Coded lens

A camera lens designed and ray-traced in code: focal length, spot size and aberrations versus targets, with ray-trace plots as goldens.

This is a CAD spec on Spekkr, published by onyx. It is listed with the other specs. Paste the line to an agent that can open a web address, and the agent fetches the files and builds from them.

Spec
onyx/coded-lens
Version
0.2.0
Published by
onyx
Updated
15 Nov 2026

Build this spec

Build the spec at https://spekkr.com/onyx/coded-lens/start.md.

Paste this to your AI agent. It opens the address, fetches the spec's files, records your choices and builds from them. Nothing to install, no account.

Published by onyx.

What does the result look like?

Coded lens: heroCoded lens: detail

What can I customize?

Tracer shipped is a small tracer in the project. external is a named open library, if present.

Your choices go into the line you copy. Anything else is yours to edit, but it leaves the tested path.

What do I get?

  • spec.md Requirements and acceptance checks.

    Coded lens

    What this builds

    A small camera lens, designed in code: a few elements, a stop, an image plane. A ray tracer (a shipped one, or an open library) reports focal length, an on-axis and an off-axis spot size, and a simple aberration or ray-fan metric, against documented targets. Ray-trace plots are committed as goldens. No glass is melted, and no bench is required. This is an original prescription, not a commercial lens's exact data presented as a product.

    Read spekkr.inputs.json first. focal and aperture are design targets. format is the image diagonal. aspheres and coatings change what the design may use. tracer chooses the engine.

    Out of scope

    • A lab bench, a real glass melt, or a commercial lens prescription copied as the product.
    • A camera body in the checks.
    • A network licence server for a CAD package.

    Requirements

    Each requirement has an id. Checks name the id they cover.

    R1. Metrics
    • The effective focal length is focal within 2%. The on-axis RMS spot size at the image plane, at the design wavelength, is under the bound in design.md (for example 15 µm at F/2.8, 50 mm, as in the file).
    • An off-axis field (0.7 of the image circle) has a spot size under its own bound. A ray-fan or Seidel-style figure is within the table in the file.
    • The prescription (radii, thicknesses, glasses or index values) is in a file the tests read back.

    Accept when: focal length, both spot sizes, and the aberration figure match the bounds, and the prescription round-trips.

    R2. Plots and files
    • Goldens: a layout (elements and rays), an on-axis spot, an off-axis spot, and a ray-fan plot, compared with a documented tolerance on plots.
    • One command traces, measures, and writes the plots. No bench, no glass, no camera.

    Accept when: the command passes for both tracers if the external one is present (otherwise shipped only), and a deliberately defocused fixture fails the on-axis bound.

    R3. Inputs
    • Numeric inputs are read from spekkr.inputs.json and must stay inside the ranges in the Inputs table in design.md.
    • A value outside a range is refused before modelling, with a message that names the key.
    • Options add or remove the features in that table. The tool swap selects CadQuery or OpenSCAD; both must meet the same numbers.

    Accept when: default and one out-of-range value are handled as above, and a second legal input set still matches the formulas in design.md.

    R4. Checks and report
    • One command builds the model at the defaults and at the person's inputs, measures it against design.md, and writes a short report.
    • A model broken on purpose, for example a wall thinner than wall, must fail the report.
    • The command exports the files this spec asks for (STL, and STEP when the tool can). No physical print is required.

    Accept when: the report says pass at the defaults and at the person's inputs, and fail on a deliberately broken model.

    Verification

    Focal length, image-plane spot size, and an aberration metric are compared to documented targets. Ray-trace plots are checked as goldens. No optical bench and no physical glass are required.

    Definition of done

    • Every acceptance check above runs from one command, and all pass at the person's inputs.
    • A README says how to run the tracer, how the targets are listed, and that a bench is not a check.
    # Coded lens
    
    ## What this builds
    
    A small camera lens, designed in code: a few elements, a stop, an image
    plane. A ray tracer (a shipped one, or an open library) reports focal
    length, an on-axis and an off-axis spot size, and a simple aberration or
    ray-fan metric, against documented targets. Ray-trace plots are committed
    as goldens. No glass is melted, and no bench is required. This is an
    original prescription, not a commercial lens's exact data presented as a
    product.
    
    Read `spekkr.inputs.json` first. `focal` and `aperture` are design
    targets. `format` is the image diagonal. `aspheres` and `coatings` change
    what the design may use. `tracer` chooses the engine.
    
    ## Out of scope
    
    - A lab bench, a real glass melt, or a commercial lens prescription copied
      as the product.
    - A camera body in the checks.
    - A network licence server for a CAD package.
    
    ## Requirements
    
    Each requirement has an id. Checks name the id they cover.
    
    ### R1. Metrics
    
    - The effective focal length is `focal` within 2%. The on-axis RMS spot
      size at the image plane, at the design wavelength, is under the bound in
      `design.md` (for example 15 µm at F/2.8, 50 mm, as in the file).
    - An off-axis field (0.7 of the image circle) has a spot size under its
      own bound. A ray-fan or Seidel-style figure is within the table in the
      file.
    - The prescription (radii, thicknesses, glasses or index values) is in a
      file the tests read back.
    
    Accept when: focal length, both spot sizes, and the aberration figure
    match the bounds, and the prescription round-trips.
    
    ### R2. Plots and files
    
    - Goldens: a layout (elements and rays), an on-axis spot, an off-axis
      spot, and a ray-fan plot, compared with a documented tolerance on plots.
    - One command traces, measures, and writes the plots. No bench, no glass,
      no camera.
    
    Accept when: the command passes for both tracers if the external one is
    present (otherwise shipped only), and a deliberately defocused fixture
    fails the on-axis bound.
    
    ### R3. Inputs
    
    - Numeric inputs are read from `spekkr.inputs.json` and must stay inside
      the ranges in the Inputs table in `design.md`.
    - A value outside a range is refused before modelling, with a message that
      names the key.
    - Options add or remove the features in that table. The `tool` swap
      selects CadQuery or OpenSCAD; both must meet the same numbers.
    
    Accept when: default and one out-of-range value are handled as above, and
    a second legal input set still matches the formulas in `design.md`.
    
    ### R4. Checks and report
    
    - One command builds the model at the defaults and at the person's inputs,
      measures it against `design.md`, and writes a short report.
    - A model broken on purpose, for example a wall thinner than `wall`, must
      fail the report.
    - The command exports the files this spec asks for (STL, and STEP when the
      tool can). No physical print is required.
    
    Accept when: the report says pass at the defaults and at the person's
    inputs, and fail on a deliberately broken model.
    
    ## Verification
    
    Focal length, image-plane spot size, and an aberration metric are compared
    to documented targets. Ray-trace plots are checked as goldens. No optical
    bench and no physical glass are required.
    
    ## Definition of done
    
    - Every acceptance check above runs from one command, and all pass at the
      person's inputs.
    - A README says how to run the tracer, how the targets are listed, and
      that a bench is not a check.
    

    Open spec.md as plain text

  • design.md Focal-length and spot-size targets, the prescription file, and ray-trace plot goldens.

    Design

    This file fixes the geometry. How the model's code is organised is up to you, as long as the numbers here hold.

    Targets (defaults)

    Focal 50 mm ± 2%. On-axis RMS spot ≤ 15 µm. 0.7-field RMS spot ≤ 40 µm. The exact bounds and the wavelength (for example 587 nm) are fixtures/lens-targets.json.

    Prescription

    Element count, radii, thicknesses, and index (or a glass name from a shipped table) are fixtures/lens-prescription.json after a design run, or a shipped starting point that already meets the bounds.

    Plot tolerance

    A pixel tolerance of 8 levels, 98% of pixels, or a numeric plot array within 1e-3 of a fixture, as the tests prefer.

    Inputs

    KeyLabelDefaultNotes
    focalTarget focal length, mm50From 24 to 85, the design target.
    apertureF-number2.8From 1.8 to 8.
    formatImage circle, mm43.3The diagonal of the image plane, default a 35 mm still frame.
    coatingsSimple coatingsoffA transmission factor in the tracer, not a physical lab coating.
    aspheresAllow an asphereonIf off, all surfaces are spherical or planar.
    tracerTracershippedshipped is a small tracer in the project. external is a named open library, if present.

    Acceptance at the defaults

    Millimetres and a unit-focal tracing. Default aperture 25, focal length 50, two elements, aspheres on, coatings off.

    QuantityValueToleranceSource
    Clear aperture25.000.05Of the first surface
    Focal length, on axis50.000.20At the design wavelength 550 nm
    On-axis spot radius0.0500.010RMS at the focal plane, 550 nm
    Field, half-angle5 deg—The off-axis check
    Aspheric term4th order only—When aspheres is on

    A ray tracer in the repo (Python is fine) traces a documented 21-ray pencil on axis and a 21-ray pencil at 5 degrees. The prescription is a table in this file or a shipped fixtures/lens-prescription.csv. No lab coating or optical bench is required.

    Screenshots

    These pictures are headless OpenSCAD renders of the reference model in reference/model.scad, at the default inputs. They are not drawings of a product that has not been modelled. The first picture is the one Spekkr shows on the spec's card. A build may ship its own goldens; those goldens are of that build and do not have to match these frames pixel-for-pixel.

    FileWhat it shows
    shots/cell.pngTwo-element cell 25 mm aperture, 50 mm focal length

    How to verify

    1. Build the model at the default inputs and at two other legal input sets.
    2. Measure every row of the acceptance table in this file; each must fall inside its tolerance. Use the solid itself or a mesh library on the exported STL, as design.md says for the chosen tool.
    3. Refuse at least one out-of-range input by name before modelling.
    4. Confirm each exported part is a manifold, watertight mesh, and that walls and print orientation hold where this file asks.
    5. Focal length, image-plane spot size, and an aberration metric are compared to documented targets. Ray-trace plots are checked as goldens. No optical bench and no physical glass are required.
    # Design
    
    This file fixes the geometry. How the model's code is organised is up to
    you, as long as the numbers here hold.
    
    ## Targets (defaults)
    
    Focal 50 mm ± 2%. On-axis RMS spot ≤ 15 µm. 0.7-field RMS spot ≤ 40 µm.
    The exact bounds and the wavelength (for example 587 nm) are
    `fixtures/lens-targets.json`.
    
    ## Prescription
    
    Element count, radii, thicknesses, and index (or a glass name from a
    shipped table) are `fixtures/lens-prescription.json` after a design run,
    or a shipped starting point that already meets the bounds.
    
    ## Plot tolerance
    
    A pixel tolerance of 8 levels, 98% of pixels, or a numeric plot array
    within 1e-3 of a fixture, as the tests prefer.
    
    ## Inputs
    
    | Key | Label | Default | Notes |
    | --- | --- | --- | --- |
    | `focal` | Target focal length, mm | 50 | From 24 to 85, the design target. |
    | `aperture` | F-number | 2.8 | From 1.8 to 8. |
    | `format` | Image circle, mm | 43.3 | The diagonal of the image plane, default a 35 mm still frame. |
    | `coatings` | Simple coatings | off | A transmission factor in the tracer, not a physical lab coating. |
    | `aspheres` | Allow an asphere | on | If off, all surfaces are spherical or planar. |
    | `tracer` | Tracer | shipped | shipped is a small tracer in the project. external is a named open library, if present. |
    
    ## Acceptance at the defaults
    
    Millimetres and a unit-focal tracing. Default aperture 25, focal length
    50, two elements, aspheres on, coatings off.
    
    | Quantity | Value | Tolerance | Source |
    | --- | --- | --- | --- |
    | Clear aperture | 25.00 | 0.05 | Of the first surface |
    | Focal length, on axis | 50.00 | 0.20 | At the design wavelength 550 nm |
    | On-axis spot radius | 0.050 | 0.010 | RMS at the focal plane, 550 nm |
    | Field, half-angle | 5 deg | — | The off-axis check |
    | Aspheric term | 4th order only | — | When `aspheres` is on |
    
    A ray tracer in the repo (Python is fine) traces a documented 21-ray
    pencil on axis and a 21-ray pencil at 5 degrees. The prescription is a
    table in this file or a shipped `fixtures/lens-prescription.csv`. No lab
    coating or optical bench is required.
    
    ## Screenshots
    
    These pictures are headless OpenSCAD renders of the reference model in
    `reference/model.scad`, at the default inputs. They are not drawings of a
    product that has not been modelled. The first picture is the one Spekkr
    shows on the spec's card. A build may ship its own goldens; those goldens
    are of that build and do not have to match these frames pixel-for-pixel.
    
    | File | What it shows |
    | --- | --- |
    | `shots/cell.png` | Two-element cell  25 mm aperture, 50 mm focal length |
    
    
    ## How to verify
    
    1. Build the model at the default inputs and at two other legal input
    sets.
    
    2. Measure every row of the acceptance table in this file; each must fall
    inside its tolerance. Use the solid itself or a mesh library on the
    exported STL, as `design.md` says for the chosen tool.
    
    3. Refuse at least one out-of-range input by name before modelling.
    
    4. Confirm each exported part is a manifold, watertight mesh, and that
    walls and print orientation hold where this file asks.
    
    5. Focal length, image-plane spot size, and an aberration metric are
    compared to documented targets. Ray-trace plots are checked as goldens. No
    optical bench and no physical glass are required.
    

    Open design.md as plain text

  • tasks.md Build tasks in order, each with a check.

    Tasks

    Work in order. Finish a task's check before starting the next.

    1. Set up. Install the tool in the tool input (CadQuery or OpenSCAD) and a mesh library for the checks. Read spekkr.inputs.json into one place at the top of the model. Check: an empty or trivial model is written and read back, and the inputs are visible in that source.
    2. Metrics (R1). Build what R1. Metrics asks for, to the numbers in design.md. In particular: The effective focal length is focal within 2%. The on-axis RMS spot size at the image plane, at the design wavelength, is under the bound in design.md (for example 15 µm at F/2.8, 50 mm, as in the file). Check: the R1. Metrics acceptance checks pass.
    3. Plots and files (R2). Build what R2. Plots and files asks for, to the numbers in design.md. In particular: Goldens: a layout (elements and rays), an on-axis spot, an off-axis spot, and a ray-fan plot, compared with a documented tolerance on plots. Check: the R2. Plots and files acceptance checks pass.
    4. Inputs (R3). Build what R3. Inputs asks for, to the numbers in design.md. In particular: Numeric inputs are read from spekkr.inputs.json and must stay inside the ranges in the Inputs table in design.md. Check: the R3. Inputs acceptance checks pass.
    5. Checks and report (R4). Build what R4. Checks and report asks for, to the numbers in design.md. In particular: One command builds the model at the defaults and at the person's inputs, measures it against design.md, and writes a short report. Check: the R4. Checks and report acceptance checks pass.
    6. Goldens and the report. Render the documented views of the default model. Run the one-command report at the defaults, at the person's inputs, and at one more legal set. Check: the report says pass throughout, and the rendered views are written next to the model.
    7. Finish. Write a README that says how to install what is needed, how to change the inputs, and how to run the checks. Run the one command from a clean folder. Check: every item under Definition of done in spec.md is true.
    # Tasks
    
    Work in order. Finish a task's check before starting the next.
    
    1. **Set up.** Install the tool in the `tool` input (CadQuery or OpenSCAD)
    and a mesh library for the checks. Read `spekkr.inputs.json` into one
    place at the top of the model.
       Check: an empty or trivial model is written and read back, and the
    inputs are visible in that source.
    
    2. **Metrics (R1).** Build what `R1. Metrics` asks for, to the numbers in
    `design.md`. In particular: The effective focal length is `focal` within
    2%. The on-axis RMS spot size at the image plane, at the design
    wavelength, is under the bound in `design.md` (for example 15 µm at F/2.8,
    50 mm, as in the file).
       Check: the `R1. Metrics` acceptance checks pass.
    
    3. **Plots and files (R2).** Build what `R2. Plots and files` asks for, to
    the numbers in `design.md`. In particular: Goldens: a layout (elements and
    rays), an on-axis spot, an off-axis spot, and a ray-fan plot, compared
    with a documented tolerance on plots.
       Check: the `R2. Plots and files` acceptance checks pass.
    
    4. **Inputs (R3).** Build what `R3. Inputs` asks for, to the numbers in
    `design.md`. In particular: Numeric inputs are read from
    `spekkr.inputs.json` and must stay inside the ranges in the Inputs table
    in `design.md`.
       Check: the `R3. Inputs` acceptance checks pass.
    
    5. **Checks and report (R4).** Build what `R4. Checks and report` asks
    for, to the numbers in `design.md`. In particular: One command builds the
    model at the defaults and at the person's inputs, measures it against
    `design.md`, and writes a short report.
       Check: the `R4. Checks and report` acceptance checks pass.
    
    6. **Goldens and the report.** Render the documented views of the default
    model. Run the one-command report at the defaults, at the person's inputs,
    and at one more legal set.
       Check: the report says pass throughout, and the rendered views are
    written next to the model.
    
    7. **Finish.** Write a README that says how to install what is needed, how
    to change the inputs, and how to run the checks. Run the one command from
    a clean folder.
       Check: every item under Definition of done in `spec.md` is true.
    

    Open tasks.md as plain text

  • reference/model.scad A parametric OpenSCAD reference of the default model, used to make the listing pictures. A build may use CadQuery or OpenSCAD and does not have to start from this file.
  • shots/hero.png Studio hero of the default model.
  • shots/detail.png Closer studio still of a joint, pocket, or surface.

Is it for me?

Kind
CAD
Tools
Python

What does it need?

Agent permissions
shell
Commands it runs
install a Python optics or ray-trace library, or a small shipped tracer install a tool that compares plots

What changed?

  • 0.2.015 Nov 2026Studio product shots from a shared Cycles rig (cyclorama, PLA materials), including a short loop when present.
  • 0.1.014 Nov 2026First version.

For agents: start.md says how to build this spec, spec.json holds everything on this page, and bundle.tar.gz holds its files.