How LoudspeakerLab measures drivers, grades measurement evidence, computes CTA-2034A curves and preference rating, and chooses crossover topologies and component values.
Minimum: on-axis FRD (SPL and phase), ZMA impedance and phase, plus manufacturer, model, and frame diameter. Optional but recommended: off-axis FRDs, distortion, nearfield bass, and T/S parameters (Fs, Qts, Vas, Sd) for enclosure design. Accepted formats are plain-text FRD/ZMA/distortion exports (REW/ARTA/DATS); max 3 MB per file.
Evidence grades measurement source and completeness, not acoustic quality. Stronger baselines: near-field scanner, free-air gated plus nearfield, free-air gated, in-cabinet, manufacturer spec sheets, then unknown. Off-axis, distortion, nearfield, and calibrated SPL can raise the grade within the source cap. A/A- high confidence; B+/B usable with limits; C+/C lower confidence.
Quality Score rates acoustic performance only within an auto-detected usable band from Fs margin, rolloff, beaming, breakup, and distortion limits. Sub-scores (0–100) cover on-axis flatness, directivity tracking, DI smoothness, and distortion. When available, a weighted composite summarizes them: above 80 is strong, above 90 exceptional, below 50 usually fights the solver.
Yes. LoudspeakerLab applies first-order edge diffraction by perimeter integration: 128 edge samples, each an impulse wavelet delayed by its distance from the driver and weighted by the angle it subtends, averaged over the cone area. It captures the ~6 dB 4π-to-2π transition plus the real interference ripple, with edge treatments, off-center drivers, and shaped baffles following their true outline.
Acoustic center is an effective crossover-band phase-center depth behind the mounting plane, not the voice-coil location. Effective depth equals mounting Z plus acoustic center; fully on-axis, less off-axis. LoudspeakerLab estimates it from frame diameter, cone area, resonance, and role; an explicit 0 mm override is valid. Properly time-referenced measured phase is not double-corrected.
For two-module stacked cabinets, each driver's response includes the neighboring module: reflection off its faces and diffraction around its edges, included automatically in previews and CTA curves and shown as a dashed curve on the Diffraction plot. The design's Assumptions detail states whether it is included.
Five: Rectangular, Tapered shoulders (45° front-corner clips that tame edge diffraction), Arched top (solid domed cap), Stepped baffle (set-back upper plate that aligns acoustic centers), and Stacked modules (separate boxes, each with its own baffle and volume). Any shape takes a front rake up to 20°, and square, rounded, or chamfered edge treatments are modeled in simulation and 3D.
A cut list with every panel's dimensions plus shaped-outline cutting notes (tapered corner clips, side notches, facet panels, arch templates), a sheet-layout nesting view with kerf and edge-trim allowances for one or more cabinets, and a driver cutout layout for routing. Copy as CSV or print.
One to three identical round ports, vertical or horizontal, placed on the back, bottom, baffle, or as a single front slot. Acoustic length is the interior run plus the panel it passes through plus end correction, so 0 mm is a bare hole through the panel. Air speed is charted with chuffing warnings and remedies.
Yes. Alignment can be sealed, vented, or passive radiator. Pick a radiator from the catalog sorted by fit, mount it on the baffle or (by default) the back, and tune with added mass in whole grams. Below the radiator's resonance, rolloff is about 12 dB/octave, gentler than a ported box. Low-displacement radiators are flagged.
Yes. The New Design flow offers custom cabinets or pre-built flat-pack cabinets with blank baffles, sorted by how well the volume fits your drivers. The box locks to the product's size and panel thickness while ports and driver layout are still designed for your drivers. The cabinet is optional and excluded from estimated cost: you only cut driver and port holes.
CTA-2034A curves include Listening Window (on-axis plus H±10/20/30° and V±10°), Early Reflections (front/side/rear/floor/ceiling averages), Sound Power from available dual-plane off-axis data, DI (on-axis−SP), SPDI (LW−SP), and Predicted In-Room PIR = 0.12×LW + 0.44×ER + 0.44×SP.
Predicted Preference Rating (PR) follows Sean Olive's 2004 AES Equation 9 adapted with smoothness metrics on a 0–10 scale. Weights: +SM_ON 2.00, +SM_PIR 2.32, −NBD_ON 2.49, −NBD_PIR 2.99, −LFX 4.31 (log10 of bass extension Hz). The with-sub variant assumes perfect bass to 14.5 Hz, removing the LFX penalty.
DI is on-axis minus sound power (dB); SPDI is listening window minus sound power. Uniform DI means consistent tonal balance as rooms change. LoudspeakerLab shows DI/SPDI in results and the solver optimizes directivity uniformity. A rising DI with frequency is expected; sudden jumps or dips near crossovers usually mean off-axis lobing.
Default priority order: on-axis flatness, CTA-2034A listening window, directivity uniformity, distortion avoidance, preference rating, crossover simplicity, amplifier-friendly impedance, and sensitivity. Higher priorities weigh more; every candidate is scored on all criteria. Priorities can be reordered in the auto-solve dialog's advanced options.
Input power is raised until a driver thermal/excursion limit (Pe, Xmax when available) or crossover part limit (resistor power, capacitor voltage, inductor current) is reached. The Max SPL curve is vs frequency; headlines use ~80 Hz for subwoofers and ~1 kHz for main speakers. Part stress uses an AES75-like music-noise spectral weight.
Yes. Advanced options in the auto-solve dialog allow pinning each handoff frequency (20 Hz to 20 kHz within driver limits) with a tolerance, setting slopes from 6 to 24 dB/octave exactly or within one order, reordering the eight objective priorities, and opting out of component families such as L-pads, Zobels, and notch filters.
The Compare button on any completed design opens a picker over your designs or public ones. Response, CTA-2034, impedance, distortion, max SPL, and polar plots overlay both designs, a metrics table highlights the better value per row (PR, slope, flatness, impedance, F3, THD, cost), and two 3D views sit side by side. The view is kept in the URL.
The header's unit picker offers Auto, Metric (mm, L), and Imperial (inches, ft³), saved per account and independent of the region that controls currency. Units apply everywhere lengths and volumes are shown or entered, including cabinet dimensions, ports, panel thickness, and the build sheet's sheet goods presets.
Per driver the solver generates many HP/LP orders (0–4th where allowed; tweeters normally at least HP2), plus attenuation, impedance compensation, traps/notches, and related aux networks, then combines them into full-system layouts. Stage 1 screens thousands of sets with coarse optimization; a simplicity objective prefers fewer parts when acoustics are comparable.
Values are optimized continuously, then snapped to E-series stock: E24 for capacitors and resistors, E12 for inductors. Final simulation includes inductor DCR and capacitor ESR. Bounds are resistors 0.1–100 Ω, inductors 1 µH–1 H, capacitors 1 nF–1000 µF.
With T/S parameters (Fs, Qts, Vas), LoudspeakerLab designs sealed, vented, or passive-radiator boxes and merges the box response into each driver before crossover solving. New designs default to a balanced B4 (Butterworth 4th-order) vented alignment. Volume and port tuning come from Qts alignment tables; shared identical drivers scale one enclosure proportionally.
Infinite baffle disables baffle diffraction correction. Use it when drivers were measured in a true infinite baffle, the speaker will be in-wall/in-ceiling, or you want to compare crossovers without baffle effects. Freestanding cabinets should leave it off and supply front-panel dimensions so baffle step is modeled.
Measured phase mixes driver acoustics with mic-path delay. Minimum phase is the phase fixed by the magnitude response via the Hilbert/Kramers-Kronig relation. LoudspeakerLab synthesizes it from SPL with a real-cepstrum method so setup delay is removed while keeping the shape the solver needs. Two-column FRDs always get synthesized phase; routine solves default to minimum phase.
No. Published drivers and designs are public to browse, view, and search without an account, including plots and schematics. An account is required to upload measurements, create designs, run the solver, star designs, vote on driver accuracy, and join community interactions.