How to Compare Crossover Slopes in dB per Octave

Compare crossover slopes by identifying the order of each individual filter flank and distinguishing its far-stopband asymptote from its actual transition response. A fixed order has the same ideal asymptotic slope magnitude regardless of alignment, while corner attenuation, phase, and acoustic summation can differ.

Open the Crossover Slope Calculator.

How to Compare Crossover Slopes in dB per Octave

An octave represents a factor of two in frequency. A slope in dB per octave describes how level changes over that ratio, but the term is often used as shorthand for a filter’s distant-stopband behavior.

The calculator returns the ideal asymptotic attenuation magnitude for one low-pass or high-pass flank from its order. It does not calculate the exact response at a selected frequency or model a real loudspeaker’s acoustic slope.

For an ordinary nth-order flank, the magnitude approaches a change of 20n log10(2) dB per octave, approximately 6.0206n. Conventional labels round this to 6n.

Use the order of the actual flank, not a generic alignment multiplier. A band-pass has separate low- and high-frequency flanks that must be considered individually.

Define Direction and the Stopband Reference

A low-pass attenuates as frequency rises into its high-frequency stopband. A high-pass attenuates as frequency falls into its low-frequency stopband.

The calculator reports a positive attenuation magnitude, so a fourth-order low-pass is described as approaching 24.0824 dB more attenuation for each frequency doubling sufficiently far into its stopband.

A signed response-versus-log-frequency plot may show that low-pass slope as negative. For a high-pass rising toward its passband, the plotted direction can be positive even though its stopband attenuation magnitude is described using the same order relationship.

Keep the direction and frequency region explicit. “24 dB per octave” alone does not identify low-pass versus high-pass, the corner, alignment, or phase.

Also distinguish an octave from a decade. A decade is a factor of ten, giving the familiar 20n dB per decade asymptote under the same model.

Calculate the Order-Based Asymptote

Far into a low-pass stopband, ideal magnitude typically behaves proportionally to f^(−n). Doubling frequency changes the magnitude by 2^(−n).

Expressing that amplitude ratio in dB gives an attenuation increase of 20n log10(2).

Order of one ideal flank Exact asymptotic magnitude per octave Conventional rounded label
First 6.0206 dB/oct 6 dB/oct
Second 12.0412 dB/oct 12 dB/oct
Third 18.0618 dB/oct 18 dB/oct
Fourth 24.0824 dB/oct 24 dB/oct
Eighth 48.1648 dB/oct 48 dB/oct

Enter four into the calculator to obtain approximately 24.0824 dB per octave. That is an asymptote, not a promise of exactly that change between the corner and twice the corner.

Linkwitz’s filter discussion lists filter orders, stages, and rounded slopes in defined crossover contexts. Alignment and stage construction must be read with the complete transfer function.

Compare Butterworth and Linkwitz-Riley at the Same Order

An ideal fourth-order Butterworth low-pass has magnitude 1/sqrt(1+x⁸), where x is frequency divided by its defined corner.

An ideal fourth-order Linkwitz-Riley low-pass has magnitude 1/(1+x⁴) under its defined alignment. Both approach the same fourth-order high-frequency asymptote, but their transition shapes differ.

At x equal to one, Butterworth magnitude is approximately 0.7071, or −3.0103 dB. Linkwitz-Riley magnitude is 0.5, or −6.0206 dB.

At x equal to two, the responses are approximately −24.0993 and −24.6090 dB respectively. Their changes from the corner to one octave above are approximately 21.0890 and 18.5884 dB, rather than the far-stopband 24.0824 dB.

This is why slope labels do not describe the entire crossover transition. Two filters can share the same order and distant slope while differing in corner behavior and phase.

Examine a Finite Frequency Interval

To measure an average slope between two frequencies, calculate the dB response difference and divide by the number of octaves between them:

Average slope = level difference / log2(f2/f1)

This describes the selected interval. If it straddles the transition region, it should not automatically be called the filter’s asymptotic slope.

For the ideal fourth-order examples, from x equal to two to x equal to four, the Butterworth attenuation increase is about 24.0655 dB and the Linkwitz-Riley increase about 23.5897 dB.

Both approach the same asymptotic value as the comparison moves farther into the stopband. The different finite-interval values do not mean the alignment changes the filter order.

Use adequate data resolution and a defined frequency interval when comparing measured responses. A single cursor point cannot establish a slope.

Count Cascaded Stages Correctly

Filter orders add when the relevant transfer functions are cascaded. Two second-order stages form a fourth-order electrical transfer, not a second-order transfer with an extra alignment multiplier.

A fourth-order Linkwitz-Riley construction can use two appropriate second-order Butterworth sections in each branch. The total branch order is four, so its asymptotic slope is conventionally 24 dB per octave.

Rane’s Linkwitz-Riley reference explains defined crossover construction and summation. Use that complete alignment rather than doubling an already stated fourth-order slope because the label says Linkwitz-Riley.

Likewise, a Bessel label changes the defined transition and timing behavior, but does not justify multiplying an nth-order asymptotic slope by an arbitrary 0.9 factor.

Read the actual circuit or DSP implementation when a product’s naming is ambiguous. A front-panel label may describe stages, total order, or a nominal acoustic target.

Keep Band-Pass Flanks Separate

A band-pass can be formed with a high-pass side and a low-pass side. Its low-frequency attenuation is governed by the relevant high-pass order, while its high-frequency attenuation is governed by the low-pass order.

For example, a defined second-order high-pass cascaded with a second-order low-pass has four poles overall, but each distant flank approaches approximately 12.0412 dB per octave under its respective conditions.

Calling both flanks 24 dB per octave because the total pole count is four would misdescribe that network.

Specify lower and upper corners, flank orders, and the actual transfer function. Interactions can affect the passband and transition when corners are close, so separate isolated-section labels may not fully describe the combined response.

The calculator is deliberately limited to one flank at a time. Use it with the relevant flank order rather than a combined band-pass label.

Distinguish Electrical and Acoustic Slopes

The acoustic branch includes driver response, enclosure, radiation, and the electrical filter. A nominal second-order electrical network does not necessarily produce a second-order acoustic response.

Where relevant responses multiply over a defined frequency region, their dB slopes can add. But the actual pole and zero behavior, frequency ranges, and driver response need to support that interpretation.

Do not apply a generic “acoustic factor” such as 1.15 to an electrical slope. Measure or model the actual branch instead.

Real passive networks also interact with frequency-dependent driver impedance and component losses. The intended acoustic target may require an electrical shape that looks different from the final acoustic branch.

Verify Summation and Output Limits Separately

Slope alone does not establish phase, polarity, driver timing, or the combined response. Equal slope labels can produce very different acoustic transitions.

Measure individual filtered drivers and the sum with a common timing reference, then inspect relevant listening angles. Check the intended alignment through the overlap rather than only distant attenuation.

Also review excursion, breakup suppression, heating, distortion, and filtering limits at the intended output. A steeper electrical label is not a universal protection guarantee.

Retain flank order, topology, alignment, corner definition, exact response where needed, finite-interval measurements, acoustic branch data, and summed verification. That record makes a dB-per-octave comparison meaningful without allowing one rounded slope label to stand in for the complete crossover.

Table of contents