I made a tutorial showing a complete two-way speaker design workflow in LoudspeakerLab:
Video:
Finished design: https://loudspeakerlab.io/designs/095973bc-6f3a-40f5-b5d1-1352c58f166c
LoudspeakerLab: https://loudspeakerlab.io
Accuracy / validation study: https://loudspeakerlab.io/accuracy
FAQ / technical details: https://loudspeakerlab.io/faq
Measurement guide: https://loudspeakerlab.io/measurements
The premise of the video is simple: start with a pair of known drivers, let the software create a reasonable enclosure and draft crossover, run the solver, and then inspect the resulting acoustic, electrical, and cabinet data. The example uses the Scan-Speak H2606/920000 tweeter and SB Acoustics SB13PFCR25-04 woofer measurement profiles from the Mechano23 project.
This is not meant to replace engineering judgment. The interesting part, at least to me, is that LoudspeakerLab keeps the whole chain in one place: driver measurement data, measurement context, baffle/box modeling, crossover synthesis, off-axis prediction/aggregation, impedance, CTA-style curves, preference rating, part values, thermal warnings, and cabinet drawings. A lot of the usual spreadsheet/tool-hopping is still conceptually present, but it is tied together in one reproducible design object.
In this particular run, the final small vented two-way came out at:
At a high level, LoudspeakerLab works by treating the loudspeaker as a combined acoustic/electrical system rather than optimizing the crossover in isolation. Driver profiles store FRD/ZMA data, optional off-axis data, distortion, nearfield bass measurements, and metadata about how the driver was measured. If a driver was measured in a cabinet, the platform can account for the measurement cabinet before applying the target design's baffle and box effects. That distinction matters: data measured in a test box is not the same thing as free-air driver behavior.
The FAQ has useful detail on this, including:
The crossover output is also not just a static drawing. On the design page you can inspect the schematic, export it as PNG/SVG, view the SPICE-style netlist, edit component values, and re-evaluate the design. LoudspeakerLab uses the driver's complex impedance in the crossover model, so the electrical transfer functions and system impedance are tied to the actual ZMA data rather than a nominal 4/6/8 ohm load.
For this design, LoudspeakerLab also flags a practical part-power warning: the 9.1 ohm resistor in the woofer branch is under-rated for the predicted dissipation. That is the kind of warning I would expect to review before building. It does not mean the design is unbuildable, but it does mean I would use a higher-power resistor or split the resistance across multiple parts.
The enclosure side is similarly integrated. The finished design includes the cabinet dimensions, internal volume, vented alignment, port dimensions, driver placement, front baffle drawing, and side cross-section. In this video I mostly accept the suggested enclosure because the point is the workflow, not manual box alignment. You can still override the enclosure type and dimensions if you want to work from a fixed cabinet or explore sealed/vented tradeoffs.
For questions about model accuracy, I would start with the Mechano23 validation writeup:
https://loudspeakerlab.io/accuracy
That page compares LoudspeakerLab predictions, VituixCAD predictions, and Klippel NFS measurements for Mechano23. The short version is that the accuracy depends heavily on input data quality. In-cabinet measurements with meaningful off-axis coverage are much better than sparse manufacturer spec-sheet data. LoudspeakerLab can synthesize missing off-axis angles and normalize/repair some measurement limitations, but it cannot make poor or incomplete measurements equivalent to a dense measurement set.
That is also why this example uses the Mechano23 measurement profiles rather than pretending spec-sheet curves are always enough. Spec-sheet data can be useful for rough design exploration, but if you want high-confidence crossover work, real measurements in a known context still matter.
I would treat this video as a quick orientation, not a deep design review. The next useful discussions are the usual technical ones: whether the topology is the best tradeoff, whether the impedance dip is acceptable for the intended amplifier, how much to value the "with sub" score, whether the directivity through crossover is good enough, and whether the power warning should push a parts change or a topology change.
Feedback welcome, especially from anyone who has built Mechano23 or worked through the same drivers in VituixCAD/XSim. I am particularly interested in how people would compare the solver's choices against a hand-tuned design when constrained to the same drivers, cabinet, and listening axis.
Video:
LoudspeakerLab: https://loudspeakerlab.io
Accuracy / validation study: https://loudspeakerlab.io/accuracy
FAQ / technical details: https://loudspeakerlab.io/faq
Measurement guide: https://loudspeakerlab.io/measurements
The premise of the video is simple: start with a pair of known drivers, let the software create a reasonable enclosure and draft crossover, run the solver, and then inspect the resulting acoustic, electrical, and cabinet data. The example uses the Scan-Speak H2606/920000 tweeter and SB Acoustics SB13PFCR25-04 woofer measurement profiles from the Mechano23 project.
This is not meant to replace engineering judgment. The interesting part, at least to me, is that LoudspeakerLab keeps the whole chain in one place: driver measurement data, measurement context, baffle/box modeling, crossover synthesis, off-axis prediction/aggregation, impedance, CTA-style curves, preference rating, part values, thermal warnings, and cabinet drawings. A lot of the usual spreadsheet/tool-hopping is still conceptually present, but it is tied together in one reproducible design object.
In this particular run, the final small vented two-way came out at:
- Preference rating: 5.1
- Preference rating with sub: 7.3
- F3: 57 Hz
- Nominal impedance: 4.7 ohm
- Minimum impedance: 3.1 ohm
- Estimated driver cost: about $150
- Solve time: 547.8 seconds
At a high level, LoudspeakerLab works by treating the loudspeaker as a combined acoustic/electrical system rather than optimizing the crossover in isolation. Driver profiles store FRD/ZMA data, optional off-axis data, distortion, nearfield bass measurements, and metadata about how the driver was measured. If a driver was measured in a cabinet, the platform can account for the measurement cabinet before applying the target design's baffle and box effects. That distinction matters: data measured in a test box is not the same thing as free-air driver behavior.
The FAQ has useful detail on this, including:
- How in-cabinet measurements are corrected before reuse in a new design: https://loudspeakerlab.io/faq
- How sparse off-axis data is extended using a fitted piston/directivity model: https://loudspeakerlab.io/faq
- How CTA-2034A curves, predicted in-room response, directivity index, and preference rating are computed: https://loudspeakerlab.io/faq
- What measurement files are accepted and how uploads are validated: https://loudspeakerlab.io/measurements
The crossover output is also not just a static drawing. On the design page you can inspect the schematic, export it as PNG/SVG, view the SPICE-style netlist, edit component values, and re-evaluate the design. LoudspeakerLab uses the driver's complex impedance in the crossover model, so the electrical transfer functions and system impedance are tied to the actual ZMA data rather than a nominal 4/6/8 ohm load.
For this design, LoudspeakerLab also flags a practical part-power warning: the 9.1 ohm resistor in the woofer branch is under-rated for the predicted dissipation. That is the kind of warning I would expect to review before building. It does not mean the design is unbuildable, but it does mean I would use a higher-power resistor or split the resistance across multiple parts.
The enclosure side is similarly integrated. The finished design includes the cabinet dimensions, internal volume, vented alignment, port dimensions, driver placement, front baffle drawing, and side cross-section. In this video I mostly accept the suggested enclosure because the point is the workflow, not manual box alignment. You can still override the enclosure type and dimensions if you want to work from a fixed cabinet or explore sealed/vented tradeoffs.
For questions about model accuracy, I would start with the Mechano23 validation writeup:
https://loudspeakerlab.io/accuracy
That page compares LoudspeakerLab predictions, VituixCAD predictions, and Klippel NFS measurements for Mechano23. The short version is that the accuracy depends heavily on input data quality. In-cabinet measurements with meaningful off-axis coverage are much better than sparse manufacturer spec-sheet data. LoudspeakerLab can synthesize missing off-axis angles and normalize/repair some measurement limitations, but it cannot make poor or incomplete measurements equivalent to a dense measurement set.
That is also why this example uses the Mechano23 measurement profiles rather than pretending spec-sheet curves are always enough. Spec-sheet data can be useful for rough design exploration, but if you want high-confidence crossover work, real measurements in a known context still matter.
I would treat this video as a quick orientation, not a deep design review. The next useful discussions are the usual technical ones: whether the topology is the best tradeoff, whether the impedance dip is acceptable for the intended amplifier, how much to value the "with sub" score, whether the directivity through crossover is good enough, and whether the power warning should push a parts change or a topology change.
Feedback welcome, especially from anyone who has built Mechano23 or worked through the same drivers in VituixCAD/XSim. I am particularly interested in how people would compare the solver's choices against a hand-tuned design when constrained to the same drivers, cabinet, and listening axis.
