Speaker Tests: Anechoic Frequency Response (Nov. 1974)

by Richard C. Heyser

THE WORD "anechoic" means free from echoes and reverberation. The anechoic frequency response measurement therefore a field measurement in which only direct sound from the speaker, with no other room reverberation, is presented. A principal purpose of this test is a standardized evaluation of the loudspeaker's ability to produce a uniform sound pressure at each frequency in the audio range when driven by a constant-amplitude, variable-frequency electrical signal. It is a strict laboratory measurement in the sense that either a special test facility or specialized data processing equipment is required to perform the test.

The recommended method of making a free-field measurement is to place the measuring microphone far enough away from the speaker so as to be in what is called the "far field" for the wavelength under test. Readings are then referred back to an equivalent distance of one meter under the assumption that the sound spreads in accordance with the inverse square law, that is, doubling the distance reduces the intensity by 6 dB.

Prior to the introduction of coherent processing techniques, the only satisfactory means of making such measurements have been either to use very expensive anechoic chambers, rooms specially constructed to minimize wall reflections, or "roof top" free-field measurements. In the Richard C. Heyser latter case, both speaker and micro phone are hoisted to a sufficient height to be well away from sound-reflecting objects. The economics of either method has led to the placement of the microphone much closer to the speaker than strict far-field conditions would dictate. An additional benefit of this closeness, from one standpoint at least, is that the response tends to become smoother. Consequently, much of the data accumulated by manufacturers and contained in their specifications and advertisements has been obtained very close to the speaker front.

For reasons of reproducibility from one speaker to the next, the anechoic measurements performed for the Audio tests are standardized at an actual distance of one meter when practical. The microphone is placed on the geometric axis of the speaker system and spaced one meter from the front-mounting surface of the forward-pointing speakers for direct-radiator systems. When common sense dictates an alternate micro phone position, such as would be necessary to measure large panels or horn-loaded systems, a more nearly far-field position is chosen and all measurements corrected to one meter. The electrical drive is maintained constant at that voltage level which would produce one watt into a pure resistor specified by the manufacturer as the speaker impedance, usually 8 ohms. The sound pressure level (SPL) is plotted in decibels relative to the standard level of 20 micro-pascals.

Audio uses a fully coherent signal-processing technique known as time-delay spectrometry for making loud speaker spectral measurements. A special class of signal is used which has a frequency domain representation closely approximating what is technically known as a rectangular function. The time domain representation of this signal is therefore technically described as a sine function, known to many as (sin x)/x. A complete description of this process may be found in the technical literature.

In this brief article we will concentrate only on the anechoic amplitude response as a function of frequency. That, after all, is what is usually called-although improperly the frequency response of the speaker. This has been a mainstay of speaker measurement for nearly 50 years.

The concept is deceptively simple. A microphone is placed at the desired position and the speaker is driven by a sine wave signal. The output of a pres sure responsive microphone is then monitored as a function of frequency.

If one wishes to know particle velocity, rather than pressure, he can use a pressure-gradient or similar "velocity" microphone. So long as far-field conditions prevail, the pressure and velocity measurements will be similar.

The first shock one gets, if he is not accustomed to such measurements, is that severe changes in frequency response can occur with minor changes in microphone position. More often than not these changes are due to acoustic interference between widely spaced drivers sharing common frequencies. The speaker manufacturer who places two tweeters several feet apart is creating a situation familiar to antenna designers as a broadside array, with many polar fingers and sidelobes.

If one is compiling data for advertising copy, he then has several alternatives. He may ignore the response irregularities and cite what these drivers are very often able to provide separately or he may smooth the response data over a sufficiently broad frequency range to minimize the condition, muttering things about critical bandwidth. Or he may measure at a point where interference effects are not prominent.

In Audio's case we are not compiling copy for advertisements, but are trying to measure speakers for objective comparison by you, the reader. That is why we have tried to pick one common spatial point we can use to measure all speakers. Our data is accumulated on a one-fifteenth octave basis with straight line interpolation between data points. We do this in order to be able to cover at least every musical note throughout the entire audio spectrum. The result is that the measurements are seldom smooth, and some manufacturers who have speakers thus tested are bent slightly out of plumb.

The amplitude plot is a touchstone of performance from the standpoint of direct sound between the speaker and yourself. It represents what the speaker is capable of doing. Because this type of test has been around for such a long time, most of the obvious timbre-related facts that one can infer from this data are well known. There are, however, a few less well known characteristics which you should be aware of.

For example, any periodicy in the SPL on a linear frequency basis is a sign of physical problems. Audio pro vides a logarithmic frequency plot be cause this is the way most users want the data. If you mentally convert the frequency readings to a linear basis where 10 kHz is halfway between d.c. and 20 kHz, then some of the defects show up as equally spaced patterns. One such defect is provided by the off-axis broadside array effect of widely spaced drivers which share the same frequency. The acoustic effect can be very unrealistic and quite disastrous to stereo imagery in some cases.

The speaker manufacturer who economized on acoustic damping material behind a wide-range direct-radiator speaker can be quickly spot ted by a periodic SPL pattern. Sound from the back of the cone, which radiates almost as well as from the front of the cone in many cases, travels through the enclosure to reflect from the back wall, then continue back to the cone. Because the speaker cone is not as efficient a wall as the cabinet, some of this first sound comes through and the rest is back-scattered to repeat the process. The energy-time plot, which we will describe in a later issue, is a dead giveaway of this behavior. However, in many cases it is also quite prominent in the SPL frequency response.

A closely allied effect is the cabinet which becomes an echo chamber for the speaker because of pinchpenny use of damping material-or design talent. Again, it shows a periodic SPL pattern on a linear frequency basis.

There are usually many peripheral humps and dips in the response which are superimposed on the periodic pattern but which a little practice you can quickly spot the trend.

Another situation to watch out for is the "over-extended woofer." A good bottom end occasionally requires a bit of mass loading of the woofer. This tends to rob some of the top end performance of that woofer if it is also expected to carry the spectrum through the upper middle frequencies. If for economic reasons the tweeter cannot come down far enough to meet the woofer, then a shallow dip in response with a number of sharp dropouts may occur near the crossover frequency.

A shallow dip may be due to a variety of good acoustical design characteristics, but one way to spot if it is due to a woofer running out of steam is to look at the dB-per-octave slope on each side of the dip. An over extended woofer usually dies at a shallower slope than the rise in acoustic response of a tweeter which is driven far below what should be its proper crossover frequency. Because of the phase behavior of a sharp drop in SPL of both woofer and tweeter, they end up cancelling and reinforcing each other in a narrow frequency range.

If a manufacturer lets two or more speakers share a common frequency range, this will inevitably show up as a number of sharp SPL peaks and dips over a much broader frequency range-sometimes as much as an octave in extent. This is one result of put ting the crossover design book aside and letting the lower frequency unit go up as high as it wants and the higher frequency unit go down as far as it can. The sonic effect can be spectacular. This type of speaker can be quickly sold to a prospective buyer in an A-B comparison with a much smoother unit by playing brass, bell, and percussive material. The smoother unit will sound dull by com parison-even if more realistic. The truth is that a large number of sharply spaced peaks and dips which change with listening position contribute to a sound best described as an "ear burner." Beside the SPL indicator, you can readily spot such a speaker by its sound as records are being played. This is a speaker that has the most apparent record background noise of ticks, pops, and scratch when balanced for the most uniform sound.

Another allied effect to watch for in the SPL measurement is any unusual peak in the response more than 3 dB above the average response in the vicinity of that peak. This is a resonance as distinct from multiple speaker reinforcement. Two speakers sharing the same frequency cannot reinforce to give more intensity than the sum of the contributions of each, although they can cancel to a complete null. This, incidentally, is the same for natural sound in a room. Because we are accustomed to such a sound pattern, we can accept it as a manifestation of reverberance. This is one reason for the observation that dips in response are less objectionable than peaks. Again, to listen for it, concentrate on the background noise to see if it is exaggerated. Who among us hasn't put a seashell to his ear to hear the "ocean." We, of course, are coupling ourselves to a resonant chamber that emphasizes the background noise which we are seldom aware of into a recognizable spectral peak. That's exactly what our "peaky" SPL speaker does and the sonic effect is the same.

As a final observation, one of the most ignored components in a loud speaker system is the physical en closure. To be sure, most designers concern themselves with enclosure volume and internal damping, but the size and shape of the "box" as well as where the drivers are placed can adversely change frequency response. A good many designers would do well to read some of the fundamental literature on this subject; for example, Dr. Harry F. Olson's "Direct Radiator Loudspeaker Enclosures," Audio Engineering, Nov., 1951, and Jour. A.E.S., Jan.,1969.

As an example, the sound pressure wave from a speaker can be visualized as an expanding "bubble" which starts from the speaker cone and grows larger in a spherical fashion. When the speaker is mounted on the front surface of an enclosure, this bubble approximates an expanding hemisphere. When an acoustic discontinuity is encountered, a new sound wave is launched from the discontinuity. Obviously the edge of the cabinet is a major contributor, as are molding trim and recessed speaker-well construction. The result of all this is that a speaker which has a very smooth response when mounted on a large baffle, as is common in anechoic chamber tests, can have a terrible looking response when mounted in a smaller enclosure. Thus, some of the disparity between measured loud speaker performance provided by Audio and the advertised performance of a loudspeaker system may be due to these effects.

Audio measures the 4n anechoic response, that is there is nothing around the speaker in the measurement. A speaker mounted against a large wall is radiating into a 2 pi r or hemispherical environment. An expanding sound wave doesn't know how big the front edge of the en closure is until it reaches the edge.

Until the edge is reached, the en closure-insofar as the sound wave is concerned-looks like a large, flat wall. If the sound wave has a high enough frequency that at any instant there is a large pressure change between the sound just starting out from the cone and the sound which has reached the edge of the enclosure, then the enclosure looks large enough to act as a wall. We say then that the front of the enclosure itself acts as a 2 n half-plane boundary for higher frequencies. For low frequencies, where there is very little pressure change across the front of the enclosure, the enclosure might as well not be there. Because of this, the high frequencies will be slightly stronger than the lower frequencies directly in front of the enclosure when measured in an anechoic environment. That is one reason the bass will come up when you properly place the speaker against a wall.

That is also why some speakers with an apparently "flat" frequency response may sound heavy in the mid-bass when you listen to them.

The anechoic frequency response is often improperly maligned by those who fail to take even these simple observations into account. The fact is that the anechoic response reveals a wealth of information when you know what to look for.

(Audio magazine, Nov. 1974; Richard C. Heyser)

Also see:

Speaker Tests Phase Response (Dec. 1974)

Breakthrough in Speaker Testing by Richard C. Heyser (Nov. 1973)

Speaker Tests--Impedance (Sept. 1974)

= = = =