[Blabber] Thermal runaway

John O'Brien jweob at cantab.net
Mon May 30 14:48:16 UTC 2016


Thanks for this James. I am going to follow your suggestion and use an
arduino clone to measure a second thermistor. I wanted to avoid anything
that relied on firmware, but since it is a separate system then I guess it
is not too bad.

I think I could have used an op-amp in a comparator circuit setup, but I
would have needed a +5V and -5V supply to be able to compare against the
0.4V.

On Sun, May 29, 2016 at 10:37 PM, James Carpino via Blabber <
blabber at list.hackmanhattan.com> wrote:

> I'm unanimous in recommending that you use a second sensor.  Otherwise, it
> isn't a failsafe, and a loose connection or failure of the original
> thermistor would lead to a failure of the failsafe.
>
> Assuming you have a separate thermistor:
> I'm pretty sure you can tweak the Schmidt trigger circuit to detect in a
> low range by adding in some voltage through a resistor from another part of
> the circuit, but I'd have to put my thinking cap on.  It might be as simple
> as a weak resistor to +5V, to change the curve coming from the thermistor
> voltage divider.
>
> But more important, keep in mind, that input is not strictly high
> impedance, and is going to pull down your thermistor value toward 0.6V
> through a resistance of about 183.9k.  That's going to severely tamper with
> the reading, since the standard thermistor is 100k.  You'd be better off
> using an op-amp to buffer that before it gets to your Schmidt trigger.  If
> you're going that far, though, you might consider a cheap ATMEGA instead
> (some arduino clones run $2-$4) and then you can then use the same
> thermistor tables as the Marlin firmware.
>
> You could also flip the thermistor divider, pulling high instead of low.
> (again, not using the same one as the Melzi).  Might want to think that
> through, in case one configuration or the other will automatically trigger
> the failsafe during power-on.
>
> Dichotomous decisions hurt my brain.
>
> James
>
>
>
> ------------------------------
> *From:* John O'Brien <jweob at cantab.net>
> *To:* Hack Manhattan! <blabber at list.hackmanhattan.com>
> *Sent:* Sunday, May 29, 2016 6:23 PM
>
> *Subject:* Re: [Blabber] Thermal runaway
>
> Following on from my printer hotend meltdown disaster, I'm trying to build
> an analog control circuit that will switch off a relay controlling the
> printer's power supply if the hotend goes above a certain temperature. If
> possible I would like to use the existing temperature control thermistor
> that is connected to the Melzi board. However I'm struggling with the
> electronic design for the control circuit. Would anyone be able to give me
> some pointers?
>
> On the input side of my planned circuit I've measured the voltage across
> the thermistor at different temperatures. This voltage is being supplied by
> the Melzi board, which biases the themistor in such a way that it gets a
> voltage it can read using the microcontroller's analog pins. At 25C it is
> 4.74V, at 200C (operating temperature) it is 0.68V. My planned cutoff temp
> is 240C. At that temperature the voltage is 0.43V.
>
> On the output side I have a relay module like this
> <http://myhowtosandprojects.blogspot.com/2014/02/sainsmart-2-channel-5v-relay-arduino.html> which
> should get a low signal (using 5V logic) to switch off the printer's 19V
> power supply. The control circuit would be on a separate 5V power supply.
>
> The desired behaviour is for the control circuit to switch off the relay
> once the thermistor voltage drops below 0.43V. It should then stay off
> until the control circuit is reset (i.e. cycle its power).
>
> My plan was to use a Schmitt trigger built from two 2N3904 BJT transistors
> (because I have them handy). I chose a Schmitt trigger because the
> hysteresis allows it to stay latched off once the threshold is reached. I
> followed the circuit diagram and design flow in this link
> <http://www.johnhearfield.com/Eng/Schmitt.htm>, but I've just realized
> that the approach won't work because the threshold voltage I need to switch
> on (0.43V) is lower than the Base-Emitter voltage needed to put the
> transistor in its active region (I think around 0.6V). How can I make a
> switch that will activate at a threshold that is this low? Or will I have
> to use a second thermistor so that I can bias it at a suitable point?
>
> On Sun, May 22, 2016 at 3:07 PM, John O'Brien <jweob at cantab.net> wrote:
>
> Thanks for the suggestions and sympathy!
>
> Guan I'll try your idea to use the existing thermistor for my cutoff
> circuit.
>
> Brad I have a raspberry pi running octopi that controls my Melzi.  I think
> I could do something similar to your ethernet idea by having the raspberry
> pi shut off the DC power to the printer using a relay if it stops getting
> responses from the Melzi.
>
> James I have a thermal cutoff like the one you linked to with a failure
> temperature of 240C. I need to work out a good place to mount it. However
> you said that this didn't solve the overall problem. I don't understand
> this part - if I run the heater cartridge in series with the thermal fuse
> then there is no reliance on the microprocessor?
>
> By the way I forgot to say that the board itself seemed ok after
> rebooting. I had to connect a working thermistor to it first in order for
> it to connect to the host, but I think that is because a short or open
> circuit on the thermistor triggers the firmware thermal runaway protection.
> Still not sure what caused it to lock up originally, so I will be cautious.
>
>
>
> On Sat, May 21, 2016 at 5:02 PM, Gilbert Gjersvik <gilbertg at gmail.com>
> wrote:
>
> The movie I'm going to watch tonight: The Straight Story, 1999, 1 hr. 51
> min
>
> On Sat, May 21, 2016 at 4:55 PM, James Carpino via Blabber <
> blabber at list.hackmanhattan.com> wrote:
>
> I want to amend that, and say explicitly that kapton/polyimide tape DOES
> melt at high temperature and may not prevent thermal runaway.
>
> There are several possible failure modes (that I may or may not have
> experienced) that can lead to an unexpected short in the head 12V->5V.
> Ideally the whole thing would be optoisolated, which still would not
> preclude runaway.  But current designs are not optoisolated,
>
> One scenario is that a thermistor wire is intermittent.  Occasional
> disconnect reads higher measured resistance than it should, indicating
> lower temperature (in the case of NTC, the most common in 3dp), causing the
> controller to heat up the heater more (incorrectly), which can lead to
> failure of the heating element's casing and other insulators, melting of
> teflon wire insulators, and melting of polyimide and any other insulators.
>
> Another scenario is any failure of the CPU leaving the system in an
> indeterminate state with the heaters enabled (even temporary).  That would
> lead to everything melting down including physical insulation.  Well, I
> think I already said this, but I needed to reiterate, just to make sure
> nobody believes I'm saying kapton tape will make things safe.
>
> James
>
>
> ------------------------------
> *From:* James Carpino via Blabber <blabber at list.hackmanhattan.com>
> *To:* Hack Manhattan! <blabber at list.hackmanhattan.com>
> *Cc:* James Carpino <jamescarpino at yahoo.com>
> *Sent:* Saturday, May 21, 2016 4:24 PM
>
> *Subject:* Re: [Blabber] Thermal runaway
>
> I'm sure it'd be easy to place a fixed temperature cutoff in series right
> there.
>
> http://www.digikey.com/product-detail/en/cantherm/SDF-DF240S/317-1142-ND/1014771
>
> But that doesn't solve the bigger problem. This situation has emerged
> because people design technology thinking that a single, central CPU can
> manage every aspect of a small device.
>
> There's also an option in Marlin to have two temperature sensors.  But it
> sounds like Marlin crashed, or the cpu or the rest of the controller
> failed, and that means all bets are off.  To me that's the bigger problem.
> We can say these things need to be monitored, but plenty of devices use
> much more severe amounts of heat and fire, like a furnace for example, and
> are customarily left unattended.  This little thing is much smaller than a
> furnace, and equivalent to a 40w light bulb in power output, and I believe
> it should be possible to make it safe.  But with all the clone parts and
> open design and bargain assembly, it's not.
>
> One very common problem is that the temperature sensor shorts to the high
> side of the 12V to the heating element.  This happens by mechanical
> failure- sometimes a screw holds a thermistor in an aluminum block, and
> that screw displaces some teflon insulation.  At the same time, there has
> to be a simultaneous failure in the heating block allowing 12V to get to
> it.  When this happens (as is remarkably often with all the print head
> motion and hackedness), 12V gets fed back into the thermistor's analog
> line, blowing the ATMEGA2560 or whatever, and probably blowing the rest of
> the board and regardless, locking the power circuitry into the ON position.
>
> That specific cause can be avoided by high quality manufacturing.  This
> hotend, for example,
>
> https://www.matterhackers.com/store/printer-accessories/b3-innovations-pico-hot-end-premium-bundle-3-nozzles-and-3-mounting-options-1.75mm
> has serious strain relief.  But it's expensive and I think the same
> prevention can be accomplished with copious use of kapton tape.  And
> anyway, if one Pololu module (or generic motor module) goes bad, it can do
> the same thing to the controller, feeding 12V into the 5V side.  Actually,
> any 5/12V failure can do it.  And it might be transient but recurring- I've
> seen a controller that functioned until it got too hot, and then when it
> got hot, the 5V supply went high (>5V). (I'm saying that might have been an
> intermittent motor module, or a regulator or any other component, I didn't
> track it down at that time.)
>
> Amazing fail.  I'm impressed.
>
> James
>
>
>
> ------------------------------
> *From:* Brad Shannon <bradjshannon at gmail.com>
> *To:* Hack Manhattan! <blabber at list.hackmanhattan.com>
> *Sent:* Saturday, May 21, 2016 3:56 PM
> *Subject:* Re: [Blabber] Thermal runaway
>
> Forgot to say, sorry about your printer! Glad it wasn't worse.
>
> On Saturday, May 21, 2016, Brad Shannon <bradjshannon at gmail.com> wrote:
>
> Some people set up thermal fuses to break and disconnect power in the case
> of thermal runaway.
>
> There are also little automatic fire extinguishers made for commercial
> kitchen fume hoods.
>
> If your electronics are on LAN, I bet you could also setup a
> LAN-controlled power strip ($~100) to ping the printer every N minutes and
> shut it off if there's no response. This is what I'm aiming to try, some
> day...
>
> On Saturday, May 21, 2016, Guan Yang <guan at yang.dk> wrote:
>
> The analog inputs on the boards should be high-impedance enough (or you
> can add a buffer) that you could just feed the thermistor output, in
> parallel, directly to a second board that can cut the power. That way you
> don't need to install a second thermistor.
>
> > On May 21, 2016, at 15:36, John O'Brien <jweob at cantab.net> wrote:
> >
> > Sharing this photo of what happens to a 3d printer hotend when the
> temperature control fails as a warning to anyone who (like me) leaves their
> printer running unsupervised.
> >
> > I left the printer running overnight and this is what I found in the
> morning. The heater block is made of aluminium, and the heat was so intense
> that it has visibly sagged and deformed.
> >
> > I have thermal runaway protection enabled in the version of Marlin I am
> running, but it seems like the board became unresponsive and got stuck with
> with the hot end on. Fortunately the heater cartridge must have burnt out
> at some point - I feel very lucky that this did not start a fire. The
> heated bed was also stuck on, but it was not powerful enough to reach a
> dangerous temperature.
> >
> > I am going to think through what safeguards I can put in to stop this
> from happening in future. Obviously the "correct" solution is to never to
> leave a printer unsupervised, but I know I will get complacent at some
> point. Any suggestions? Current plan is to install a second thermistor in
> the hotend attached to an analog circuit that will cut power to the printer
> above a certain temperature.
> > <IMG_5130.jpg>_______________________________________________
> > Blabber mailing list Blabber at list.hackmanhattan.com
> > https://list.hackmanhattan.com/listinfo/blabber
>
> _______________________________________________
> Blabber mailing list Blabber at list.hackmanhattan.com
> https://list.hackmanhattan.com/listinfo/blabber
>
>
>
> --
> brad shannon
> 202-997-5299
>
>
>
> --
> brad shannon
> 202-997-5299
>
>
> _______________________________________________
> Blabber mailing list Blabber at list.hackmanhattan.com
> https://list.hackmanhattan.com/listinfo/blabber
>
>
>
> _______________________________________________
> Blabber mailing list Blabber at list.hackmanhattan.com
> https://list.hackmanhattan.com/listinfo/blabber
>
>
>
> _______________________________________________
> Blabber mailing list Blabber at list.hackmanhattan.com
> https://list.hackmanhattan.com/listinfo/blabber
>
>
>
> _______________________________________________
> Blabber mailing list Blabber at list.hackmanhattan.com
> https://list.hackmanhattan.com/listinfo/blabber
>
>
>
>
> _______________________________________________
> Blabber mailing list Blabber at list.hackmanhattan.com
> https://list.hackmanhattan.com/listinfo/blabber
>
>
>
> _______________________________________________
> Blabber mailing list Blabber at list.hackmanhattan.com
> https://list.hackmanhattan.com/listinfo/blabber
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://list.hackmanhattan.com/pipermail/blabber/attachments/20160530/a1cdf0fe/attachment.html>


More information about the Blabber mailing list