PT 1000 question

wilriker

@deckingman AFAIK the accuracy for PT1000 is within +-2°C (or maybe even +-1°C). It could be improved by replacing the reference resistor on the board e.g. from the original 4.7k to something like 1k but personally I would be very reluctant to solder on my Duet.

EDIT: The Maestro has a lower reference resistor (I think 2.2k) by default that's why it is more accurate on these.

gavatron3000

If it helps you I've got one on my Hotend. Didnt notice any real difference on the change over. It's accurate enough that it seems the same as a thermistor. I'm using it on an E3d and duet wifi
Gav

Veti

i have a maestro and a pt1000 both on the hotend and the bed.
It works well and i find the temperature seems to be more accurate compared to random NTC100k B3950 thermistors.

deckingman

@wilriker said in PT 1000 question:

@deckingman AFAIK the accuracy for PT1000 is within +-2°C (or maybe even +-1°C). It could be improved by replacing the reference resistor on the board e.g. from the original 4.7k to something like 1k but personally I would be very reluctant to solder on my Duet.

EDIT: The Maestro has a lower reference resistor (I think 2.2k) by default that's why it is more accurate on these.

Manuel,

Is that a stab in the dark, an educated guess or could you put your hand on your heart and say with 100% conviction that if I were to connect say 8 PT1000s to the Duet/Duex5 Ethernet and have them arranged to measure the same point, that they would read within 2 deg C of each other over a temperature range of say 100 deg C to 250 deg C?

@all For info, I have in front of me a batch of 10 NTC 100k thermistor cartridges, all bought at the same time, from the same supplier. At normal ambient the resistance of one of them is 94.4K and one is open circuit. Ignoring those two, the rest fall within a range of 102.8 to 109.1k. So in the order of 7% variation and who knows what they would be like at say 190 deg C. That's unacceptable so the reason why I need to look at something else.

Veti

The PT1000 sensors can be bought with different accuracies.

Class B is ±0.3 °C at 0 °C
Class A is ±0.15 °C at 0 °C

deckingman

@veti Understood. But I think that the accuracy of the sensor is only part of the equation because the Wiki indicates that different boards give different results. The rest of the equation must be something to do with reference resistors and DACs or some such, in order for this statement from the Wiki to be true -

quote..... "The accuracy of PT1000 sensors should be very good on the Duet 2 Maestro and generally good on the Duet 2 Wifi and Ethernet."

So I'm still seeking a quantifiable definition of "generally good" with respect to the Duet Ethernet board. Is it within 1%, 2%, 2 degrees C or some other value?

I'd be more than happy with 1%, so say 2 degrees at around 200 degrees C but as I'll likely be buying a batch of 10, I need to be reasonably sure before parting with my hard earned.

wilriker

@deckingman said in PT 1000 question:

Is that a stab in the dark, an educated guess or could you put your hand on your heart and say with 100% conviction that if I were to connect say 8 PT1000s to the Duet/Duex5 Ethernet and have them arranged to measure the same point, that they would read within 2 deg C of each other over a temperature range of say 100 deg C to 250 deg C?

What I know for sure is that accuracy (per sensor) depends on the reference resistor. I think I read the number of 2°C somewhere but since I cannot find the source for that anymore I would grade that part down to "stab in the dark", sorry.

EDIT: There is also this page in the wiki but the Special Considerations row of the table in the PT1000 column does not make very much sense to me...

dc42

I would expect the accuracy of PT1000 sensors to be better than thermistors because they are generally made to closer tolerances. But resolution may be an issue depending on the application. On the Duet WiFi/Ethernet the native resolution of the ADC gives a temperature resolution of about 0.9C at room temperature and a little worse at high temperatures. On the Maestro it's a little better, 0.4C at room temperature and 0.7C at higher temperatures. We use 2-bit ADC over sampling to try to improve the resolution by a factor of 4, but I don't know how reliable that is because it depends on the profile of the noise seen by the ADC.

The reference resistors have a tolerance of 1% and that is equivalent to a 0.4C change in the temperature reading.

On the Duet WiFi/Ethernet the resistance of the VSSA fuse introduces an additional small error, but it is the same for all channels. On the Maestro it is measured and compensated

deckingman

OK so no definitive answer.

Moving on then, is it possible to calibrate each sensor? What I have in mind is to get a block made that will take a heater and two sensors. One of which would be a reference such as a PT100. Then I could plug in each of the PT1000s into the other hole and tweak something like the L and / or H values until the PT1000 reads the same as the PT100. The Wiki isn't clear (to me anyway) as to whether L and H are applicable to PT1000s. If not L and H, then are there any other parameters that could be tweaked?

Or could I use the same approach with thermistors instead of PT1000s? I'm assuming that the resistance vs temperature is a linear relationship and I don't have mess around with least squares curve fits to generate polynomial equations.

dc42

You can adjust the R parameter to compensate for differences in the PT1000 sensors. The change in resistance with temperature isn't quite linear but the firmware compensates for that.

deckingman

@dc42 Brilliant!