Nibbling on the Audio Food Chain
Audio Specifications: Loudspeaker Frequency Response
Article By Dick Olsher

  Let us get ready to slice and dice one of the most venerated loudspeaker specifications of all time. Frequency response is probably the first piece of information the eye is drawn to when perusing a manufacturer’s brochure. It appears to be objective, informative, and duly scientific as it invokes technical terms that might seem esoteric to the general public. Is there any reason to question the usefulness and validity of this spec? The answer will surprise you. I submit that it is in fact one of the most deceptive descriptors ever invented by the audio industry.

The most basic form of the specification is a frequency range, followed by a sound pressure level (SPL) tolerance. For example, 40Hz to 20kHz (+/- 3dB) denotes a bandwidth range of 40 Hz to 20 kHz and a tolerance band of 3dB. The measurement normally refers to a distance of one meter from the front baffle, and although often unstated, it is implicit that the reference angle is 0-degrees or normal to the front baffle.

The 3 dB tolerance band is pretty much standard, and represents a factor of two variation in intensity from the mid line. If this sounds like a lot to you, well… it should. Intensity variations of even less than a dB may be audible – depending on the associated bandwidth. This alone should give you an inkling that there’s more to frequency response than is captured by this simple specification and that by itself it does little to accurately describe a speaker’s real-world tonal balance. It is fair to say that this basic specification is ambiguous, and in subtle ways that may not be obvious to a non-technical audience. Unfortunately, the audio industry is pretty much stuck with it, so let’s dig a bit below the surface and explore its hidden nuances. Let’s first consider exactly what the spec doesn’t tell you, and then go on to discuss frequency response measurements in general and their utility in defining the in-room performance of loudspeakers.

Flat is Beautiful
In engineering terms, a loudspeaker is a transducer, converting electrical and mechanical energy into acoustical energy. If the intensity of the acoustic signal exactly tracks that of the electrical input with frequency, the conversion is deemed to be perfect. This is exactly what a flat frequency response implies. In other words, flat is perfect. Cleavage is not a desirable feature in a loudspeaker response curve. But we’ve yet to see a speaker design grace this planet that is without some curvature or other irregularities in its response curve. The audio gods have made it hard to scale mount Olympus. Not a problem, at least if the manufacturer only has meet the basic response spec. After all, the speaker is allowed to swing a total of 6 dB - a factor of four in intensity - measured between the response minimum and maximum. Let’s take a look at some of the response possibilities embraced by such a specification.

The following figures (1 – 4) illustrate the frequency response of four loudspeaker models – all of which meet the basic frequency response spec of 40 Hz – 20 kHz, +/- 3 dB. The key point, however, is that the sound of all of these speakers is drastically different.

Speaker 1:

The midrange is recessed relative to the bass and treble. One would expect the tonal balance to emulate the classical British sound as embodied by the QUAD-57 ESL, that is, laid back and polite,

Speaker 2:

In contrast to speaker 1, the midrange is emphasized by a couple of dB relative to the bass and treble. The expected tonal balance would be opposite to that of speaker 1: forward, more immediate, and possibly a bit “honky” in its vowel colorations.

Speaker 3:

The tonal balance is seen to rise slowly from the upper base to the lower treble. The impact of such a response characteristic is likely to be a bright and thin balance, where upper harmonic detail is artificially enhanced. The rising upper octave response might also be indicative of what I call “sushi” (read: raw) treble. Such a sound is not my cup of tea, but there appears to be a fair number of audiophiles (judging on the basis of audiophile speaker sales) who find brighter timbres, added crispness if you will, to be a mark of excellence.

Speaker 4:

This speaker appears to be well behaved except for a broad peak centered in the presence region (4kHz to 6kHz). This could be indicative of a big problem in the time domain such as a significant resonance in this range. The resultant sound quality may well be the least listenable in the bunch: harsh, sibilant, and aggressive.

The moral of the story is that the conventional frequency response spec is unable to convey the essence of a speaker’s tonal balance. The question of how to objectively judge a particular speaker’s tonal balance and sound quality requires much more information. We’ll have to delve deeper into the inner detail of frequency response measurements in order to find more reliable answers.

The Art and Science of Measurement
Let’s assume for the moment that we have at our disposal a properly calibrated sound measurement system and that access to a suitable anechoic chamber isn’t a problem. So we plunk down the measurement mike at one meter from the front baffle and let the system rip. Out of curiosity we decide to repeat the measurement on the axes of both the tweeter and midrange. We discover that both of these response curves don’t look all that smooth, especially in the transition region between these drivers. The problem is that the tweeter and midrange interfere with each other in the frequency range covered by both drivers. Because the drivers are not coincident on the front baffle, the relative distance from each driver to the mike depends on the mike’s position. Small shifts in mike position can make for significant changes in the response. If we were to collect enough data to get a 3-D map of its output, we would discover that the speaker actually radiates acoustic energy in lobes. At lower frequencies, the response tends to be quite uniform in space. With increasing frequency, the response begins to become directional, beaming along the axis of the driver. And in the frequency range around the crossover between the woofer/midrange and tweeter, the response really becomes ragged as a function of position with severe notches in the response corresponding to small changes in mike position.

Out of all these response curves, there’s probably one, a spot somewhere along the front baffle that looks best, and it’s a safe bet that that’s the one curve that will be latched onto by the marketing and promotion department to illustrate the performance of the product. Marketing works along fairly predictable channels, and it would also not be unusual to see a smoothed or artistic version of the response curve inserted in the product’s glossy brochure. This audio cynic in me would even go one step further and point out that some “audiophile” speakers are offered for sale without even a basic frequency response specification. Either these speakers were designed strictly by “ear” without recourse to any response measurements or the response is too “weird” to allow public exposure. Designing a speaker strictly by ear is a dangerous practice that is almost guaranteed to fall short of proper integration. This type of design is in essence a personal statement that reflects the designer’s musical taste or lack thereof.

The measurement distance is also an important technical issue that often gets ignored. A one meter distance is conventionally used, but it may not be appropriate in all cases. No one I know sits that close to a loudspeaker. The definition of a loudspeaker’s near field is the distance within which intensity falls off more slowly than 6dB for each doubling of distance. For large planar speakers, a 1-meter spacing is clearly still in the near field throughout the midrange. This may also be the case for a large conventional box speaker with multiple drivers. And since the typical listening seat is on the order of three meters from the speaker, in the far field, how closely will a conventional measurement describe the response at the listening seat.

Finally, we get to the real thing, and that is the actual sound field at the listening seat. If you agree that you hear the response that is actually produced at your ears, you will have to accept the room as part of the reproduction chain. The room accounts for over 50% of the soundfield at the listening seat. And this basic fact has been known for 25 years! In a typical domestic environment, a spacing of 3-meters from the speaker already places the listener in a sound field where over 50% of the energy (at least up through the midrange) comes from room reflections. The room’s reverberant response then becomes a primary factor in defining the tonal balance at the listening seat. To estimate the reverberant field’s tonal balance requires full knowledge of the speaker’s 3-D frequency response – also known as power response.

Take a look at Figure 5 (John M. Eargle, Handbook of Recording Engineering, Van Nostrand Reinhold, 1992)

What we see here is the power response (in the horizontal plane) of a woofer mounted in a large flat baffle. At low frequencies, where the woofer’s diameter is less that ¼ wavelength, the radiation pattern is uniform in all directions. With increasing frequency, the woofer’s diameter becomes equal to and then exceeds the wavelength of the radiated sound. As a consequence, the sound begins to beam. This is denoted be a higher Q or directivity. If we were to measure the frequency response of this driver, everything would look cool right on axis. We might even feel comfortable in crossing over this driver to a 1-inch tweeter at 3kHz. This would be a mistake, and unfortunately it’s all too common a mistake with many two-way speaker designs. The problem is that not all is well off axis. The power response goes to hell in the midrange as the woofer becomes beamy. This means that there is a lot less midrange energy going into the room’s reverberant field relative to the bass. Therefore, the room’s contribution at the listening seat is deficient in midrange. Despite the fact that such a speaker might measure extremely flat on-axis in an anechoic chamber, it will surely sound recessed and lifeless under most real-world listening conditions.

Harmonic Distortion: A Final Caveat
Amplifiers of all shapes and sizes are typically decorated with a total harmonic distortion (THD) specification. Solid-state amps usually do really well here, sporting THD figures of 0.1% or less. Tube amps are sometimes belittled by techno heads for living dangerously at a THD of 1% or more. It is curious, however, that THD is rarely quoted for loudspeakers. You certainly won’t find that sort of measurement in Stereophile Magazine. The reason is that the figures are pretty bad. There are many speakers out there that routinely exceed 5% THD or even 10% when pushed hard. That’s one of the reasons I’m drawn to horn-loaded designs. They at least don’t sound grainy, rough, or harsh when going from soft to very loud. Lack of distortion, relaxed musical textures, and purity of tone count for a lot in my book. That sort of information, unfortunately, is lacking in frequency response measurements. The speaker may measure pretty flat, but sound like crap. This is really the best reason for using one’s ears to evaluate a loudspeaker. Not all harmonic distortions are born equal. Some are consonant and others (in much lesser percentages) may be very dissonant. Hence, only the auditory system can reliably determine what is or is not objectionable.

A small piece of the big picture can be a dangerous thing. Next time you examine those frequency response specifications or anechoic on-axis response measurements, I hope that you’re better able to read between the lines.