10.2.2 Detailed Design Procedure

The biasing resistors (R_PU and R_PD) serve two purposes. The first purpose is to set the common-mode voltage of the thermocouple to within the specified voltage range of the device. The second purpose is to offer a weak pullup and pulldown to detect an open thermocouple lead. When one of the thermocouple leads fails open, the positive input will be pulled to VDD and the negative input will be pulled to GND. The ADC consequently reads a full-scale value, which is outside the normal measurement range of the thermocouple voltage, to indicate this failure condition. When choosing the values of the biasing resistors, care must be taken so that the biasing current does not degrade measurement accuracy. The biasing current flows through the thermocouple and can cause self-heating and additional voltage drops across the thermocouple leads. Typical values for the biasing resistors range from 1 MΩ to 50 MΩ.

Although the device digital filter attenuates high-frequency components of noise, TI recommends providing a first-order, passive RC filter at the inputs to further improve performance. The differential RC filter formed by R_DIFFA, R_DIFFB, and the differential capacitor C_DIFF offers a cutoff frequency that is calculated using Equation 5. While the digital filter of the ADS1118 strongly attenuates high-frequency components of noise, TI recommends to provide a first-order, passive RC filter to further suppress high-frequency noise and avoid aliasing. Care must be taken when choosing the filter resistor values because the input currents flowing into and out of the device cause a voltage drop across the resistors. This voltage drop shows up as an additional offset error at the ADC inputs. TI recommends limiting the filter resistor values to below 1 kΩ.

Two common-mode filter capacitors (C_CMA and C_CMB) are also added to offer attenuation of high-frequency, common-mode noise components. TI recommends that the differential capacitor C_DIFF be at least an order of magnitude (10x) larger than these common-mode capacitors because mismatches in the common-mode capacitors can convert common-mode noise into differential noise.

The highest measurement resolution is achieved when the largest potential input signal is slightly lower than the FSR of the ADC. From the design requirement, the maximum thermocouple voltage (V_TC) occurs at a thermocouple temperature (T_TC) of 1250°C. At this temperature, V_TC = 50.644 mV, as defined in the tables published by the National Institute of Standards and Technology (NIST) using a cold-junction temperature (T_CJ) of 0°C. A thermocouple produces an output voltage that is proportional to the temperature difference between the thermocouple tip and the cold junction. If the cold junction is at a temperature below 0°C, the thermocouple produces a voltage larger than 50.644 mV. The isothermal block area is constrained by the operating temperature range of the device. Therefore, the isothermal block temperature is limited to –40°C. A K-type thermocouple at T_TC = 1250°C produces an output voltage of V_TC = 50.644 mV – (–1.527 mV) = 52.171 mV when referenced to a cold-junction temperature of T_CJ = –40°C. The device offers a full-scale range of ±0.256 V and that is what is used in this application example.

The device integrates a high-precision temperature sensor that can be used to measure the temperature of the cold junction. The temperature sensor mode is enabled by setting bit TS_MODE = 1 in the Config register. The accuracy of the overall temperature sensor depends on how accurately the ADS1118 can measure the cold junction, and hence, careful component placement and PCB layout considerations must be employed for designing an accurate thermocouple system. The ADS1118 Evaluation Module provides a good starting point and offers an example to achieve good cold-junction compensation performance. The ADS1118 Evaluation Module uses the same schematic as shown in Figure 50, except with only one thermocouple channel connected. Refer to the application note, Precision Thermocouple Measurement With the ADS1118, SBAA189, for details on how to optimize your component placement and layout to achieve good cold-junction compensation performance.

The calculation procedure to achieve cold-junction compensation can be done in several ways. A typical way is to interleave readings between the thermocouple inputs and the temperature sensor. That is, acquire one on-chip temperature result, T_CJ, for every thermocouple ADC voltage measured, V_TC. To account for the cold junction, first convert the temperature sensor reading within the ADS1118 to a voltage (V_CJ) that is proportional to the thermocouple currently being used. This process is generally accomplished by performing a reverse lookup on the table used for the thermocouple voltage-to-temperature conversion. Adding these two voltages yields the thermocouple-compensated voltage (V_Actual), where V_Actual = V_CJ + V_TC. V_Actual is then converted to a temperature (T_Actual) using the same NIST lookup table. A block diagram showing this process is given in Figure 51. Refer to the application note, Precision Thermocouple Measurement With the ADS1118, SBAA189, for a detailed explanation of this method.

Figure 51. Software Flow Block Diagram

Figure 52 and Figure 53 show the measurement results. The measurements are taken at T_A = T_CJ = 25°C. A system offset calibration is performed at T_TC = 25°C that equates to V_TC = 0 V when T_CJ = 25°C. No gain calibration was performed during the measurements. The data in Figure 52 are taken using a precision voltage source as the input signal instead of a thermocouple. The solid black line in Figure 53 is the respective temperature measurement error and is calculated from the data in Figure 52 using the NIST tables. The solid black line in Figure 53 is the measurement error due to the ADC gain and nonlinearity error. The dashed blue lines in Figure 53 include the guard band for the temperature sensor inaccuracy (±0.5°C), in addition to the device gain and nonlinearity error. Note that the measurement results in Figure 52 and Figure 53 do not account for the thermocouple inaccuracy that must also be considered while designing a thermocouple measurement system.