Temperature is the most commonly measured physical property, and thermocouples are inexpensive, versatile, and widely used temperature sensors. Many temperature-measurement applications require attachment of thermocouples to metal objects or surfaces energized with voltages greater than the millivolt levels associated with thermocouple measurements. Thermocouples are available in a variety of types with varying usable temperature ranges. The output voltages are nonlinear functions of the temperature in the range of 0-70 mV. But most temperatures are not at the type-extremes, so typical readings are often 10mV or less. The measurement input circuitry must amplify these low voltages to achieve reasonable data resolution. Unfortunately, voltages externally superimposed on a thermocouple will be amplified as well, resulting in problems ranging from noticeable measurement errors to the inability to read a temperature. In particular, the presence of common-mode voltage with respect to earth ground will interfere with most thermocouple measurement techniques because a portion of the common-mode voltage adds to the thermocouple output (see Photo 1). Fundamental Concepts Common-mode voltage is the voltage that is measured with respect to a common-mode reference point and present on (or common to) both sides of the signal voltage. In a complete system—consisting of a temperature-measuring device and one or more thermocouples measuring electrically energized objects—there are two common-mode reference points. The common-mode reference point for an amplifier IC is generally the analog common midway between the bipolar power supply rails. For this reason, the maximum tolerable common-mode voltage for most signal amplifier ICs is <15 V. Most frequently, the common-mode reference point for an overall system is earth ground. If a system common-mode requirement is low, it’s permissible to have analog common nominally at earth-ground potential.
There are many applications in manufacturing and research in which the temperature of electrically energized metal must be measured. Sometimes, the voltages are pure DC, but more often, they are AC or have an AC component. Examples of DC common-mode environments are heatsinks in regulated power supplies, power amplifiers, and DC motor drives. Some examples of AC common-mode environments are operating AC switchgears, motor or generator windings, and electric heating elements in household appliances. In many DC applications, the pure DC contains AC ripple, which turns the less troublesome DC environment into the much more difficult AC environment. Methods of Measuring The simplest way of measuring a thermocouple connected to an electrically energized surface is with a self-contained instrument that is either self-powered or does not require any power. An analog meter movement calibrated for a type of thermocouple and a range of temperature requires no power source and works well at elevated temperature's but lacks sensitivity at ambient levels. A handheld, battery-powered temperature meter or a digital multimeter with a thermocouple adapter overcomes the sensitivity drawback but requires regular replacement of the batteries. Seemingly, a thermocouple formed from two dissimilar wires and connected to a third metal in contact with both wires will short itself out. However, this does not happen because the voltage measurable between the other ends of the wires is developed over the length of the two wires rather than at the junction. If the two ends of a single conductor are placed at two different temperatures, a net emf is generated. The emf is called the absolute Seebeck emf. A practical thermocouple measurement is made by measuring the difference between two absolute Seebeck emf’s for two dissimilar metals. Process Transmitters. These signal conditioners convert thermocouple outputs to high level voltage or current signals. The devices perform optimally when located close to the thermocouple. Most transmitters are isolated to several hundred volts by optical isolators or transformers and have built-in linearization circuitry. Models with 4-20 mA current outputs offer high noise immunity for reliable transmission over substantial distances. Voltage outputs of 0-10 VDC are also available but are better suited for localized T/C-to-volts conversion. The drawbacks of using these devices are the high cost per channel and the need for periodic calibration. The major benefit is the option of using a less expensive, nonisolated data acquisition device if most of the other system signals do not require isolation. Insulated Thermocouples. These devices often give a user a false sense of security. If physically connected to energized objects, the thermocouples should be treated as if they were uninsulated because the unit’s protection tube, or housing, is generally metallic and therefore energized. Even if the housing is not energized, the thermocouple itself is capacitively coupled to the energized device. If the applied voltage is AC, a substantial amount of voltage can couple with the amplifier input, even though there is no DC path. Digital Panel Meters for Thermocouples. These meters, generally powered by AC line voltage, are inherently isolated to a few hundred volts, usually limited by the spacing of edge connector pins and internal transformer insulation. The instruments are commonly type-and range-specific. Because digital panel meters use dual-slope, integrating A/D converters, the AC line noise not rejected by the isolation is averaged out in the typical reading time of 250 ms. As with transmitters, the cost per thermocouple is a drawback. And if you have to interface these instruments with a computer system, the price increases because the digital interface must also be isolated from the A/D converter. Digital Multimeters with Relay-Type Scanners. These instruments can measure multiple energized thermocouples of varying types. The instruments scan quickly enough to monitor a significant number of temperatures, and the integrating A/D converters digitize slowly enough to average out most of the line noise that gets through the isolation and rides on the thermocouple voltages. Keep in mind, though, that if two sequentially read channels are connected to widely disparate voltages (e.g., opposite polarities of 200 VDC with respect to earth ground), the first reading will charge the internal capacitance to earth ground to one polarity, and the second reading will close a relay across a 400 VDC potential difference. This can damage the relays or reduce their useful life substantially (see Figure 3 page 42). Multiplexed High-Speed A/D Converters. These A/D converters are inherently isolated from the host computer, but the channel-to-channel isolation is an issue if solid-state multiplexers are used instead of relays. If relays are used — solid-state, reed, or armature types — the channel-to-channel isolation can reach several hundred volts. Unfortunately, the speed at which A/D converters must operate to scan hundreds of channels precludes such systems from using integrating-type A/D converters. Line-cycle averaging—which is a process that basically synchronizes a series of readings with the power line frequency and averages the series—nulls out the effect of most AC line noise on the signals. The systems have excellent cost per channel, and they can support mixed types and ranges of thermocouples, gathering data into computers for convenient analysis on spreadsheets and graphs. Additionally, the calibration of these systems is localized and typically done by computer, with minimal external equipment. Nothing’s Perfect There is no perfect system for measuring electrically energized thermocouples. The highest attainable common-mode rejection ratios, or CMRRs, do not eliminate all significant noise or errors. In large systems, where hundreds of thermocouples are connected, there is considerable economic incentive to use equipment that is least capable of rejecting noise by the bandwidth-limiting effects of integration or averaging. But some system design compromises can substantially improve results. For example, a system measuring the temperatures of connection lugs on energized heating elements would be more easily monitored if the grounded neutral end of each test element were monitored instead of the hot end. If it is necessary to measure a series of thermocouples at high potential voltages, grouping sequentially scanned channels at the same potentials—instead of alternating them between AC phase lines or opposite DC power supply polarities—will increase relay life. When only a few thermocouples in a system are measuring electrically energized surfaces, locally isolating the channels will minimize the problems that energized channels add to the overall system. The trick is to develop realistic expectations about CMRRs and what they really mean. Higher is always better, but the highest numbers attainable can still result in noticeable errors in measured signals. Use any averaging capabilities the system offers to minimize AC noise in readings and reduce data-storage memory requirements. This is easily justified if you remember that actual temperatures do not change instantaneously or flicker back and forth. Consider and allow for all possible error sources. Thermocouples themselves are good, but they’re not perfect. |