A stadium sells beer at 3 °C to seventy thousand people across four hours. Getting there means a chain of refrigerated steps stretching back weeks, each one owned by a different party, each handoff a place where responsibility — and temperature — can be dropped.

The engineering lesson generalises well beyond drinks. Every cold chain fails the same way: not in the middle of a monitored space, but at the moments when the product moves between owners and nobody is measuring.

Six handoffs, six sensing problems

Notice the pattern. Accuracy dominates at the ends — where the product is made and where it is delivered. In the middle, topology and evidence dominate. That distinction should drive the component choice, and it usually doesn't.

The cold store: why 1-Wire wins here

A cold store needs many points and modest accuracy. That is precisely the DS18B20's case. From the Focusensing FST DS18B20 datasheet: 9–12 bit user-selectable resolution, ±0.5 °C from −10 to +85 °C, 1-Wire open-drain interface, 12-bit conversion in 750 ms maximum, supply 3.0–5.5 V. (Focusens FST DS18B20 datasheet.)

The number that matters is not the ±0.5 °C. It is the word open-drain. Many addressable sensors share one data line — twenty probes, one pin, one cable run. Where the wiring is the expensive part and ±0.5 °C sits comfortably inside the product's tolerance band, that topology is worth more than accuracy you would never act on.

• 5 mm tubular DS18B20 probe — space temperature
• Waterproof tubular DS18B20 probe — washdown areas
• Thread-mount DS18B20 probe — where it must not move
• Flange-mount DS18B20 · all digital assemblies

Choosing between a digital probe and a plain thermistor for a pack of points? DS18B20 vs NTC thermistors works the comparison through in detail.

The evaporator and the defrost problem

A refrigeration evaporator ices up. Ice insulates, capacity falls, the system works harder to do less. So it defrosts — and then must decide when to stop. Terminate too early and ice remains and accumulates cycle over cycle. Terminate too late and you have heated the product for nothing.

That decision is made on a sensor reading, which makes defrost the one place in a refrigeration system where thermal lag is a product-quality parameter. The controller ends the cycle on what the sensor reports, not on what the coil is doing. If they disagree by two minutes, the two minutes come out of the product.

The dispense line

You cannot put a probe into a product line. Beyond the hygiene argument, every penetration is a leak path and a cleaning problem. So the line gets a clamp-on sensor — surface reading, no penetration, fitted in minutes to a live line.

Accept what that means: you are measuring the pipe, not the beer. On a thin, well-insulated line the offset is small and stable. On a bare line running through a warm corridor, the sensor is reporting a blend of product and corridor. If the number is a compliance record rather than a control input, insulate over the sensor — the cheapest accuracy improvement available in the whole chain.

Where the chain actually breaks

Not the cold store. The cold store is instrumented, alarmed and owned. It breaks on the dock — the twenty minutes when a pallet sits between a truck and a door, belonging to nobody, measured by nothing. Every serious cold chain review lands on the same conclusion: instrument the handoffs, not just the boxes.

Request the datasheet → · Browse digital assemblies →