How Do You Use the MAX31865 RTD-to-Digital Converter with a PT100 Sensor?

Detailed Explanation

The MAX31865 from Analog Devices (formerly Maxim Integrated) replaces the discrete analog measurement chain required for RTD interfacing. Where a custom design needs a precision current source, low-drift reference resistor, instrumentation amplifier, and high-resolution ADC, the MAX31865 integrates all of these into a single 8-pin IC with an SPI interface. The host microcontroller configures it over SPI, triggers or schedules conversions, and reads back a 15-bit resistance code that maps directly to RTD resistance.

For the theory of PT100/PT1000 sensors — excitation current, 2/3/4-wire connections, the IEC 60751 standard, and when to choose an RTD over a thermistor — see PT100 and RTD fundamentals. This page covers the MAX31865 IC implementation specifically.

How the MAX31865 Works

The IC operates on a ratiometric principle. Bias voltage from the internal FORCE outputs drives current through an external reference resistor (R_REF) and the RTD — both in the same series current path. The 15-bit ADC measures the voltage across the RTD (RTDIN+/RTDIN−) and the voltage across the reference resistor (REFIN+/REFIN−), computing their ratio:

RTD_CODE = (V_RTD / V_RREF) × 2¹⁵
         = (R_RTD / R_REF) × 32768

Because both voltages share the same excitation current, any variation in that current cancels. Measurement accuracy depends only on the reference resistor's precision and the ADC linearity — not on a stable current source.

---

Hardware Connections

Reference Resistor and Excitation Path

The reference resistor connects between FORCE1 and REFIN+. REFIN− connects to RTDIN+ (the high side of the RTD), and RTDIN− connects to FORCE2, completing the excitation current loop:

FORCE1 → R_REF (430 Ω) → REFIN+
                          ↓
                         REFIN− = RTDIN+ → [PT100 RTD] → RTDIN− = FORCE2

Reference resistor selection:

Use a precision metal-film resistor (e.g. Vishay MRS or Panasonic ERA series). A ±1% resistor introduces approximately ±2.6°C systematic error at room temperature — the reference resistor is the primary accuracy bottleneck in the signal chain.

2-Wire, 3-Wire, and 4-Wire RTD Connections

2-wire (avoid for accuracy-critical applications): RTDIN+ and RTDIN− connect directly to the two RTD terminals, sharing the same wires as the excitation current. Cable resistance adds directly to the RTD reading, causing a positive temperature offset. Acceptable only for cable runs under approximately 0.5 m at low accuracy requirements.

3-wire (standard industrial practice): Three wires run to the RTD — one shared force/sense wire and two individual sense wires. The MAX31865 compensates for lead resistance by measuring it separately via FORCE2 during the conversion cycle. Set bit 4 of the configuration register to enable 3-wire compensation. This assumes all three leads have equal resistance, which is a valid assumption for matched cables.

4-wire (best accuracy): Two dedicated force wires carry the excitation current (via FORCE1 and FORCE2 through the RTD); two separate sense wires connect RTDIN+/RTDIN− directly at the RTD terminals, carrying negligible current. Lead resistance of the sense wires causes no voltage error. Use 4-wire for cable runs over approximately 5 m, or when accuracy better than ±0.5°C is required. This mode uses the same configuration register setting as 2-wire (bit 4 = 0).

SPI Interface Connections

The MAX31865 supports SPI Mode 1 or Mode 3 (CPHA = 1 in both cases — data clocked on the falling edge of SCLK). Maximum SCLK is typically 5 MHz per the datasheet. Pull CSB low to begin a transaction. If reads return 0xFF despite correct wiring, an SPI mode mismatch is the most likely cause — see diagnosing SPI CPOL/CPHA faults for a logic-analyser-based diagnosis approach.

Mount a 100 nF ceramic decoupling capacitor between VREG and GND as close to the IC as practical.

---

SPI Registers and Configuration

Read addresses are the register index (0x00–0x07); write addresses are the register index with bit 7 set (0x80–0x87).

Configuration register (read: 0x00, write: 0x80):

Recommended configuration values:

For battery-powered products, use single-shot mode: enable VBIAS, wait at least 10 ms for settling, trigger a conversion (write bit 5), wait approximately 75 ms (50 Hz filter), read the result, then disable VBIAS to eliminate quiescent current through R_REF.

RTD data registers (0x01 and 0x02):

A sequential two-byte read beginning at 0x01 returns the full 16-bit RTD register:

Bits [15:1] = 15-bit RTD ADC code (MSB first)
Bit  [0]    = Fault bit (1 = fault detected — check fault status register)

Fault status register (0x07):

---

Reading and Converting Temperature

Step 1 — Read the RTD Code

// Read 2 bytes from register 0x01 (RTD MSB) and 0x02 (RTD LSB)
uint8_t msb = spi_read_register(0x01);
uint8_t lsb = spi_read_register(0x02);

uint16_t raw = ((uint16_t)msb << 8) | lsb;

if (raw & 0x0001) {
    // Fault bit is set — read fault status register for diagnosis
    uint8_t fault = spi_read_register(0x07);
    handle_fault(fault);
    return;
}

uint16_t rtd_code = raw >> 1;  // 15-bit ADC code (fault bit removed)

Step 2 — Convert Code to Resistance

#define R_REF   430.0f   // Reference resistor in ohms (PT100)
#define R0      100.0f   // PT100 nominal resistance at 0°C

float r_rtd = ((float)rtd_code / 32768.0f) * R_REF;

Expected values for PT100 with a 430 Ω reference:

Temperature resolution: approximately 0.034°C per LSB (430 / 32768 Ω ÷ 0.385 Ω/°C), which is sufficient for virtually all PT100 applications.

Step 3 — Convert Resistance to Temperature

Linear approximation (0°C to approximately +200°C, typically within ±0.5°C):

// T ≈ (R_RTD - R0) / (R0 × α)
// α = 0.00385 Ω/Ω/°C (IEC 60751 alpha coefficient for standard platinum)
float temp_c = (r_rtd - R0) / (R0 * 0.00385f);

Callendar-Van Dusen polynomial (IEC 60751, better than ±0.1°C above 0°C):

For T ≥ 0°C, the IEC 60751 relationship is:

R(T) = R0 × (1 + A×T + B×T²)

where A = 3.9083 × 10⁻³ °C⁻¹ and B = −5.775 × 10⁻⁷ °C⁻².

Solving quadratically for T:

#define CVD_A   3.9083e-3f
#define CVD_B  -5.775e-7f

float temp_c = (-CVD_A + sqrtf(CVD_A * CVD_A - 4.0f * CVD_B * (1.0f - r_rtd / R0)))
               / (2.0f * CVD_B);

For temperatures below 0°C, an additional C term (−4.183 × 10⁻¹² °C⁻⁴) is required per IEC 60751. For sub-zero accuracy, a pre-computed IEC 60751 lookup table is the most practical firmware approach.

---

Fault Detection

Check the fault bit on every reading and inspect the fault status register when it is set. Latched faults persist until explicitly cleared — write bit 1 of the configuration register:

uint8_t config = spi_read_register(0x00);
spi_write_register(0x80, config | 0x02);  // Set bit 1 to clear fault status

---

Design Considerations

• Power-on settling: After enabling VBIAS (bit 7 of Config = 1), wait at least 10 ms before triggering a conversion. The bias circuit requires time to settle before the reference resistor voltage is stable enough for an accurate reading.
• 50 Hz filter for Australian designs: Australia operates on 50 Hz mains. Set the filter bit (bit 0 = 1 in Config) to reject 50 Hz interference from nearby power circuitry and long RTD cable runs. The 50 Hz filter setting extends the effective conversion time to approximately 75 ms.
• Reference resistor placement and quality: Mount the reference resistor close to the MAX31865. Use a precision metal-film resistor rated ±0.1% with a temperature coefficient of ≤ 25 ppm/°C. Carbon-film resistors (typically ±100–200 ppm/°C) will dominate the measurement error. Do not use a general-purpose 470 Ω or other E24-series value in place of 430 Ω — the mismatch shifts every temperature reading proportionally.
• Cable shielding for long RTD runs: RTD cables over approximately 3 m in industrial environments pick up 50/60 Hz and RF interference. Use shielded twisted-pair cable and connect the shield to a single ground point at the measurement electronics end. The MAX31865's built-in noise filter provides rejection of in-band mains noise. For broader guidance on reducing noise in sensor interfaces, see sensor signal conditioning basics.
• Single-shot mode for power-sensitive designs: In battery-powered products, disable VBIAS between readings. The quiescent current flowing through R_REF in continuous mode contributes to battery drain. In single-shot mode, current is only drawn during the brief conversion window.

For complex sensor integration — including RTD interface design, multi-channel data acquisition, PCB layout, and embedded firmware — Zeus Design provides end-to-end electronics engineering services.

Common Mistakes

• Wrong reference resistor value: The 430 Ω reference resistor is not interchangeable with the nearest E24-series value (470 Ω). A 10% mismatch shifts the RTD code and produces a proportional temperature offset that cannot be corrected by firmware calibration alone without also updating the R_REF constant. Use 430 Ω precisely, or update both the resistor and the software constant to match.
• Not enabling VBIAS before conversion: If a conversion is triggered with bit 7 of the configuration register at 0, the excitation circuit is inactive. The RTD code returns near zero or sets a fault. Always enable VBIAS and wait 10 ms before the first conversion.
• Mismatched wire mode configuration: A 4-wire RTD operated in 3-wire mode (or vice versa) produces inaccurate readings or fault conditions. Verify the physical wiring and confirm bit 4 of the configuration register matches it.
• Ignoring the fault bit: The LSB of the 16-bit RTD register is a fault flag, not part of the resistance code. Reading the raw 16-bit value without right-shifting includes this bit in the calculation, producing a resistance value about twice the real one at the fault boundary. Always right-shift by 1 before using the code.
• Using the linear approximation below 0°C: The approximation T ≈ (R_RTD − 100) / 0.385 is accurate to within approximately ±0.5°C from 0°C to +100°C but diverges significantly at sub-zero temperatures. Use the Callendar-Van Dusen polynomial or an IEC 60751 lookup table for measurements below 0°C.