Are thermocouples more accurate than thermistors?
-
The Duet5+ mini says :
"3 Thermistor/PT1000 inputs and support for 2 PT100 or thermocouple inputs via the optional daughterboards."
Do thermocouple provide an inherently more accurate temperature measurement compared to pt100? I am interested in the range 200 - 260C.
The reasons is I don't trust the temp reading of my current system (which uses a thermistor) and trying to calibrate it with an external K type sensor/meter attached to the heating block gives readings that are about 15C lower than what the duet reports (15C lower at 240C, both agree at room temperature) and I don't now if it's because of an actual error or my measurement method.
Having a nozzle sensor that I can trust out of the box would be great.
-
A thermocouple should be far more accurate than a generic thermistor, as long as it's used with a properly designed interface / display.
The connecting wires and any connectors must be the correct thermocouple type all the way through to the interface for proper accuracy; eg. a type K thermocouple must be matched with type K cable, connectors and interface.
Using any other wire or connector type compromises the accuracy, as every dissimilar metal joint also acts as another thermocouple.
The most practical option, far simpler for a Duet, is a PT1000 - they should be extremely accurate and as a resistance type device, no special wiring requirements other than avoiding proximity to such as stepper cables.
Or, PT100 via the add-on interface, which should have similar accuracy. Four wire mode with a PT100 should be better still.
-
@zapta Thermocouples are accurate to +/- 3C (at room temp). That accuracy will degrade over time because the construction inherintly suffers from galvanic corrosion.
RTD's, like a PT100 or PT1000, at the lowest class, are accurate to +/-0.5C at around 250C.
Which is a long way to say: RTD's are significantly more accurate than thermocouples.
Although both are much more accurate than thermistors (which I have seen be +/-20C at 250C).RTD's have a longer service life, use normal wiring and are less prone to electrical noise than thermocouples. Basically you use thermocouples when you need to measure very high temps or a very wide range. Otherwise RTD's are the better choice. Thermistors are for low temps and really low costs.
-
Thanks @rjenkinsgb and @theruttmeister. PT100 seems to be the most promising for my application. I read a little bit about it and having a couple of questions
-
Is it OK to connect with regular non shielded two wires? (going through the Voron cable chain it's easier to reuse the existing thermistor wires)
-
Where can I find a V6 PT100 cartridge with known datasheet of the internal sensor? (I believe from the duet documentation and other sources that knowing it's temperature/resistance table helps calibrating it).
-
-
@zapta
If it's a standard size one like in a V6 etc., E3D do them & they are available from many distributors.
https://e3d-online.com/products/pt100-temperature-sensorAnother example:
https://reprapworld.co.uk/extruder/extruder-parts/temperature-sensors/pt100-sensor-with-cartridge/?ds=trueAny high spec PT100 should be accurate to better than 0.2% and not need any additional calibration. Cheaper ones may be something like 0.5% off, but unlikely to have calibration info.
This is typical temperature vs resistance chart:
https://www.tnp-instruments.com/sitebuildercontent/sitebuilderfiles/pt100_385c_table.pdfIf you want to use four wire connections but the PT100 has two wires, connect two wires to each sensor lead, as near as practical to it.
Power feeds via one set of connections and the voltage is read back via the other, to eliminate voltage drop that would otherwise occur in the shared power+sense connections. -
Thanks @rjenkinsgb. The E3D two wires PT100 seems to be a good choice. I will look for it.
Duet's documentation has this about calibrating PT100 inputs. Is it worth doing it? https://duet3d.dozuki.com/Wiki/Calibrating_thermistor_and_PT1000_readings
Also, my wires will be AWG24 1.5m each direction so if I got the math correctly this should add about 0.7C to the reading. Is there a way to compensate for it at the gcode level? (I want to stay with two wires).
-
@zapta said in Are thermocouples more accurate than thermistors?:
Duet's documentation has this about calibrating PT100 inputs. Is it worth doing it? https://duet3d.dozuki.com/Wiki/Calibrating_thermistor_and_PT1000_readings
I believe this pertains to PT1000, not PT100.
-
@zapta said in Are thermocouples more accurate than thermistors?:
Is there a way to compensate for it at the gcode level? (I want to stay with two wires).
Not that I know of. What reason for staying with only 2 wires?
https://www.omega.ca/en/resources/rtd-2-3-4-wire-connections
-
@phaedrux said in Are thermocouples more accurate than thermistors?:
What reason for staying with only 2 wires?
Simplifying the wiring. I also have 2 wires going through the drag chain.
Another quest, duet also support PT1000 (e.g. by E3D) which doesn't require a special daughter board and are probably less sensitive to wiring resistance. Anybody knows what are the pro and cons of PT100 vs PT1000?
https://duet3d.dozuki.com/Wiki/Connecting_thermistors_or_PT1000_temperature_sensors
-
@zapta PT100 are used in serious industrial environment, where the very low impedance of the 100 ohm sensor keeps noise down. However, it absolutely requires 4-wire connection for accuracy, since even modest wiring resistance can through them out of spec. 1 ohm of wiring puts it 3 degrees off! Also, Pt100 can be wound with heavier wire, so they are more mechanically robust, too, and less sensitive to metallurgical issues. 'Standards Grade' platinum sensors (called SPRTs) are all Pt100 because of these issues.
On the other hand, PT1000 are much easier to read out, since the 1k resistance is more compatible with a typical A/D inputs, and the wiring resistance is only 1/10 as much a problem. They are higher impedance, which could mean bigger noise pickup in heavy industrial environments, but this isn't an issue mostly with short wiring runs on a 3d printer. The wire inside is finer, making them less stable at the highest level of fussiness. I suspect that the modestly priced ones we are buying for 3d printers are actually thin film, not wire wound anyways, although I have not investigated this.
My recommendation: use Pt1000. They are trivial to use, and can provide better than 1C temperature measurement over the whole range from room temp to 350C (or 500C on special ones).
-
@mendenmh said in Are thermocouples more accurate than thermistors?:
My recommendation: use Pt1000.
Thanks @mendenmh, this is also supported by the Duet's documentation "The accuracy of PT1000 sensors should be very good on the Duet 3". I tend to go with an E3D PT1000 cartridge.
Edit: Just ordered an E3D PT1000 from Filastruder.
-
I've no doubt that I'll get flamed for the following but I'm too old to care so here goes. If you look at the data sheets for filaments, you'll see that the recommended print temperature is typically within a range of 20 to 30 Deg C. E.g 190 to 220. Given that most people accept that every hot end is different (because of the separation between the sensor and the noozle tip), it is also accepted that one should print a temperature tower to find the best value to use on one's machine. Such towers usually use increments of 5 degrees and it is often difficult to discriminate between the "best" temperature and those that are 5 deg C either side. So one can then say that the optimum temperature for ones own particular machine has a tollerance of +/- 5 deg C. So why then do people become obsessed with finding a sensor which has an accuracy of 1 Deg C or less when this level of accuracy is much higher than needed?
-
@deckingman I won't flame you... What I am looking for is interchangeability of sensors. If one breaks, and I can put in a new one and have everything stay the same, it is good. The variations between thermistors is HUGE under the conditions they are used in 3d printers. They are standardized at 25C. However, tiny variations in beta can result in mismatches between thermistors of tens of degrees at 250C. Besides, I am a thermometry nerd. In real life, I require absolute temperature measurements of a few thousands of a degree.
-
@mendenmh
I've just ordered a couple of the E3D ones myself, as i suspect the temperatures showing on my recently built machine are a bit off.Having something known to be accurate is useful for keeping compatibility between printers, rather than having to individually calibrate the temperatures on each.
-
This post is deleted! -
I could also add that, before I came to my senses, and when I had less experience of actually printing things, I too have used PT100s and t/couples. I too got sucked in by the hype. After many hours of chasing down problems caused by electrical noise which resulted in erroneous readings with these "more accurate" types of sensor, I realised that the quality of the signal between the sensor and the control board has a much greater influence over how the machine behaves, than the accuracy of the sensor itself. What good is having a sensor with claimed 1 degree accuracy if it is overlaid with +/- 2 degrees of EMF induced noise? After having changed the 2 wire PT 100s to 4 wired devices, then re-routing the cables, then changing all that wiring to screened cable and still having issues, I have reverted back to using simple but good quality thermistor cartridges and now have stable and reliable readings, which IMO is much more important than sensor accuracy as quoted in data sheets.
-
@deckingman said in Are thermocouples more accurate than thermistors?:
So one can then say that the optimum temperature for ones own particular machine has a tollerance of +/- 5 deg C.
@deckingman, having +/-5C error is reasonable as long as the error is stable over time. The question, is, how do you determine that your tolerance is 5C and not 10C or 20C?
EDIT: I looked for liquids with high temperature boiling point (above water's 100C, closer to the 230C range) and found this one https://www.amazon.com/gp/product/B08RCT4YNZ at 188.2C. Any thoughts?
-
@zapta said in Are thermocouples more accurate than thermistors?:
@deckingman, having +/-5C error is reasonable as long as the error is stable over time. The question, is, how do you determine that your tolerance is 5C and not 10C or 20C?
I never said that having a +/- 5 degree error is reasonable or otherwise. I only said that one can assume a +/- 5 degree tolerance is acceptable. Measurement tolerance and measurement errors are completely different things.
And just because one might use an accurate sensor, does not mean that measurements taken with that sensor will be accurate. With 3D printers, the sensor is often some distance away from the part that we are interested in. Usually hot ends have a sensor somewhere in the heater block but not in the nozzle itself. From the work I have done measuring temperatures inside the nozzle, I can pretty well guarantee that sensor in the heater block will not show an accurate temperature of the nozzle itself - regardless of how accurate that sensor is.
Take a look at this write up that I did https://somei3deas.wordpress.com/2020/05/21/the-effect-of-deflected-part-cooling-air-on-brass-and-steel-nozzle-temperatures/ and you'll see that the temperature inside a brass nozzle could be around 1.7 degrees cooler than the nozzle block without a fan running, and as much as 8.5 degrees cooler with some deflected air from a part cooling fan. And for a hardened steel nozzle, the difference is much more dramatic.
So I say again, what is the point of using a sensor with a claimed accuracy of 1 deg C or better, when there could be measurement errors of 8 degrees or more which depend on the position of that sensor in relation to the nozzle tip?
With regard to temperature drift over time, once I've established a particular temperature to use for a particular filament, I've never had recourse to re-evaluate it. And I can't say that I've ever seen any evidence that temperature sensors used in 3D printers will drift sufficiently over their lifetime to require re-evaluation. But I have no doubt that if you search the internet long enough and hard enough, you'll find "evidence" to the contrary.
-
@deckingman said in Are thermocouples more accurate than thermistors?:
So I say again, what is the point of using a sensor with a claimed accuracy of 1 deg C or better, when there could be measurement errors of 8 degrees or more which depend on the position of that sensor in relation to the nozzle tip?
Because you use a sensor/electronics/firmware that gives you 10C error than the reading can be much more than 8C away from the nozzle's temp.
BTW, I am using platted copper 0.4mm nozzle with flow < 10mm^2/sec and material fan (which is weak in the first place) at 50%. What's your guts feeling about the nozzle temp drop in my system (nozzle vs heating block temp), when printing at ~230C?
-
When you are shipping thousands of machines, the support burden of in-accurate sensors is non-trivial. Plus finding a good thermistor for temps above 250C is still a challenge.
But for the solo hobbyist, no it's not critical.
@zapta PT100's are more readily available, although the sensors for FDM are already a custom part so that's kinda academic.
Isolate, substitute, verify.
