How to measure loudspeaker frequency response

How to measure loudspeaker frequency response

Frequency response measurement is the foundation of loudspeaker evaluation and design. A measured frequency response reveals what a loudspeaker is actually doing across the audible spectrum — peaks, dips, breakup modes, and resonances that cannot be heard in isolation but significantly affect perceived quality. This guide covers the practical measurement of loudspeaker frequency response using industry-standard techniques.

What you are measuring

Loudspeaker frequency response is the ratio of acoustic output pressure level to input voltage as a function of frequency, typically expressed in dB. A flat frequency response (0 dB ± a few dB across the frequency range) indicates neutral tonal balance. Peaks indicate resonances or modal behaviour; dips indicate cancellations or absorption. Upper midrange frequency peaks can sound aggressive or fatiguing; dips can sound dull or lacking presence. Extra care should be taken with the overall balance between upper and lower frequencies to achieve a neutral sounding speaker, with particular attention given to the frequency range that human hearing is most sensitive to.

Frequency response varies significantly with measurement distance, angle, and room acoustics. A speaker that sounds flat on-axis may be severely coloured off-axis. A measurement taken in a reflective room typically captures both direct sound and room reflections, making it difficult to evaluate the loudspeaker alone. The techniques below isolate the loudspeaker from these confounding factors.

Equipment required

Microphone — a measurement microphone with a flat response, typically cardioid or omnidirectional. Consumer microphones (USB mics, headset mics) can introduce significant coloration, if not chosen carefully. Professional measurement microphones can cost significant sums to achieve a flat and consistent response, wide dynamic range (low noise) or high SPL. The most common type of measurement microphones are 1/4- and 1/2-inch condenser capsules, coupled to a preamplifier and audio interface.

Microphone preamplifier and audio interface — if using a professional measurement microphone, a preamplifier with a low-noise output is required. The audio interface must have low noise, low distortion, and adequate input gain. Built-in PC sound cards are generally too noisy for accurate measurements, however, a modern (half decent) external sound card is generally good enough for most common measurements of loudspeakers.

Measurement software — Room EQ Wizard (REW) is a very popular solution, available free for Windows, Mac, and Linux. Alternatives include Acourate, Smaart, and open-source tools like JMESS. REW combines ease of use with professional-grade analysis. Professional measurement software is generally rather expensive and out of reach for the enthusiast but tend to offer features such as more repeatable measurement via calibration that is required for quality consistency.

Loudspeaker under test — powered loudspeaker, amplifier + passive loudspeaker, or headphone driver. The measurement approach is the same regardless.

Playback device — laptop or desktop computer capable of running measurement software and outputting to an audio interface.

Measurement techniques

Near-field measurement

Near-field measurement places the microphone very close (typically 10–30 cm) to the loudspeaker driver, within the acoustic near field where the wavelength is larger than the measurements distance (a definition of the near field and far field can be found here). The near-field measurement captures the output of that driver with minimal contamination from room reflections and other drivers in a multi-driver system, due to the level being so much higher from the driver than the reflected sound.

Procedure:

1. Position the microphone close to (<15 cm) the centre of a driver
2. Take a measurement using swept-sine (chirp) excitation (see below)
3. Ensure that the stimulus level is within the linear range of the microphone (too high and the microphone can clip, too low and there will be too much noise)
4. Repeat at the same distance for other drivers
5. If desired, one can overlay or splice the measurements to obtain the combined nearfield response.

Near-field measurements isolate each driver but do not account for the acoustic interference (cancellations and summations) that occur between drivers at different distances. They are invaluable for identifying driver-level problems — breakup modes, defects, or anomalies — but do not represent the listener's experience of the assembled loudspeaker.

Far-field measurement

Far-field measurement places the microphone at a practical listening or standard measurements distance (typically 1–2 metres), on the loudspeaker's primary radiation axis. The measurement captures the combined output of all drivers and includes acoustic interference effects at the microphone location.

Room reflections contaminate far-field measurements in untreated spaces. To suppress reflections, use time-windowing (gating) of the impulse response: capture the full impulse response, then apply a time window that includes the direct sound (and possibly early reflections) but excludes the diffuse reflections that arrive later. The FFT of the windowed impulse response gives a frequency response that is less sensitive to room acoustics.

Procedure:

1. Position the microphone at the intended distance, on-axis with the loudspeaker.
2. Position the microphone height at the intended listening height (typically the tweeter axis or midway between the tweeter and the midrange driver (in a 3 or 4 way speaker) or low frequency driver (in a 2 way speaker)
3. Measure the impulse response using swept-sine, or MLS noise burst excitation (depending on the software used)
4. In the measurement software, apply a time window (gate) that captures approximately the first 10–20 ms of the response, depending on room size and reflections.
5. The software computes the FFT of the windowed response, yielding the gated frequency response
6. Repeat off-axis (e.g. 15° lateral, 15° vertical) if directivity is of interest

Time-windowing is typically effective for frequencies above approximately 200 Hz. Below 200 Hz, the wavelength is long and early reflections cannot be cleanly separated from direct sound; far-field low-frequency response remains contaminated by room modes even with gating.

Swept-sine (chirp) excitation

Swept-sine measurement uses an exponential sine wave that sweeps logarithmically from low to high frequency over a fixed duration (typically a few seconds). The frequency increases exponentially so that each octave takes the same amount of time.

The recorded output is deconvolved with the input signal to extract the impulse response, from which frequency response is computed via FFT. Swept-sine offers slightly better signal-to-noise ratio compared to pink noise or pseudorandom (MLS) signals, particularly at low frequencies where noise and room modes are problematic.

Practical measurement procedure for REW

In REW: Generate → Sweep Signals, set the frequency range (typically 20 Hz to 20 kHz) and duration (~5 seconds). Play the sweep through the loudspeaker, record the acoustic output with the microphone, import the recording into REW, and the software performs the deconvolution automatically.

1. Choose the measurement location — ideally an acoustically treated space, or at minimum a room with minimal early reflections. A large, empty room is preferable to a small furnished room. At the very least, conduct measurements at the intended listening position to capture the acoustic environment where the loudspeaker will be used.
2. Position the loudspeaker — on a stable speaker stand, pointing toward the measurement microphone. Avoid corners and walls — place the loudspeaker as far from boundaries (including the floor) as possible.
3. Position the microphone — at ear height, on the loudspeaker's on-axis direction, at the listening distance of interest (typically 1–2 m for studio monitors).
4. Check levels — play the sweep at the intended playback level. In REW, record the output and check that the peak level is at least 20 dB above the background noise floor and well below clipping (use at least 3 dB of headroom).
5. Acquire the measurement — play the sweep and the software will simultaneously record the microphone output.
6. Process the data — REW deconvolves the recording with the known sweep signal to extract the impulse response. Verify that the impulse response is clean and peaks near the beginning; a smeared or noisy impulse response indicates a problem with the measurement or signal routing.
7. Apply windowing — use a time-gated window to suppress room reflections. Start the gate at the impulse peak (time = 0) and extend it 10–20 ms forward, depending on how close the loudspeaker is and how reflective the room is. Longer gates include more of the acoustic environment; shorter gates isolate the loudspeaker but may miss high-frequency detail due to spectral leakage.
8. Examine the frequency response — plot the magnitude response from 20 Hz to 20 kHz. Look for peaks (resonances), dips (cancellations), and overall trends. Smooth the response using 1/3-octave smoothing for a perceptually relevant view; use narrow (unsmoothed) response to identify narrow peaks that may indicate driver breakup or enclosure resonances.
9. Repeat measurements at different positions — take on-axis, 15° off-axis lateral, and 15° off-axis vertical measurements to assess directivity. Professional loudspeakers should maintain relatively consistent response within ±15° of on-axis.

Common measurement errors

Room reflections dominating the measurement — Reflections arriving after the direct sound contaminate the frequency response, particularly below 500 Hz where wavelengths are long. Solution: Use time-windowing more aggressively, move to a larger or more treated room, or move the measurement position closer to the loudspeaker.

Insufficient signal-to-noise ratio — If the background noise is high relative to the signal, the measurement becomes unreliable. Solution: Increase playback level, use a longer sweep duration (30–60 seconds), or improve the microphone or preamp signal chain.

Microphone proximity effects and cable handling — Holding the microphone by hand induces small movements and reflections from the hand; handling or moving the microphone cable induces noise. Solution: Mount the microphone on a stand or boom, keep cables away from the measurement area, and avoid touching the microphone during measurement.

Incorrect microphone orientation — A cardioid microphone's response varies significantly off-axis; pointing it at an angle to the source introduces errors. Solution: Ensure the microphone's axis is perpendicular to the loudspeaker surface (for near-field) or aligned with the on-axis direction (for far-field). Omni-directional microphones are not affected by orientation effects.

Inadequate impulse response length — If the captured impulse response is truncated before the late decay has fully decayed, the FFT shows artifacts. Solution: Extend the IR capture time or reduce the sweep duration so that the full impulse response fits within the recording.

From measurement to understanding

A frequency response curve is a starting point, not an end in itself. Interpretation requires listening:

• A narrow peak (Q > 5) at high frequencies often sounds more problematic than a broad 3 dB peak
• A smooth presence peak (4–8 kHz) can enhance perceived detail and presence without sounding aggressive
• A dip in the presence region sounds dull and lacking projection
• Bass response below 100 Hz is difficult to evaluate from measurement alone in small rooms; listening in different rooms provides crucial context

Use measurement to identify problems, but always validate perceived sound quality through careful listening in the intended application.