How to measure a loudspeaker: a beginner's guide

In short: A calculation tells you how a loudspeaker should play. A measurement tells you how it really plays. All you need is a measurement microphone for about €100, a computer and a free program. The three most important measurements are: the frequency response of each driver in the finished cabinet, the impedance of the finished speaker, and a check once the crossover is connected.

Why measure at all

Every loudspeaker design program, Loudsmith included, works with a model. The model is based on the driver manufacturer's data and on physics formulas. Reality differs from it for several reasons:

Driver parameters vary. Two units of the same driver commonly differ by 10 to 20 % in resonance and in other parameters. On top of that, catalogue data is often measured differently from the way you will use the driver.
The cabinet is not ideal. Damping material, braces, air leaks and the space taken up by the drivers change the volume and the tuning.
The edges and shape of the cabinet change the sound in the treble and midrange in ways that cannot be calculated exactly.

An example from practice: the Danish designer Troels Gravesen measured the same bass reflex cabinet empty and with dense filling. The tuning dropped from 37 Hz to 33 Hz, just because of the damping material. In his finished designs, which are usually heavily damped, the real tuning comes out about 10 % lower on average than the bare formula gives. With SEAS kits and designs with ordinary damping, the Loudsmith calculation is accurate to within a few percent. A measurement tells you where your speaker stands.

The good news: measuring is neither expensive nor complicated today.

What you need

Item	What for	Price
A USB measurement microphone, for example the miniDSP UMIK-1 or Dayton UMM-6	frequency response, with a calibration file	around €100
A computer with REW (Room EQ Wizard) or the speaker measurement page on Loudsmith	plays the test signal and analyses it	free
An amplifier and cables	drive the driver during the measurement	you already have them
A microphone stand	keeps the microphone in a fixed position	€10 to €20
An impedance jig: a 100 Ω resistor and a sound card with an input and output, or a ready-made Dayton DATS	impedance, port tuning, driver parameters	€10 to €150

A phone microphone is not good enough for measuring a loudspeaker. It is inaccurate, especially in the bass and the treble. You can use it for a rough comparison of spots in the room in the room planner.

How a frequency response measurement works

The program plays a rising tone through the speaker, from the deepest bass to the highest treble. The microphone picks it up and the program compares what it played with what it heard. The result is a curve showing how loud the speaker plays each note. Ideally it would be a straight line.

The problem is that the microphone does not hear only the speaker. It also hears reflections from the floor, ceiling, walls and furniture. They turn the curve into a comb of peaks and dips that have nothing to do with the speaker.

The solution is a simple trick: a time window. The direct sound from the speaker reaches the microphone first. The floor reflection arrives a few milliseconds later, because it travels a longer path. The program uses only the start of the recording, before the first reflection arrives, and throws away the rest. That gives you a measurement as if it had been made in a large anechoic chamber.

This comes at a price. The shorter the window, the worse it captures low notes. Roughly: the lowest frequency the window captures reliably is 1 divided by the window length. A 5 ms window is therefore valid from about 200 Hz up. That is why the bass is measured differently, close up.

Step 1: measurement from a distance (mids and treble)

Put the speaker in the middle of the room, as far from the walls as possible. Ideally on a stand or a chair, so that the tweeter is at a height of about 1 m or more.
Place the microphone 50 cm to 1 m away, at tweeter height or between the drivers. When measuring a single driver, put it on that driver's axis.
Set the volume so that the test sounds clear but not unpleasant. At one metre that is roughly 80 to 85 dB.
Measure each driver separately and do not move anything in between. Neither the microphone nor the speaker may move, otherwise the timing between the drivers, which the crossover needs, will be wrong.
If you want to design the crossover properly, measure off axis too: 15°, 30°, 45° and 60° to the side. This shows how the speaker plays into the room, not just straight at you.

From the distance and height you enter, the speaker measurement page calculates when the floor reflection arrives and from which frequency the measurement is valid.

Step 2: near-field measurement (bass)

When you put the microphone right next to the cone, its sound is so strong that the room reflections get lost against it. This is how the bass is measured without a window.

Place the microphone 1 to 2 cm from the centre of the woofer cone, without touching it.
Turn the volume down. Up close the sound is very loud and could overload the microphone.
With a bass reflex, also measure the port opening up close. With a passive radiator, measure its cone.
The program adds the port and the driver together according to their areas. A larger area radiates more, so the port is scaled by the ratio of the areas when they are added.

A near-field measurement is accurate up to about the frequency at which the wavelength is similar to the size of the cone. For a 17 cm driver that is a few hundred hertz, which is enough, because above that the measurement from a distance takes over.

Step 3: joining the two measurements

The near-field bass and the mids and treble from a distance are joined into one curve. They are joined in the range where both measurements are valid, typically between 200 and 400 Hz. The near-field measurement does not capture what happens further from the cone: the front baffle stops supporting the bass. This is called the baffle step: a bass loss of up to 6 dB once the wavelength is larger than the front of the speaker. When joining, the near-field bass is therefore adjusted according to the baffle width.

On Loudsmith this happens automatically. You save the result as an FRD file, a text file with frequency, level and phase that VituixCAD, XSim and Loudsmith can all read.

Step 4: impedance and port tuning

Impedance shows how much the driver in its cabinet resists the current at different frequencies. At first glance it looks dry, but it is the best check of whether the cabinet came out as calculated.

A sealed box has one peak. Its frequency is the resonance of the driver in the box, fc. If it is clearly higher than the calculation shows, the box is smaller or less damped than you think.
A bass reflex has two peaks with a dip between them. The dip is the port tuning, fb. If it is higher than you wanted, make the port longer. If it is lower, make it shorter. Tip: make the port a little too long at first and shorten it step by step; that is easier than lengthening it.
A passive radiator behaves much like a bass reflex. You change the tuning by adding or removing mass on the radiator.

The impedance is saved as a ZMA file. It is essential for a passive crossover, because the crossover components respond to the driver's real impedance, not to its nominal 4 or 8 Ω.

A simple jig is enough to measure impedance: a 100 Ω resistor in series with the driver and a sound card that measures the voltage before and after the resistor. REW has a step-by-step guide for this in its help. More convenient is the ready-made Dayton DATS, which measures both the impedance and the driver parameters at the press of a button.

Step 5: driver parameters

If you want certainty before you build the cabinet, measure the Thiele-Small parameters of the driver itself. The added-mass method and the formulas are in the article Where to get driver data. First let a new driver play bass at medium volume for a few hours. The suspension loosens a little and the resonance drops, so a new driver should only be measured after this break-in.

Step 6: checking the finished speaker

Once the crossover is connected, measure the whole speaker from a distance, on axis and at several angles. Compare the curve with the design. If it differs by more than 1 to 2 dB over a wider band, something is wrong: a driver connected with the wrong polarity, a component with a different value or drivers in a different position.

Checking the polarity is easy: at the crossover frequency the drivers should add up. If there is a deep dip there, one driver is probably connected the wrong way round.

The most common mistakes

An overloaded microphone or sound card. The curve then looks oddly flat or full of nonsense. Turn the volume down.
A missing microphone calibration file. Every measurement microphone has its own file for its serial number. Without it the treble can be off by several dB.
A microphone moved between driver measurements. The crossover then works with the wrong timing.
Too long a window. The curve is jagged in the mids and treble because of reflections. Shorten the window.
A speaker on the floor or next to a wall. The reflection arrives too early and the window has to be very short.
Noise during the measurement. A fridge, a car outside or footsteps show up in the bass. Measure in silence.
Comparing curves with different smoothing. 1/3-octave smoothing hides what 1/24 octave shows. Compare like with like.

How Loudsmith uses the measurements

In the Drivers step you load the FRD and ZMA for each driver. The app then calculates with the real measurement instead of the model.
It designs the crossover from the measured curves, including the real behaviour at the cabinet edges and the real impedance.
For an active crossover it creates DSP filters made exactly for your drivers, for example for the Hypex FusionAmp.
You can enter the port tuning from the impedance measurement back into the design, and the app recalculates the port length.

Small glossary

Frequency response: a curve showing how loud the speaker plays each note.
Time window (gating): only the start of the recording is used, before the reflections arrive.
Near-field measurement: the microphone right next to the cone, so that the room has no influence.
Impedance: the resistance the speaker presents to the amplifier, different at every frequency.
FRD and ZMA: text files with the measured frequency response and impedance.
Baffle step: a bass loss of up to 6 dB when the front baffle of the speaker stops directing the sound forwards.
Port tuning (fb): the frequency at which the port plays the most and the driver moves the least.
Calibration file: a correction for the deviations of a specific microphone unit.