The circuit shown in Figure 1 is a complete isolated 4-channel temperature measurement circuit optimized for performance, input flexibility, stability, and low cost. It supports all types of thermocouples (with cold junction compensation) and any type of RTD (resistance temperature detector, two-, three- or four-wire connection configuration) with resistance up to 4 kΩ.

RTD excitation current can be programmed for optimal noise and linearity performance.

RTD measurement accuracy is 0.1°C (typ) and Class K thermocouple measurement accuracy is 0.05°C (typ) because the 16-bit digital temperature sensor ADT7310 is used for cold junction compensation. This circuit uses a 4-channel, 24-bit, Σ-Δ ADC AD7193 . This device integrates a PGA on-chip and has high precision and low noise characteristics.

Input transient and overvoltage protection is provided by low-leakage transient voltage suppressors (TVS) and Schottky diodes. The SPI-compatible digital inputs and outputs are isolated (2500 V rms), and the circuit operates from a fully isolated power supply.

Application Note AN-0971 Introduction to Temperature Measurement

Thermocouples and RTDs (Resistance Temperature Detectors) are the most commonly used sensors for temperature measurement in industrial applications. Thermocouples can measure extremely high temperatures, up to around +2300°C, and have fast response times (measurements are completed instantly). RTDs are more accurate and stable than thermocouples, and the resistance of long wires (hundreds of meters) to remote RTDs can be compensated for using a three-wire or four-wire connection.

A thermocouple consists of two different metal wires connected at one end. Place the connected end where the temperature needs to be measured, which is called the measurement node. The other end is connected to a precision voltage measurement unit. This connection is called the reference junction, or cold junction. The temperature difference between the measuring junction and the cold junction produces a voltage (called the Seebeck effect voltage). The value is related to the temperature difference between the two junctions. The signal generated by the temperature difference usually ranges from a few microvolts to tens of millivolts. , depending on the temperature difference.

For example, a Class K thermocouple measures −200°C to +1350°C and has an output range of approximately −10 mV to +60 mV. For the signal chain, it is important to keep the impedance as high as possible and the leakage current as low as possible so that the measured voltage has the highest accuracy. To convert this voltage to an absolute temperature, the cold junction temperature must be known accurately. Generally speaking, 1°C to 2°C is sufficient. Although the absolute temperature error will directly increase due to the cold junction temperature measurement error, higher cold junction temperature measurement accuracy is beneficial.

RTDs are made of pure materials, such as platinum, nickel, or copper, and have predictable changes in resistance as temperature changes. The most common RTD material is platinum (Pt100 and Pt1000).

One way to accurately determine resistance is to measure the RTD voltage produced by a constant current source. Errors in the current source (such as ratiometric measurements) can be eliminated by reducing the measured value to the voltage across a reference resistor (driven by the same current). Minimizing leakage current in the current path is important to achieve high accuracy, since the excitation current is typically only a few hundred microamps to prevent self-heating.

For industrial field applications, high performance and protection against high voltage transient events and DC overvoltage conditions are important design considerations.

Working principle of this circuit

The circuit shown in Figure 1 is designed for precision temperature measurement applications in industrial field environments and is optimized for flexibility, performance, stability, and cost. This circuit uses the low-noise, 24-bit Σ-Δ ADC AD7193 to ensure that the entire circuit has high resolution and linearity.

The AD5201 33-bit digital potentiometer , AD8603 operational amplifier , and single-channel switch ADG702 form a simple programmable current source and bias voltage buffer for RTD and thermocouple measurements. The ADG738 routes current sources to active RTD channels, allowing wire resistance compensation for three-wire RTD configurations.

The ADT7310 digital SPI temperature sensor has ±0.8°C maximum accuracy over the −40°C to +105°C temperature range (with a +5 V supply) and is used for cold junction compensation in thermocouple measurements. The ADR3440 is a low noise, high accuracy, 4.096 V reference connected to REFIN1(+)/REFIN1(−) of the AD7193 for thermocouple measurements.

Analog to Digital Converter

The AD7193 is a low-noise complete analog front end suitable for high-precision measurement applications. It integrates a low-noise, 24-bit Σ-Δ analog-to-digital converter (ADC). The ADC features high resolution, low nonlinearity and low noise performance, as well as extremely high 50 Hz/60 Hz rejection. The data output rate can vary from 4.7 Hz (24-bit effective resolution, Gain = 1) to 4.8 kHz (18.6-bit effective resolution, Gain = 1). An on-chip low-noise PGA amplifies small differential signals from thermocouples or RTDs with gains of 1 to 128, allowing direct interfacing. The gain stage buffer has a high input impedance and limits input leakage current to ±3 nA (max). The gain of the AD7193 must be configured appropriately based on the temperature range and sensor type. An on-chip multiplexer allows four differential input channels to share the same ADC core, saving space and cost.

Programmable current sources for RTDs and bias voltage generation circuits for thermocouples

RTD measurements require a low-noise current source to drive the RTD and reference resistor. Thermocouples, on the other hand, require a common-mode bias voltage to convert the smaller thermocouple voltage into the input range of the AD7193. The circuit shown in Figure 2 meets both requirements and uses the AD8603 low-noise CMOS rail-to-rail input/output op amp, which has a maximum input bias current of only 1 pA and a maximum offset voltage of 50 μ V; at the same time, It also uses a single-channel CMOS low-voltage 2Ω single-pole single-throw (SPST) switch ADG702, and an 8-channel matrix switch ADG738.

With the ADG738 turned on and the ADG702 turned off, the AD8603 can be used as a low-noise, low output impedance unity-gain buffer in thermocouple applications. The voltage from the AD5201 digital potentiometer is buffered and then applied to the thermocouple common mode voltage, which is typically half the supply voltage, which is 2.5 V. The AD5201 33-bit digital potentiometer is driven by the ADR3440 low-drift (5 ppm/°C) 4.096 V reference to obtain the required accuracy.

When ADG738 is turned on and ADG702 is turned off, AD8603 generates RTD excitation current, that is, I _EXC = V _W /R _REF .

Temperature measurement is a high-precision, low-speed application, so there is sufficient settling time to switch a single current source among all 4 channels, providing excellent channel-to-channel matching, low cost, and small PCB size.

The ADG738 is an 8:1 multiplexer that switches current sources between channels. To support two-wire, three-wire, and four-wire RTD configurations, two switches are required for each of the four channels.

In many applications, the RTD may be placed remotely from the measurement circuitry. Long lead resistors can produce large errors, especially when used with low resistance RTDs. To minimize lead resistance effects, a three-wire RTD configuration is supported, as shown in Figure 3.

关断ADG738的S1，同时打开S2，则AD7193输入端的电压为 V₁。打开S1，同时关断S2，则AD7193输入端电压为 V₂。RTD传感器两端的电压为V_RTD，而电流源的激励电流为I_EXC。V₁ 和 V₂包括引线电阻产生的误差，如下所示：

假定 R_W1 = R_W2 = R_W3，然后结合等式1、等式2和等式3，可得：

等式5表示三线式配置需要分别进行两次测量 (V₁ 和V₂) 才能计算R_RTD，因此输出数据速率有所下降。在很多应用中，这并不是个问题。

四线式RTD连接要求具有两个额外的检测线路，但对导线电阻不敏感，且仅需进行一次测量。

图4总结了双线式、三线式和四线式RTD和热电偶应用的连接器配置和跳线位置。

保护电路

在制造过程中和现场使用时，都有可能产生瞬变和过压条件。为了获得较高的保护水平，有必要使用外部保护电路，补充IC的内部集成保护电路。外部保护功能会增加额外的电容、电阻和漏电流。这些效应应当仔细考虑，以获得高精度水平。额外保护电路如图5所示。

漏电流会对RTD测量造成巨大影响，应仔细考虑。当较长的热电偶引线具有极高电阻时，漏电流也会对热电偶测量产生一些误差。

本电路中，PTVS30VP1UP瞬变电压抑制器(TVS)可快速箝位任何瞬变电压至30 V(25°C时典型漏电流仅1 nA)。选择30 V TVS，以便支持30 V直流过压。使用1.69 kΩ电阻，后接低泄露BAV199LT1G肖特基二极管，用于在瞬变和直流过压事件发生时将电压箝位至5 V供电轨。在30 V直流过压条件下，1.69 kΩ电阻将流过外部二极管的电流限制为15 mA。为了确保供电轨能够吸取该电流，可使用齐纳二极管将供电轨进行箝位处理，以保证它不超过连接电源的任意IC的绝对最大额定值。选择5.6 V齐纳二极管(NZH5V6B)实现这一目的。300Ω电阻可进一步限制有可能进入AD7193或ADG738的电流。

隔离

ADuM5401和ADuM1280使用ADI iCoupler®技术，在测量端和电路控制器端之间提供2500 V rms隔离电压。ADuM5401还提供隔离电源，用于电路的测量端。ADuM5401采用了isoPower 技术，该项技术使用高频开关元件，通过变压器传输电力。设计印刷电路板(PCB)布局时应特别小心，必须符合相关辐射标准。有关电路板布局建议，请参考应用笔记AN-0971。

热电偶配置测试结果

电路的性能高度依赖于传感器和AD7193的配置。K类热电偶输出变化范围为−10 mV至+60 mV，对应温度范围为−200°C至+1350°C。AD7193 PGA配置为G = 32。PGA电压摆幅范围为−320 mV至+1.92 V，即2.24 V p-p。斩波使能时，50 Hz/60Hz噪声抑制使能，滤波器字FS[9:0] = 96，1024个样本的噪声分布直方图如图6所示。

AD7193分辨率为24位，即2²⁴ = 16,777,216个代码。AD7193的全动态范围为2 ×V_REF = 2 × 4.096 V = 8.192 V。位于PGA之后的热电偶输出电压仅为2.24 V p-p，并且不会完全占据AD7193的所有动态范围。因此，系统范围以2.24 V/8.192V系数降低。

噪声分布约为40个代码峰峰值。2.24 Vp-p测量范围内的无噪声代码分辨率为：

K类热电偶的满量程温度范围为−200°C至+1350°C，即1550°C p-p。因此16.8位无噪声代码分辨率相当于0.013°C无噪声温度分辨率。热电偶配置测试结果电路的性能高度依赖于传感器和AD7193的配置。K类热电偶输出变化范围为−10 mV至+60 mV，对应温度范围为−200°C至+1350°C。AD7193 PGA配置为G = 32。PGA电压摆幅范围为−320 mV至+1.92 V，即2.24 V p-p。斩波使能时，50 Hz/60Hz噪声抑制使能，滤波器字FS[9:0] = 96，1024个样本的噪声分布直方图如图6所示。

热电偶测量线性度

热电偶测量线性度图7显示K类热电偶系统的近似线性度。该曲线中，“冷结”温度为0°C。

Fluke 5700A校准仪提供分辨率为10 nV的高精度直流电压源，用于校准以及测试。图8中的电压误差位于0.2μV理想范围内，相当于大约0.004°C。该结果是系统在25°C时校准后的短期精度，此时没有温度漂移效应。本电路的主要误差来源于冷结补偿测量。在本电路中，ADT7310用于冷结补偿，典型误差为−0.05°C，采用5 V电源时，在−40°C至+105°C温度范围内的最差情况误差为±0.8°C。若使用3 V电源，则器件在该温度范围内具有±0.4°C的最大误差。

RTD配置测试结果

The default ADC gain setting is G = 8 for the Pt100 RTD and G = 1 for the Pt1000 RTD. The ADC's reference voltage is equal to the voltage across the 4.02 kΩ reference resistor. The temperature coefficient of Pt100 RTD is approximately 0.385Ω/°C and the resistance can be as high as 400Ω at +850°C. With a default excitation current of 400 μA , the maximum RTD voltage is approximately 160 mV. The ADC reference voltage is 4.02 kΩ × 400 μ A = 1.608 V. For G = 8, the maximum RTD voltage is 160 mV × 8 = 1.28 V, which is approximately 80% of the usable range.

For Pt1000 RTD, the maximum resistance at +850°C is approximately 4000Ω. The default excitation current is 380 μA , resulting in a maximum RTD voltage of 1.52 V. The reference voltage of the ADC is 4.02 kΩ × 380 μ A = 1.53 V. With the default gain setting of G = 1, the RTD maximum voltage can utilize almost all of the available range.

The general expression of RTD resistance R in terms of ADC code (Code), resolution (N), reference resistance (R _REF ) and gain (G) is as follows:

Leakage current from TVS, diodes, clamping diodes and ADCs is the largest source of error in RTD measurement circuits, although nanoamp devices are used in this design.

The total leakage current per input is 9 nA (3 nA from the AD7193 with buffer on; 5 nA from the clamping diode; 1 nA from the TVS diode). Therefore, all 4 channels will produce a maximum leakage current of 36 nA. The feedback loop in Figure 2 maintains a constant current through the reference resistor. This means that leakage current affects the RTD excitation current, creating errors. The default excitation current is 400 μA (Pt100) and 380 μA (Pt1000). For a Pt100 RTD, the approximate worst-case system error due to leakage current is:

For a Pt100 with a measurable range of −200°C to +850°C, this equates to a system accuracy of approximately:

The amount of error depends on the configuration of the inputs. After completing the input configuration, room temperature calibration can be performed to further reduce errors.

Experimentally show leakage current effects. Each channel is first configured as a four-wire RTD. The 100Ω fixed resistor is connected to channel 1 at the RTD location. The 0Ω resistors are connected to the inputs of the other 3 channels.

The gain is set to G = 1 and the excitation current is 380 μA (Pt1000 configuration).

Collect data, then remove the jumpers connecting channel 4, channel 3, and channel 2 to collect data for each condition. The results are shown in Figure 9.

The ADC code changes from approximately 437,800 to 437,600 and the corresponding measurement changes from 104.9015Ω to 104.8627 or 0.0388Ω. This represents a measurement error of approximately 0.1°C; however, the error can be eliminated by calibrating at room temperature with a fixed input configuration.