PSLE Science Reality Lab Vol No.295 | “Sensor Sensitivity = 20 mV/°C” — Is the Sensor 20% Accurate?

Series ID: PSLE-SCI-REALITY-0295

Wait, What? A Sensor Can Be Very Sensitive and Still Be Wrong

A product sheet for a fictional temperature sensor says Sensitivity: 20 mV/°C. One learner looks at the number 20 and says, “So it is 20% accurate.” Another says, “No, sensitivity means it can detect tiny changes, so it must be more accurate than a sensor with 10 mV/°C.” Both answers sound scientific because they use familiar words. Neither is justified by the specification.

In measurement science, sensitivity describes how much the indication or output changes when the quantity being measured changes. If a temperature sensor has a sensitivity of 20 millivolts per degree Celsius, then a 1°C change in the relevant range is associated with about a 20 mV change in the stated output. That tells us about the relationship between input change and output change. It does not, by itself, tell us how close the reported temperature is to the true value.

This is a powerful Reality Lab problem because everyday language makes “sensitive” sound like “good at noticing” and therefore “accurate.” Scientific specifications are stricter. A reader has to identify exactly what quantity the number describes, what units it carries, what conditions apply, and which tempting conclusions still need separate evidence.

Quick Answer

Sensitivity tells you how much a measuring system’s indication changes for a corresponding change in the measured quantity.
20 mV/°C is not 20%. The units already tell you it is a change in voltage per change in temperature.
A large sensitivity value does not automatically mean high accuracy.
Sensitivity is not the same as resolution, response time, detection limit, repeatability or measurement uncertainty.
The value may apply only over a stated measurement range and under stated operating conditions.
A sensor can respond strongly to temperature changes yet have an offset, calibration error, noise or other influence that makes a reading inaccurate.
To judge a scientific claim, separate the response relationship from the quality of the final measurement.

The Exact Learner Job This Volume Owns

This volume owns one evidence-transfer job: how to evaluate a sensor sensitivity number in a datasheet, graph, product comparison or science report without turning that response slope into a percentage accuracy claim.

It does not re-teach the general owners for measurement accuracy, precision, resolution, uncertainty, variables or graph reading. It applies those ideas to one real-world communication object: a sensitivity specification written as output change per input change.

Why This Fits the Current PSLE Science Frame

The 2026 PSLE Science assessment objectives include interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning. The 2023 Primary Science syllabus also encourages healthy scepticism, attention to assumptions and uncertainty, and the use of evidence when building explanations. A sensor datasheet is ideal practice because the learner has to read the evidence object before deciding what it proves.

No examiner trick is required. The scientific habit is simple: name the quantity, read the units, reconstruct the comparison, then limit the claim to what that evidence actually measures.

Start With the Units Before You Start With the Number

The label says 20 mV/°C. Read that aloud as “twenty millivolts of output change for each degree Celsius of temperature change,” within the conditions for which the specification is valid. That sentence is already more informative than saying “sensitivity twenty.”

Specification	What the units suggest	What not to assume
20 mV/°C	Output voltage change per temperature change	20% accuracy
0.5°C	A temperature-sized quantity	Automatically sensitivity
2 s	A time-sized quantity	Automatically accuracy
0.1°C resolution	Smallest displayed or distinguishable step under the stated definition	Every reading is correct to ±0.1°C

Units are not decoration. They tell you what kind of scientific object the number is. If the unit is millivolts per degree Celsius, a percentage interpretation has already failed its first check.

Rebuild an Original Sensor Case

Imagine a school greenhouse uses a fictional sensor called T-Leaf. Its datasheet states:

Measurement range	10°C to 40°C
Nominal sensitivity	20 mV/°C
Output at 20°C	1.00 V
Accuracy specification	±0.8°C under stated laboratory conditions
Response time	12 s under the stated airflow condition

If temperature rises from 20°C to 25°C, the change is 5°C. Using the nominal sensitivity, the expected output change is 5 × 20 mV = 100 mV. So the output would be expected to move from about 1.00 V to about 1.10 V, assuming the stated relationship applies.

Notice what we did not calculate. We did not say the sensor is 20% accurate. We did not say its uncertainty is 20 mV. We did not say it responds in 20 milliseconds. The sensitivity helps connect temperature change to output change; the other specifications answer different questions.

Observed, Specified, Calculated and Claimed

Layer	Example	Learner question
Observed	The output voltage changed by 98 mV while a reference temperature changed by 5°C	What actually happened?
Calculated	98 mV ÷ 5°C = 19.6 mV/°C	What relationship did the data support?
Specified	Nominal sensitivity 20 mV/°C	Under what range and conditions does this specification apply?
Claimed	“This is a highly accurate sensor”	What additional evidence is required?

A strong science answer keeps these layers apart. The fact that a measured slope is close to the nominal sensitivity may support the response relationship. It does not automatically prove that every reported temperature is close to the true temperature.

Worked Case 1: Same Sensitivity, Different Accuracy

Two fictional sensors both have a sensitivity of 20 mV/°C. Sensor A is correctly zeroed. Sensor B has a constant offset: its output is always 80 mV too high. When temperature increases by 1°C, both outputs still rise by 20 mV.

Their sensitivities can therefore be the same even though their indicated temperatures are not equally accurate. Sensor B responds by the correct amount to a change, but starts from the wrong baseline. This is the cleanest way to see why sensitivity and accuracy are different jobs.

Worked Case 2: High Sensitivity Can Magnify Noise Too

Sensor C produces a large voltage change for a small temperature change, which can make the signal easier for electronics to detect. But its output also jitters because of electrical noise. A strong response slope does not erase that variation. The quality of a measurement system depends on more than one number.

If an advertisement says, “Twice the sensitivity means twice the measurement quality,” ask what “quality” means. Accuracy? Resolution? Stability? Signal-to-noise behaviour? Response speed? Without a defined outcome and evidence, the statement travels too far.

Worked Case 3: Sensitivity Can Change Across the Range

An original light sensor gives these outputs: 0 units of light → 0.20 V; 100 units → 0.60 V; 200 units → 1.00 V; 300 units → 1.32 V. The first two 100-unit increases each produce a 0.40 V change, but the next produces only 0.32 V.

A single sensitivity number may therefore describe a nominal value, a local slope or a stated operating region rather than every possible input. The BIPM/JCGM vocabulary notes that sensitivity can depend on the value of the quantity being measured. The learner should ask: which range does this number describe?

Worked Case 4: A Steep Graph Is Not Automatically a Better Sensor

Sensor D changes by 50 mV for each degree. Sensor E changes by 10 mV for each degree. A graph of D is steeper. A student declares D “five times better.” That conclusion is unsupported because “better” has not been defined.

If the measurement electronics can read both outputs well, E could still be more accurate, more stable, faster, less affected by humidity or better suited to the required range. The slope answers one question. Product or experiment suitability requires several relevant criteria.

Worked Case 5: Sensitivity Is Not Response Time

Two thermometers both change 20 mV/°C. One reaches its new stable indication quickly after a temperature change; the other takes much longer. They can share sensitivity while differing in response behaviour. A value describing “how much output changes” is not the same as a value describing “how long the system takes to respond.”

The Representation Check: Read the Axes

Datasheets often show a graph instead of a sentence. The horizontal axis may be temperature. The vertical axis may be voltage. Sensitivity is then related to how much the vertical quantity changes when the horizontal quantity changes. Before calling a line “steep,” read both scales and units. A graph can look steep or flat merely because an axis has been stretched.

Constructed example: Graph A rises 1 V over 50°C. Graph B rises 0.2 V over 5°C. On a small screen A may look steeper, but the relevant ratios are 20 mV/°C and 40 mV/°C. Scientific comparison uses the quantities, not the visual angle of the line.

The Baseline Check: Where Does the Output Start?

A sensitivity tells you about change, not necessarily the starting output. Two lines can have the same slope and different intercepts. If the baseline is shifted, converting the output into a measured value may require an offset correction or calibration relationship. This is why “correct slope” does not guarantee “correct reading.”

The Method Check: How Was Sensitivity Obtained?

Was a trusted reference used for the input quantity?
Were several points measured across the stated range?
Was the sensor allowed to stabilise at each point?
Were environmental conditions controlled or recorded?
Was the relationship approximately linear in the range where one sensitivity value is claimed?
Were repeated measurements used to see whether the response was stable?
Was the direction of change relevant?
Was the sensitivity measured on the actual unit or quoted as a nominal design value?

These questions do not turn Primary Science into engineering. They teach the PSLE habit of checking how evidence was produced before extending a claim.

The Comparison Check: Compare the Same Quantity

A product table lists Sensor F sensitivity as 25 mV/°C and Sensor G accuracy as ±0.5°C. Which is better? The question cannot be answered from those two entries because they are not the same property. It is like comparing a length with a time and asking which number wins.

A fair comparison needs matching evidence: sensitivity with sensitivity under comparable conditions; accuracy with accuracy under comparable definitions; response time with response time; and so on. One of the most useful Reality Lab habits is to stop before comparing numbers that merely happen to appear in the same table.

Alternative Explanations for a Changed Output

Suppose a sensor output rises by 100 mV. The advertised sensitivity is 20 mV/°C, so a student immediately says the temperature rose 5°C. That may be the intended interpretation, but a careful scientist asks whether other quantities can influence the sensor. Supply voltage, humidity, strain, ageing, contamination or electronic drift could matter for some systems.

The correct response is not to invent problems. It is to check the datasheet and method for known influence quantities. If the experiment controlled them or the device is designed to compensate for them within a stated range, confidence grows. If they were uncontrolled and large enough to matter, the conclusion needs a boundary.

What Evidence Would Strengthen an Accuracy Claim?

Comparison with traceable or otherwise appropriate reference values across the needed range.
Small differences between indicated and reference values under relevant conditions.
Repeated measurements that show stable behaviour.
Evidence about temperature, humidity or other influence quantities relevant to the device.
Calibration or verification information appropriate to the claimed use.
A clearly stated accuracy or uncertainty specification rather than sensitivity being used as a substitute.

What Evidence Would Strengthen a Sensitivity Claim?

Several known input values and corresponding output indications.
A response relationship consistent with the stated slope over the claimed range.
Units that make the relationship explicit.
Repeat measurements showing that the output change is not a one-off fluctuation.
A statement of the conditions under which the sensitivity was measured.

How Far Can the Conclusion Travel?

A bounded conclusion might be: “Within the stated operating range, this sensor has a nominal sensitivity of 20 mV/°C, so a 1°C change is associated with about a 20 mV output change under the specified conditions.”

That conclusion should not silently grow into: “The sensor is 20% accurate,” “it detects every 0.05°C change,” “it responds in 20 milliseconds,” “it is twice as good as a 10 mV/°C sensor,” or “every reading is correct.” Each of those statements needs evidence for a different measurement property.

Tempting but Invalid Reasoning

“Twenty means twenty percent.” The stated unit is mV/°C, not %.
“Bigger sensitivity means better accuracy.” Different properties.
“Steeper graph means more correct.” Slope is not closeness to a reference value.
“Sensitivity tells the smallest change detectable.” That may involve resolution, noise and threshold behaviour; sensitivity alone is insufficient.
“Sensitivity is constant everywhere.” It can vary with the measured quantity or operating conditions.
“The datasheet number proves this particular experiment worked.” A specification does not replace checking the actual set-up and data.

PSLE-Style Transfer Case: The Soil-Moisture Probe

A fictional probe produces 0.70 V in dry test soil and 1.10 V after the reference moisture value is increased by 20 units. A learner calculates a response of 0.40 V ÷ 20 = 0.020 V per unit, or 20 mV per unit.

Question 1: What does the calculation describe? The change in output per change in the reference moisture quantity over that interval.

Question 2: Does it show that the probe is 20% accurate? No. Accuracy requires comparison between indicated and reference values under an appropriate definition.

Question 3: What would you check before using the same sensitivity outside the tested interval? Whether the response remains suitable and approximately consistent across the wider range and conditions.

Question 4: If the output jitters by ±30 mV, why does that matter? The noise may make small changes hard to distinguish even though the nominal sensitivity is unchanged.

Explained Practice

1. Datasheet: sensitivity = 5 mA/kPa. What changes? The electrical output current changes by a stated amount for a change in pressure, under the stated conditions.

2. Can 5 mA/kPa be rewritten as 5% accuracy? No. The dimensions and scientific meanings are different.

3. Two sensors have equal sensitivity but different offsets. Can their readings differ? Yes.

4. One sensor has twice the sensitivity. Must it detect changes half as small? Not necessarily. Resolution and noise also matter.

5. What is the first check when a technical number looks impressive? Read its quantity name, units, reference conditions and comparison basis before deciding what it means.

Delayed Independent Return: Four Cards, Four Jobs

Tomorrow, make four cards labelled sensitivity, accuracy, resolution and response time. On separate slips, write: “change in output per change in measured quantity,” “closeness under a stated accuracy definition,” “smallest indicated or distinguishable step under a stated definition,” and “how rapidly the system responds under stated conditions.” Match the slips to the cards, then explain why one device can be strong on one property and weak on another.

A Second Return: Build a Sensitivity Graph From Original Data

Use the fictional pairs (10°C, 0.40 V), (20°C, 0.60 V), (30°C, 0.80 V), (40°C, 1.00 V). Plot temperature horizontally and voltage vertically. Find the output change for each 10°C interval. Then shift every voltage upward by 0.15 V and plot a second line. The slopes remain the same while the baseline changes. That visual difference is the heart of the sensitivity-versus-offset distinction.

Useful eduKateSengkang Routes

Parent and Tutor Teaching Guide: Make the Units Do the First Half of the Work

Give the learner six invented specification cards: 20 mV/°C, ±0.5°C, 0.1°C resolution, 8 s response time, 0–50°C range and 5 V supply. Ask them to sort the cards before explaining anything. The sorting task forces them to stop treating a specification sheet as a row of competing scores.

Next, draw two straight lines with the same slope but different starting heights. Ask whether their sensitivity is the same. Then draw two lines that start together but have different slopes. Ask what changed. Only after the learner can describe the graph should you introduce the formal word sensitivity.

Finally, ask for a bounded sentence. A strong learner should be able to say, “The sensor’s output changes by about 20 mV for each 1°C change within the stated range; this number alone does not tell us the accuracy of its temperature reading.” That sentence is far more valuable than memorising “sensitivity is not accuracy” without understanding why.

Authoritative Sources

The international metrology vocabulary defines sensitivity through the ratio between a change in indication and the corresponding change in the quantity being measured. It also notes that sensitivity can depend on the value of the measured quantity. That is why a sensitivity number should be read as a measurement relationship, not promoted into an accuracy percentage.

The Quiet Rule to Keep

When a scientific specification sounds impressive, do not ask whether the number is big. Ask what changes, compared with what, in which units, under which conditions. Sensitivity is the response relationship. Accuracy is a different question. The moment you separate those jobs, a technical datasheet becomes much easier to read—and much harder for a misleading claim to misuse.

Small Group Tutorials