Experiments
DraftAn experiment is work where something is actually done and measured, rather than read about or computed. The record exists so that the result can be trusted later and repeated if necessary.
What counts
| Type | Examples |
|---|---|
| Hardware | Building and testing a circuit, an antenna, a sensor, a mechanism |
| Ground station | Receiving satellite downlinks, decoding telemetry, measuring link performance |
| Observational | Telescope observation, photometry, satellite tracking |
| Software | Benchmarking an implementation, validating an algorithm against a known case |
| Environmental | Thermal, vacuum or vibration testing of a component |
Software work counts as an experiment when the point is to measure something. Writing a propagator is code; measuring its accuracy against a reference ephemeris is an experiment.
Structure
The order matters, and the first two fields must be written before the result is known.
| Field | Content |
|---|---|
| Objective | What is being determined, in one sentence |
| Hypothesis | The expected outcome, and the reasoning behind it |
| Prediction | A quantitative expectation where possible, with a tolerance |
| Setup | Equipment, configuration, versions, environment, dates |
| Procedure | What was done, in enough detail to repeat |
| Raw observations | What was measured, before any interpretation |
| Analysis | Processing applied, and why |
| Result | The finding, with uncertainty |
| Conclusion | Whether the hypothesis held, and what is now known |
| What went wrong | Everything that did not work |
Writing the prediction before the measurement is the single most valuable discipline here. It is uncomfortable and it converts an experiment from a demonstration into a test. Without it, a result is almost always interpreted as consistent with whatever was expected, because the expectation is reconstructed after the fact.
Uncertainty
A measurement without an uncertainty is not a measurement. Recording where the uncertainty comes from is as important as its size.
- Statistical
- From repeated measurement. Reduced by taking more data, and usually the easy part.
- Systematic
- From calibration, method or apparatus. Not reduced by more data, and usually dominant.
- Quote both separately
- A single combined figure hides which one limits the result and therefore what to improve.
- Significant figures
- Do not report more precision than the uncertainty supports. A result of 3.47821 plus or minus 0.2 is claiming precision it does not have.
Recording failure
An experiment that did not work is worth recording in full, and the instinct to omit it should be resisted.
A failed setup documents a route that does not work, which saves repeating it. A result that contradicts the hypothesis is more informative than one that confirms it. And an experiment abandoned partway records what blocked it, which is often the useful finding.
The pattern worth watching for is an experiment quietly not written up because the result was uninteresting. That is how a personal record acquires a survivor bias that makes it useless for calibration.
Reproducibility
The dashed line is the important constraint. Raw observations are stored unmodified and analysis operates on copies, so that a later change of method can be applied to the original data. Overwriting raw data with processed data is irreversible and it happens routinely.
Safety and legality
Some experiments in this field have real constraints worth stating.
Radio transmission requires a licence in nearly every jurisdiction. Receiving is generally unrestricted; transmitting is not, and amateur bands have specific rules.
Optical hazards apply to any laser work, including rangefinding and communications experiments.
High voltage and stored energy appear in propulsion and power experiments and deserve the treatment they would get in any laboratory.
Chemical propellants are outside the scope of amateur work in most jurisdictions and are dangerous in ways that are not obvious.
References
Experiments cite their own sources and equipment. This page is a convention.