NightShieldA small experiment in light pollution

EARTH FORWARD THEME · DESIGN BRIEF

Light the path.
Limit the spill.

A tool for student environmental clubs to design and test ways to keep useful light and reduce side spill—with a small LED and a cardboard hood.

Try the 30-second example ↓Theme fit

A prototype for tabletop experiments. Its effect on real outdoor lighting has not been measured.

Select the three scenes to compare

Is a simple hood enough?

Actual LED performance: unmeasured
Calculated light paths
Green: path to keep litPink: direction to keep darker

Dots are independent measurement positions, not a building or a wall.

Path light kept—

Light left at side points—

Each point starts at 100% under the original lamp. The numbers average those point-by-point ratios. They are not total light output or safety ratings.

Mean of each sampled point’s new light divided by that point’s original light. Bar length is capped at 100%; the displayed number is not capped. The lowest path-point value is reported separately.

How this fits an environmental hackathon theme

Turn unwanted light into a design that can be tested.

NextStep’s Earth Forward theme asks for technology that addresses environmental problems. NightShield investigates light reaching unwanted directions.

Environmental problem

Light reaches places it is not needed.

Artificial light at night can affect surrounding ecosystems. Limiting where it shines is an approach supported by environmental research and lighting principles.

What it does

Compare hood shapes and lamp positions.

Calculate different changes to the same lamp. Compare path light with side spill, then find a small change that meets the goals within the tested candidates.

What the student tests

Build it. Measure it. Revise it.

Build the basic hood and the suggested design, then compare them. When predictions fail, investigate why and show how that changes the design.

Read the official challenge ↗

Theme fit is not proof of environmental benefit. Physical measurements and user observations are still pending. Wildlife protection and energy savings have not been demonstrated.

Test your design

A tabletop experiment with a small LED and a cardboard hood

  1. Check the LED’s light pattern

    Check the actual light source. Save the design and prediction before collecting performance readings.

  2. Compare four conditions

    Original lamp, basic hood, fixed-position design and app suggestion. Rebuild independently for four comparison blocks.

  3. Explain and revise

    Show improvements and failures together. Keep the raw readings and document what you changed.

Try your own lighting setupDimensions · detailed calculations · build sheet · measurement records

The full experiment tools use the same calculation as the example above. Editing dimensions or goals updates the example too.

FULL EXPERIMENT WORKBENCH · READINGS 0
01
cm · tabletop scale
Original setupPREDICTION
Path — keep lit Protected patches — reduce spillReceiver probes · no wall occlusion
Mean path-point retention—mean of each point’s ratio
Lowest point retention—of that point’s original light
Mean protected-point reduction—from mean per-point retention
Use your own dimensions+

Enter measured tabletop dimensions. The protected patches are independent receiving points, not a solid wall.

THE DECISION, WITH ITS COST

What changes. What you keep.

All percentages reference the original lamp.
Hood and movement: fixed source output. Dimming is an analytic control.

InterventionLamp moveMean path retentionLowest retentionMean protected-point reductionAll targets

TRY TO BREAK THE PREDICTION

Does the answer hold up?

Numerical check pending
Sampling, margins and small changes +

These are numerical checks of a simplified model. A stable calculation is not a physical measurement. Design targets are chosen for this experiment, not ecological or lighting-safety thresholds.

03 TAKE IT OFF THE SCREEN

Build the change.
Challenge the prediction.

Make a low-voltage LED model with marked positions and a cardboard hood. Save the prediction before measuring. A disagreement is a useful result.

Includes geometry, targets, search scope and per-probe predictions. Measurements stay separate.

YOUR PROPOSED HOODModel dimensions

Cross-section guide · not to scale. Use a ruler. Front/back leakage and material reflections are not modeled.

PREDICTION ≠ MEASUREMENT

Let the experiment answer.

No measurements uploaded

Bring your measurements

One CSV. Repeated readings. Original baseline at every probe. Everything stays on this device.

or drop a CSV here · up to 1 MB

Paste CSV instead +
0measured trials

The model has made a prediction.
The physical result is still yours to discover.

  • Same marked probe before and after
  • Keep camera exposure or meter setup fixed
  • Reassemble and repeat; keep failed trials
DarkSky & IES lighting principles ↗

Uploaded readings are user-provided, not independently verified. Camera brightness supports a direction-of-change comparison, not a calibrated optical percentage. A few measured points do not validate the whole simulated scene.

BEHIND THE PREDICTION

One baseline.
A falsifiable prediction.

  1. Can a hood do enough? Test aperture widths, depths and directions with the lamp fixed.
  2. Would a small move help? Test allowed horizontal positions. Find the least displacement among feasible designs on that grid.
  3. Does it happen on the table? Save the prediction, build the geometry, and compare repeated readings at the same marked probes.

Every hood on the finite grid is compared at the same numerical resolution, including candidates that would fail a coarser calculation. The selected candidate is checked again at higher resolutions and under named small changes. A failed or unresolved convergence check blocks its build recommendation.

“Mean path retention” is the equally weighted mean of each sampled point’s new light divided by that point’s original light. “Mean protected-point reduction” is one minus that mean for protected points. Neither is total emitted flux, average lux, energy saved or a wildlife response.

The source model is an explicit assumption: uniform directions, or a downward Lambertian pattern whose intensity follows the emission cosine. The source pattern is applied to both the original baseline and each changed design. Comparing these two idealizations does not calibrate a real LED.

The source uses equally weighted samples, direct visibility, inverse-square distance and receiver incidence. Moving the lamp never changes the reference: each result is divided by that probe's original unshielded reading.

The model is a 2D cross-section. It does not calculate reflections, real fixture photometry, certified lux, energy savings or ecological outcomes. Protected-area markers are independent probes, not an opaque wall. Professional lighting design requires richer models and measurements.

Initial software and documentation were made with AI assistance. Actual student experiments, revisions and learning must be recorded separately.
WHY THIS EXPERIMENT EXISTS

Environmental purpose.
Measurable first step.

The research. A 2024 field study in Germany found fewer insects attracted to tailored, shielded luminaires than to conventional lighting in its studied settings. This motivates testing where light goes; it does not establish an ecological benefit from NightShield. Read the primary study ↗

The proposed user. A student team investigating outdoor light spill, starting with a controllable tabletop model. Whether this tool improves their decisions over a simple hood activity is a hypothesis to test.

The comparison. Shield-making and professional lighting simulation already exist. NightShield's proposed contribution is the complete decision trail: test simple alternatives, compare a finite set of changes, expose unstable predictions and check the selected design against repeated readings. DIALux ↗ · Existing classroom activity ↗

The competition. NextStep Hacks 2026 asks for technology addressing environmental problems under Earth Forward. Its criteria are originality, theme fit, completion, learning, design and technology. The event closed on September 13, 2026. This page uses its theme as a design brief and does not report a hackathon entry, score or award. Official brief ↗

Software predictions: available. Physical comparisons: pending. User observations: pending. Student experiments and revisions: pending. A tabletop test cannot establish wildlife benefit, outdoor compliance or energy savings.
Review the build sheet