Ask the light directly — Parallax Sketchbook

SKETCHBOOK / 05 The Lantern Room · 2-week chapter

Page 1 of 3: An emitter is geometry and a distribution

Emission / direct sampling

An emitter is geometry and a distribution

A small bright emitter can be difficult for a diffuse random walk to discover. Next-event estimation explicitly samples a light at a surface and tests visibility. It changes how you estimate a contribution; it does not remove the need for indirect paths.

The reference already has a two-sided emissive sphere reached by the diffuse walk. Give your room one explicitly sampled area emitter before adding a light-selection system. State whether it emits from one side or both, how points are sampled, and the returned density; visible geometry and the sampling representation must describe the same emitter.

A shadow segment must stop at the selected point rather than accepting blockers behind the light. Record the origin offset, endpoint policy, and occlusion result in one debug sample. The root renderer's directional shadow is useful background, but finite area-light sampling is a new learner feature.

TRY IT IN YOUR STUDIO

Design one rectangular emitter and calculate the area PDF for a chosen point. Draw the bounded visibility segment from a surface point.

From the page to your own work

Your studio assignment.

Illuminate your Lantern Room with one area emitter and an explained direct-light estimator.

A finish line you can see

An area-light study, a visibility test, a PDF calculation, and a bounded MIS extension or worked estimator trace.

Check the area-to-directional PDF conversion independently.
Test an occluder before and behind the sampled light.
Avoid adding unweighted emitter hits on top of the same direct estimator.
Report sample budget and estimator limitations.

Keep the evidence in your weekly notebook ↗

Explain it like an engineer

Try answering aloud before opening the review notes. Use your own implementation as evidence.

What could make an image twice as bright after next-event estimation is added?

The same direct contribution may be counted through two strategies without consistent weighting.
Compare PDF measure and light-selection probability.
Use a single-emitter diffuse scene to isolate the issue.

These prompts support practice; they are not an automatic assessment of your work.

A tip for this exercise

Use one diffuse surface and one area emitter until a selected light point can be explained on paper. Save its position, distance, light normal, visibility result, and PDF in the same trace.

Hints, when you need them.

Try a first step, then reveal as much help as you need. Each hint gets a little more specific.

1Hint 1 · Find a starting point

Label the units of each PDF before combining strategies. Draw the finite visibility segment and decide whether an obstacle beyond the selected emitter point can affect it.

2Hint 2 · Choose an approach

Convert the area density using squared distance and the absolute light-side cosine. Include light-selection probability when applicable, then account for light-sampled contributions and BSDF emitter hits consistently.

3Hint 3 · Check your reasoning

With area density 0.25, distance 2, and light-side cosine 1, the directional density should be 1 per steradian. Move an occluder from before the light to beyond it and check the visibility result changes.

A small creative variation ↗

Move the same emitter farther away without changing its area or emission; predict the geometric effect.

Optional. Save it for later if this chapter already feels full.

Coming back after a break? ↩

Disable the second sampling strategy and recover the previous estimator before comparing weights.

Books & principles behind this study +

Original Parallax explanations and assignments, with these works for further reading.

YOUR MARGIN NOTES

Leave yourself a good question.

What changed your mind? What would you try next?

Notes for Ask the light directly

Notes are private to this browser and saved as you write.

A CHECKPOINT, NOT A SCORE

Can you explain the idea and try the exercise yourself?

Completion records your own assessment. It does not compile or test your C++.

Companion source 10 files