough the device at the extremes)

For a 20 mm MOV:
2000 A / 20 µs: 10 pulses
500 A/ 20 µs: 10K pulses
Infinite life: 25A / 20µs

The data assumes enough time to cool down between pulses.

That's how I conclude that there can't be a MOV in those Vishay TVS diodes and why I think your facepalm was inappropriate.

Again, what power line?

that's the point and problem in this discussion which has been skipped by the OP.

I found a PDF version of the Harris book:
http://www.bitsavers.org/components/harris/1994_Harris_Transient_Voltage_Suppression_Devices.pdf

Page 67

[T3sl4co1l]

Thanks for the clarification.

The other way you can tell is the peak voltage is low.  Anything sold as a MOV, with comparable ratings, has a Vpp in the 100, 150V range.  Usable with a voltage limiter circuit, you can get some pretty extraordinary robustness that way; but not alone, with a regulator like in question here.

Tim

[Marco]

I see discussion of 8/20µs pulses which is not relevant to load dumps which are 100-400ms.

https://www.vishay.com/docs/87003/vishayloaddumptvsseriesfor24vpowertrains.pdf'

PS. that's for higher clamping voltage than desired here though.

[tszaboo]

MOV manufacturers give curves for infinite lifetime.  Only undersized MOVs, or excessive surge conditions, cause significant deterioration.

It seems I should have made that more clear.

No you shouldn't have, and that's my point. People who spend their entire life online will find a way to find any teeny-tiny "mistake" in your generally true comment, because then they can point it out and then you are wrong sir.

Thanks for the clarification.

The other way you can tell is the peak voltage is low.  Anything sold as a MOV, with comparable ratings, has a Vpp in the 100, 150V range.  Usable with a voltage limiter circuit, you can get some pretty extraordinary robustness that way; but not alone, with a regulator like in question here.

Tim

Haha, you see Tim is wrong here. You are wrong sir. There are MOVs with lower voltages.
https://www.bourns.com/docs/product-datasheets/mov14d.pdf

So you are wrong and I'm right because I found a teeny-tiny part number that is not according to your broad generalization.
 

[bdunham7]

https://www.vishay.com/docs/87003/vishayloaddumptvsseriesfor24vpowertrains.pdf'

PS. that's for higher clamping voltage than desired here though.

OK, thanks for the additional detail.  To make those work completely reliably, you'd need quite a bit of source impedance, which means series resistance in this case as we've both mentioned previously.  IIRC the OP's requirements were for 3A of current, so that's a big, warm 5R ceramic resistor.  Or perhaps many parts in parallel.

[langwadt]

https://www.vishay.com/docs/87003/vishayloaddumptvsseriesfor24vpowertrains.pdf'

PS. that's for higher clamping voltage than desired here though.

OK, thanks for the additional detail.  To make those work completely reliably, you'd need quite a bit of source impedance, which means series resistance in this case as we've both mentioned previously.  IIRC the OP's requirements were for 3A of current, so that's a big, warm 5R ceramic resistor.  Or perhaps many parts in parallel.

buck converter rated for 3A _output_ current, it probably makes 5V so input current less that 1A

[T3sl4co1l]

Haha, you see Tim is wrong here. You are wrong sir. There are MOVs with lower voltages.
https://www.bourns.com/docs/product-datasheets/mov14d.pdf

So you are wrong and I'm right because I found a teeny-tiny part number that is not according to your broad generalization.

I know you like to be argumentative for the sake of it, but if you would notice the pulse current, and scroll down to the plots, you'll see the point.  You'd probably choose a 390 for 24V automotive use, which has a peak voltage around 88V at 10A, 110V at 100A.  Which isn't taking much of a bite out of a say 150V load dump, and doing basically nothing at all at lower currents.

Actually the 820 isn't much worse, clamping ~160V at 100A, despite more than double the nominal rating.  The catch is, low voltage MOVs have even poorer performance, they're actually formulated differently; so it's all the more reason to prefer TVS here.  Note also the test current in the table, all the low ones are a measly 10A.

Voltages on the lowest rated parts are a bit better than I hinted, but leakage is pretty high as well, and the curve isn't any sharper, making TVS all the more attractive; you really don't use MOVs down here except in special cases.

Tim

[temperance]

So you are wrong and I'm right because I found a teeny-tiny part number that is not according to your broad generalization.

Sorry to inform you but Tim is talking about a MOV's Vpp clamping voltages at comparable current ratings which are much higher than those of TVS diodes.

However, I wholeheartedly apologize if what I wrote earlier made you nervous and revert to belittling.

[Marco]

I have never done it, but is it really not possible to make a reasonably efficient SEPIC converter with say LM3488 and say TPH5200 MOSFET and a 250V film capacitor for Cs?

It won't care about load dump, it won't care about cold start (buck boost) and with a Schottky on the battery input reverse polarity is handled too.