When to use MOV, TVS or GDT?

[MikeEagle95]

Hi

I realized that in most cases when I design input protection, I use MOVs, sometimes TVS diodes and almost never GDTs.
It's well known that TVS diodes reacts much faster than the MOVs, but MOVs have higher voltage rating.
But what should anyone consider choosing input protection method and why?
For example, when to use GDT instead of MOV?

[T3sl4co1l]

Is it?

MOVs are basically thicker, looser TVSs on legs.  They are both semiconductor devices.  They have the same speed and the same limitations: C and ESL.

MOV leads and bodies do tend to be longer, so that's a thing.  They might be used at lower impedances as well, exacerbating ESL.

Thyristor TVSs and GDTs have a definite delay or rate limitation, however.

Low capacitance TVSs also have a slight delay, though to a lesser extent.  The trick is, they get the low capacitance by using a PN diode in series with a biased TVS (usually the TVS is connected to VCC).  The PN diode exhibits forward recovery, which looks like ESL but is actually due to voltage drop across the junction.

Just FYI

So, that said, when to use GDTs (and similarly, thyristor TVSs)?  Whenever you need to shunt a lot of surge current, and when there is not a continuous power source connected.

Classic use case: telephone lines.  These are subject to induced lightning surge, but supply little power (under a watt) and limited current -- not enough to keep a GDT conducting.  The line also has pretty high resistance, so the surge current is fairly modest, and all the surge voltage gets dropped across that resistance, instead of across your circuit.

MOVs are used where high energy capacity is required.  Ever checked the price on mains-capable TVSs?  Wheee-hew!  GDTs must NOT be used on mains, because of the power supply -- once it sparks over, it's going to keep going until destroyed.  This also means the MOV must drop a large voltage (about 2-3 times nominal mains voltage) during the surge, making it all the more critical that it be a big dumb lump of cheap semiconductor material that can absorb a shitload of energy, quickly.

There is, actually, one trick where GDTs can be used near the mains.  It is for line-to-ground protection.  We do not normally expect neutral-to-ground voltage, so a GDT could be used here...BUT, we must always be weary of the case where the wiring is cross-wired, or supplied from commercial three-phase (e.g., 208V in the US, where each is 120V to GND).  The trick is this: you put a MOV in series with the GDT, so the mains voltage is dropped, preventing the GDT from exploding.  What this saves you is ground leakage current, which can be significant for an MOV already, but worsens with age.  The GDT remains off until breakdown voltage is seen, and then it switches in the MOV.

Some other examples:
- General purpose, low voltage facility wiring, e.g., furnace or doorbell wiring, alarm system wiring, industrial Modbus, etc.
- High frequency signals (the GDT has very low capacitance), e.g. Ethernet (which can be run at great enough lengths -- between facilities, say -- that it can become a surge hazard)

Note that it takes length (actually loop area, but, assuming a worst-case condition of shitty grounding...) to get significant induced surge voltages and currents.  You wouldn't bother with this on short runs, unless you're REALLY worried about EMP (fast EMP, like bombs and nukes).

(And even then, nuke EMP is probably too fast for a GDT, you'll really need a combination of shielding, filtering and TVSs I think?)

Tim

[David Hess]

I have the Motorola version of this application note, A Review of Transients and Their Means of Suppression, on my desk at the moment which covers zeners (TVSes), MOVs, GDTs, and some other less common devices.

The thing which always surprises me how the peak power capability of zener diodes and TVSes.

[T3sl4co1l]

I have the Motorola version of this application note, A Review of Transients and Their Means of Suppression, on my desk at the moment which covers zeners (TVSes), MOVs, GDTs, and some other less common devices.

The thing which always surprises me how the peak power capability of zener diodes and TVSes.

Oh yes, for short pulses.  It's, not quite proportional to time, actually, more like sqrt(t), thermal diffusion or something.  Even for planar chips I guess.

The peak power goes way up at short times, of course.  Not way way up because of this, but still, lots of kW for a little SMT!

Related:

I built a soft-starting and current-limiting module, where the lost power is dissipated in a stack of TVSs.  Nominal rating 30V 20A, and using three SMDJ ("3kW") type TVS diodes.  It current-limits for 150ms, burning up to 600W * 0.15s = 90J.  The diodes are perfectly happy with this, and their surface temperature step-changes (such that it can step, as measured on the surface -- within a second or two of the event) by 10-20°C.

So you can hit the "start" button repeatedly, and crank up the diode temperature quite a lot.  They start to smoke a little (that's my filthy fingerprints evaporating, not the magic smoke!), then the thermal cutout stops it (which is delayed by a few seconds, due to thermal conduction over the PCB).

I was quite satisfied with the performance.  I want to build an even bigger one, say for industrial application (400V 30A?).  Unsure if I want to just use a bigger stack of diodes for that (it's only about $30 worth -- not all that much considering!), or a couple of MOVs in parallel (which would be well within their ratings then, and have a reasonable clamping voltage).  Biggest problem is simply dissipating the heat once it's there: even at one pulse per minute, that's an average of 30W!

Tim

[David Hess]

Avalanche rated rectifiers and TVSes but not necessarily zener diodes are fabricated to produce a uniform junction to prevent hot spots increasing their avalanche capability.  Any exposed surface of the junction is also passivated to prevent surface effects.

[PartialDischarge]

(And even then, nuke EMP is probably too fast for a GDT, you'll really need a combination of shielding, filtering and TVSs I think?)

Some of you may appreciate this old document, A guide to the use of spark gaps for EMP protection, where shielding and other hardening techniques are also mentioned

https://www.scribd.com/document/382246366/A-Guide-to-the-Use-of-Spark-Gaps-for-EMP

[Diablo2813]

There is, actually, one trick where GDTs can be used near the mains.  It is for line-to-ground protection.  We do not normally expect neutral-to-ground voltage, so a GDT could be used here...BUT, we must always be weary of the case where the wiring is cross-wired, or supplied from commercial three-phase (e.g., 208V in the US, where each is 120V to GND).  The trick is this: you put a MOV in series with the GDT, so the mains voltage is dropped, preventing the GDT from exploding.  What this saves you is ground leakage current, which can be significant for an MOV already, but worsens with age.  The GDT remains off until breakdown voltage is seen, and then it switches in the MOV.

Hi,
I'm looking for a surge protection method that matches IEC 61000-4-5 (4kV surge).
It also have to be sturdy enough to pass some others certification requirements.
It looks like the serial association of MOV and GDT is the right way to go.
I would appreciate more information in order to understand how this is supposed to works.

Precisely, i don't get how the GDT could turn on if the MOV is blocked and no current is passing through.
How should i size them then ?

I'm also still open to any other suggestions.

Thanks in advance,
Benjamin.

[MagicSmoker]

...
Precisely, i don't get how the GDT could turn on if the MOV is blocked and no current is passing through.
How should i size them then ?
...

The sum of the MOV and GDT breakdown voltages must be less than the impulse voltage that needs to be shunted. For example, a MOV with a 375V breakdown voltage in series with a GDT with a 1kV flashover voltage will (theoretically) protect against impulses of 1.375kV or higher across 240V nominal mains. The parenthetical "theoretically" is there because MOV breakdown voltage is a bit sloppy while GDT flashover voltage is really, really sloppy (and worse is that it increases with dV/dt of the impulse).

[wraper]

When in series, discharge should start to happen when GDT breakdown voltage is reached. MOV has significant leakage current while GDT does not.

[Diablo2813]

Thanks for your helps.

From what you both said i understand that discharge could start to happen from GDT breakdown voltage to maximum GDT+MOV breakdowns. That seems foggy but ok.

Though it appears I'm still missing something.
If the GDT is still blocked, i don't understand how the MOV can be active in anyway.

What is going on exactly at the point when this serial association reach his breakdown voltage ?
Are you saying MOV is passing first ? But what's the point if GDT is blocked ?

Even if i think this is how it's working, i don't get how the GDT is supposed to help once they are both passing.

I see no synergy there.

[T3sl4co1l]

The MOV is a passive device, in that it acts like a zener diode, or a wonky resistor.  Normally its voltage drop is zero, because the GDT's leakage current is very nearly zero.

The GDT is an active device, in that it acts like a switch, triggered by a high voltage (in excess of the rated voltage).  When this happens, as long as the current remains above the holding current, the voltage drop will be very low, perhaps 20V or so.

Say the MOV is 300V and the GDT is 500V.  Say we apply anything up to 500V: nothing happens, because the MOV voltage drop is zero, the GDT voltage drop equals the input, and breakdown does not occur.

Say we apply something more than 500V, from a low impedance surge generator: now the GDT breaks down, and goes from dropping 500V+ to ~20V, and suddenly 480V+ is applied to the MOV, which now draws a large current from the surge generator's source impedance.  Say the surge peak voltage was 1kV in this case, and say the MOV breakdown voltage is 600V.  That drops 400V across the surge generator's source impedance, and if that is 2Ω, the current is 400V / 2Ω = 200A.  This is in the MOV's surge area (where it may be dropping even more voltage, like 700-800V), and more than enough current to keep the GDT turned on at a low voltage drop.

When the surge current disappears, the MOV voltage drop falls as well, but so does its current flow.  As this drops to, say 1mA, this may be below the GDT's holding current, and its voltage drop rises.  As the voltage drop rises and current falls, the situation returns to normal.  This may be delayed by line current (one reason why they test with surges timed to the peaks and valleys of the mains), which can keep the MOV and GDT conducting for longer; at worst, they remain conducting for about a quarter cycle (~5ms), as the line voltage crosses through zero and the GDT turns off for sure.  (In that case, it's probably that the MOV has failed shorted, and mains fault current flows through the pair, blowing the fuse in front of the protection circuit.)

The time required for these changes is fractional microseconds.  Typical surge pulses have a rise time of a few µs, and a pulse width of 10s of µs.  The GDT may take a microsecond or more to break down, when the applied voltage is low and slow; or it may take mere ns when the voltage is high and fast (say, ESD, which delivers many kV in a few ns).  The MOV's response is largely instantaneous, delayed only by its capacitance, which is relatively large (~nF usually).

Tim

[MagicSmoker]

The MOV is a passive device, in that it acts like a zener diode, or a wonky resistor.  Normally its voltage drop is zero, because the GDT's leakage current is very nearly zero.

The GDT is an active device, in that it acts like a switch, triggered by a high voltage (in excess of the rated voltage).  When this happens, as long as the current remains above the holding current, the voltage drop will be very low, perhaps 20V or so.
...

Sorta maybe probably not. The GDT (literally) insulates the MOV from leakage current flow, so that the MOV's breakdown voltage adds to the GDT's.  That said, it's more accurate to model both components as capacitors in their pre-breakdown state, with the MOV handily beating the GDT in raw capacitance so that when presented with a high dV/dt transient it may very well breakdown last, even if it has a lower breakdown voltage than the GDT.

That's my intuitive sense of how things would behave, but I confess I haven't actually tested this out even in SPICE, much less the real world (that said, the neon bulb model in LTSpice is surprisingly good at mimicking the behavior of spark gaps and GDTs when the various parameters are tweaked).

[coppercone2]

when you feel like getting into a fight with cost people

[wraper]

Sorta maybe probably not. The GDT (literally) insulates the MOV from leakage current flow, so that the MOV's breakdown voltage adds to the GDT's.

Nope, voltage across MOV is zero due to it's leakage until total voltage reaches GDT breakdown voltage. Then GDT becomes basically short with very low voltage across it and all voltage drop is then across MOV. Nothing adds ups. It would add up if you connected two MOVs in series.

That's my intuitive sense of how things would behave

Many things are completely counter-intuitive. For example, for flight stability rockets need to have center of mass towards it's top, not bottom as you might think.

[Diablo2813]

Thank you all for this explanation. That's much clearer now.

Do you think this kind of association could be suited to protect from 4k surge voltage ? rising 1us and falling 100us.

NB:
At the moment I'm using a gas discharge tube simulation model from TDK EPCOS. It's available on demand. Looks like its working well for now.

Thanks, thanks again,
Benjamin.

[MagicSmoker]

Sorta maybe probably not. The GDT (literally) insulates the MOV from leakage current flow, so that the MOV's breakdown voltage adds to the GDT's.

Nope, voltage across MOV is zero due to it's leakage until total voltage reaches GDT breakdown voltage. Then GDT becomes basically short with very low voltage across it and all voltage drop is then across MOV. Nothing adds ups. It would add up if you connected two MOVs in series.

Yeah, ok. Treat the MOV as a very high value resistance in its pre-breakdown state that is in series with the even higher resistance of the GDT; all the voltage gets dropped across the latter. In any event, thinking about this - even by putting forth a wrong argument - was illuminating.

[T3sl4co1l]

Yes, that will be fine for ring wave.  The GDT may even stay ionized for the duration of the wave, letting the MOV do its job that much better.

Mind that SPICE models of gas discharge elements are notoriously bad, or conversely, notoriously difficult to make accurate.  I don't know offhand which case that one is.  I should check it out...

Tim

[MagicSmoker]

...
Mind that SPICE models of gas discharge elements are notoriously bad, or conversely, notoriously difficult to make accurate.  I don't know offhand which case that one is.  I should check it out...

Yes, modelling an arc discharge in an ionized gaseous medium is challenging, but I've managed to get the "neon bulb" model (under Misc) in LTSpice to do a reasonably good job of mimicking the real world behavior of an automotive spark plug and a massive carbon electrode spark gap (Ipk >30kA). I haven't yet ascertained if it gets the negative resistance characteristic right, but I suspect it doesn't because it wants you to specify the voltage across the arc (Vhold) as well as its impedance (Zon) - but at least it lets you set the gas ionization time (Tau). Right click on the symbol to set all of the above as well as the striking voltage (Vstrike) and minimum arc current (Ihold).

For example, to model a spark plug I set Vstrike to 3600, Vhold to 300, Zon to 0.1, Ihold to 1, and Tau to 1u. For a larger spark gap Tau might increase to 100u to 1m, while Zon will generally fall to 10m to 1m. Ihold and Vhold seem to be less critical as far as getting the simulation to match the real world.

[Diablo2813]

Sorta maybe probably not. The GDT (literally) insulates the MOV from leakage current flow, so that the MOV's breakdown voltage adds to the GDT's.

Nope, voltage across MOV is zero due to it's leakage until total voltage reaches GDT breakdown voltage. Then GDT becomes basically short with very low voltage across it and all voltage drop is then across MOV. Nothing adds ups. It would add up if you connected two MOVs in series.

Yeah, ok. Treat the MOV as a very high value resistance in its pre-breakdown state that is in series with the even higher resistance of the GDT; all the voltage gets dropped across the latter. In any event, thinking about this - even by putting forth a wrong argument - was illuminating.

So where are you wrong ? Does the resistance value of the GDT falls when the breakdown voltage is reached ?

Otherwhise, i was looking for a deeper explanation. I found this : .

Point D is relative to breakdown voltage.
So the glow discharge behaviour is what justify GDT one ?

If this is the case, i think the leakage current is involve next D.
Then when the current goes up, voltage falls until the GDT is really opening.
So the difference of voltage between points D and G is critical in order to get a significant voltage across the MOV ?

Thanks

[wraper]

That's neon bulb, not GDT suppressor.
Here is GDT (page 4): https://www.mouser.com/pdfdocs/bourns_gdt_white_paper.pdf

[T3sl4co1l]

But yes, that's the basic idea.  Surge currents will push it into the arc range, not the glow range.  Both have a negative resistance (switching) characteristic, going from high resistance (dark discharge, low leakage) to low (glow or arc discharge).

Tim

[Diablo2813]

That's neon bulb, not GDT suppressor.
Here is GDT (page 4): https://www.mouser.com/pdfdocs/bourns_gdt_white_paper.pdf

But yes, that's the basic idea.  Surge currents will push it into the arc range, not the glow range.  Both have a negative resistance (switching) characteristic, going from high resistance (dark discharge, low leakage) to low (glow or arc discharge).

Tim

Thanks.
 
To summarize,

General information :

MOV and GDT are both able to short a surge when a specific breakdown voltage is reached.

GDT acts like a switch and do not significantly dissipate. It also do not limit the current and induce a risk to burn the lines.
A low DC voltage across the GDT may prevent it from returning to his normal state. Using it alone to the ground isn't an option.
GDT switch delay is faster as dV/dt rise and can be few us for lower dV/dt.
In the other hand, the GDT have zero leakage current and is much more durable than a MOV.

MOV acts like zener and does dissipate. It's able to limit current using his "zener" voltage only.
It doesnt care much about DC voltage. But it also have a low but not zero leakage current.
MOV switch delay is low and constant.
The MOV have a lower life expectancy and is likely to burn over the uses and induce a short-circuit.

A serial fuse should be use in any case.

Serial association :

Now, if i understand right, the only synergies i see are :
from adding a serial GDT to a MOV

• To modify global breakdown value so the MOV get to support fewer energy.
  
• To prevent leakage current from the MOV (but that's not something that important ?).

and from adding a serial MOV to a GDT is :

• To secure the return of the GDT to his blocking state

In my opinion

So i actually don't see much interest in adding a serial GDT in my application. (4kV surge voltage)
Some SMT MOV even got decent sizing close to GDT ones. I would rather add another MOV than a GDT.

[wraper]

Some SMT MOV even got decent sizing close to GDT ones.

Then look at their ratings and think again. Small MOV do explode quite often when actually suppressing surge.

[T3sl4co1l]

You got it

Leakage is important for some standards, like IEC 60950-1 which prohibits MOVs directly from line to ground (for most equipment), however a MOV+GDT is allowed.  This is for ground leakage reasons -- on equipment where the ground leakage could potentially be exposed to the user (an insulation fault in a double-insulated product, or via enclosure / connections if the ground pin gets lifted), those MOVs would be a big problem just to begin with, but also as they age and decay, leakage current increases and also breakdown voltage may fall (or one outright shorts due to a large enough surge, or enough hits), and that's where the GDT comes in, blocking that leakage until surge voltage is applied.

SMTs, are simply rated what they are rated.  If that's adequate for your application, great!

I wouldn't recommend them for mains (IEC 61000-4-5, 2kV and 2Ω, say), but they may be adequate for ring wave, and are definitely adequate for ESD and EFT purposes.  Indeed, you can get tiny (0603 and such) chips for ESD and minor surge purposes -- they handle more energy than TVS diodes of the same size, though again with the same rather lax voltage drop, characteristic of any MOV.

Note that low voltage MOVs have somewhat poor energy ratings, partly because not much material thickness is required.  MOVs below 80V also use a different formulation, with less voltage drop per thickness but much higher internal resistance --> higher peak voltage drop.  In fact, manufacturers usually recommend that, if you need a 60 or 70V MOV, just go ahead and choose the 80V part, because it will actually have lower clamping voltage.  (40V parts I think are about a wash, usually, while lower voltages are fine.)

But yeah, if we're talking mains voltages here, whatever is fine.  A typical 20mm, 300V MOV can gobble up over a hundred joules in its final heroic act, and can do it repeatedly for smaller surges.  If you only need ring wave that's not too strenuous, maybe a 7 or 10mm part is fine, or even an SMT?

The other downside to SMTs, they seem to be more expensive.  Go figure?

Tim

[Diablo2813]

In the end, i voted for a MOV+GDT serial association.

I'm willing to pm my setup if anyone is able to check it out.
That would be very helpful.

Thanks all of you,
Benjamin.