Thermal conductivity errors, also called stem effects, from your temperature sensor could be giving you a false sense of security about the accuracy of your temperature measurements.

Figure 1. A controlled experiment consisting of a container of hot water surrounded by a reservoir of cold water measures errors caused by the stem effect.

Many factors determine the errors produced by your temperature sensor. Inherent errors caused by the elements, insulation, sheath material, junction style, etc., are fixed and can be determined by calibrating your thermocouple against a standard reference. Just because your sensor has been calibrated against a standard traceable to the National Institute of Standards and Technology (NIST), however, does not mean that your sensor is reading the temperature of your application accurately. Other factors such as radiation, convection and conduction also create errors. Because conduction tends to create the largest errors, this article deals with conduction errors and how to eliminate them. It is oriented specifically toward thermocouples although many of the comments also apply to resistance temperature detectors (RTDs), thermistors, bimetal thermometers and other sensing devices.

Contact temperature sensors such as thermocouples, RTDs, thermistors and bimetal thermometers measure only their own temperatures. Thus, the object of the design engineer is to select the correct sensor design and install it such that it measures the temperature of interest.



Figure 2. Three conventional thermocouple designs -- exposed, grounded and insulated junction -- were tested in the controlled experiment to determine stem effects on temperature measurement accuracy.

A controlled experiment was designed to measure the errors caused by "stem effect" with various types of thermal junction styles. The stem effect is defined as the error caused by heat conduction along the sensor's stem via the steel probe, wires and insulation. The experiment consisted of a container of hot water surrounded by a reservoir of cold water (figure 1). The hot water was kept hot by an immersion heater and the cold water was kept cold by a simple circulating pump and cooling radiator. A gradient of about 80oF (44oC) was thus maintained over an extended period of time.

Four thermocouples, each with a different style of thermal junction, were installed in the wall of the hot water tank and aligned flush with the inner surface of the tank.1 Thus, the thermocouples were in contact with the hot water. All thermocouples had iron/constantan elements, and their outputs were monitored by the same meter through a thermocouple selector switch. In addition, all thermocouples were made from the same lot of wire, and resistances were matched. The four thermal junction styles used were:



Figure 3. The sensing tip and the adjacent ribbons are parallel to the plane of heat flow, and because the ribbons on both sides of the junction are heated simultaneously with the junction, the error caused by the stem effect is eliminated.

Thermocouple 1 (T1). The first thermocouple used an exposed-bead weld junction (figure 2).

Thermocouple 2 (T2). The second thermocouple in the experiment used a grounded junction. In this junction system, the sheath is welded closed and the thermal junction is electrically in contact with the sheath.

Thermocouple 3 (T3). The third thermocouple in the experiment used an insulated or ungrounded junction. In this junction style, the sheath is welded closed but the thermal junction is electrically isolated from the sheath. All RTDs and thermistors are similar in design to this insulated junction. The sensing element is electrically insulated and then enclosed in a metallic sheath that protects the sensing element from environment abuse such as corrosion, oxidation, erosion and chemical attack.

Thermocouple 4 (T4). The fourth thermocouple in the experiment used a right-angle ribbon junction. In this configuration, the right-angle ribbon sensor has a junction thickness of 0.003".2 The extension leads in the vicinity of the junction also are ribbons and lie in the same plane as the sensing junction for a distance of at least 20 times the thickness of the junction. Thus, the sensing tip and the adjacent ribbons are parallel to the plane of heat flow, and because the ribbons on both sides of the junction are heated simultaneously with the junction, the error caused by the stem effect is eliminated (figure 3).



Table 1. The experiment showed stem effect errors on some thermocouple junctions.

Measurements were taken from these four thermocouples over a 16-hr period (table 1). Errors of from 10 percent to 66 percent were observed using conventional-style thermocouples. The ribbon thermocouple design eliminated the conduction error caused by the stem effect. This was attributed to its design wherein the thermal junction and the ribbon leads in the vicinity of the junction are in the same plane and heated simultaneously. Also, the distance from the junction to the longitudinal axis of the probe divided by the thickness of the junction forms a ratio of 20 to 1 or greater.

Applying this requirement to conventional temperature sensors dictates that the sensor be installed such that it is parallel to the plane of heat flow for a distance of at least 20 times its diameter. For example, a 0.125" round stainless steel probe must be positioned in the oven so that it lies parallel to the plane of heat flow for 2.5". This can be accomplished by one of two methods: Make a 90o bend in the probe 2.5" from its tip; or position the mounting so that the probe is parallel to the plane of heat flow.

When installing the sensor parallel to the plane of heat flow, it still must be immersed a distance of at least 20 diameters. For a 0.375" dia. probe, this distance is 7.5".



Figures 4 and 5. This schematic (left) is of an oven with heating elements at the bottom. The isotherms are shown as broken lines. The sketch on the right shows the isotherms in an oven with heaters on all sides. In all instances, the correct orientation requires that the thermocouple be installed parallel to the isotherms for a length equal to 20 times the probe diameter. A right-angle thermocouple can be installed without any 90o bends because the probe incorporates this requirement in its design.

Figure 4 is a schematic of a an oven with heating elements at the bottom. The isotherms are shown as broken lines. Figure 5 is a similar sketch showing the isotherms in an oven with heaters on all sides. In all instances, the correct orientation requires that the thermocouple be installed parallel to the isotherms for a length equal to 20 times the probe diameter. A right-angle thermocouple such as the one used in the test described can be installed into an oven without any 90 obends because the probe incorporates this requirement in its design.

A simple test can be performed to determine if your sensor is producing a significant error caused by the stem effect. Replace your current temperature sensor with a similar sensor that has a much smaller diameter. Insert this sensor to the same depth as the original sensor. Compare the indicated temperatures of the two sensors. If the sensor with the smaller diameter indicates a higher temperature than the larger diameter, you have a significant error.

Minimizing errors caused by conduction along the stem of the sensor will greatly improve the accuracy of the measured temperatures.



Links