Skip to content

docs(research): depth fade vs deep catalogs (Milliquas) investigation#187

Open
rulkens wants to merge 1 commit into
mainfrom
claude/explore-milliquas-dataset-rzgXs
Open

docs(research): depth fade vs deep catalogs (Milliquas) investigation#187
rulkens wants to merge 1 commit into
mainfrom
claude/explore-milliquas-dataset-rzgXs

Conversation

@rulkens
Copy link
Copy Markdown
Owner

@rulkens rulkens commented May 21, 2026

Summary

Doc-only. Adds an investigation memo diagnosing why the default-on Depth fade toggle erases almost all Milliquas quasars past z ≈ 0.2, and maps the fix onto the points pipeline. No code changes — the memo is written to feed directly into a follow-up implementation plan under docs/superpowers/plans/.

New file: docs/superpowers/research/2026-05-21-depth-fade-deep-catalog.md

What the memo covers

  • Diagnosis. Depth fade is 1/(1+(camDist/1000Mpc)²) (vertex.wesl:184-188) — a galaxy-density anti-saturation hack tuned to the GLADE volume. With the camera near the origin it acts as "attenuate by redshift", so by z ≈ 0.5 quasars sit at ~18% brightness and by z ≈ 1 at ~5%. A hard invisibility cull (vertex.wesl:224, intensity < 0.005) means deep quasars past z ≈ 1.5–2 are not just invisible but non-pickable — the name sidecar / InfoCard can never fire for them.
  • Two distinct defects: D1 — wrong scale for a sparse, deep catalog; D2 — camera-absolute fixed scale dims the whole field (galaxies included) once you zoom out to frame deep objects.
  • Four ranked options, each with a code sketch and exact file/line touchpoints, no committed direction:
    • A — per-source falloff scale
    • B — skip the fade for sparse (quasar) sources
    • C-lite — focal-relative adaptive scale (one float into the free _pad4 slot; galaxy view unchanged via a max(1000, focalDist) floor)
    • C-full — true view-axis focal plane (needs a vec3 → struct resize)
  • Plumbing facts that constrain any fix: byte-matched uniform layout across io.wesl / pointRenderer / pickRenderer, the one free 4-byte slot at offset 124, automatic pick parity via the shared uniform buffer, and cam.distance as the focal-distance source.

No on-disk .bin format impact; all options touch only runtime uniforms + the shader.

Test plan

  • Read the memo end-to-end and decide which option(s) to pursue (A/B/C-lite/C-full compose; they don't exclude)
  • Confirm the §3.1 redshift→fade numbers match the intended diagnosis
  • When a direction is chosen, spawn a TDD plan under docs/superpowers/plans/ (the §5 options are written to feed directly into that)

https://claude.ai/code/session_01H9ifaFejATHQ2qWcPVTUrn


Generated by Claude Code

Memo diagnosing why the default-on depth fade erases Milliquas quasars
past z~0.2 (galaxy-tuned 1000 Mpc scale + camera-absolute falloff), with
four ranked fix options (per-source scale, skip-sparse, focal-relative
adaptive, focal-plane DoF) and the points-pipeline plumbing facts each
would touch. No committed direction.

https://claude.ai/code/session_01H9ifaFejATHQ2qWcPVTUrn
@cloudflare-workers-and-pages
Copy link
Copy Markdown

Deploying with  Cloudflare Workers  Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status Name Latest Commit Preview URL Updated (UTC)
✅ Deployment successful!
View logs
skymap ca17f1b Commit Preview URL

Branch Preview URL
May 21 2026, 02:16 PM

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants