K-type thermocouple & MAX6675 module

Other Hardware Sensors
1

My k-type thermocouple is giving me a reading that is roughly 5 degrees Fahrenheit higher than actual.
So, I've read that the accuracy can be up to +-6 C, and I also read that with most thermocouples the temperature voltage relationship is almost completely linear. Can anyone attest to this? If I just use an offset of -5 F, would I be ok for readings down to -70 F?
Why that low?
About a year ago I launched a high altitude balloon that, among a barrage of other sensors, had a DHT22 onboard. The sensor reached its lower limit of -40 degrees F. Therefore, my next balloon will have a thermocouple instead. Bit of an overkill? Not really, stratosphere temperatures reach about -65 to -70 degrees Fahrenheit (depends on time of year, latitude, altitude, etc.). So, a LM60 or a TMP36 would not work.
Any input, or tips on calibration are appreciated.

Andres

2

To calibrate it I would use melting ice, and boiling water to give you 2 readings for 0°C and 100°C.

I would have thought that any error would be due to the cold junction compensation not working correctly, which leads me on to believe that:

"If the thermocouple, and the MAX6675 are both at the temperature of the upper atmosphere, (which I assume they will be), then the output of the thermocouple will be zero, and that any reading that you get will be that provided by the IC's cold junction compensation circuit."

I'm not sure whether that is true or not, it was just a thought experiment that I had.

Maybe someone else can confirm or deny my theory.

A thermocouple is normally used to measure the difference in temperature between ambient (or some other reference) and the temperature of the object being measured.

3

So, you're saying there must be a temperature differential between the MAX6675 and the thermocouple in order to get an accurate reading? I apologize if the question sounds ignorant....but, I am rather new to this.
The module will be inside the capsule and the thermocouple will be outside. There's a roughly 60 degree F differential between interior and exterior temperatures, with the differential narrowing down to about 40 degrees at apogee. It sounds like too much, but confirmed by multiple sensors inside the capsule on my previous launch. This is mostly due to all the heat generated from 16 AA batteries, and all of the modules inside.

4

aparis1983:
My k-type thermocouple is giving me a reading that is roughly 5 degrees Fahrenheit higher than actual.
So, I've read that the accuracy can be up to +-6 C, and I also read that with most thermocouples the temperature voltage relationship is almost completely linear. Can anyone attest to this? If I just use an offset of -5 F, would I be ok for readings down to -70 F?
Why that low?
About a year ago I launched a high altitude balloon that, among a barrage of other sensors, had a DHT22 onboard. The sensor reached its lower limit of -40 degrees F. Therefore, my next balloon will have a thermocouple instead. Bit of an overkill? Not really, stratosphere temperatures reach about -65 to -70 degrees Fahrenheit (depends on time of year, latitude, altitude, etc.). So, a LM60 or a TMP36 would not work.
Any input, or tips on calibration are appreciated.

Andres

You are going to have to insulate the MAX6675 from the cold temperatures, it is only rated down to -20C, if your outside temp is -70F (-59C) you are outside the envelope for the MAX6675.

the 'Cold Junction' is where the K-type wires terminate. each connection of the thermocouple wires form a temperature sensor. So, A simple K-type has actually three thermocouple joints. One at the measurement point (welded Junction), one on the T+ junction, one on the T- Junction. The MAX6675 assumes the T+ and T- junction are at the same temperature as it's die. It (the MAX6675) uses an internal reference table to subtract it's temperature equivalent voltage from the voltage it is measuring from the K-type thermocouple. It then translates this 'corrected' voltage to the probe temperature.

assume the MAX6675 is at 15C and the Probe Junction is at -40C, the Raw voltage at the MAX6675 input is:
'cold Junction' 15C = 0.597mV
Probe (welded Joint) -40C = -1.527mV

Raw voltage at input of MAX6675 = -0.93mV or (-24C using K-Type conversion table for -0.93mV);
Correction factor because the MAX6675 measures it's internal temp of 15C= -0.597mV
Net Voltage after correction is -1.527mV (-40C)

Now lets say the 'cold Junction' is not actually at 15C, but because it is actually colder because it is not as well insulated, or has no active heating, or is in the airstream. And it's actual temperate is -20C.

Cold Junction at -20C = -0.778mV
Probe (welded Joint) -40C = -1.527mV

Raw voltage at input of MAX6675 = -2.305mV (-62C )
Correction factor because the MAX6675 internal temp is 15C =-0.597mV
Net Voltage after correction is -2.902mV (-79.5C)

As this example shows, you need to make sure the 'cold junction' is at the same temperature as the MAX6675.

You should calibrate the sensor against a know reference, (freezing water, accurate thermometer).

Make sure the 'cold Junction' and the MAX6675 are at the same temperature, Insulate both in a common compartment.

There does not have to be any difference between the 'cold junction' temperature and the 'joint' temperature, BUT, there CANNOT be any temperature difference between the 'cold Junction' and the MAX6675.

Chuck.

5

ChuckTodd, this was very helpful. Thanks!
I don't see it being much of a problem to keep the MAX6675 above -20 C. Usually, the lowest internal temperature stays above -5 C.

I assume the cold junction is close to the T+ and T- junctions? I'm at work right now (can't check) but I would think it's marked somewhere? Or does the cold-junction occur on the module itself?

6

Just to be clear, the 'Cold Junction' (or Reference Junction) is where the K-type wires turn into copper traces that connect to the MAX6675. That Junction needs to be at the same temperature as the MAX6675 because an on-chip temperature sensor is available, but I forget if it automatically corrects the measurement or if you have to read it and do some math. At any rate, the K-type wires need to run the full distance between the thermocouple and the connection to copper where the measurement chip is and that connection must be at the measurement chips temperature.

7

aparis1983:
ChuckTodd, this was very helpful. Thanks!
I don't see it being much of a problem to keep the MAX6675 above -20 C. Usually, the lowest internal temperature stays above -5 C.

I assume the cold junction is close to the T+ and T- junctions? I'm at work right now (can't check) but I would think it's marked somewhere? Or does the cold-junction occur on the module itself?

As Ron Sutherland said:

ron_sutherland:
Just to be clear, the 'Cold Junction' (or Reference Junction) is where the K-type wires turn into copper traces that connect to the MAX6675. That Junction needs to be at the same temperature as the MAX6675 because an on-chip temperature sensor is available, but I forget if it automatically corrects the measurement or if you have to read it and do some math. At any rate, the K-type wires need to run the full distance between the thermocouple and the connection to copper where the measurement chip is and that connection must be at the measurement chips temperature.

The point were the special wires (Chromel and Alumel) connect to copper wire (or PCB Trace) is the cold Junction. Those connections must be at the same temperature as the MAX6675.

Chuck.