Przejdź do treści
Biblioteka skilli
Nauka i badania · Analiza

design-optical-systems

/design-optical-systems

Analizuje lokalną recepturę soczewek w formacie ZMX przy pomocy Optilandu i raportuje ogniskową, plamkę, PSF i MTF w trybie tylko do odczytu.

Licencja
MIT
Ostatnia zmiana repozytorium
06.11.2026

Ocena i bezpieczeństwo

100/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
5
Skrypty
1
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 06.11.2026, bez odwołań do wycofanych modeli i API.

Licencja

MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.

Co sprawdzić przed instalacją
  • 1 skrypt w paczce. Przejrzyj je przed instalacją, bo Claude może je uruchomić.

Skrypty: scripts/optiland_regression.py

Co robi i kiedy

Analizuje lokalną recepturę soczewek w formacie ZMX przy pomocy Optilandu i raportuje ogniskową, plamkę, PSF i MTF w trybie tylko do odczytu.

Oryginalny opis (EN)

Analyze a local sequential lens prescription (absolute-path .zmx) or the built-in Cooke/Double-Gauss/Hubble samples with the pinned Optiland 0.6.3 runtime, reporting EFL/F-number, spot, OPD, scalar FFT PSF/MTF and a three-point radius sensitivity, read-only. Use when an agent needs reproducible geometric/wavefront metrics or Optiland runtime qualification. Not for wave-optics propagation or inverse design (analyze-optical-designs), independent cross-software checks (cross-check-optical-designs), or editing/optimizing in OpticStudio (ai-zemax-optical-design).

Jak działa

  1. Purpose
  2. Workflow
  3. Adjacent Skills
  4. Commands
  5. Required Interpretation
  6. Acceptance Gate
Sekcje z nagłówków SKILL.md, w oryginale · 61 linii, 762 słów
Podgląd SKILL.md
---
name: design-optical-systems
description: Analyze a local sequential lens prescription (absolute-path .zmx) or the built-in Cooke/Double-Gauss/Hubble samples with the pinned Optiland 0.6.3 runtime, reporting EFL/F-number, spot, OPD, scalar FFT PSF/MTF and a three-point radius sensitivity, read-only. Use when an agent needs reproducible geometric/wavefront metrics or Optiland runtime qualification. Not for wave-optics propagation or inverse design (analyze-optical-designs), independent cross-software checks (cross-check-optical-designs), or editing/optimizing in OpticStudio (ai-zemax-optical-design).
---

# Design Optical Systems

## Purpose

Generate reproducible optical-design metrics while keeping prescription source files read-only and separating numerical self-consistency from cross-software validation.

## Workflow

1. Read `references/runtime-lock.md` and use only the pinned environment.
2. Preserve the source prescription, reject URLs and symlinks, and record its SHA-256 digest.
3. Run `--self-test` only when you need environment qualification by itself. To analyze a ZMX, run `--input-zmx` directly: it includes the same qualification and skips user-data analysis if qualification fails, so do not also run a separate self-test. Keep the built-in acceptance gate intact.
4. For a local absolute-path `.zmx`, run the script with `--input-zmx` and write to a new JSON output. UNC paths and any symlink/junction path component are rejected; the report can never replace the source, including with `--force`.
5. Review every `checks` entry and the `limitations` section. Treat a failed or skipped metric as unresolved rather than filling it with an estimate.


## Adjacent Skills

- This skill: geometric ray tracing of sequential prescriptions and the scalar FFT PSF/MTF derived from those traces.
- `analyze-optical-designs`: wave-optics propagation defined on pupils or fields, and inverse design.
- `cross-check-optical-designs`: the independent second opinion. This skill does not compare itself with other software; Optiland vs. OpticStudio discrepancies are judged there.

## Commands

Install the pinned environment once from this skill's `requirements.txt` (see `references/runtime-lock.md`). `<python>` is that environment's interpreter and `<skill-dir>` is this skill's folder.

```bash
<python> <skill-dir>/scripts/optiland_regression.py --self-test --output /absolute/new-result.json
<python> <skill-dir>/scripts/optiland_regression.py --input-zmx /absolute/source.zmx --output /absolute/new-result.json
```

The self-test uses Optiland's synthetic/reference Cooke triplet, Double-Gauss, and Hubble telescope. It also checks immutable Hubble constants measured independently in OpticStudio 2023 R1.02. It is software verification, not a measured optical-system acceptance test.

## Required Interpretation

- Prescription distances and radii are millimetres; wavelengths are micrometres; spatial frequency is cycles/mm; OPD RMS is waves at the stated wavelength.
- Spot metrics are geometric ray metrics. FFT PSF/MTF are scalar calculations in the pinned NumPy backend; every report records requested and actual pupil sampling, FFT grid, strategy, and tilt handling.
- Sensitivity uses a deterministic three-point radius sweep and verifies that the nominal prescription is restored. It is not a manufacturing Monte Carlo or a tolerance yield claim.
- JSON equality and ZMX import/export are self-round-trips. They do not demonstrate semantic equivalence in Ansys Zemax OpticStudio.
- The pinned writer preserves binary64 numeric precision and emits Nd/Vd so the reader can disambiguate duplicate glass names inside declared `GCAT` catalogs. The regression requires Cooke catalog identity, actual n(λ) at all system wavelengths, and EFL to survive the self-round-trip. This does not guarantee every third-party importer will select the same catalog.
- Independent Hubble values must never be regenerated from Optiland. If they fail, investigate the implementation, definition/convention, or external reference before changing a constant or tolerance.
- Hubble RMS wavefront reduction is still sampling-limited, and the tangential field-curvature comparison remains a strict upstream `xfail`. Both are accepted long-term as non-blocking findings: keep them visible, do not loosen their thresholds or drop them to obtain a green report, and re-evaluate them at the next Optiland upgrade.
- `--input-zmx` results cover only the surfaces and keywords Optiland parses; unsupported features may be silently ignored. Compare the reported `surface_count`, radii, thicknesses and stop with the source file before relying on the metrics.
- External compatibility remains version-specific. The repository's sample references came from OpticStudio 22.2.2 and were reproduced in OpticStudio 2024 R1, while the immutable upstream Hubble constants came from OpticStudio 2023 R1.02.
- Stray light, ghost analysis, non-sequential propagation, optomechanical coupling, coating manufacturing data, and full commercial-file feature coverage are outside this regression.

## Acceptance Gate

Accept the runtime baseline only when:

- the source commit, installed-code manifest hash/count, and Python platform are recorded and the installed manifest matches the frozen value;
- Cooke prescription, EFL, F/#, spot arrays, OPD RMS, PSF, and MTF checks pass;
- the tolerance sweep is finite and restores the nominal radius;
- Optiland JSON round-trip is identical;
- Hubble and Double-Gauss ZMX self-round-trips preserve paraxial quantities at `rtol=1e-12` and real-ray image intercepts at `atol=1e-9`;
- Cooke surface-group EFL/ABCD composition passes without treating isolated group powers as additive system power;
- Cooke glass identity, actual n(λ), and EFL survive ZMX round-trip, and the float formatter recovers hard binary64 values exactly;
- Hubble first-order quantities, six real-ray intercepts, and two-field spot radii agree with immutable OpticStudio 2023 R1.02 references at their stated tolerances;
- the unresolved Hubble wavefront-sampling and tangential-field-curvature findings remain present and non-blocking; if either unexpectedly passes, treat it as a behavior change and investigate before updating documentation;
- the report retains all compatibility limitations.
Pliki w paczce (5)
  • SKILL.md 6 KB
  • requirements.txt 1 KB
  • agents/openai.yaml 1 KB
  • references/runtime-lock.md 3 KB
  • scripts/optiland_regression.py 39 KB

Wersje

W bazie nie ma innych skilli o nazwie design-optical-systems.

Migawka bazy z 13.11.2026 · wersja paczki 4a4dbe96f081 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 100cross-check-optical-designs
    Sprawdza tylko do odczytu projekt optyczny Zemax .zmx trzecim silnikiem RayOptics i porównuje ogniskową oraz liczbę F z raportem.
  • 100diagnose-optical-discrepancies
    Wyjaśnia, dlaczego silniki Optiland i RayOptics różnie liczą ten sam układ optyczny z pliku Zemax .zmx.
  • 100matlab-model-optics
    Buduje, importuje, analizuje i optymalizuje układy optyczne w Optical Design and Simulation Library: śledzenie promieni, import z Zemax, powłoki optyczne i tolerowanie.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.