Speaker building FAQ
Ten questions we are asked repeatedly, answered properly. Every answer cites its sources, and where an answer depends on your room or your design we say so instead of giving you a number that will mislead you. Where a widely repeated claim did not survive checking, we say that too.
Most of what circulates about speaker building is repeated rather than checked. We went through the common claims against primary sources — AES papers, Thiele/Small theory, published bench measurements and vendor documentation. Some held. Several did not.
What happens if I build the cabinet smaller than the designer specified?
For a sealed box the alignment changes in two ways at once. Total system Q rises as Qtc = Qts × √(Vas/Vb + 1), and system resonance Fc rises by the same factor. So a smaller box does not simply lose bass — it raises the resonant frequency and adds a peak just above it. That is what builders are describing when they call a speaker thumpy or one-note.
The volume needed for a target Qtc is Vb = Vas / [(Qtc/Qts)² − 1]. The alignments available to you are therefore bounded by the driver's own Vas and Qts, and no box, however large, gets Qtc below Qts. Choosing the driver chooses most of the outcome before you cut any wood.
Two corrections to that formula in practice. It assumes a lossless box — real leakage and absorption losses, plus series resistance from the crossover and the amplifier, raise effective Qts. And fibrous stuffing increases the apparent volume by roughly 15 to 20 percent, so a stuffed box behaves larger than its internal dimensions suggest.
For a vented box the consequences are worse, because box volume and port tuning interact. An undersized vented box is not simply a smaller vented box; it is a different and usually worse alignment.
Sources: R. Small, “Closed-Box Loudspeaker Systems”, JAES vol. 20 no. 10 (1972); V. Dickason, Loudspeaker Design Cookbook. Confidence: high.
Sealed, vented or quarter-wave — what is the real difference?
Below its cutoff a sealed box rolls off at roughly 12 dB per octave; a vented box below its tuning frequency rolls off at roughly 24 dB per octave. That single fact drives most of the practical difference between them.
Room gain in a typical domestic room rises at something like 6 to 12 dB per octave as frequency falls. It can partly offset a 12 dB per octave rolloff. It cannot offset 24. That is the mechanism behind the reputation sealed designs have for smoother in-room extension.
Three qualifications, because this comparison is routinely oversimplified. Near and above its tuning frequency the vented box has more output than an equivalent sealed box, not less; the curves only cross lower down. The 12 dB per octave figure assumes an ideal second-order alignment, and a high-Qtc sealed box peaks before it rolls off. And “smoother in-room” is a statement about room interaction, not an inherent property of the alignment.
Sources: Parts Express Tech Talk; Arendal Sound. Confidence: high on the slopes, medium on in-room behaviour.
How much baffle step compensation does a speaker actually need?
Less than the textbook figure, in almost every real room. Baffle step is a rise of up to about 6 dB in on-axis output as wavelengths become short relative to the baffle width and the speaker transitions from radiating into full space to half space. The transition frequency can be estimated as f3 = 115 / baffle width in metres, or 380 divided by the width in feet.
Here is the part most builders get wrong. Rod Elliott: “In a normal room, it is unlikely that the step will be greater than 3 dB, unless the speakers are a considerable distance from the walls.” Nearby boundaries restore the low-frequency energy the baffle step removed. Apply full 6 dB compensation to a speaker sitting close to a wall and you get a bass-heavy speaker. Common practice is 3 to 4 dB.
This matters more here than in most markets. Singapore and Malaysian rooms rarely allow speakers a metre or more into the room, so designs voiced for free-space placement routinely measure and sound heavy when they arrive in an HDB or condominium living room. If you are building from a European or American design, assume you will need less compensation than the plans call for.
Note also that the on-axis rise is about 6 dB while the power response rises only about 3 dB; the transition spans roughly two octaves rather than being a sharp knee; and diffraction ripple is superimposed on it. The closed-form number is an estimate, not a target.
Sources: Elliott Sound Products; PURIFI tech note; f3 formula attributed to John Murphy. Confidence: high.
Why can I not just calculate a crossover from the standard formulas?
Because Butterworth, Bessel and Linkwitz-Riley alignments assume an ideal driver: flat response and a purely resistive, frequency-independent impedance. Real drivers have neither. Feed textbook component values into a real driver and you do not get the intended acoustic slope.
Driver impedance variation alone can shift a passive network's effective corner frequency by as much as roughly 0.4 octave. That is precisely why Zobel and impedance-compensation networks are standard practice rather than optional refinements.
Linkwitz-Riley is an electrical target function. What you hear is that electrical filter convolved with the driver's own natural rolloff, plus baffle step and diffraction from the cabinet you built. Hence the standard distinction in DIY practice between the electrical slope and the acoustic target slope. They are rarely the same, and tools such as VituixCAD and XSim exist to reconcile them.
None of this makes the formulas useless. They give the intended result into a resistive, flat load and they are the right starting point. But the finished network has to be simulated against measured driver data and then verified by measurement on the actual baffle.
Sources: Rane Note 160, “Linkwitz-Riley Crossovers: A Primer”; Elliott Sound Products. Confidence: high.
Should I model my box from the datasheet, or measure the drivers first?
Measure. Manufacturers publish nominal parameters with production tolerance around them, and independent bench testing regularly finds real differences on real samples.
The clearest documented example we found: audioXpress Test Bench measured the SB Acoustics Satori MT19CP-8 in a sealed box and obtained F3 of 76 Hz at Qtc 0.70, where the factory figures for the same volume gave 86 Hz at Qtc 0.55. That is the difference between an overdamped alignment and a maximally flat one — audible, and entirely down to which parameter set you modelled from.
There is a second reason to measure that has nothing to do with tolerance: parameters shift with temperature and with use, and a driver's Fs and compliance when it arrives are not necessarily what they settle at. If your alignment only works at one exact Qts, it is a fragile design.
An impedance measurement rig such as a Dayton DATS is the cheapest meaningful improvement you can make to build accuracy — cheaper than a single pair of premium crossover capacitors, and it affects the result far more.
Source: audioXpress Test Bench, Satori MT19CP-8. Confidence: high on the figures; the practice recommendation is ours.
Can I substitute a different driver into a published design?
Not without redesigning the crossover, and the vendors of these designs will not do it for you. Jantzen Audio, who supply the kits for Troels Gravesen's designs, put it plainly: “No, we cannot help with changing the crossovers to accommodate using different drivers than specified, different driver placement on the baffle or changes to cabinet design or volume.”
The reason is technical rather than commercial. A published crossover is not a filter calculated from datasheet parameters. It is a network tuned against the measured amplitude, phase and impedance of those specific drivers mounted on that specific baffle. Change the driver and you change its rolloff, its impedance curve and its acoustic centre. Change the baffle and you change diffraction and baffle step. The published component values then target a response that no longer exists.
This is one named vendor's stated policy, not a universal industry rule — but the engineering behind it is general. If you want to swap drivers, budget for a measurement rig and a crossover redesign, or pick a different published design that already uses the drivers you want.
Sources: Jantzen Audio kit FAQ; Jantzen custom kits FAQ. Confidence: high.
Is an active DSP crossover better than a passive one?
We are not going to tell you it measurably sounds better, because we could not find independent measured evidence for that. The advocacy that exists in public is vendor opinion. Tortuga Audio, for instance, states that active crossovers sound better because they eliminate the disadvantages of passive crossovers, but publishes no frequency-response or distortion measurements to support it. Treat that as an attributed opinion from a party with a commercial interest.
What is certain is the cost structure, which is arithmetic. Stereo active operation needs one amplifier channel per driver per side: four channels for a two-way, six for a three-way, eight for a four-way. That said, channel count no longer means box count — plate amplifiers such as the Hypex FusionAmp range, multichannel professional amplifiers and AVR pre-outs collapse eight channels into one or two units and remove the extra interconnects.
The honest case for going active is capability rather than inherent sound quality: arbitrary filter shapes, independent delay per driver, correction of baffle and room effects, and no power lost in passive components. Those are real advantages when a design needs them. They are not an argument that every speaker should be active.
Sources: Tortuga Audio (vendor opinion, 2017); miniDSP. Confidence: high on channel counts, low on any sound-quality claim.
Does tweeter dome material matter — beryllium, aluminium or TeXtreme?
Less than the marketing suggests, and the public evidence is thinner than you would expect. The most substantial comparison we could find is informal sighted listening by Vance Dickason: TeXtreme sounded more detailed with more high-frequency content than aluminium, and close to beryllium in detail and definition, with beryllium coming across warmer and TeXtreme more forward. The same article positions TeXtreme as a less costly substitute for beryllium.
Read that for what it is. Dickason himself calls the comparison informal — one listener, sighted, not level-matched under controlled conditions, magnitudes unstated. It is a directional impression, not a measurement.
What dome material genuinely determines is where the first breakup mode sits and how well behaved it is. That governs how low you can cross the tweeter and how much breakup energy lands in band. It is a crossover-design constraint, not a ranking of materials — which is why the same material can be excellent in one design and a poor choice in another.
Source: audioXpress, Dickason, 19 November 2020. Confidence: medium — attributed subjective impression, not measurement.
How does a line array work, and how many drivers does it need?
In its near field a line source loses about 3 dB per doubling of distance, against 6 dB for a point source in the far field. The whole design goal is to make the line long enough that the listener sits inside that near field. The transition distance is approximately d = 1.5 × f × h², with d and h in metres and f in kHz.
That formula carries a qualification usually left out of enthusiast write-ups: the transition distance scales with frequency and halves for every octave down. A roughly 2 metre domestic array holds 3 dB per doubling behaviour only in the upper midrange and treble. At low frequencies, at a 3 metre listening distance, it is already behaving as a point source. The benefit is band-limited, not a blanket property of the format.
Driver count and spacing are then set by comb filtering. Centre-to-centre spacing must stay under one wavelength at the highest frequency the array runs, or under half a wavelength on the stricter criterion. Beyond one wavelength comb lines form; at two wavelengths you get the first cancellation. This directly bounds the crossover point — 130 mm drivers at 130 mm centres hold increasing directivity only to about 2.6 kHz, with first cancellation near 5.2 kHz, so a third or fourth-order slope is required.
In short, the driver count is not chosen for its own sake. It falls out of the line height you need for your listening distance, divided by the spacing your intended crossover frequency permits.
Sources: Ureda, “Analysis of Loudspeaker Line Arrays”, JAES 52(5) 2004; Urban, Heil & Bauman, “Wavefront Sculpture Technology”, 111th AES Convention 2001; via Griffin, Near Field Line Array white paper (2003). Confidence: high.
Is building from a kit actually cheaper than buying a commercial speaker?
Usually yes, but we will not quote you a multiplier, and you should be sceptical of anyone who does. You will see claims of the form “this kit equals a retail speaker at three times the price” everywhere. When we went looking for a defensible figure, the available anecdotes spanned roughly two times to ten times. A range that wide is evidence that no fixed multiplier holds.
What the saving actually depends on: how you finish the cabinet, since veneer and paint can exceed the driver cost; the tier of crossover components you specify; whether you value your own labour at zero; and whether you have to buy tools for the job. In Singapore and Malaysia, shipping and duties on driver orders are a real line item that flattens the advantage on smaller builds.
The more reliable argument for building is not price at all. It is access to driver quality and design approaches that do not exist at any commercial price point, and the ability to size and voice the speaker for the room you actually have rather than the room a designer in Denmark assumed you had.
Sources: CSS Audio; Parts Express Tech Talk. Confidence: medium — only the negative claim is defensible.
Still deciding what to build?
See builds people have actually finished for worked examples of most of the above, or which speaker is right for me if you are choosing between formats. If you want a specific design checked against your room before you order drivers, talk to us — we are in Singapore and we would rather you built the right thing once.