What are my options for thermal protection?

[sparkydog]

I've been looking into how I might protect a project (that's going to be inside a sealed, metal box) from overheating.

I've seen a lot of references to "thermal fuses" but actually finding components has proven a little more difficult.

This particular project is a relay-controlled load (that generates a fair bit of heat on its own under normal operation) plus some cooling fans, a 12V-to-5V LDO regulator, and an MCU (ATtiny) with built-in AC-to-DC (probably going to use an IRM-10-12). I could rig something with some form of thermal sensor, a bunch of resistors, and a comparator that would interrupt the relay, but I'm very limited on board space and couldn't fit this without making the board bigger. (Which might be possible, but certainly isn't ideal.)

(Note: the ATtiny is only mentioned because it's part of the list of "stuff generating heat"; all of the pins are already used for other things. It isn't available to help with thermal protection, and in particular there is no way to use it to trigger an interrupt on the AC side of things.)

What I'd really prefer is a single component that is either reusable (resettable), or user-replaceable. However, I've only found two possible components: leaded fuses that would have to be desoldered to be replaced if they ever trip (and, at least during the prototyping phase, that's not ideal), or parts made for battery systems that are only rated for at best 9V.

Another option I looked at is a "thermal breaker" (I have a teensy bit more room for something wired directly into the AC that isn't actually on a PCB), but despite the name, those seem to trip based on current, not temperature.

I do have current fuses on both the AC side (0.25 A per the PSU's spec) and load side (also 0.25 A / 3 W, IIRC), both in sockets/holders so they are easily replaceable. Should I even be worrying about thermal protection or is it exceedingly likely that I'd trip one of the current fuses first? Is it better to break the AC side (so that the PSU is also protected), or the load side (so that the fans can keep running)? Any recommendations for something that doesn't need a solder-job to replace if it trips?

[TimFox]

Depending on how fast a response you need, and if you have a surface whose temperature is a good signature for thermal protection, you might try a "Klixon" temperature switch.
It has two terminals, has substantial hysteresis between high and low temperature thresholds, and is available with a large discrete set of setpoints.
Both normally-open and normally-closed switches are available.
Above 0^o F, the stock values are spaced by 10 F^o, and the range is from -65^o to +200^oF.
I have used them on heatsinks for power transistors to remove power or load when the temperature is too high, and the manufacturer of my gas furnace uses them as thermostats for cycling the fan, etc.
https://www.sensata.com/products/temperature-sensors-and-switches/precision-thermostats  is the current vendor, replacing Honeywell.
https://www.sensata.com/sites/default/files/a/sensata-4344%20series-hermetic%20sealed%20thermostats-datasheet_2.pdf  for a specific series.

[luudee]

Couple of options come to mind:

1. Prevent overheating in the first place
Since this is in a metal box, use it as a giant heat sink. Make sure
the components that generate the heat, make good physical contact
with the box.

2. If you have a I2C bus, you could use something like a LM75 to
measure temperature, and have the ATiny decide when to power
things of.

3. Last but not least, use a NTC with an OpAmp to control a switch.

Good Luck!

luudee

[Vovk_Z]

What I'd really prefer is a single component that is either reusable (resettable), or user-replaceable.

Look at LSD9700 KSD9700 series of self-resettable fuses. It has a metal case. And there is similar type but with a plastic case - BH-B2D or BH-B-B2D. There are another similar series: B-1009N, ST-22.
They are made with NO and NC contacts, so you may chose what is more convenient for your design.
There are different form-factors, e.g. TO-220 case BTC-220 series, etc.
It depends on a rated current too, e.g. KSD301 series for large currents - 10 A rated.

[sparkydog]

1. Prevent overheating in the first place Since this is in a metal box, use it as a giant heat sink. Make sure the components that generate the heat, make good physical contact with the box.

Well, yes, that's the hope! I think "good contact" should be straight forward (note to self: consider using thermal paste / pads / compound), but, while I should have lots of mass, I'm not sure how well it will radiate. (My heat load is continuous over potentially many hours, so I expect it to hit thermal equilibrium.) Unfortunately, the best place to put heat sinks is inside the box, pointed either downward, or sideways on the side opposite the exhaust vent, neither of which is ideal. (Plus, said locations are exactly where I'd be putting a thermostat, so I'm probably looking at a "choose one and only one" situation.)

On the plus side, I think most of my heat generation will have a slab of metal between it and my electronics. (Which won't keep the load from cooking itself, but at least the fire risk if it does should be lower than otherwise.)

2. If you have a I2C bus, you could use something like a LM75 to measure temperature, and have the ATiny decide when to power things of.

3. Last but not least, use a NTC with an OpAmp to control a switch.

I don't have free pins on the ATtiny I'm already using, and for a safety system, I'd prefer sticking with a plain comparator (I think that's what you mean for (3)?) for the triggering device. However, I don't have board space for either of those options.

I did finally manage to dredge up the 67L080. Unlike the stuff meant for batteries, it's rated for up to 48 V. For now, I've slapped one of those on the load side; if it trips, it will cut out not only the load, but will also open the relay (and it won't re-close on its own, so the load will stay off even if the 67L080 resets). The plan is for it to just be waving in the air, out of the direct path of my fans. IIUC, the load needs to be kept under 85°C, so this isn't exactly protecting that (for that, I'd need something in contact with the case, if not the load itself), but I'm hoping that will at least guard against the PCB catching fire, and somewhat protect the 75°C rated PSU? (That said, if (1) works well enough — and I hope to do some testing to that end before I've committed to building the whole project — I'll use a 67L070 or lower.)

Look at [​list of parts].

Right. Of those I could actually find, they look to be essentially the same as the "Klixon"s Tim mentioned... probably what I'm looking for, but figuring out how to order them looks like fun, unless I'm willing to go to Amazon or eBay. But... yeah, that seems to be the best option for AC side protection and/or reacting to case surface temperature. (Do you have a link for "LSD9700"? I failed to find that with either Google or Octopart.)

Apparently "snap action thermostat" is the magic query. What do you think of the NTE DTO series, e.g. DTO170? They're cheap and seems to be exactly what I need... and have clear part numbers with readily available suppliers; things that seems to be lacking from Honeywell and Sensata. (I might prefer a Sensata, but if I can find part numbers at all, they're coming up in the $100 range or higher. Honeywells range from $15 to $50 but they have way too many part numbers that I haven't been able to figure out what's different, and supply seems to be dodgy.)

[TimFox]

The DTE series from NTE seems to be a Celsius-calibrated equivalent to the Klixons, although they do not seem to have the very low-temperature (< 0^o F) versions.
https://www.nteinc.com/circuit_protect/disc_thermostats.pdf
DigiKey's "Marketplace" channel will ship from NTE at $2.74 in singles.

[Vovk_Z]

(Do you have a link for "LSD9700"? I failed to find that with either Google or Octopart.)

It has to be KSD9700 but not LSD9700, that's my mistake.
Just google '9700 thermal relay'.

[Jeroen3]

Maybe some PTC's just like in motor winding protection? You can put those in series to protection multiple points in your design.
The resistance just goes up steep at a certain point. Thermal trip doesn't care about accuracy.

You can use trivial and small analogs to make it work, maybe even an TL431, and then hold down the ATTiny reset line?

[sparkydog]

Maybe some PTC's just like in motor winding protection? You can put those in series to protection multiple points in your design.
The resistance just goes up steep at a certain point. Thermal trip doesn't care about accuracy.

You can use trivial and small analogs to make it work, maybe even an TL431, and then hold down the ATTiny reset line?

The trouble with a PTC is that it isn't an on-or-off device... and it turns into a heater in its own right when it "trips". In order to use something like that, I'd also need another transistor, as well as something (or multiple components!) to turn the PTC's linear function into a digital signal.

The ATtiny doesn't control my load... hmm... the ATtiny puts out a 'healthy' signal that does cut the relay if it goes low / open... but relying on the MCU reset to bring that low on reset seems riskier than just interrupting said signal. To be fair, I hadn't thought of before, and I could use one of those those battery-doohickey breakers for that... except, at 8.0 mm × 3.2 mm, it won't fit on the MCU daughter board. So, it's not much improvement over interrupting the relay / load directly with a 67L080. 🤷

[Infraviolet]

"couldn't fit this without making the board bigger"
What about the other side of the board? Might not be ideal, but sometimes making some space on the undersidefor one or two extra parts is the easiest way round, double sided boards can still be reflowed by using an adhesive as well as solder paste when populating them.

Then you could have a temperature sensor of some form, with a comparator, extra ATTiny, or some other means monitoring its analogue output and cuting off power if it goes too high for too long. If you want it to be fast acting I think your issue is mroe about how rapdily the temperature sensor responds to heat flowing from somewhere than how fast other things act from the electrical signal representing the temperature. Might need to put the temperature sensor at the end of a wire so you can have it thermal-pasted right on to the parts you want to check the temperature of.

Also, for monitoring that irst ATTiny specifically, don't they have an internal temperature sensor "wired" to be a selectable source for the ADC? ATMEGA328 chips certainly do.

[sparkydog]

Oh, good! It looks like 1NT01L-7939 is a viable alternative to DTO170...

Is it better to break the AC side (so that the PSU is also protected), or the load side (so that the fans can keep running)?

No one seems to have said anything on this... 😢

Actually, a related question would be what makes more sense as a trip threshold at each side? There are two failure cases; PSU overheating / something on fire (air temperature matters), and load overheating (case surface temperature matters). There are two points at which I can interrupt power; the load, and the whole thing. Which combination(s) make(s) sense? (And how worried am I about the "something on fire" possibility if I kill the load at, say, a case temperature of 80°C? Given active cooling, is that low enough for the PSU that's only rated to 75°C?)

"couldn't fit this without making the board bigger"
What about the other side of the board?

The other side is full of traces. 🙂

What, you want my traces on the front with the components? Okay, the other side is full of ground plane¹. 🙂 (I really do have all the traces on the back. This particular board is 100% THT², so it doesn't really matter which side they're on, except that being on the unpopulated back means they're just as easy to inspect with the board populated as unpopulated.)

More seriously, it's a few mm from the case and has zero airflow; I'd be worried about anything there getting cooked, and I'd still have to dodge lots of THT pins. Well, okay, there's maybe space under the relay and PSU... but a 67L0xx is a single component that physically breaks the circuit. Squeezing one of those in was actually pretty easy, not to mention that the schematic hardly notices the change. Simple is good!

Hmm, actually, now I'm debating mounting that on the back.

(¹ ...and, no, I don't really want to pay 250% more, best case, for a 4-layer board. That's with JLC; the markup from PCBWay is over 400%.)

(² Almost half the components either aren't available as SMD; are headers, for which I wouldn't use SMD versions for mechanical reasons anyway; or just aren't really an improvement over the THT components. In some cases, having pads further apart actually helps with routing.)

Also, for monitoring that irst ATTiny specifically, don't they have an internal temperature sensor "wired" to be a selectable source for the ADC? ATMEGA328 chips certainly do.

They do! Here is someone else's code for using it. Alas, my poor MCU is busy wrangling interrupts from not one, but two fans, and that code is apparently pretty slow. Still, I don't suppose there's any reason I need to use every possible tach interval to adjust the fans' PWM; alternating between PWM adjustment and reading the temperature is probably fine. It's an interesting idea I may look into as a backup, or even just because the fan controller having an idea of the temperature is obviously useful for other reasons.