• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

An automated solver for passive crossovers and boxes + CTA-2034A integrated driver database - LoudspeakerLab

wigginjs

Member
Joined
Mar 26, 2026
Messages
12
Likes
40
Evaluating passive crossover topologies is traditionally a manual, iterative process. I wanted to see what happens when you let a solver evaluate thousands of combinations to find the optimal mathematical fit for on-axis, listening window, preference rating, directivity, sensitivity, impedance, and distortion.
The result is a free, browser-based tool called LoudspeakerLab. It handles the computational heavy lifting of crossover design and provides a standardized, public community database for driver measurements and for sharing designs.

How the Solver Works​

You select the drivers, approve the enclosure alignment, and run the solver. It automatically enumerates filter topologies per driver, evaluating 1st through 4th order HP/LP filters, Zobels, parallel RL, and shunt RLC compensation networks. For a typical 2-way, it searches thousands of combinations.
Every candidate topology is simultaneously scored against a comprehensive set of metrics:
  • On-axis flatness and Listening Window
  • Directivity uniformity across 19+ angles
  • Crossover-region lobing and distortion avoidance
  • Minimum impedance limits and preference ratings
  • Overall component count
Baffle diffraction uses a phase-coherent model with 128 edge sample points. Driver acoustic centers are estimated and included in inter-driver phase alignment. T/S parameters are used to automatically design the enclosure (B4, QB3, or SBB4) with port tuning adjusted to fit the cabinet.

Human vs. Machine​

To validate the solver, I ran it against two designs I built manually for my own home theater a few years ago. In both cases, the solver found a mathematically superior compromise in a fraction of the time.
Test 1: 3-Way In-Wall Main (SB26STAC, RS52AN, SB23NACS45)
The solver evaluated the space in roughly 19 minutes (1172s). It significantly flattened the response while pulling the minimum impedance up to a much safer load.
MetricMy Manual DesignLoudspeakerLab Solver
Preference Rating (w/ Sub)6.1 (7.9)6.3 (8.4)
On-Axis Pk-Pk / RMS3.5 / 1.0 dB2.1 / 0.5 dB
LW Pk-Pk / RMS4.0 / 1.2 dB2.6 / 0.6 dB
Nom. / Min Impedance5.0 Ω / 1.9 Ω6.1 Ω / 2.9 Ω
Screenshot 2026-04-17 144643.png


Test 2: Coaxial Surround (SB16PFCR25-4-COAX)
I use these for surrounds. The solver evaluated the space in about 11 minutes (641s), halving the peak-to-peak variance while maintaining a safer minimum impedance.
MetricMy Manual DesignLoudspeakerLab Solver
Preference Rating (w/ Sub)1.0 (3.7)1.4 (4.6)
On-Axis Pk-Pk / RMS13.5 / 2.4 dB6.5 / 1.3 dB
LW Pk-Pk / RMS13.8 / 2.5 dB6.5 / 1.2 dB
Nom. / Min Impedance4.0 Ω / 2.0 Ω8.8 Ω / 2.4 Ω
Screenshot 2026-04-17 145128.png

Output & Analysis​

Once a solution is found, you get a full suite of interactive plots: full CTA-2034A curves, directivity index, horizontal and vertical contour maps, polar response, and per-driver contributions. You can adjust the listening axis via sliders and watch the plots update live, or tweak component values directly in the schematic. You can also input or export your own crossover design via a unique SPICE-netlist language.
The tool generates an exportable crossover schematic (SVG/PNG) and a parts list with vendor links, helping you find the most cost-effective source for components globally.

Standardized Community Database​

The other half of the project is the measurement database. Everything is public, with no paywalls or locked features.
  • All data is standardized to a common frequency grid (48 points per octave, 20 Hz to 20 kHz).
  • Multiple profiles can be uploaded for the same driver.
  • When uploading in-cabinet measurements, you provide the dimensions, and the platform extracts the measurement baffle's diffraction before applying your new design's correction.

Limitations​

It does not model cabinet resonances, panel vibrations, or room acoustics. For drivers with limited angle data, off-axis estimation uses a piston model fit. It is strictly for passive networks (no DSP). Finally, the solver traverses a massive search space to find strong solutions, but those solutions are ultimately constrained by the quality of the input measurement data.
The tool is free and ad-free. You can try it out and browse the database here: loudspeakerlab.io
 
Last edited:
Back
Top Bottom