How to choose TVS diodes to dispel flyback surges from relay coils?

[calzap]

In the past, I've used simple diodes for DC relays and MOVs for AC relays.  These appear to have disadvantages.  A simple diode can slow the opening of the contacts leading to arcing and premature wear.  With repeated conduction, MOVs can eventually fail open.

So, I thought I'd give bidirectional TVS diodes a try on some 24VAC relays and contactors.   The devices to be protected are reasonably robust ... voltage monitors, LEDs, and timing relays.  I'm guessing a good choice of reverse standoff voltage spec for a TVS diode is about twice AC RMS voltage ... so about 50 V in this case.  But what about the peak pulse power spec?  Again guessing ... maybe 5-10 times inrush VA of the relay.  Largest coil in present design has inrush VA of 70.  So maybe use a Littefuse P6KE series TVS diode.  These have a peak pulse power rating of 600 W.  The P6KE62CA has reverse standoff V of 53, breakdown V of 59-65, and max clamping V of 77.

Advice appreciated.

Mike in California

[Wolfram]

For inductive loads, the peak dissipation the TVS will see is the coil current at the moment of turn off, multiplied by the clamping voltage. The energy the TVS needs to dissipate is approximately equal to the energy stored in the coil, 0.5*L*I^2 . Note that inductance will be different when the relay armature closes the magnetic circuit of the coil, so the coil inductance when the relay is closed is good to use as a worst case value. For the voltage rating of the TVS, you need to make sure that it is comfortably above the maximum supply voltage. Note that the circuit that drives the coil will see the sum of the clamping voltage and your supply voltage, which would be 24 + 77 = 101 V with the P6KE62CA.

[jonpaul]

Use ordinary diode, 1N4004 for small relays, not a transient absorbing Zener

Jon

[MagicSmoker]

MOVs eventually fail from the cumulative energy dissipated in them, so most will handle millions of "surges" from a relay/contactor opening.

As for selecting a bidirectional TVS diode for clamping instead, I generally go with twice the supply voltage or half the switch voltage rating, whichever is less. Even a SMA package TVS is going to be robust enough for all but the largest relays/contactors, but if you want to take the belt-and-suspenders approach then just ensure the peak current rating of the TVS exceeds the current drawn by the relay/contactor. Unless you are using an SOD-123 package TVS to protect a huge 3ph. motor starting contactor I doubt there will be an issue.

[T3sl4co1l]

Yeah, won't take much, a P6KE will certainly do.  Voltage rating should be high enough not to run afoul of line swell + modest surges, for which double is probably alright.

Noteworthy that larger surges will still be absorbed by it; if this is 24VAC from a modest sized transformer, that should be okay as the leakage inductance of the transformer acts to limit peak surge current.  That's where you'd need to match up peak power ratings.

For the coil voltage itself, the pulse length will be quite long (some ~ms), and the peak power quite low (about 2 * V * I, as you guessed), giving plenty of time to absorb that energy.

Tim

[NiHaoMike]

For an AC relay coil, add a capacitor in parallel. One that gets the relay+capacitor close to a unity power factor is a good start.

[trobbins]

calzap, have you suffered a number of relay or protection device failures in that application, or are you just fishing for an alternative?

MagicSmoker, I think you will find that MOV's don't degrade unless they are hit above a significant surge level above a certain repetition rate - this relay application is power limited (compared to a primary side mains application), and I'd be surprised if the application pushes the MOV.

[calzap]

Thanks for the replies.

I became interested in using TVS diodes because bidirectional ones can be used with AC or DC relays and because of reading some articles that claimed they were better than simple diodes or MOVs for many applications.   And partly just out of curiosity and willingness to try something different.

 Here are some links:

https://www.te.com/commerce/DocumentDelivery/DDEController?Action=srchrtrv&DocNm=13C3264_AppNote&DocType=CS&DocLang=EN

https://m.littelfuse.com/~/media/electronics_technical/application_notes/varistors/littelfuse_transient_suppression_devices_and_principles_application_note.pdf

https://m.littelfuse.com/~/media/electronics/application_notes/littelfuse_tvs_diode_overview_application_note.pdf.pdf

Relay specs don't usually include coil inductances.  I can measure them on relays easily enough without moving the action.  For others, getting inside the case to move the action manually is problematic.  So, I got out my DER EE LCR meter and checked a few where the action could be easily moved manually.  First up was an A-B contactor, rated 2 HP @240VAC, 120VAC coil, M/N 700-HG47A1.  Like many (most?) relays, it has a fixed core.  When energized, it attracts a metal bar with contacts on it.  Inductance was 829 mH until the metal bar was pushed against the core, then it increased to 836 mH.  Next was a Schneider relay, 1 HP @240VAC, 24VAC coil, M/N 782XBXM4L-24A.  It also had a fixed core.  249 mH without the metal bar pressed against core; 271 mH with the metal bar against the core.  Next was an ACI contactor, 3HP @240VAC, 24VAC coil, M/N CK16.310-024.  This one has a moving core.  14.5 mH with the core out of coil; 114 mH with the core in the coil.

Not surprisingly, if a relay doesn't have a moving coil core, coil inductance doesn't really change much when energized.  If it has a moving core, there's a huge change in inductance when energized.  Interesting that the moving-core ACI has a lot less inductance (with the core in) than the fixe-core Schneider even though the ACI is moving larger contacts (3 HP rated vs 1 HP rated).  However, the sealed power consumption of the ACI is quite a bit more than the Schneider (12 VA vs 1.4 VA).

Both the ACI and Schneider relays will be used in the current project, which is a pump controller.  Relays will change state fewer than 100 times per day.  Power supply for the relays is a transformer, SquareD, M/N 9070TF50D23, primary 120/240 VAC (using 120 VAC arrangement), secondary 24 VAC, 50 VA.

In the case of the ACI contactor, coil has a sealed VA of 12.  At 24 VAC, that's 0.5 A.  Coil inductance is 0.114 H when activated.  So using the formula 0.5*L*I^2, that's  0.5*0.114*0.5^2 =  0.0142 watt*seconds stored energy.  So, how does this relate to the power rating of the TVS diode?  The max clamping voltage is 77 and wattage is 600, so current limit is 7.8 A.  The energy, hopefully, will be expended by heating the coil as opposed to heating the TVS diode or other components in the circuit.  So the coil resistance plays a role in dissipating the energy.  Coil resistance is VA/A^2  = 24/0.5^2 = 96 ohms.   Once power is removed from the coil and voltage is clamped at 77 volts, current will be 77/96 = 0.80 A ... well under 7.8 A.  The TVS diode should be quite comfortable.

As long as voltage stays at 77, wattage will be  77*0.80 =  62.   At that rate of energy expenditure, the whole lot will be expended in 0.0142/62 = 0.000229 sec or 229 usec ... pretty quick.  Will actually be somewhat longer because of coil reactance, and some of the energy will be expended after the voltage drops below 77.

For comparison, the Schneider relay has a sealed current of 1.4/24 = 0.06 A and a stored energy of 0.5*0.271*0.06^2 =  0.000488 watt seconds.  Comparing with the ACI contactor, notice how the higher inductance is way more than offset by the lower current.   The coil resistance is even greater than that of the ACI (1.4/0.06^2 = 389 ohms), so the wattage expenditure at the clamping voltage will be even less (77^2/389 = 15).   Time to dissipate energy will be about 0.000229/15 = 0.000015 sec or 15 usec.

I may be starting to get the hang of this.

Wolfram:  I'm trying to conceptualize why the supply voltage and clamping voltages should be added to get the voltage spike max.  Wouldn't the supply voltage disappear when the switch controlling the relay is opened?  At that point, the only energy source is the magnetic field around the coil.

 Mike in California

[David Hess]

Use ordinary diode, 1N4004 for small relays, not a transient absorbing Zener

The low forward voltage drop of a normal diode extends the decay time of the inductor current which delays and slows opening of the relay contacts increasing damage to the contacts.  The whole purpose of using a zener or avalanche diode to control the flyback pulse is to increase the voltage across the inductor during flyback so that the magnetic field collapses more quickly.

[trobbins]

Do you know the inrush of the motor controller and whether the controller has some form of soft-start of the motor, or is this a direct-on-line motor start situation?

Is the relay contact for single or DC power connection to the motor controller, and is just one pole or multiple poles used?

Is the motor controller 'under load' or otherwise require a clean disconnection from power source?

[BravoV]

... slow the opening of the contacts leading to arcing and premature wear. ...

On preserving contacts ... abit related, a cap (+resistor) across the switch form a snubber that suppresses arc and prolong relay contact life -> Keysight AppNote : Maximizing the Relay Life Span

PS : It used to be a direct download for the "5988-6917EN.pdf", now Keysight needs registration. 


Edit :

Alternate download link ->

https://www.newark.com/pdfs/techarticles/agilent/MaximizingtheLifeSpanYourRelays.pdf

[T3sl4co1l]

Heh, interesting case I'm working with, a customer has relay contacts that often switch inductive loads (larger relays, solenoids) on a 24V DC system.  They're getting random resets when the contacts open.  Okay, sounds like EFT, let's put an RC across them.  Wait, there's already 4.7R + 1uF there?  That should slow things nicely...

As it happens, the 1uF's are failing shorted, possibly from manufacture (we've had issues with probably-cracked caps elsewhere on the board), but much more likely it's from overvoltage.  The solenoids are probably something like 10mH, so at a couple amperes, they store a good slug of energy, compared to that poor ceramic chip cap.  The 4.7R then burns open because it can't handle the load current.  End result, no RC, contacts arc.

All of this is made more awkward on account of the assembly being fully potted.

Should be fine with something that can absorb the energy -- a TVS or MOV.

In short, the biggest problem with an RC (or other snubber network), the value must be suited to the load.  If you can't account for the worst-case load, you will eventually encounter problems.  (For AC loads, part of the limitation is leakage across the switch -- sure you can put on like 1R + 100uF, but it's practically never going to open.  The leakage eventually goes away for DC loads, but the inrush as the RC discharges through the contacts can be just as problematic!)

Conversely, a TVS or MOV is inflexible with respect to voltage.  In an automotive environment, you might be fine with a 18V TVS, in the average case, but what about load dump?  At mains voltage, you might be fine with a 360V MOV, but what about surge?  If it's acceptable for the load to turn on only momentarily, you can place the limiter across the contacts.  If not, you may need transient limiting at the source as well as contacts (a trio of MOVs between source, load and contacts, perhaps), or something more complicated to address the situation.

Tim

[Yansi]

For an AC relay coil, add a capacitor in parallel. One that gets the relay+capacitor close to a unity power factor is a good start.

Pretty bad idea, actaully - at least for the other contacts switching the coil with the cap in parallel.

Use series RC instead.

[calzap]

trobbins,

Motor is 240 VAC, 1 Ph, 3 HP, direct-to-mains via 2 poles of the ACI contactor.  A description of a failed example of the motor is here:
https://www.eevblog.com/forum/projects/pump-motor-postmortem/msg2856420/#msg2856420

Motor is driving a two-stage centrifugal pump ... usually starting against 45 PSI pressure.  In-rush for ACI contactor is 70 VA.   System receives heavy use 8 months of the year.  In 11 years of operation, just about everything has failed at least once.  Biggest problem was repeated failure of the line voltage monitor that protects other controller components and  the pump.   Pretty much solved that by increasing ventilation of cabinet and increasing spike protection on the line-side of the voltage monitor.  But you can only do so much to shield from the sins of the power company.

Mike in California

[trobbins]

Hi Mike, always good to appreciate the scope of the application, and not just zero in on a particular point.

When you say in-rush of contactor is 70VA, are you referring to the contactor coil or the 2 pole contacts that feed the motor?   Have you assessed or measured the motor full-load or locked rotor current, and related that to the contactor contact rating for say AC3 specification ?

As a first-in comment, it may be that the contactor contacts are the weak link in the chain and no tweaking of how the coil is managed will remove that weak link, or significantly alleviate the stress on the contacts (other than to insert an electronic motor drive in to a problematic high temperature environment).

Are you using a mains monitor relay with contacts that are used to turn off the contactor coil (and hence the motor) and other sensitive devices when mains AC voltage is either too high or too low ?  I've come across differences in how much internal power dissipation occurs in different such monitors (especially at 60Hz), and the max ambient temp rating for the devices - and had to swap over to another brand for 60Hz operation in one application.

Ciao, Tim

[calzap]

Tim,

Thanks for your comments.

70 VA is the in-rush of the contactor coil (24 VAC).   Contactor specs meet those of the motor for HP,  FLA and LRA.  In 11 years, the contactor failed once (after a couple of years).  That's when I noticed the 24 VAC transformer had been spec'd for the sealed VA of the contactor rather than in-rush VA.  Put in a bigger transformer, and no contactor problems since.  I'll  never say who specified the too small transformer. 

The voltage monitor is an ICM Controls ICM491.  It monitors the 240 VAC supply from the power company for high/low.  If supply is out of range, the monitor allows no 24 VAC to reach the contactor coil.   This component was failing every other year.  Finally, I put a couple of MOVs and a couple of RC snubbers across its 240 V terminals and increased ventilation through controller cabinet.   Seems to have worked ... no recent failures ... knock on wood!

Right now there are no real problems with the system.  But the controller is being upgraded to have additional safety features for the motor and pump.  These include flow switch on the intake pipe, float switch on the intake pipe (in a side-arm), temperature sensors on the motor and the pump, and water detector under the pipes.  Maybe an airflow sensor for the motor.  These sensors will interact with the 24 VAC supply to the contactor via relays.  Hence, my interest in TVS diodes for flyback.  Everything will be on DIN rail with indicator LEDs and easy means of bypassing individual sensors for diagnostic purposes.

Mike

[DBecker]

Conversely, a TVS or MOV is inflexible with respect to voltage.  In an automotive environment, you might be fine with a 18V TVS, in the average case, but what about load dump?  At mains voltage, you might be fine with a 360V MOV, but what about surge?  If it's acceptable for the load to turn on only momentarily, you can place the limiter across the contacts.  If not, you may need transient limiting at the source as well as contacts (a trio of MOVs between source, load and contacts, perhaps), or something more complicated to address the situation.

Automotive inductive outputs have changed from snubber networks, which are large, expensive and (slightly) unreliable, to clamped protected drivers.  These have a "clamp" that turns on the transistor slightly at 40V-70V e.g. a zener diode to the gate to limit the shut-off voltage excursion.

Typically relay drivers are clamped at 40V, and injectors/actuators at 70V.  The result is a clean, fast turn-off of the injector or relay, with little risk of the driver being damaged by over-voltage if the inductance is higher than expected or from a snubber capacitor degrading.

This clamped driver approach works best with low voltage DC.  Where it does apply, it's far superior to snubber networks and much easier to design with.

[trobbins]

Good stuff Mike.  Yes some level of protection/filtering of the 240Vac coming in to the control devices (but not the motor feed itself) would seem very appropriate, to help constrain stress on local control devices.  Whether that also supresses any local disturbances such as from the control relay for the contactor coil I guess depends on the type of protection/filtering deployed.  Did you have something in mind that may add some transient impedance between the main AC feed and the control AC supply, and protect the control AC side (such as MOV or TVS device) ?

Yes, nothing like some extra process protection devices to remove system and pump/motor stresses for the known about faults/blockages that can occur over years/decades.  Those stresses could well be a major cause of past failures.

The motor itself may be inducing transients, both from it's own inductive energy, and from line inductance as motor current starts/stops, so some form of MOV module with health indicator and even fault contacts, would appear reasonable when connected across main incoming connection (live/neutral/PE), although such protective devices would also bear the brunt of mains feed transients being drawn in to your location.

With respect to 24VAC control, are other devices being powered from that other than the contactor coil?  I guess your protection contacts are.   A TVS has a much more controlled operating range, so a better way to go than a MOV if coil energy is to be constrained right at the AC coil, but maybe even a TVS needs to be softened so as to avoid TVS damage if mains supply tolerance and surges etc get through to the 24VAC level.

[Circlotron]

Have you considered using an opto coupled triac for switching the contactor coil? They switch off at the zero current crossing so the energy stored in the coil is essentially zero at that point. Snubbing it should be very easy.

[calzap]

The SquareD transformer that supplies the 24 VAC is pretty hefty and robust for its VA.  It's fused on both primary and secondary sides.  Short of a lightning strike on the lines, I don't think the power company can hurt it.  It probably acts at a pretty good choke on spikes from the input side as well.  But it would be a good  idea to put transient protection on its output as well.  So, I'll do that.

Circlotron:  I know essentially nothing about practical use of opto-coupled triacs.   I'll have to educate myself.  Thanks for the suggestion.

Mike in California

[NiHaoMike]

I know essentially nothing about practical use of opto-coupled triacs.   I'll have to educate myself.  Thanks for the suggestion.

Those are better known as SSRs.

[calzap]

I know essentially nothing about practical use of opto-coupled triacs.   I'll have to educate myself.  Thanks for the suggestion.

Those are better known as SSRs.

OK.  I was sleepy and should have realized that.     Those I've used before.

[DBecker]

I know essentially nothing about practical use of opto-coupled triacs.   I'll have to educate myself.  Thanks for the suggestion.

Those are better known as SSRs.

'SSR' usually refers to an assembly that includes an opto-isolated triac triggering a power triac, along with a handful of other components (input resistor, noise filtering, snubber).

[ffrige]

@calzap: just curious, did you end up putting those TVS diodes parallel to the load or parallel to the switch? I wonder if the second option is better, because it keeps the voltage across the switch lower. Thanks!

[wraper]

Use ordinary diode, 1N4004 for small relays, not a transient absorbing Zener

Jon

Using regular diode has a huge downside. Relay turns off much slower, thus you will get way more contact arcing/ wear. https://www.te.com/commerce/DocumentDelivery/DDEController?Action=srchrtrv&DocNm=13C3311_AppNote&DocType=CS&DocLang=EN