LoopStringLoopString
All Posts
Temperature probes compared, DS18B20 to RTD and thermocouple
Buyer's Guides6 min read1,158 words

Temperature Probes: From DS18B20 to Thermocouples — The Complete Quality-Tier Buyer's Guide

By LoopString Team


Quick picks by tier. Hobby → DS18B20 (1-wire, waterproof) or an NTC thermistor for fast/cheap. Maker/Prosumer → a PT100/PT1000 RTD on a MAX31865. Commercial → an industrial RTD probe in a thermowell with a 4-20 mA transmitter. Hardened industrial / extreme heat → a Type K/N/R/S thermocouple + transmitter. Why each wins is below.

This guide is the companion to our temperature & humidity sensor guide — that one covers ambient air and relative humidity; this one covers contact temperature probes for liquids, surfaces, immersion, and extreme ranges from cryogenic to furnace, where a board sensor can't go.

The thing that actually bites you: range, wiring, and the cold junction

Temperature looks like the easy measurement — every part claims a tenth of a degree — and then three things quietly ruin the reading.

First, range dictates technology, not preference. A DS18B20 is wonderful up to +125 °C and useless in a kiln. A thermocouple reads to +1700 °C and is hopeless at resolving a fridge to 0.1 °C. You don't pick the best sensor; you pick the one whose range brackets your job, then optimise accuracy within it.

Second, wiring error swamps sensor accuracy. A 2-wire RTD adds the resistance of its own lead wires to the measurement — a few ohms of cable can read as several degrees of error. RTDs need 3- or 4-wire connections to cancel it. Thermistors and RTDs also self-heat: the excitation current you push through them to measure resistance warms them up, so a sensor in still air or a small thermal mass reads high.

Third, and the classic: thermocouples need cold-junction compensation (CJC). A thermocouple only measures the difference between the hot tip and the connector where it meets copper. Get the cold-junction temperature wrong — or run plain copper wire instead of matching thermocouple extension wire — and the whole reading shifts. Polarity matters too; a reversed thermocouple reads backwards.

So the real question isn't "how accurate is the sensor?" — it's "does its range bracket my job, and have I wired it so the cable and cold junction aren't lying to me?"

The decision axes

  • Range — the first filter: −55…+125 °C (DS18B20) → −40…+150 °C (thermistor) → −200…+600 °C (RTD) → −200…+1700 °C (thermocouple).
  • Accuracy & stability — RTDs are the most accurate and stable mid-range; thermocouples trade accuracy for reach; thermistors are precise in a narrow band.
  • Linearity — DS18B20 and RTDs are near-linear; thermistors are steeply nonlinear and need Steinhart-Hart correction.
  • Wiring — 1-wire digital → analog (ADC) → 3/4-wire RTD (MAX31865) → thermocouple amp with CJC (MAX31855/31856) → 4-20 mA transmitter.
  • Response time & thermal mass — bare thermistors are fast; sheathed probes and thermowells are slow but rugged.
  • Enclosure — bare bead → waterproof sheath → thermowell (lets you pull the probe without breaching the process).
  • Traceable calibration — the audit requirement for cold-chain, pharma, and process safety.

Tier

Probe

Element

Range

Accuracy

Linearity

Interface

Price (USD)

Best for

Hobby

DS18B20

1-Wire digital

−55…+125 °C

±0.5 °C

near-linear

1-Wire

$2–5

Waterproof, multi-drop, convenient

Hobby

NTC thermistor 10k

Thermistor

−40…+150 °C

±0.1–1 °C

nonlinear (Steinhart-Hart)

Analog (ADC)

$1–5

Fast, cheap, narrow band

Prosumer

PT100 / PT1000 RTD + MAX31865

Platinum RTD

−200…+600 °C

±0.15–0.5 °C

linear

SPI (3/4-wire)

$10–50

Accurate, stable, drift-free

Commercial

Industrial RTD probe + 4-20 mA transmitter

Sheathed RTD, thermowell

−200…+600 °C

±0.1–0.3 °C

linear

4-20 mA

$60–300

Process, harsh, long runs

Industrial

Thermocouple Type K/N/R/S + transmitter

Thermocouple (CJC)

−200…+1700 °C

±1–2 °C / 0.75%

nonlinear

4-20 mA / SPI

$20–500+

Furnaces, kilns, extreme heat

Walking up the ladder

Hobby — DS18B20 ($2–5). A 1-wire digital probe, ±0.5 °C from −10 to +85 °C, available in a waterproof stainless sheath, and — the killer feature — you can multi-drop dozens on a single GPIO because each has a unique address. Perfect for fridges, tanks, brew kettles, and soil/water temperature. Stop here if your range fits within −55…+125 °C and ±0.5 °C is fine.

Hobby+ — NTC thermistor ($1–5). Cheaper, faster, and more sensitive than a DS18B20 in a narrow band, but steeply nonlinear (needs a Steinhart-Hart curve) and analog (needs an ADC). Great where response speed matters — a hotend, a fast-moving liquid. Stop here if you need speed and low cost over absolute accuracy.

Maker/Prosumer — PT100 / PT1000 RTD + MAX31865 ($10–50). This is the accuracy-and-stability tier: platinum RTDs are linear, drift-free, and accurate to ±0.15–0.5 °C across −200…+600 °C. The MAX31865 handles 3/4-wire sensing so lead resistance stops mattering. Stop here if you need real precision in the −200…+600 °C band and care about long-term stability.

Commercial — industrial RTD probe + 4-20 mA transmitter ($60–300). A sheathed RTD in a thermowell with a head-mounted transmitter outputting 4-20 mA you can run hundreds of metres. The thermowell lets you replace the probe without opening the process — essential in food, pharma, and pressurised systems. Stop here if you have a real process line, wash-down, or long cable runs.

Hardened industrial — thermocouple Type K/N/R/S + transmitter ($20–500+). When the range exceeds what RTDs survive — furnaces, kilns, exhaust, molten processes — thermocouples reach +1700 °C (platinum R/S/B) and tolerate brutal environments. You trade accuracy (±1–2 °C or 0.75%) for reach, and you pair them with a transmitter that does CJC and can ship traceable calibration. You need this tier if your process runs hot enough to destroy an RTD, or an auditor needs a cert.

Interface & wiring notes (per tier)

  • DS18B20 (1-wire): one data line + 4.7 kΩ pull-up; parasitic-power mode works but external 3.3 V is more reliable on long buses. Each probe is individually addressed for multi-drop.
  • Thermistor (analog): read through an ADC in a voltage divider; keep excitation current low to limit self-heating, and linearise with Steinhart-Hart in software.
  • RTD (3/4-wire, MAX31865): always use 3- or 4-wire — 2-wire adds lead resistance as error. Match the chip's reference resistor to PT100 vs PT1000.
  • Thermocouple (MAX31855/31856 + CJC): use matching thermocouple extension wire, never copper, all the way to the cold junction; observe polarity; pick the type for your range (K general, N harsh/high-temp, T cryogenic, R/S/B extreme).
  • 4-20 mA transmitters: loop-powered, immune to distance; head-mount on the probe so only a clean current loop runs back to the Pi's 4-20 mA HAT.

How LoopString reads every tier — in one dashboard

A Raspberry Pi running LoopString's Node-RED templates reads a $3 DS18B20 over 1-wire and a $300 thermowell RTD transmitter over 4-20 mA into the same dashboard — same live trend, same threshold and rate-of-change alerts, same compliance log for cold-chain and process records. Prototype on a DS18B20, move to an RTD or thermocouple loop for production, and your dashboards and history don't change. Wire your probe to a Pi and see it live at app.loopstring.io.

Useful next reads: the PID temperature control guide, the Raspberry Pi industrial monitoring guide, and the Raspberry Pi automation guide.

Frequently asked questions

Two wiring traps cause it. A 2-wire RTD adds the resistance of its own lead wires to the measurement — a few ohms of cable can read as several degrees of error — which is why RTDs need 3- or 4-wire connections (e.g. via a MAX31865) to cancel it. Both RTDs and thermistors also self-heat: the excitation current you push through them to measure resistance warms them up, so a probe in still air or a small thermal mass reads high. Keep excitation current low.