Skip to main content
Skip table of contents

Thermocouples (App Note)

If you are measuring in the range of -50 to +150 °C, consider using a silicon type temperature sensor rather than a thermocouple.  See the Temperature Sensors application note.

Understand the difference between resolution and accuracy.

Silicon Type Sensors

In the range from -50 to +150 °C, silicon temperature sensors are generally cheaper, easier to use, and more accurate, than other types of temperature sensors. With no or minimal extra components, they provide a high-level linear voltage output that connects directly to a LabJack's analog inputs.

  1. The EI-1034 is a silicon based temperature probe made by Electronic Innovations and sold by LabJack.  It uses an LM34CAZ sensor element from National Semiconductor.  The LM34 provides an easy-to-use 10 mV/°F.

  2. The EI-1022 is a silicon based temperature probe made by Electronic Innovations and sold by LabJack.  It uses an LM335A sensor element from National Semiconductor.  The LM335A provides an easy-to-use 10 mV/°K, but needs a resistor also.


Thermocouple Basics

Thermocouples are little more than two wires made of different conductive materials clamped or soldered together. A small voltage is established between the two wires as described by the Seebeck effect. At 0 °C, thermocouples will not set up a thermo-voltage (0V reading).

Since thermocouples are essentially just wires, the equivalent impedance of a thermocouple is small. Additionally, there should be very little (or ideally no) current in a thermocouple measurement circuit. As such, thermocouples are quite prone to EMI and errors due to ground loops as discussed in the thermocouple complications section below.

Thermocouple Selection

Primary considerations when selecting a thermocouple should be:

  • Type

  • Limits of Error

  • Junction Type

Type and Limits of Error

The thermocouple type might be selected based on your measurement range, measurement media, and the sensor cost. Type K thermocouples are commonly used due to their low cost and wide measurement range.

Thermocouples are commonly designated as having either standard or special limits of error where the special limits of error sensors have tighter accuracy tolerances.

The following table has additional information to consider about different thermocouple types and limits of error. Tolerance class 1 in the table is sometimes described as having special limits of error when ordering. Also see the types section further up the page:

Junction Type

Thermocouples are available with the hot junction grounded, exposed, or ungrounded in probes. They can also be purchased as bare wire.

  • Grounded probes have the hot junction grounded to the probe. Exposed probes have the hot junction exposed outside of the probe. These probe choices and bare wire can provide better response times, but could be more prone to noise issues such as ground loops (see the complications section below).

  • Ungrounded probes have the hot junction disconnected from the probe, which worsens response time, but can help avoid noise issues.

Voltage to Temperature Conversion

The relationship between thermcouple voltage and temperature is non-linear. Lookup tables and/or polynomials that describe the relationship are commonly used to convert voltage measurements to temperature.

One well trusted source for this information is the National Institute of Standards and Technology (NIST). Our thermocouple AIN_EF (T7 only) and TCVoltsToTemp functions typically use the polynomials described by the "Inverse Coefficients" tables on the NIST website. These tables also describe the expected error due to the conversion:

Thermocouple Junctions and Cold Junction Compensation (CJC)

Except at 0 °C, a voltage will be established at any points in a thermocouple connection where one conductive material transisitions to another. This includes the junction between the two thermocouple materials at the remote end (hot junction) and in the connection between each thermocouple material and extension wire or terminal connection (cold junctions). The only voltage of interest is the one at the hot junction, so the voltage at the cold junction(s) will appear as an error in the measurement.

Thermocouple Junctions

Cold junction effects are commonly reduced or eliminated using one of two methods:

  1. A temperature sensor near the cold junction is used to measure the cold junction temperature. The temperature is converted to an equivalent thermocouple voltage to be removed from the thermocouple voltage measurement. This is cold junction compensation (CJC).

  2. The cold junction is placed in an ice bath such that it is at 0 °C. No thermo-voltage will be generated at 0 °C, thereby eliminating any cold junction effects.

Most LabJack devices have built-in temperature sensors that can be used for CJC. See your device datasheet for further information.

Hardware Recommendations

  • The T8 is the best LabJack device for thermocouple measurements.

  • The U3 or T4 with LJTick-InAmps are the least expensive way to get decent readings from thermocouples, but would provide lower resolution than any of the T8, T7, or U6 devices.

  • The 24-bit sigma-delta converters available on the T8, T7-Pro, and U6-Pro provide higher resolution than the 16-bit converters on the base U6 and T7. Perhaps more importantly, they also provide outstanding noise rejection. In particular, at lower sample rates the sigma-delta converters reject 50 and 60 Hz noise, which is a common problem with thermocouple signals.

Thermocouple Measurements with LabJack Hardware

  • U3/T4:  The U3 is a USB-only device that when combined with LJTick-InAmps provides the minimum cost solution for 4 or less thermocouples.  The T4 is similar but has USB and Ethernet.  See the U3/T4 Thermocouple Tutorial after reading the information below.

  • U6:  The U6 (or U6-Pro) is a USB-only device.  Thermocouples can be connected directly with better performance than the U3 or T4 and LJTick-InAmp combination.  See the U6 Thermocouple Tutorial after reading the information below.

  • T7:  The T7 (or T7-Pro) has the same analog input system as the U6 (or U6-Pro), but provides Ethernet and USB connectivity (and also WiFi on the T7-Pro). The AIN_EF system can be used to perform all thermocouple math in T7 firmware so you can get direct readings of temperature.  See the T7 Thermocouple Tutorial after reading the information below.

  • U6-Pro/T7-Pro:  Same as the U6/T7 info above, but the 24-bit sigma-delta ADC on the Pro versions have better resolution and noise rejection.

  • T8: The T8 benefits from high-speed acquisition of up to 50 ksamples/second with a 24-bit sigma-delta ADC, simultaneous measurements across all 8 analog inputs, and it has isolated inputs. Isolated inputs can help avoid complications such as ground loops. See the T8 Thermocouple Tutorial after reading the information below.

  • UE9 and U12:  New applications should first consider one of the devices described above.  The UE9-Pro is good for thermocouples, while the UE9 is okay, providing a resolution of roughly 1 °C. The U12 can only resolve raw thermocouple voltages within 10s of °C, and thus generally requires an amplifier such as the EI-1040. Also note that the UE9 variants and EI-1040 are at end of life.

Thermocouple Complications

Thermocouples are not particularly accurate, and sometimes can be more difficult to use, but they are a very common way to measure temperature.  Some applications with extreme temperatures or specific mechanical requirements might require thermocouples.

Thermocouples have 1 main issue and other minor issues:

  1. Small Output Voltage:  The small output voltage of a thermocouple makes it difficult to get good temperature resolution.  The common K-type only provides around 40 µV/°C, thus for a temperature resolution of 0.1 °C you need a voltage resolution of 4 µV.  This is the reason the U3/T4 requires LJTick-InAmps, whereas the U6/T7 devices have the necessary resolution and amplification built-in.

  2. Cold Junction Effects:  See the information in the thermocouple basics section above. Note that our thermocouple AIN_EF (available on some T-series devices) and TCVoltsToTemp functions can handle calculations for CJC given a cold junction temperature. If connecting to a breakout board such as the CB37 rather than directly to the LabJack terminals, you might want a silicon based sensor such as the LM34CAZ to measure the cold junction temperature.

  3. Non-Linear Output:  The output of a thermocouple is non-linear.  NIST provides tables and equations to convert a thermocouple voltage to a temperature.  The LabJack UD or M libraries provide convenient functions that use the NIST equations to handle the conversion, and DAQFactory has built-in conversion functions of its own.

  4. Poor Inherent Accuracy:  Thermocouples have errors due to the consistency of the thermocouple materials themselves.  There is a standard called IEC 584-2 that limits the allowable error for different thermocouple types at different temperatures, and these allowable errors are typically ±1.0 to ±2.5 °C at moderate temperatures, but can be as high as ±9.0 °C at 1200 °C.  See the "Thermocouple comparison" table on the Wikipedia thermocouple page.

  5. Bad Ground Loops:  This is a common and complex problem that our customers encounter.  It occurs when bare thermocouple wire is used, or a metal thermocouple probe where one of the thermocouple wires inside is connected to the probe (often called a "grounded" probe).  If multiple of these are connected to a common conductor (e.g. a system of connected metal pipes), and connected single-ended to the LabJack, you can get ground loops and unexpected thermocouple junctions.  The typical fix is to use differential inputs with a resistor (100k is typical) from the negative input to GND.  As noted in the Differential Readings Application Note, the resistor is sized to be low enough to provide a path for bias currents but high enough to prevent ground loop issues.  Differential connections usually prevent this problem, but sometimes further steps are needed (e.g. channel-to-channel isolation).  The best way to make sure this will never be an issue is to install thermocouples such that this situation does not exist:

    1. Use some sort of substance, between the thermocouple metal and test specimen, that has good thermal conduction but poor electrical conduction.  Epoxy, tape, etc.

    2. Place the thermocouple near, but not actually touching, the test specimen.

    3. Use probes that are called "ungrounded". A very common option.

    4. Use thermocouples with an electrically insulating layer between the thermocouple metal and test specimen.  One example is the SA1 series of self-adhesive thermocouples from Omega.

  6. Ground Offsets:  Care must be taken to avoid ground offset errors with single-ended measurements.  Error will result if the negative lead of the thermocouple is connected to a GND terminal that is at a different voltage than ground at the A/D chip.  See the "Ground Offsets" section on the Mux80 Datasheet.  A quick summary of this information would be that differential measurements (with ~100 kΩ resistor from negative AINx to GND) are recommended if using thermocouples with the Mux80.

  7. EMI Susceptibility:  Thermocouples produce weakly driven, small voltages.  EMI can easily introduce noise on long thermocouple wires. For example, an arc furnace near the thermocouple could cause noise issues.

Connectors and Connections

If your LabJack has built-in amplification, we recommend using it over our LJTick-InAmp.

Built-in Amplification Connections

For a single-ended connection (good if you do not have complications #5 or #6 above) connect thermocouple+ to any AIN terminal and thermocouple- to GND.

For a differential connection, connect + to positive AINx (ex: AIN0) and - to negative AINx (ex: AIN1), and you almost always need to add a resistor (100k is typical) from negative AINx to GND per the Differential App Note.

Differential Thermocouple Connection

LJTick-InAmp Connections

When using the LJTick-InAmp, connect the thermocouple positive lead to an IN+ terminal and thermocouple negative lead to the corresponding IN- terminal. Also add a resistor from IN- to GND.

LJTick-InAmp Thermocouple Connection

General Connection and Connector Information

The LabJack analog inputs are accessed through screw terminals or DB connectors.  If your thermocouple has plain wire ends, just clamp them right in screw-terminals and know that screw terminal is your cold junction.

If your thermocouple has a 2-pronged connector at the end such as the Amazon B00OK6CBMU, you can simply remove a screw or 2 and remove that connector, or get a female connector such as the REED LS-182 and add a couple pigtail wires.  Note that this connector is the cold junction, and also note that per the law of intermediate metals the type of this connector should not matter; you should be able to use a "type K connector" with any type of thermocouple.

Thermocouple Connector

Single-Ended or Differential

The previous section describes how to connect a thermocouple for single-ended (SE) or differential (DIFF) acquisition.  For most signals, all specifications (noise, accuracy, etc.) are the same for single-ended or differential, so why use an extra channel for a differential connection?  The Differential Readings App Note provides 3 typical reasons for using differential readings, but here is a list of reasons specifically for thermocouples:

  1. Bad Ground Loops:  Some systems create shorts between multiple thermocouples.  Differential connections can often solve this problem, but the best fix is to modify the thermocouple isolation to avoid the shorts.  See item #5 in the "Thermocouple Complications" section above.

  2. Ground Offsets:  If connecting thermocouples to the Mux80, substantial ground offsets can occur and differential connections are an easy fix.  See item #6 in the "Thermocouple Complications" section above.

  3. Rejection of Common-Mode Noise: See "Why use differential?" item #3 from the Differential Readings App Note

Per the Differential Readings App Note, analog inputs cannot be totally floating, so a differential thermocouple connection will require the addition of a resistor from the negative channel to GND.  100 kΩ is a typical value.

Troubleshooting Tips

Too much noise

If you see too much noise (notable swings) in your readings, try the following.

Range and Resolution configuration settings: The most common problem leading to too much noise is not having the correct range and resolution settings (on the U6/T7 in particular). To test:

  1. disconnect the thermocouple and jumper the analog input to GND (or both analog inputs to GND if using a differential input).

  2. Open LJControlPanel and go to the test panel (U6) or go to the Analog Inputs tab in Kipling (T7).

  3. Set the range and resolution as needed and look at the noise level on the applicable channels and compare to the specified typical noise for the LabJack. See the specifications in the appropriate device datasheet(s):

    1. Appendix A of the U3 Datasheet

    2. Appendix B of the U6 Datasheet

    3. Appendix B of the UE9 Datasheet

    4. Appendix A-3-1 of the T-Series Datasheet

    5. Appendix B of the LJTick-InAmp Datasheet

      With the default thermocouple settings of Range=0.1 and ResolutionIndex=0, the noise level of a U6/T7 will be perhaps 1 or 2 microvolts.  After confirming the proper noise level in LJControlPanel/Kipling, go to your actual software and look at the raw voltage to confirm the proper noise level there.

Floating differential input: If using a differential input you need a resistor from the negative channel to GND.  See the Differential Readings App Note.

Removing Environmental factors: See the Bad ground loops, ground offsets, and EMI susceptibility complications described above.

  • Try removing all external connections from the LabJack except one thermocouple and testing. This will help ensure that external connections are not affecting the reading.

  • If your thermocouple is mounted to an electrically conductive surface, try a test measurement with a thermocouple that is not mounted (in free air). This could help if there is an issue with a ground loop from your mounting.

  • If possible, try testing with the LabJack and thermocouple in a different environment. This can help establish whether something in your test location is causing the issue.

Wrong temperature

If the issue appears to be a static error, for example regularly reading 5 above the expected temperature, there is typically a problem with either the CJC setup, some issue with the thermocouple hardware setup, or some issue with the conversion math.

Is the CJC temperature correct?

See the Cold Junction compensation thermocouple basics section above.

If using the AIN_EF feature on a supported T-series device, read AIN#_EF_READ_C for the CJC temperature.

If there is a +10 degree error in the measurement of the cold junction, you will see a roughly +10 degree error in the calculated value for absolute temperature of the remote end of the thermocouple.

  1. If there appears to be a static error, you may be able to simply add an offset to your CJC temperature reading to account for it.

  2. If doing CJC, note that the internal temperature sensors on LabJack devices sometimes do not properly reflect the cold junction temperature, and an external CJC temperature sensor such as an LM34 may be useful. For example, when using an expansion board such as the CB37, the cold junction may be at room temperature whereas the LabJack internal temperature sensor is a few degrees warmer than room temperature due to self heating from the LabJack. These errors in the cold junction temperature reading may be worse in environments with a significant temperature gradient.

Is the raw thermocouple voltage as expected?

If using the AIN_EF feature on a supported T-series device, read AIN#_EF_READ_B for the raw thermocouple voltage.

Test against a 0 V condition: Bring the remote end (hot junction) of the thermocouple near the local end (cold junction) to check the condition where both ends are at the same temperature. You should read very close to 0 volts, within a few hundred microvolts typically.  A thermocouple gives you a voltage related to the difference in temperature between the 2 ends, so if the cold junction and hot junction are at the same temperature, the voltage difference measured should be close to 0. This is a good way to test the thermocouple setup if you are uncertain about the cold junction and/or hot junction temperatures.

Test any Arbitrary temperature: To check at some arbitrary temperature difference, see the NIST tables linked in the thermocouple basics section above. For example, say you have a Type K thermocouple where the cold junction is at an expected temperature of 25 °C and hot junction at 35 °C:
The voltage at the 25 °C end is 1.000 mV and the voltage at the 35 °C end is 1.407 mV, so the difference is 407 µV. If the LabJack is reporting 407 µV, then continue to step 3.  If not, use a DMM to measure across the screw-terminals where the +/- leads are connected to see if the DMM reading is different or agrees with the LabJack.

If the DMM does not agree:

  • There may be a ground offset issue (see the complications section above).

  • If you are taking a differential measurement, ensure that you have a reference to GND (see the connections section above).

  • If using a differential connection, there could perhaps be an issue with common mode voltages; check the voltage on each AIN input compared to GND.  Both should be near 0 and within the typical measurement range of your device.

Is the math correct?

If the CJC temperature is right and the raw thermocouple voltage is right, then there may be something wrong with the conversion math. Note that our conversion functions typically only support the ranges (and associated reverse coefficient polynomials) described on the NIST website for conversion.

LJTick-InAmp Troubleshooting

Check that the LJTick-InAmp offset is correct and that the noise level is as expected.

With signals connected to the LJTIA, a DMM can be used to measure the differential input INA+ versus INA-, and also measure the output (versus GND), and confirm that you see Vout = (gain*Vdiff) + offset.

See Appendix C.1 of the LJTIA Datasheet.

JavaScript errors detected

Please note, these errors can depend on your browser setup.

If this problem persists, please contact our support.