The baffle step explained – and how to compensate it

Where the baffle step comes from

When the wavelength is small compared with the baffle, the driver radiates forward only – into half space. When the wavelength grows larger than the baffle, the sound wraps around the cabinet and spreads into full space. In half space the baffle acts like a mirror: the sound that would otherwise go backwards adds in phase to the front, and the on-axis pressure is twice that in full space. Without this doubling, up to 6 dB are missing on axis.

The transition is a step over roughly two octaves, not an edge. By the common rule of thumb its centre is at f3 ≈ 115 / W, with the baffle width W in metres: 20 cm → about 575 Hz, 30 cm → about 380 Hz.

Diffraction: the ripple on top

The baffle edges act as a second, delayed sound source. Depending on frequency it adds to the direct sound or partly cancels it – the response ripples around the baffle step, typically by a few dB. Centred drivers on narrow baffles show it most; offset drivers and rounded edges smooth it. The rule of thumb knows only the width, not the ripple.

Compensating it – passively

The classic solution is a network before the woofer: an inductor with a resistor across it. Below the step the inductor passes everything; above it the resistor limits the current and the woofer plays quieter by the chosen amount. Often a larger low-pass inductor, whose roll-off starts earlier, is enough.

The price: above the step the woofer loses level. The tweeter and midrange have to be lowered by the same amount (e.g. with an L-pad), so the whole speaker becomes less efficient.

How much to compensate depends on placement: free-standing, almost the full 6 dB are missing; near the wall behind the speaker part of it comes back. Many designers choose 3–4 dB and then listen.

The transition frequency and starting values for the inductor and resistor come from the baffle step calculator.

Why a simulation with measurements is more accurate

If you measure the driver in its finished cabinet (gated, on axis), the baffle step above the window is already in the FRD – including the diffraction ripple of your baffle. Below it the window is too short; there many add a near-field measurement whose level is matched to the gated part. In the simulation the network then works with the woofer's real impedance (ZMA), and the inductors' DC resistance (DCR) counts too – both change the result of the rule of thumb.

Create FRD and ZMA files · How to set coil DCR and parts in Crover: parts in the manual