I am working in a lab and would like to build a temperature controller for a heater. The heater runs off 120V AC and is 15W. (I could find a different heater if this one makes things difficult). I need the controller to be a PID and maintain a steady temperature (42.5C) +/- 0.1. I am using type T thermocouples for temperature feedback.
Is this a feasible project with an arduino microcontroller? If so, does anyone have a recommendation as to which controller I should purchase? Also, is this a difficult project or fairly straightforward?
A good project.
Basicly you need to read the thermocouples value (that is a voltage ins't it?), and send a signal to a driver for the heater. A SSR would be a good choice.
Any arduino board can do the task. I have been using leonardo boards, and happy with them.
What output does your thermocouples put out? what is it at 42.5C ?
Personally, I would use a semiconductor temp sensor such as LM34 or LM35 or DS18B20 rather than a thermocouple, since the temperature is so low. If you must use a thermocouple, use a MAX31855 to interface it to an Arduino (take care to get the right version of the MAX31855, depending on your thermocouple).
I agree with jackwp, use an SSR to control the heater. As the power is so low, a 1A or 2A SSR will suffice. If you are comfortable doing mains wiring, then you could even use a power opto triac instead of an SSR (you can get power opto triacs rated at 0.9A, for example Vishay VO2223).
I need the controller to be a PID and maintain a steady temperature (42.5C) +/- 0.1. I am using type T thermocouples for temperature feedback.
With the thermocouples (or solid state devices) do you need the PID?
I also like the idea of a solid state temperature sensor. They are cheaper and a lot easier to use than a thermocouple. Thermouples put-out a tiny voltage that has to be amplified, and they are non-linear (or maybe it's the typical bridge circuit that makes them appear non-linear). In scientific applications thermocouples do have an advantage of working from physical principals/constants so they "stay calibrated". But, the amplifier & voltmeter still have to be periodically checked/calibrated. (I'm not suggesting that the solid state sensor will drift out of calibration, I'm just saying that the laws of nature never drift. )
and maintain a steady temperature (42.5C) +/- 0.1.
I have a couple of concerns... The temperature is probably not constant across the heating-pad surface
That's a lot of precision... 0.1 out of 40 is 1/4%. With the LM35, 0.1 degree is 1mV. 1mV of noise or drift in the signal or your measurement system will cause a 0.1 degree error
Using the Arduino's 10 bit (1024 count) analog-to-digital converter and the 1.1V reference gives you a resolution of about 1mV... That's cutting it close if you need 1mV accuracy.
DVDdoug:
That's a lot of precision... 0.1 out of 40 is 1/4%. With the LM35, 0.1 degree is 1mV. 1mV of noise or drift in the signal or your measurement system will cause a 0.1 degree error
Good point - 0.1C is going to be difficult to achieve, either with a thermocouple or with a semiconductor temp sensor.
The OP is not around, so we can debate it. I thought they said they were going to use the thermocouple (not, should I use it?)
I am not sure of the most accurate sensor, but I think a LM34 is a bit more accurate than the LM35. The OP does not need a lienar reading, just a one temp reading, 42.5C (108.5F).
Do all your thermometers in the lab have that much (0.1C) accuracy?
Does your thermocouple have that accuracy?
The lm35 specs say
0.5°C Ensured Accuracy (at +25°C)
Do you need that much accuracy? You can get 0.5/C more easily, but maybe that will not be good enough for your project.
jackwp:
The OP is not around, so we can debate it. I thought they said they were going to use the thermocouple (not, should I use it?)
Yes, but I thought it worth pointing out that it might not be the best choice. The decision may not be set in stone.
jackwp:
I am not sure of the most accurate sensor, but I think a LM34 is a bit more accurate than the LM35. The OP does not need a lienar reading, just a one temp reading, 42.5C (108.5F).
Do all your thermometers in the lab have that much (0.1C) accuracy?
Does your thermocouple have that accuracy?
The lm35 specs say
0.5°C Ensured Accuracy (at +25°C)
The MAX31855 (thermocouple interface chip) datasheet quotes at best +/- 2C accuracy. The LM35 is better at +/- 0.5C as you quoted. The LM34 is slightly better than the LM35 at +/- 1F; however it main advantage is that it produces 10mV per degF instead of per degC, so it provides better resolution at the same (or no) amplification.
If the OP wants 42.5 +/- 0.1C, then I think he will have to calibrate whatever sensor he uses against a precision lab thermometer.
jackwp:
The OP is not around, so we can debate it. I thought they said they were going to use the thermocouple (not, should I use it?)
Yes, but I thought it worth pointing out that it might not be the best choice. The decision may not be set in stone.
jackwp:
I am not sure of the most accurate sensor, but I think a LM34 is a bit more accurate than the LM35. The OP does not need a lienar reading, just a one temp reading, 42.5C (108.5F).
Do all your thermometers in the lab have that much (0.1C) accuracy?
Does your thermocouple have that accuracy?
The lm35 specs say
0.5°C Ensured Accuracy (at +25°C)
The MAX31855 (thermocouple interface chip) datasheet quotes at best +/- 2C accuracy. The LM35 is better at +/- 0.5C as you quoted. The LM34 is slightly better than the LM35 at +/- 1F; however it main advantage is that it produces 10mV per degF instead of per degC, so it provides better resolution at the same (or no) amplification.
If the OP wants 42.5 +/- 0.1C, then I think he will have to calibrate whatever sensor he uses against a precision lab thermometer.
For the cost of it why not just put in 3 or 4 DS18B20 sensors and average the readings to get the level of accuracy he desires ?
craigcurtin:
For the cost of it why not just put in 3 or 4 DS18B20 sensors and average the readings to get the level of accuracy he desires ?
Because there is no guarantee that the average of 3 or 4 sensors with +/- 0.5C accuracy will be within 0.1C of the correct value, especially if they are all from the same manufacturing batch. They might all be 0.5C out in the same direction.
dc42:
Because there is no guarantee that the average of 3 or 4 sensors with +/- 0.5C accuracy will be within 0.1C of the correct value, especially if they are all from the same manufacturing batch. They might all be 0.5C out in the same direction.
If 0.1°C is critical then they're going to need to calibrate whatever sensors they use, regardless of type, against a known standard.
But, I am not sure this accuracy is practical. If it was in a 1 cubic foot container, the temp in one section would likely be more than .1c from another section a few inches away. But, only the OP can give us the details.
)