It’s possible they’re using the NTC resistance of the heater coil itself to sense temp. The heater resistance drops slightly at higher temps. But I don’t think you need to – merely populating a “power vs temp” lookup table with empirical values per printer would get you close enough for basic purposes.
@Ryan_Carlyle It would have to be pretty well insulated for ambient temperature not to be a factor in that. Note that they not only have to know the plateau temperature, but also the rate of heating. Presumably you don’t want to wait forever for it to reach its maximum temperature for a given heater power, and would want to put in more power while heating from cold. Running it open-loop like that could get dangerous if the machine resets and assumes that it’s at room temperature when it’s still hot.
@Whosa_whatsis It definitely is well-insulated – the heater sits inside a big silicone shell, inside a plastic shroud. This a super simple first order process, so there ARE open-loop control schemes that would have no issues preheating and hitting a target +/-5C in the absence of major system disturbances. You just run full power for a predetermined period of time to get close, then switch to the calibrated power output, and call it good. That works fine when you print slow and don’t have major airflow changes. The reported temperature would then be predicted from inputs / history matching rather than measured.
That said, I don’t actually think they’re using open-loop control, because they do have some means of identifying that the heater is not receiving power. That will throw an error message. But the disassembly photos clearly show no thermistor in the wiring harness and the heater troubleshooting docs make no mention of temp sensors. So I think they’re either measuring heater resistance, or measuring heater current and performing a simple energy balance calculation to predict temperature. (More likely the former since most people aren’t process control nerds like I am.)
Barring those options, I’d have to look at something more exotic, like constructing the heater wires out of TC metal and then rapidly switching back and forth between sending power and sensing millivolts. (Which is kind of an awesome idea, now that I think about it.)
@Whosa_whatsis Also, check out this line in the troubleshooting pages: “We cannot send a replacement heater element because the calibration values would not match and would potentially cause your printer to not work properly.” SOMETHING interesting is going on there.
@Ryan_Carlyle that sounds to me like they might be using current sensing to determine the temperature after all, because I would expect that to be more inconsistent within a batch than just the heater power (because current is a function of the square of resistance). Regardless, it definitely sounds like they’re not using anything that was manufactured to be a temperature sensor.
@Whosa_whatsis Now I’m really curious how steep the NTC curve must be on those M3D heaters. I wouldn’t really have expected it to be large enough to get adequate sensing resolution/accuracy. Makes me wonder if we could cobble together an equivalent bolt-on current-sense board that would output a readable signal for open-source controllers, like with a custom thermistor table or something. (I wouldn’t want to use that for primary temp control, but it’s an INSANELY good redundant data channel for an independent heater safety board. If heater resistance drops below X, kill heater. Covers severe over-temp events, some ground faults, some wire shorts, etc.)
I suppose that, depending on how instantaneous your current reading is, heater coil induction might also be a factor. This is particularly true if you’re measuring while doing PWM, but even magnetic fields near the hot end could have an effect, and I wouldn’t be surprised to see a 60hz oscillation in the readings.
You might choose anti-seize compounds over thermal paste since they have much higher thermal ceilings. They’re usually a metal suspended in a lubricant that eventually goes away, like aluminum, copper, graphite, moly, or nickel. Graphite and moly have the advantage of being non-reactive, and I think their temp ceilings are over 2000F. The stuff’s available at most auto parts stores, though I only remember seeing aluminum & copper varieties. I found at least one forum post where someone was using it as CPU thermal compound.
Interesting approach.