Short (1ms) high current (~200A) current pulses - relay rating?

[max_torque]

Hi All,

I am building a surge / spike generator for a project to meet a rare std for which you currently cannot just buy a stuitably rated COTS device (if there was one, i'd buy it!)

The spike generator basically charges up a capacitor to a voltage (typically around 250V) and dumps that into the power supply lines of the DUT.  Overall energy per pulse is reasonably low, around 200mJ. but the pulses must be applied repetitively (1Hz repetition rate)

I want to put in a suitable output isolating relay / contactor to ensure HV safety, ie the devices Estop must be set, otherwise the output terminals are physicaly disconnected from the energy storage system, which necessarily uses semi-conductor switching to meet the timing requirements (pulse slope/ length etc).


So, to the question, does anyone have experinence of electro mechnical relays used to carry such a pulse, and how to take a guess at the rating for that relay?  Note the relay does NOT open or close under any current, it must have a suitable opening gap, and maybe be of a "force guided" multipole archtitecture to allow contact weld detection.

It's hard to know exactly what the relays contacts will experience during a 200A peak 1ms pulse?  The overal heating is pretty low, so my thoughts are to find a relay with large contacts, so thermally they can experience the brief resistive heating and not overheat?

ideally the relay will be PCB mount btw   

[coromonadalix]

i would use  igbt  parts instead   .... you're gonna burn the relay contacts sooner or later

[Psi]

There's plenty of large relays that can switch 200A. Most of them are low voltage, like 12V for car starters. The arc from connecting/disconnecting 200A at 250V would destroy the contacts. But since you don't need to make/break 200A you can probably overrate the relay voltage a lot.
A 250A 12V relay might work fine, assuming the insulation & clearance is ok

Finding a 250A relay that's PCB mount is going to be tricky though, all the relays I've seen that will do that current are chassie mount.  Any relay that large mounted to a PCB is probably going to need to be socketed, (eg socket is pcbmount and not the relay). And then there's the issue of making a socket that can handle 200A. Probably why its not usually done.  Chassie mount just makes more sense.

You could have multiple smaller PCB mount relays in parallel and share the load between them, since you're not switching active power with them you have time for all of them to fully connect/disconnect before a change in current flow. That makes it easier to parallel them safely and you could use 250 volt rated relays.   
If doing E-stop it might be a little tricky, since you can't stop a pulse that is already underway and have to wait for the pulse to be over before cutting the power. But it sounds like the pulses are so short that wouldn't be an issue.
If you go down this multiple relay route you'll probably need those large press-fit M6 or M8 through-hole ring terminal blocks on the PCB to get the 200A into and out of the PCB.

It's defiantly much easier if you can drop the PCB mount requirement . 

[moffy]

We used to pass 1500A through a motor reversing contactor with no issues as long as it was zero current switching, shouldn't be a problem for a suitably rated contactor.

[mag_therm]

I peak: 200 Amp assumed rectangular
Time On: 1 ms
Period:  1000 ms
Switch duty: Off Load only
Overall run time of the pulse string: not stated **
Circuit Voltage: not stated **

From my Resistance Welding days:Calc:
Dominant heating frequency is approx 1/ (2*1e-3) : ~ 500 Hz
Duty_Cycle= 1/1000 = 0.001

Equiv_RMS_I= (200^2)*0.001 = 6.3 Amp (6.3 Amp AC at 500 Hz) Edited, was 40A and was corrected by jbb
This may be too much practically for a pcb relay and pcb header etc

For small single or double pole relays, look at , for example, https://deltrol-controls.com/sites/default/files/pdf/data-sheet/900%20Series%20102011.pdf
The basic contacts there are UL listed for 30 Amp and motor start
There is also a 50 Amp contact set rated at 50 Amp thermal continuous.

** You would need to rfq the relay supplier with your above ratings, and overall run time of the test, and the circuit voltage.

[SeanB]

Yes a 25A or 32A 4 pole contactor will work fine there, as the spike is not going to be long enough to give rise to thermal heating of the contacts, and they also do have 2kV isolation, and double acting contacts as well. Pretty much any industrial supplier has these off the shelf, DIN rail mount or using 2 screws, and a range of coils from 12VDC up, though you probably will want to use a 24VAC coil as that is reasonably common, cheap and provides good contact force.

[mag_therm]

Contacts WILL heat based approximately 0n the integral of the contact heating power:  energy= I_equiv_rms^2 * R_contact * Overall_Run_Time_Of_Test
That is why Max needs to include Overall_Run time in spec.

[Wolfram]

I would look at hydrogen contactors. Something like an LEV100 or LEV200 should do the job for less than 100 dollars. You can probably find Chinese equivalent parts for less than 30 if the application is price sensitive.

[Gyro]

The only problem I can see is that the contacts are going to be closing on a 'dry' circuit. Contacts of that rating are going to be made of some sort of Silver alloy material, such as Ag/Cd, which will build up a surface film (maybe only overnight). You can avoid this by adding a highish valve resistor to the GND rail to provide a wetting current prior to the high current pulse.

[ejeffrey]

I peak: 200 Amp assumed rectangular
Time On: 1 ms
Period:  1000 ms
Switch duty: Off Load only
Overall run time of the pulse string: not stated **
Circuit Voltage: not stated **

From my Resistance Welding days:Calc:
Dominant heating frequency is approx 1/ (2*1e-3) : ~ 500 Hz
Duty_Cycle= 1/1000 = 0.001

Equiv_RMS_I= (200^2)*0.001 = 40 Amp (40 Amp AC at 500 Hz)
This may be too much practically for a pcb relay and pcb header etc

Since the OP says it will be a capacitive discharge with a 200 mJ total pulse energy, I think the RMS current is going to be considerably less than 40 A, so a 20-30 A relay may be sufficient.

[ejeffrey]

In fact, if we take the stated pulse energy of 200 mJ, and peak voltage of 250 V, the capacitance is only ~6 uF.  At 200 amps peak, that implies a time constant under 10 microseconds, not 1 ms, and the RMS current / heating potential would be relatively tiny.

[max_torque]

Sorry, yes, 200A is the expected PEAK current of the pulse. As the system is oscilatory, the current bounces backwards and forwards in decreasing cycles between the holding capacitor and the impedance of the DUT.  Total energy, stored in a 8.2uF cap charged to 250Vdc is around 250mJ, but that energy will travel a few times over the relay contacts, depending on the complex impedance of the DUT etc.

The cap is discharged with a semi-conductor switch (back to back NFETs), this relay is simply a added mechanical break in the output path to ensure that the unit can be made "safe" as necessary.  I really want a PCB mount relay in order to keep the inductance low, as i have rate of rise targets to meet as well as the total energy release ones, so adding a remote relay with loops of wire is likely to add some significant inductance in the output path.

I might try to use faston terminals soldered to the pcb to robustly "socket" a standard automotive type high current relay.  I'm happy to replace / life this relay, the control system can integrate total energy relased and calculate a life value for the relay, beyond which the relay can be routinely changed.

I'm very familar with the typical high current DC rated contactors used for battery applications (i design BEV powertrains for a living) but i don't really want to add this level of cost, size and specific-ness if i can help it. Not having to switch the relay under any load, or use it to break the carrying current ought to make it's life a lot better.

I could parallel multiple smaller relays, and i do wonder about using a very small X class cap (small nF size) across the relay to carry the very short pulse energy part of the spike?  Would have to size that cap to ensure the emitted energy is low enough to not be hazardous under worst case (max operational voltage)


I think probably the best option will be to make a test pcb with a couple of relay options on it and see which one actually performs the best 

[mag_therm]

With 8.2 uF the calcs are  like ejeffrey's at 10 usec, and not like mine at 1 MILLIsec.

However the heating frequency of the 200 A pulses is now up around 50 kHz. (loop L of 1.2 uH  and R of say 0.5 Ohm would be about 10 microsec damped)
The current can not flow across the relay contact area, it will crowd into annulus of about 300 micrometre width.
Relay suppliers would probably refer to an RF relay -large, costly.

If this is only for safety compliance, can you add a small conventional pcb relay just switching, say, the power supply that charges the cap?

[jbb]

Hold on…

200 A * 1ms gives 200^2 * 0.001 = 40 Amps squared

So actually more like 6 Amps RMS. But you still want nice chunky connections to handle the peaks. And not add too much series resistance.

Edit: mag_therm raised a good point: as this is intended for safety, there may be other ways to achieve the result. They suggested disconnecting the capacitor charging supply. I’d like to extend that to 1) disconnecting the capacitor charger and 2) discharging the capacitor through a suitably sized discharge resistor

[mag_therm]

Thanks jbb, I corrected my error

The addition of one contact discharging through resistor is a good suggestion, as it is a safety relay.

[coppercone2]

1) mount to PCB sheet
2) add brass or bronze pins with staking and solder
3) have PCB mount part. either socket or solder on


The only hard part is 1) needs to be thick

[T3sl4co1l]

The pulse duration is so small I wouldn't worry about it.  Probably a 15A relay would suffice; these are typical currents for mains inrush for example (which is 1/2 or 1 or several cycles at mains frequency -- much longer duration), and comparable to normal (lightning induced) surge but usually not quite as long (that'd be more like a telecom 10/1000us pulse, but that isn't usually used with 2 ohm source impedance as mains would be).  You do occasionally see surge ratings on relays, which are usually in this range when seen, so, you might just shop around for one with adequate ratings and call it good.

The biggest risk I suppose, is poorly-settled/wiped contacts with high initial resistance, and then the pulse comes in and welds them together.  You might want two in series, or a higher reliability sort of contactor, or something that can read and confirm that the contacts have opened, or to monitor that the circuit has been deenergized.

Note that most mains contactors, motor starters, etc. only push the contacts together; independent springs open them back up.  One or more circuits can get stuck in place, while the armature returns to resting position.  Or the armature may get stuck in place, but not necessarily all contacts (including the side / sense / NC control contact) will be held in that case.  (I haven't seen this many times myself; I did once weld a contactor and the armature was stuck, but I didn't happen to check the state of all the contacts.  Designs may vary, anyway; most are of the contact-pair with armature pushing down a bridging part design, but there are probably others.)

Continuous plus switching rating is most certainly not required, but you can go the maximum-confidence route that way if needed.

Heh also fancy exotica, like, a mercury displacement relay comes to mind.  No contact to weld if the contact is already molten.  Usually used for soft and reliable switching (internal arcing can only produce more mercury vapor, at least until the housing melts or explodes), and obviously there's no contact wear to speak of.  Probably not worthwhile here, but may be of interest.

Tim

[max_torque]

Update: Change of plan! 

Going to use a crowbar relay that is not is series with the output, but in parallel with it. So when the Estop is hit (or power is removed from the device), not only will i de-power the switch mode convertor that generates the 250Vdc, i'll drop a crowbar across the output capacitor to ensure it is discharged and cannot be charged.  As the device will be connected to DUTs that run at up to 33vdc, i'll probably use some big 40V TVs diodes and a low value power resistor to quickly drop the output voltage below 50Vdc (as required by the Low Voltage directive)

This means my output path from the pulse caps is now just across the output control switch which is a semi-conductor switch

[johansen]

No contact to weld if the contact is already molten. 

heh.. heh.. no contacts to weld.

would it be a sin if i find out how much I^t it takes to make a mercury thermostat explode?

[jbb]

…
would it be a sin if i find out how much I^t it takes to make a mercury thermostat explode?

Your insurance company would probably think so…

[Smokey]

Altran AEV250 Series:
https://www.digikey.com/en/products/detail/altran-magnetics-llc/AEV250-M/9748544
https://mm.digikey.com/Volume0/opasdata/d220001/medias/docus/3717/AEV250_2.pdf

900V, 500A.

If you put big holes in your PCB, you can screw the studs right through the board, so I guess that's "PCB Mount".

[T3sl4co1l]

Altran AEV250 Series:
https://www.digikey.com/en/products/detail/altran-magnetics-llc/AEV250-M/9748544
https://mm.digikey.com/Volume0/opasdata/d220001/medias/docus/3717/AEV250_2.pdf

900V, 500A.

If you put big holes in your PCB, you can screw the studs right through the board, so I guess that's "PCB Mount".

The fun part about test equipment is, it's a one-off, we've collectively spent orders of magnitude more labor / project timeline discussing candidates, when a worst-case option is readily available for hardly an hour of shop time.

Tim

[Smokey]

Altran AEV250 Series:
https://www.digikey.com/en/products/detail/altran-magnetics-llc/AEV250-M/9748544
https://mm.digikey.com/Volume0/opasdata/d220001/medias/docus/3717/AEV250_2.pdf

900V, 500A.

If you put big holes in your PCB, you can screw the studs right through the board, so I guess that's "PCB Mount".

The fun part about test equipment is, it's a one-off, we've collectively spent orders of magnitude more labor / project timeline discussing candidates, when a worst-case option is readily available for hardly an hour of shop time.

Tim

You could have just asked me from the beginning

[Berni]

Update: Change of plan! 

Going to use a crowbar relay that is not is series with the output, but in parallel with it. So when the Estop is hit (or power is removed from the device), not only will i de-power the switch mode convertor that generates the 250Vdc, i'll drop a crowbar across the output capacitor to ensure it is discharged and cannot be charged.  As the device will be connected to DUTs that run at up to 33vdc, i'll probably use some big 40V TVs diodes and a low value power resistor to quickly drop the output voltage below 50Vdc (as required by the Low Voltage directive)

This means my output path from the pulse caps is now just across the output control switch which is a semi-conductor switch

That is indeed a better solution.

Tho the thing to watch out for is closing that relay while the capacitors are still very full. Be it because a wire going to your DUT fell off, or the DUT failed open quickly. Closing a relay directly across the charged capacitors has a very high chance of welding relay contacts with the huge initial pulse current. The relays are most vulnerable to being damaged during switching (due to switch bounce), once it is fully closed, it can handle very large peak currents.

Perhaps it is adequate for your solution to have the relay connect a low value resistor across the capacitors. Then pick a value that will discharge the capacitor with a few 10s of Amps just below what a regular small relay might survive without welding, yet have the resistor be low enough in value that the charging circuit is not powerful enough to charge up the capacitor to any meaningful voltage with the resistor applied.

Nice thing is also that if the relay welds, it fails stuck into the safe state

[max_torque]

Yup, looking to use something like a 40V bidirectional TVS (to prevent current flow when the output is connected to the DUTs normal supply voltage) and say a 10ohm power resistor (or resistor array).  Should limit peak discharge currents to 20A or so.  I'll also put  a high value discharge resistor across the caps that is permanently connected, and normally, the shut down / turn off will be handled by the control micro, that can turn off the boost converter and just wait a bit till the output cap has mostly discharged across that fixed R.  Only on the occasion where the Estop is pressed during normal test running or during supply failure (ie someones tripped over extension lead and yanked the kettle lead out the back........) will the crow bar circuit operate with the output cap charged.