Recipe

Calibrate a sensor with two reference points

Use when: a sensor's readings are steadily off, and you can put it beside two known values

Goal: a gain and an offset that turn every raw reading into a corrected one

Taught in Calibration

1. 1

   Choose two known values near the ends of the range you care about. Use a trusted reference: ice water for 0 °C, a calibrated thermometer, a multimeter for volts.

   Example

   Ice water (0.0 °C) and warm water beside a reference thermometer (40.0 °C).

   Why: the farther apart the points, the less a small error in either one tilts the line.

2. 2

   Wait until both sensor and reference stop changing. Average several raw readings at each point.

   Example

   Raw averages 0.8 °C in the ice water and 41.1 °C in the warm water.

   Why: a sensor reading on its way to a new temperature isn't at the reference value yet.

3. 3

   \text{gain} = (\text{ref}_2 - \text{ref}_1)/(\text{raw}_2 - \text{raw}_1), in decimals.

   Example

   (40.0 - 0.0)/(41.1 - 0.8) = 40/40.3 \approx 0.9926.

   Why: the gain is how much the true value changes for each unit of raw change.

4. 4

   \text{offset} = \text{ref}_1 - \text{gain} \times \text{raw}_1.

   Example

   0.0 - 0.9926 \times 0.8 \approx -0.794.

   Why: it puts the line through the first point; the gain already makes it pass through the second.

5. 5

   Apply the correction at a value you didn't calibrate with, and compare with the reference.

   • within the reference's own accuracy — done; record the constants, the date and the references used.
   • still off — the error isn't a straight line, or a reading wasn't settled; repeat, or calibrate a narrower range.

   Why: two points always fit a line exactly, so only a third point tests it.