Sensors and data loggers
A sensor measures something and turns it into an electrical signal. Examples: temperature, light, sound, pH, motion, pressure, heart rate.
A data logger is a device (or an app) that records the sensor readings again and again with the time, and stores them. You do not have to stand and watch.
Good things about loggers: many readings, same time gap, no reading mistakes, can run for hours or days, and can measure fast changes that people cannot read by eye.
Sampling rate: how often to measure
The sampling rate is how many readings are taken each second (or the time gap between readings). Fast changes need fast sampling. A slow cooling cup is fine with one reading every few seconds. A bouncing ball needs many readings each second.
A simple rule: measure at least about twice per cycle of the fastest change you want to see. Too slow, and you miss what happened between the dots.
Number of readings = (total time ÷ time gap) + 1. For 80 s with a reading every 2 s, that is 80 ÷ 2 + 1 = 41 readings.
Spreadsheets, graphs and simulations
- A spreadsheet stores data in rows and columns and can add, average and draw graphs.
- A graph turns numbers into a picture so patterns jump out.
- Noise is small random error in readings. Do not panic: look at the overall trend.
- A simulation is a computer model of a real thing. You can change settings and see the result without doing the real experiment.
- Coding (for example in Python) lets you clean, plot and analyse thousands of readings fast.
AI and models: finding patterns and predicting
A model is a rule or formula that fits the data. AI (machine learning) finds the rule by itself from many examples. It can spot patterns in photos (such as a sick leaf), sort star types, or predict weather.
A model fitted to dots inside a range is usually good inside that range. Going outside the range is called extrapolation, and it is risky. In the 3D, the straight line did well for the data but gave an impossible answer later.
Using digital tools wisely
- Check the answer: does it match science and common sense? (A cup cannot get colder than the room.)
- Garbage in, garbage out: bad data gives bad results. Calibrate sensors, repeat trials, record units.
- Bias: if the examples were unfair or one-sided, the AI will be too.
- Chat AI can make mistakes and sound very sure. Verify facts with a book or a trusted site and name your sources.
- Privacy and ethics: do not share personal data and always give credit.
- Back up your data.
Try it: log a cooling cup
Pour hot (not boiling) water in a cup and put a thermometer or phone sensor app in it. Write the temperature every 30 seconds for 5 minutes in a table (your own hand-made "logger"). Draw the graph on paper or in a spreadsheet. Before you finish, predict the temperature at 10 minutes. Check with one more reading. Was your prediction right? Then use the 3D to compare many readings.
Key formulas and definitions
- Number of readings = total time ÷ time gap + 1
- Sampling rate (Hz) = readings per second = 1 ÷ time gap (s)
- Rule of thumb: sample at least twice per cycle of the fastest change
- Hot object cools quickly at first, then slowly towards room temperature (curve, not a straight line)
- Always check: does the prediction obey the science?
Worked examples
1. A logger records every 5 s for 60 s, starting at 0 s. How many readings are there?
Readings = 60 ÷ 5 + 1 = 13 readings.
2. A sensor reads 4 times each second for 2 minutes. How many readings is that?
2 minutes = 120 s. Readings = 4 × 120 = 480 readings.
3. A straight-line model of a cooling cup says 3 °C at 2 minutes, but the room is 25 °C. What is wrong?
A cup cannot cool below the room temperature (25 °C). Cooling slows down as the water gets near the room temperature, so a curve fits better than a straight line.
4. A student reads a thermometer every 30 s. A bubble of heat comes and goes in 10 s. Will the student see it? What should change?
No, the 30 s gap is too long and misses it. Use a logger with a gap of 2 s or less (at least two readings per 10 s event).
5. A phone app uses AI to say a leaf is "healthy" from one photo. Give two ways to check it.
Take more photos in good light and compare; check with a teacher or a plant guide; look for the same signs in a few leaves.
6. You collect 41 readings but a few dots are far from the others. What do you do?
First look for a reason (a loose wire, a bump). If it is a clear error, mark it and leave it out; if it might be real, keep it and repeat the experiment. Never delete data without saying so.
Common mistakes
- Trusting an AI or a computer answer without checking it against science.
- Using too slow a sampling rate and missing fast changes.
- Using a straight line for something that curves, then predicting far outside the data.
- Thinking noise means the experiment failed. A little noise is normal; look at the trend.