The circuit shown in Figure 1 is a complete thermocouple system based on the AD7793 24-bit Σ-Δ ADC . The AD7793 is a low-power, low-noise, complete analog front-end suitable for high-precision measurement applications. It has a built-in PGA, reference voltage source, clock and excitation current, thereby greatly simplifying thermocouple system design. System peak-to-peak noise is approximately 0.02°C.

The AD7793's maximum power consumption is only 500 μA, making it suitable for low-power applications such as smart transmitters where the power consumption of the entire transmitter must be less than 4 mA. The AD7793 also has a shutdown option. In this mode, the entire ADC and its auxiliary functions are shut down, reducing the maximum power consumption of the device to 1 μA.

The AD7793 provides an integrated thermocouple solution that interfaces directly with thermocouples. Cold junction compensation is provided by a thermistor and a precision resistor. The circuit only requires these external components to perform cold junction measurements, and some simple RC filters to meet electromagnetic compatibility (EMC) requirements.

This circuit uses a T-type thermocouple. Constructed of copper and constantan, this thermocouple has a temperature measurement range of −200°C to +400°C and produces a typical temperature-dependent voltage of 40 μV/°C.

The transfer function of a thermocouple is not linear. The response is very nearly linear over the temperature range of 0°C to +60°C. However, over a wider temperature range a linearization procedure must be used.

The test circuit does not include linearization; therefore, the useful measurement range of this circuit is 0°C to +60°C. Within this temperature range, the thermocouple produces a voltage from 0 mV to 2.4 mV. An internal 1.17 V reference voltage is used for thermocouple conversion. Therefore, the AD7793 is configured for a gain of 128.

The AD7793 operates on a single power supply, and the signal generated by the thermocouple must be biased above ground to be within the range supported by the ADC. For a gain of 128x, the absolute voltage at the analog input must be in the range GND + 300 mV to AVDD – 1.1 V.

The AD7793's on-chip integrated bias voltage generator biases the thermocouple signal to a common-mode voltage of AVDD/2, ensuring that the input voltage limits are met with considerable margin.

The thermistor has a value of 1 kΩ at +25°C, a typical value of 815 Ω at 0°C, and a typical value of 1040 Ω at +30°C. Assuming a linear transfer function from 0°C to 30°C, the relationship between cold junction temperature and thermistor R is:

Cold junction temperature = 30 × (R – 815)/(1040 – 815)

The 1 mA excitation current of the AD7793 is used to power the thermistor and the 2 kΩ precision resistor. The reference voltage is generated using this 2 kΩ external precision resistor. This architecture provides a ratiometric configuration where excitation current is used to power the thermistor and generate a reference voltage. Therefore, deviations in the excitation current value do not change the accuracy of the system.

The AD7793 operates at a gain of 1 when sampling the thermistor channel. For a maximum cold junction temperature of +30°C, the maximum voltage developed across the thermistor is 1 mA × 1040 Ω = 1.04 V.

The selection condition of the thermistor is: the maximum voltage generated on the thermistor multiplied by the PGA gain is less than or equal to the voltage generated on the precision resistor.

For the converted value of ADC_CODE, the corresponding thermistor value R is equal to:

R = (ADC_CODE – 0x800000) × 2000/2 ²³

The output compliance voltage of the AD7793 IOUT1 pin also needs to be considered. When using 1 mA excitation current, the output compliance voltage is equal to AVDD – 1.1 V. From the above calculation, we can see that the circuit meets this requirement because the maximum voltage of IOUT1 is equal to the voltage on the precision resistor plus the voltage on the thermistor, which is equal to 2 V + 1.04 V = 3.04 V.

The AD7793 operates at an output data rate of 16.7 Hz. For every 10 thermocouple conversion results read, 1 thermistor conversion result is read. The corresponding temperature is equal to:

Temperature = thermocouple temperature + cold junction temperature

The conversion results of the AD7793 are processed by the analog microcontroller ADuC832 and the resulting temperature is displayed on the LCD display.

This thermocouple is designed to be powered by a 6 V (2 x 3 V lithium battery) battery. A diode reduces the 6 V voltage to a level suitable for the AD7793 and the analog microcontroller ADuC832. There is an RC filter between the ADuC832 power supply and the AD7793 power supply to reduce the power supply digital noise entering the AD7793.

Figure 2 shows the voltage developed on a T-type thermocouple as a function of temperature. The area within the circle is from 0°C to +60°C, where the transfer function is nearly linear.

When the system is at room temperature, the thermistor should indicate a value at room temperature. The thermistor indicates the relative temperature relative to the cold junction temperature, that is, the temperature difference between the cold junction (thermistor) and the thermocouple. Therefore, at room temperature, the thermocouple should indicate 0°C. .

If you place the thermocouple in an ice bucket, the thermistor still measures the ambient (cold junction) temperature. The thermocouple should indicate the negative of the thermistor value such that the total temperature equals 0.

Finally, for an output data rate of 16.7 Hz and a gain of 128, the rms noise of the AD7793 is equal to 0.088 μV. Peak-to-peak noise is equal to:

6.6 × rms noise = 6.6 × 0.088 μV = 0.581 μV

If the thermocouple's sensitivity is exactly 40 μV/°C, the thermocouple's temperature measurement resolution is:

0.581 μV ÷ 40 μV = 0.014°C

Figure 3 shows the actual test board. The system was evaluated as follows: the thermistor temperature, thermocouple temperature, and resolution were measured at room temperature and with the thermocouple placed in an ice bucket. The results are shown in Table 1.

From Table 1, we can see that the thermocouple reports the correct temperature, but the thermistor has an error of 0.3°C. This is the system accuracy without linearization. If the thermocouple and thermistor are linearized, the system accuracy will be improved and the system will be able to measure a wider temperature range.

If the difference between the minimum and maximum temperature readings is calculated every 10 readings, the peak-to-peak noise in temperature is 0.02°C. Therefore, the actual peak-to-peak resolution is very close to the expected value.