A Resistance Temperature Detector (RTD) is a precision platinum resistor whose resistance changes predictably with temperature. A Pt100 is 100 Ω at 0 °C; a Pt1000 is 1000 Ω at 0 °C. The temperature coefficient (α) is roughly 0.00385 Ω/Ω/°C — meaning a Pt100 changes only ~0.385 Ω per °C. That tiny change is why the excitation and wiring topology matters more than the amplifier does.

The excitation problem. You must push a known current through the RTD to convert its resistance into a measurable voltage. Too much current causes self-heating (the RTD warms itself and lies about ambient temperature). Rule of thumb: keep dissipation below 1 mW for a standard element in still air. For a Pt100, that limits excitation to about 1 mA; for a Pt1000, about 0.3 mA. Use a precision current source (a REF200 or an op-amp/JFET Howland).

2-wire is a trap. Lead resistance adds directly to your reading. A pair of 10 m copper leads at ~0.1 Ω/m adds 2 Ω — that's 5 °C of error on a Pt100 before your signal even reaches the ADC.

3-wire compensation. Run two matched leads to one side of the RTD and one lead to the other. Use two identical current sources: I1 flows through RTD + lead A, I2 flows only through lead B. Measure the difference. If both leads have equal resistance R_lead, the lead drops cancel. This is what most industrial transmitters (like the MAX31865) use — cheap and good to ±0.1 °C if leads are truly matched.

4-wire (Kelvin) is the gold standard. Force current through two leads, sense voltage across the RTD with two separate high-impedance leads. Since no current flows in the sense leads, their resistance is irrelevant. Use this for lab-grade or long-cable installations.

Worked example. Pt100 at 100 °C with 1 mA excitation:

• R(100 °C) ≈ 100 × (1 + 0.00385 × 100) = 138.5 Ω
• V_sense = 1 mA × 138.5 Ω = 138.5 mV
• Span from 0 °C to 100 °C: only 38.5 mV — feed this into an instrumentation amp with gain ~50 to hit a 0–2 V ADC input.

Don't forget linearization. The Callendar–Van Dusen equation adds a quadratic term; above ~150 °C, ignoring it costs you 0.5 °C. Do the correction in the MCU with a lookup table or polynomial.

Also: use a ratiometric ADC reference derived from the same excitation source. Any drift in the current cancels out — the ADC sees a ratio, not an absolute voltage.