Why tweeters get padded
Sensitivity is how loud a driver plays for a fixed input, measured at 1 meter. Datasheets give it in dB/W/m (decibels for 1 watt at 1 meter) or per 2.83 V at 1 m. Check that both drivers are quoted the same way before you compare them (how to read a driver datasheet). Tweeters are often more sensitive than the woofer they’re paired with. Rod Elliott of Elliott Sound Products (ESP) puts the rule simply: the least sensitive driver sets the whole system’s efficiency, and every other driver has to be padded back to match it. ESP’s crossover tables say the same from the design side. An L-pad is generally needed to bring the tweeter down to the woofer’s level, so make sure the woofer is the least efficient driver or the levels will be hard to get right. Colin Pykett describes what a mismatch sounds like. His rebuilt two-way speakers sounded too bright and were tiring to listen to. When he measured them, the tweeter was at least 6 dB hotter than the woofer, and he settled on 8 dB of padding.
What an L-pad is
An L-pad is two resistors: one in series with the tweeter and one in parallel with it. Together they cut the level and still present the crossover with the impedance it was designed for. ESP says a properly set-up L-pad makes sure the crossover network “sees” the desired impedance. That’s why a single series resistor isn’t the same thing. ESP doesn’t recommend series-only padding for design work. Pykett explains why: crossover values are worked out for a resistive load, and a lone series resistor changes that load. In his speaker it would have put a hump of about 4 dB just above the crossover frequency and softened the filter’s knee. He says he checked that prediction by experiment. An L-pad also helps in a second way. The parallel resistor caps how high the tweeter’s resonance peak can climb, so the network sees a steadier load (ESP). Pykett adds that the extra resistance isolates the crossover from the tweeter’s own impedance quirks. Neither author says this replaces impedance correction. ESP is clear that no passive crossover performs properly if the driver’s impedance changes with frequency (layout and wiring mistakes are a separate problem).
Working out the values
ESP’s L-pad page has two calculators, and it also prints the formulas so you can put them in a spreadsheet. Here they are as ESP prints them, with Z as the speaker impedance in ohms and A as the attenuation in dB:
- Vr = 10^(A / 20)
- Rs = Z × ((Vr − 1) / Vr) (series resistor)
- Rp = Z × (1 / (Vr − 1)) (parallel resistor)
- A = 20 × log(Rs / ((Z × Rp) / (Z + Rp)) + 1) (checks the attenuation of a pad you already have)
Pykett’s appendix uses the same math written another way, and he warns that many online calculators get the dB conversion wrong by mixing up power and voltage ratios. Both methods give the same resistor values. Before the tweeter, look at the woofer’s coil. ESP calls this the step that’s “almost always forgotten”. The woofer’s low-pass inductor has resistance, which turns the woofer down slightly, so the tweeter needs that much extra padding. ESP gives the loss as dB = 20 log(Ri / Z + 1), where Ri is the coil’s resistance. Worked example (illustrative, using ESP’s own numbers). These figures are ESP’s example values, not a real driver pair:
- Woofer 89 dB/W/m, tweeter 95 dB/W/m, so the tweeter needs 95 − 89 = 6 dB.
- ESP’s sample coil is about 600 µH at about 0.53 Ω, feeding a corrected 6 Ω woofer: 20 log(0.53/6 + 1) ≈ 0.74 dB. The woofer loses about 0.74 dB, so the tweeter pad grows to about 6.74 dB.
- Plug 6.74 dB into the formulas, assuming the corrected tweeter also measures 6 Ω (ESP’s typical figure for an ‘8 ohm’ driver after correction): Rs ≈ 3.24 Ω and Rp ≈ 5.12 Ω. The tweeter plus pad still looks like 6 Ω to the crossover.
- Round to resistor values you can actually buy, then run the “existing network” formula to see what attenuation the rounded pair really gives.
ESP’s calculators show three decimals, and he notes you’ll rarely need more than one decimal place. Voice coils heat up and driver parameters drift with age, so he says precision is unrealistic anyway.
Use measured impedance, not the “8 ohm” label
ESP says the speaker impedance in the calculation must be the actual measured value, including any impedance-correction networks. If you use the nominal figure, the attenuation can come out wrong. After correction, he expects a “typical” 8 Ω driver to sit around 6 Ω, a little above Re, the voice coil’s DC resistance. ESP’s crossover tables say the same: never design with the nominal impedance. If you haven’t measured your tweeter yet, start with speaker impedance curves, explained.
Resistor power ratings
ESP’s calculator deliberately skips power ratings, and he says some online calculators give “grossly inflated” figures. His guidance for tweeters: a 100 W system’s tweeter will generally take no more than about 10 W of continuous program material, so no resistor in the pad can exceed 10 W, and 5 W will most likely do. He prefers 10 W parts for the margin. A midrange pad in a three-way is a different case. ESP says it can see considerable power and you’ll almost certainly have to verify it by measurement.
When the woofer is the louder driver
An L-pad doesn’t fix that. Pykett says a woofer has to be driven from a very low source impedance or the bass turns boomy and poorly damped, so it can’t take a passive pad. ESP says an L-pad should not be used on a mid-bass driver. The way out is bi-amping, with level set before the power amps (active or passive crossover). Baffle step also changes how loud the woofer ends up compared with the tweeter, and in a passive crossover the traditional fix is to lower the tweeter level. Read baffle step explained before you settle on a pad value.
Measure after
ESP’s crossover tables say to measure each driver’s sensitivity in the enclosure it will live in, because everything makes a difference. Without that measurement, he says, it’s very hard to get the L-pad right. Pykett is blunter: listening tests are “hopelessly inadequate” for judging a sensitivity mismatch. He found the tweeter in his second cabinet was about 3 dB more sensitive than the first, so each speaker needed its own pad. So pad, then measure: a gated sweep at the listening axis, plus a fresh impedance sweep, because the pad changes both (REW basics). If the summed response isn’t right around the crossover, change one thing at a time (first vs second-order crossover).
Checklist
Woofer impedance rise is a different fix: what a Zobel network does.