speaker-design

Speaker Design

Reference knowledge and tools for loudspeaker engineering. Use this skill whenever a question touches transducers, enclosures, crossovers, acoustic measurement, or audible behavior of physical sound sources.

Three layers

1. Physics of sound — wave behavior, SPL, radiation, room interaction.
2. Acoustic properties — measurable system behavior (FR, distortion, directivity, impedance) and the parameters that drive it.
3. Materials and construction — what a speaker is built from, and how each part's physical properties surface in the acoustic result.

Routing table

Load the reference closest to the user's question. Skim only — don't preload everything; this skill is structured for progressive disclosure.

User is asking about	Load
sound waves, SPL math, room modes, near/far field	references/physics-of-sound.md
frequency response, distortion, directivity, Q	references/acoustic-properties.md
Fs, Qts, Vas, BL, Xmax, datasheet interpretation	references/thiele-small.md
sealed / ported / bandpass / horn / TL tuning	references/enclosures.md
crossover order, slope, alignment, BSC, Zobel	references/crossovers.md
cone, surround, magnet, voice coil, cabinet stock	references/materials.md
compression, ribbon, AMT, planar, electrostatic	references/driver-types.md
REW, gating, Klippel, polar, T/S extraction	references/measurement.md
subwoofers, multi-sub, room gain, sub integration	references/subwoofers.md
active speakers, FIR/IIR DSP, room correction	references/dsp-and-active.md
baffle step, edge diffraction, panel/internal modes	references/baffle-and-cabinet-acoustics.md
closed-box geometry, shape, Vb, driver placement	references/closed-box-geometry.md
room modes, image method, FEM/FDTD, multi-sub sim	references/room-response-simulation.md
horn flare, waveguide, OS, tractrix, CD, Hornresp	references/horns-and-waveguides.md
compression-driver phase plug, coaxial bullet, tangerine	references/phase-plugs.md
point source, coaxial, line array, CBT, splay	references/point-source-and-line-arrays.md
listening triangle, toe-in, height, first-reflection	references/listening-setup.md
absorbers, bass traps, diffusers, LEDE, T_60	references/acoustic-treatment.md
spinorama (CTA-2034), Harman target, standards	references/standards-and-targets.md
impulse / step response, group delay, audibility	references/time-domain.md
over-ear, IEM, planar, Harman headphone target	references/headphones.md
terminology lookup (cross-skill)	references/glossary.md
foundational papers, books, standards, web resources	references/bibliography.md

Calculation tools

Executable helpers live in tools/. Run them from this directory. Prefer them to manual arithmetic — they bake in unit handling and end corrections you will otherwise forget.

Tool	What it computes
tools/ts_from_added_mass.py	T/S parameters from free-air + added-mass Fs
tools/sealed_box.py	F_c, Q_tc, -3 dB from T/S and box volume
tools/ported_box.py	B4/QB3/C4 box and Fb from T/S
tools/port_length.py	Port length from area, Vb, target Fb
tools/crossover_lr.py	Linkwitz-Riley filter component values
tools/xmax_spl.py	Displacement-limited SPL vs. frequency
tools/room_modes.py	Axial/tangential/oblique modes, Schroeder freq
tools/group_delay.py	Filter group delay vs. audibility threshold
tools/diffraction_olson.py	Baffle step + edge diffraction prediction

All scripts take CLI args, print results in SI, and warn when an input violates a model assumption (e.g. ka > 1 for piston-band T/S).

Worked examples

cookbooks/ carries end-to-end design walkthroughs. Read one before the user's first real project — they show how the pieces compose.

• cookbooks/bookshelf-2way.md — 6.5" + 1" dome bookshelf, ported, LR4.
• cookbooks/sealed-subwoofer.md — 12" sealed sub with Linkwitz transform and a single PEQ for room mode 1.
• cookbooks/three-way-tower.md — 10" + 5" + 1" floor-stander, ported LF, sealed mid, optional active LF.
• cookbooks/open-baffle-dipole.md — H-frame dipole mid/HF with boxed LF assist; Linkwitz-school architecture.
• cookbooks/active-studio-monitor.md — DSP-active 6.5" + 1" with waveguide, Linkwitz transform, FIR option.

Design workflow

When the user is actually designing (not just asking a conceptual question), walk them through:

1. Targets — bandwidth, max SPL, distortion ceiling, size, placement, room. Without targets, every parameter is arbitrary.
2. Topology — number of ways, driver sizes, transducer types. Sanity-check against target SPL and bandwidth.
3. Alignment per driver — Thiele/Small for LF; horn/waveguide for HF. Predict response, excursion, port velocity.
4. Crossover — pick acoustic target slopes (LR2, LR4 are common), derive electrical components accounting for driver impedance (Zobel, LCR notches), check on/off-axis blend.
5. Baffle and room — baffle step compensation (BSC), edge diffraction, room gain below Schroeder, boundary loading.
6. Measure and iterate. Simulation is a starting point, not an answer. Gated quasi-anechoic + nearfield merge gives a usable in-room prediction.

Output discipline

• Show formulas with variables defined, in SI units, with assumptions noted (e.g. "valid for ka < 1", "small-signal only", "lossless box").
• Cite the formula a value came from, not just the number.
• Prefer plots/diagrams for transfer functions, impedance, polar response; if the environment can't render, describe the shape (slope, peak frequency, Q) explicitly.
• Distinguish electrical, mechanical, and acoustical domains when discussing impedance, Q, or power. Mixing them silently is the most common error in DIY speaker writing.

Pitfalls and red flags

If you see any of these in a user's design or question, flag it explicitly:

• Tuning a ported box below Fs without a steep electrical high-pass — guarantees cone unloading and mechanical failure on low content.
• Crossing a tweeter within 1× of its Fs at a slope < LR4 — burns voice coils, raises distortion above audibility.
• Reading datasheet sensitivity as "loudness" — it ignores baffle step, room boundary, and impedance dips. The real in-room number is often 3–6 dB lower than the spec.
• Designing the crossover from electrical filter formulas alone without measured driver response — the acoustic slope is what matters, not the filter slope.
• Believing a port works below its tuning frequency — below Fb the port unloads the cone, output drops 24 dB/oct, and excursion peaks.
• Stuffing a port — kills the resonance you designed for.
• Using "more damping = better" — over-stuffed boxes lose efficiency and bloat the effective volume unpredictably.
• Quoting Q without specifying which Q — Qts, Qes, Qms, Qtc, Q of a filter, Q of a notch are all different things.
• Conflating efficiency and sensitivity — see acoustic-properties.md.