Safest way to switch a resistive 2kW mains load?

[tld]

Hey,

I feel like it's easy enough to find a way to solve a problem, but it's a lot harder to settle on the way to solve something.  I suspect it's a learning-ability vs. experience thing, I'm sure I'm not alone in this.  With all things safety-related, the lack of experience worries me at times.

Now, I'd like to switch a 2kW space-heater, being about 9A @ 230Vac, that's enough to worry me, when it'll be running 24/7, unattended, when sleeping etc.

It would be easy enough to dive into it and make a circuit myself, using TRIACs probably, but I'm very tempted to use a zero-crossing solid state relay (SSR) instead.  Zero-crossing seeming nice for everything from EMI and load on the circuit, to connecting/disconnecting when carrying as little current as possible.

SSR over Triac is tempting to have a "package" that I didn't make myself, being able to just talk to it over optoisolated safe nice familiarly comfortable TTL-levels.

I emailed an SSR-vendor, asked about safety and failure-modes, and got back:

"Hi, the relays can fail either open or closed and they can catch on fire but that is not normal. They do not have any kind of over temp protection."

That's fair enough, and somewhat comforting, but not exactly obliterating any worries I might have.

I could use a 16A rated wireless lamp-switch thing, but it's really just a mechanical relay, and those can also fail over time, poor contacts leading to heating and whatnot.

Okay, I might be a bit paranoid here, but I'm curious about this for the educational value of it as well.

Any wise words on safes way to do this?

tld


(If I do end with an SSR, it'd be easy enough to include temperature-control of it in the low-voltage side, but if it fails short while overheating... )

[Rerouter]

though its a big resistive load, at those wattage's it picks up a fair amount of inductance through the way its built, so whatever your switching device you may want to include a snubber across its contacts to minimize a medium sized spike,

as for the switching element, no matter what it is, fit a fuse, this will ensure you never over shoot and helps if you risk it being shorted out,

[SeanB]

Trundle over to an electrical wholesalers and get a mains rated contactor, normally you can get a 24 VDC coil unit with a 4 pole contact arrangement with double break contacts rated at 16A resistive load that is not terribly expensive. This will fail safe ( open circuit) as it has double contacts, if you are feeling paranoid wire 2 of the contacts in series so you have 4 contacts to break the current. A little on the large side ( about the size of a closed fist with the clearances) but will last essentially forever, at least 10 years is easy to reach.

[Skimask]

Wanted to turn on the car's block heater remotely using a pair of cheap 433Mhz TX/RX units..
Used a servo connected to a pair of heavy duty lamp switches in parallel via a chunk of plastic with holes drilled thru the lamp switch "handles", powered the MCU/servo/receiver from off a wall wart on the input side, mounted the whole thing in a plastic box, sealed it up, added a 15amp fuse, good to go.  Programmed the MCU's to off the heater after a few hours, and only turn on the switches after receiving a long burst of the correct message from the transmitter, powering up the servo, moving it, then powering it down again.  Same thing with manually turning off the heater, but would shut off after receiving only one correct message, thereby making it 'fail-safe-er' than if it would've waited for the long burst.

And in this case, I believe it met all three requirements...Cheap, Fast, AND Good...although I'm sure every regulatory agency out there would've found a million-and-a-half things wrong with the design, it worked for me, and worked well...until I ran the box over! 

[Spawn]

I have been using different brand relays and temperature controllers over the years and I like to work with Telemecanique relays, I believe it is sold under name Schneider-Electric in the states.
They don’t come cheap but if you want complete solution check this catalogue, they have great temperature controllers (bit industrial I know) on page 15 you will find their solid state relays till 45A, so you can combine those with their temperature controllers.

http://static.schneider-electric.us/docs/Machine%20Control/Relays%20and%20Timers/Relays-Control%20and%20Measurement/REG%20Zelio%20Temperature%20Controllers/8430CT1001.pdf
Be warned it is 4.2mb

About being reliability of the solid state relays, I never seen one getting stuck in closed position, whenever I needed to replace one it was always in open position, but that doesn’t guarantee they can’t fail in closed position, it just what I have experienced, our solid states where 20A but I can guarantee you those last a lot longer than normal relays.

[SamD]

Trundle over to an electrical wholesalers and get a mains rated contactor, normally you can get a 24 VDC coil unit with a 4 pole contact arrangement with double break contacts rated at 16A resistive load that is not terribly expensive. This will fail safe ( open circuit) as it has double contacts, if you are feeling paranoid wire 2 of the contacts in series so you have 4 contacts to break the current. A little on the large side ( about the size of a closed fist with the clearances) but will last essentially forever, at least 10 years is easy to reach.

Sean,

We've disagreed… you hit the nail square… Straightforward, solid solution. 

Sam

[poorchava]

You could use A bridge rectifier + a mains-rated n-mosfet. On the bright side: such arrangement will usually fail open in case of overcurrent, so that's good. Downside is that a surge/spike conditions are likely to cause short-mode failure (but those can be prevented using MOVs/TVSs). Another downside is that it will need a small powersupply (may as well be transformerless capacitor-based one) to control the mosfet via an optocoupler.

[SeanB]

I have had optoswitched fail short circuit, like any other triac this is a common failure mode. They can also go intermittent and arc internally.

[tld]

Hey,

Many thanks to all of you!

To sum up, it seems like I have two good options:
 * Contactor
 * Accept that an SSR isn't perfect, and deal with that.

An SSR in a carefully thought through solution is looking a bit like a possible solution.  I could add multi-step temperature-protection, starting out with a DS18B20 connected to the MCU and/or a NTC/PTC pulling the control-line of the SSR to shut off if temperature rises, adding in a thermal-fuse to permanently open the circuit if my own temperature-control fails, or in the event that the SSR turns bad (fails short, with temperature-rise).  If nothing else, that should leave the circuit electrically dead if temperature gets out of hand.

Again, many thanks!

tld

[Paul Moir]

If the failure mode is catastrophic, a thermal fuse is a very good idea.  If you look at drip coffee makers for example, they'll often be fitted with several thermal fuses in case one doesn't go.  Far better to have a dead coffee maker from the occasional overly sensitive thermal fuse than to have a melt down in someone's home.

Mechanical contactors can fail shorted too; say the spring breaks and it's oriented so gravity will close it or you get contact welding for some reason.  Or a bug gets in it and causes heating that warps the contactor housing and prevents it from disengaging.  You could employ two contactors in series for redundancy but even that can't get you 100%.  And of course then there's your control circuit which no doubt has many ways it could fail contactor closed.

I guess what I'm saying is that for these situations you do a risk assessment and choose your level of protection based on it.  Sometimes you can't fail safe and so you have to go for redundancy.  Sometimes you use both if failures are bad but a fail-safe is possible.
 

[tld]

If you look at drip coffee makers for example, they'll often be fitted with several thermal fuses in case one doesn't go.

That's what I'm figuring.  Something like:

When all else fails, make sure the product is (electrically) dead.

SSRs and Contactors seem to be made of fire-retarding materials and the like, and as such shouldn't be the biggest risk, but if you keep adding energy (electricity), I'm thinking they could still very easily be a problem, so not doing that in times of crisis seems like a good thing.

Best way to add a thermal fuse though?

My current thinking is to run the mains-wire over the heatsink and back again, adding a (different) thermal-fuse in each direction.  That way there's two of them that can fail, and they're sure to not be from the same batch.  Epoxy them down to provide electrical isolation, yet secure good (or decent) thermal transfer to them from the heatsink.

Using an MCU-based thermothing as well (DS18B20) mounted in a similar manner, and I should be able to set things up so that I turn off the relay at 70-80C perhaps, and have one fuse open at about 120C, the other at 140C.

Other than the obvious "use extreme caution when soldering", and good advice on using these?

The L50 series from Cantherm seems easier to work with than the "normal" types, but also about 20 times more expensive.

L50: http://www.digikey.no/scripts/dksearch/dksus.dll?FV=fff4000a%2Cfff80150%2Cfffc013d&vendor=0&mnonly=0&newproducts=0&ptm=0&fid=0&quantity=0&chp=0&PV-5=33480

Again, thanks for all the help guys!

tld

[SeanB]

Most thermal fuses are crimped to the wiring, and then covered with a woven glass cloth shield, often a siliconised sleeving. They are normally not bonded but attached with a metal clip to a grounded surface. as well include an ordinary ceramic overcurrent fuse before the thermal one.

[tld]

Most thermal fuses are crimped to the wiring, and then covered with a woven glass cloth shield, often a siliconised sleeving. They are normally not bonded but attached with a metal clip to a grounded surface. as well include an ordinary ceramic overcurrent fuse before the thermal one.

Crimping does sound like a very good idea, and would solve a lot of issues, provided I can crimp securely and safely enough for the current.  On the up-side, if the crimps are not perfect, leading to a temperature-rise sometime in the future, well, there'll be a thermal-fuse there.

My first thought when you mentioned woven glass cloth shield was to wonder if that wouldn't hinder thermal-transfer to the fuse, and be a problem.  On second thought, I realized that that's probably the point, as the temperature-rise would get to the fuse through the wires?

As in a similar way, as long as I keep the fuse right next to the SSR, I might not need to struggle for perfect thermo-coupling to the heatsink, as the wires will do the job?

Crimping lends itself perfectly to quick-connects, which would make a lot of things a lot easier as well.

Heaps of thanks everyone.

Terje

[Paul Moir]

Crimping does sound like a very good idea, and would solve a lot of issues, provided I can crimp securely and safely enough for the current.

That should be no problem.  People field crimp connectors that handle a lot more than 9A every single day.  The only concern I would have is the temperature if you use copper crimp connectors, which can loosen in really hot environments.  But you should be well below the annealing temperature for copper.  If you were expecting elevated temperatures you can use brass or steel crimp connectors.  Most el-cheapos are steel; you can ID them with a magnet.

Don't worry about the fiberglass insulating the fuse.  It'll get to the temperature of what it's mounted to quickly.  Fiberglass on it's own isn't that great a thermal insulator; it's the trapped air in batts that does the trick.

[ejeffrey]

I can say that I know of several people with high end electric stoves with an SSR that have failed ON.  In two cases, that happened when the stove was off and (presumably) a voltage spike destroyed the SSR and caused it to fail short.  One of those people was on vacation at the time and it caused several thousand dollars worth of damage to his kitchen -- all the cabinets and floor near the oven had heat damage.  The same happened to my brother, but luckily he was in the room when his oven magically switched on and wouldn't turn off.  There are two morals to this story:  the UL are a bunch of incompetent hacks, and thermal fuses are good -- use them.

[nukie]

See how air con compressor works.

[smashedProton]

Use a crowbar! 

[TerraHertz]

The problem with SSRs, is they always include a snubber circuit. Which consists of a capacitor and resistor in series, across the triac.
Why that's potentially a problem, is that even when the triac is off, the snubber puts a small AC current through the load. Which means the load is never fully isolated from mains, and so if you want to do any work on it you have to remember that it's unsafe even when 'off', and isolate it somewhere else.

A contactor is definitely the way to go. SSRs are best used where rapid, long lifetime on-off cycling is needed, and mechanical contacts would wear out.

[David_AVD]

Why that's potentially a problem, is that even when the triac is off, the snubber puts a small AC current through the load. Which means the load is never fully isolated from mains, and so if you want to do any work on it you have to remember that it's unsafe even when 'off', and isolate it somewhere else.

Attempting to work on a circuit that you deem to be "not energised" (instead of physically disconnected) is just silly.  A relay and an SSR are no different from that perspective.