by VICTOR BROCINER
The existing U.S. standards of measurement are quite old, and do not apply to complete speaker systems. Several technical committees in the United States and Europe are at work revising the standards. The most recently issued standard, Publication 200 (1966) is marked "International Electrotechnical Commission Recommendation"; however, it also deals only with "single direct-radiator electro-dynamic loudspeakers of the moving-coil type." Some of the tests outlined are applicable to speaker systems. To quote, "For lack of internationally agreed knowledge and experience, certain important measurements have been deferred for future study . . . " The omissions include:
- Harmonic Distortion
- Intermodulation
- Transient Distortion
- Efficiency
- Distortion-limited power handling capacity
- Objective tests in more typical surroundings
- Subjective tests
It is not surprising, then, that speaker manufacturers tend to be rather vague in their specifications. This is more true of advertising than of sales literature. A random selection of speaker advertisements yielded the descriptions and claims that are listed below.
Manufacturer A:
New advanced form of acoustic coupling
Size and weight
True-pitch bass down to 12 cps
Extended hi fi frequency dispersion
Manufacturer B:
Woofer: size; high compliance, with soft rubber annulus
Low-resonance wide dispersion midrange tweeter with fiberglass pad to smooth frequency variations, and controlled perforations in the cone to eliminate low frequency peaks.
Variable high-frequency control
Impedance
Crossover frequency Response 35 to 20,000 Hz±3 dB
Extra bracing in cabinets
Dimensions
Manufacturer C:
Speaker sizes
Number of speakers
Mylar-domed tweeter
High-compliance woofer
Dimensions
Manufacturer D:
High efficiency system
Response 30-20,000 Hz
Woofer size
Horn-loaded tweeter
Dimensions
Manufacturer E:
New surround material on woofer permits large excursions with linearity
Smooth frequency response from below 30 Hz to beyond 20,000 Hz
Dimensions
Manufacturer F:
Heavy large (size specified) woofer in front-loaded exponential horn
High-frequency driver with cast aluminum horn works from lower mid-range (specified) through wide angle
Precision, two-section crossover network
Dimensions
Manufacturer G:
Controlled impedance to complement solid-state amplifiers
Reproduce original sound with no added coloration
Air suspension enclosures
Manufacturer H:
3-way bookshelf system
Woofer and mid-range sizes specified
Cone tweeters plus horn tweeter covers 25-25,000 Hz
Watts peak power input specified
Dimensions
Manufacturer F:
Very compact bookshelf speaker
Elliptical woofer
Matching tweeter
LC network Dimensions
Manufacturer J:
High-compliant high-linearity woofer (size specified) with optimum damping
Big power handling capacity, particularly at extremes of audio spectrum.
Very smooth response
3-position treble switch to match room acoustics
Manufacturer K:
Frequency response: ±4 dB from 50 to 18,500 Hz
Dimensions (small)
Manufacturer L:
Omnidirectional design eliminates hot spots, pinpointed directionality, gritty ear-shattering highs by diffusing the sound over the entire room. 360° of sound. The music extends beyond the room.
Manufacturer M:
Direct/reflecting speaker
Proper balance of directed and reflected sound, as ... in the concert hall
Multiple same-size, full-range speakers, internally coupled to eliminate audible resonances
Active equalization for smooth power output throughout the spectrum
Flat power output to the room
Aside from the cabinet dimensions, which are definite, the only other conclusion that can be drawn from advertising is that hi fi speaker systems, regardless of size, price, and design, all cover most if not all of the range of human hearing. Detailed sales literature contains more information, but this requires even more interpretation, in which it is hoped this article will help the reader.
As to performance and measurements, a proposal made to amplify the scope of the IEEE standard now being evolved is: "The characteristics to be considered are those that are customarily measured and that are deemed significant in terms of loudspeaker performance." The types of speakers included in the standard are classified according to transducing principles and radiating systems. In the first category, speakers used in hi fi stereo comprise dynamic and electrostatic units. Radiating systems are of two classes: direct radiators and horns.
Frequency Response--The frequency range covered by a speaker system is usually the first thing one thinks about. This is most generally expressed in words. Quite often, what is specified is " Usable Frequency Range" or "Response from X Hz to 20 kHz." This is so vague as to be meaningless.
What is "useful"? How much "response"? There is "X to Y Hz within 10 dB". Figure 1A shows the different things this can mean. "Plus or minus 5 dB" is definite, but it reveals nothing of the shape of the curve, whether there are peaks or valleys or shelves or slopes. See Figure 1B.
Fig. 1A-Frequency response within 10 dB, X to Y Hz.
Fig. 1B-Both speakers have a response of ±5 dB, but (A) will sound dull and boomy compared with (B).
Curves of output vs. frequency are usually on-axis response curves, taken in an anechoic chamber (reflection less room) . Is this complete information? The off-axis responses are also important. Not only do we usually listen off-axis in stereo, but also the off-axis response largely determines the nature of the sound reflected from the room boundaries, which comprises the major portion of the sound energy reaching our ears.
We can measure the total sound power output of the speaker. This is most conveniently done in a reverberation chamber, which gathers up the sound put out at all angles by the speaker, the way an optical integrating sphere measures the total light output of a light source. For a speak-quencies, this is a pretty good criterion of what we hear. Unfortunately, few, if any, speakers approach this ideal. With speakers that become directional as the frequency increases, the on-axis response can be made to increase with frequency so that the total sound-power output remains constant, but the listening quality would be unbearable. Some speakers are designed with a slight rise as a compromise measure. The sound quality seems to be determined by the on-axis response, the first reflections, and the reverberant (multiply-reflected) sound. In the bass range, the listening room and the location of the speaker in the room have a large effect on the response, raising the bass output above that obtained in an anechoic chamber. This is sometimes measured by using a multiple-microphone system in a "typical" living room.
These factors explain why manufacturers are reluctant to publish a "frequency response curve." As you can see, there is really no such thing as the frequency response of a speaker. This is particularly true of some of the recently developed types that utilize wall reflections. A series of curves that are smooth, and show wide-range response off-axis as well as on axis, indicate that the speaker is probably pretty good. A curve in a reverberant room is simpler to interpret. The power response should fall off somewhat at the high end unless the speaker is perfectly omnidirectional.
Distortion--If the instantaneous relationship of sound output to electrical input is not linear, the output waveform of a speaker differs from the input. Non-linearity has two effects.
When the input is a sine wave, harmonics are generated. These multiples of the original frequency can be measured individually with a wave analyzer and so specified, or their sum can be measured by means of a distortion meter as Total Harmonic Distortion (THD). Some authorities maintain that since the higher harmonics are more objectionable to the ear, they should be given more weight when the harmonic components are combined, but this is inconvenient and not usually done. Except at low frequencies, distortion in speakers is quite difficult to measure in a meaningful manner, because of irregularities in the frequency response and variations of spatial distribution with frequency. Low-frequency measurements do provide a reliable guide to the ability of a speaker to reproduce bass tones.
When the input consists of two or more sine waves, non-linearity produces not only the harmonics of the original waves, but their sum and difference frequencies as well, plus all kinds of combinations of sums and differences of harmonics and fundamentals. This is Intermodulation Distortion. Most of the intermodulation distortion products are not harmonically related to the original waves and consequently sound far more objectionable than harmonic-distortion products alone. They make music sound harsh, muddy, and poorly defined. When sustained, heavy bass tones occur simultaneously with treble tones, the latter are modulated by the low frequencies, producing a characteristic vibrato effect.
Another IM distortion measurement is that of the CCIF (Comite Consultatif International Telephonique) [Don't ask why "F" stands for "Telephonique"] which uses two frequencies quite close together. This has the advantage that one can traverse the whole frequency range and plot distortion vs. frequency. This type of distortion is demonstrated when a chorus starts and is accompanied by a low-frequency "grunt" by the speaker.
There are other schemes using more than two frequencies and also bands of noise, to simulate program material more closely. They are not in general use.
Two rather special kinds of distortion are sub-harmonic generation and Doppler Distortion. The first involves the creation of additional frequencies below that of the fundamental.
Doppler Distortion is related to the Doppler effect, commonly illustrated by the variation in pitch of a train whistle as the train comes toward us and then moves away from us. Suppose a loud bass note is producing very vigorous cone motion while a high note is also being reproduced.
The large back-and-forth motion of the cone at low frequency will cause a wavering pitch or frequency modulation of the high note. There are differences of opinion as to the importance of this effect. It is used as an argument against very small full-range speakers that have large excursions, and in favor of large horn-loaded speakers.
Efficiency--the ratio of the useful power output to the power input, expressed in per cent. It expresses the amount of sound (acoustic output) you get for a given electrical power input. However appealing the term, it has nothing to do with the sound quality of a speaker. A few speakers have extremely low efficiency and require amplifiers in the high-power rating category. Because of irregularities in frequency response that occur in even the best speakers, efficiency should be stated over a frequency range rather than at one frequency. This avoids specifying efficiency at a peak or dip in response.
Horn speakers are very efficient but they are most often used for the midrange and treble only, when they have to be "padded down"--used with reduced inputs--to avoid drowning out the woofer, so their efficiency is not used, whatever other virtues they may have. Direct-radiator speakers that reproduce the extreme bass range in fairly compact cabinets inherently have low efficiencies. The level of low-frequency response of a direct radiator with respect to its mid-frequency response is determined by the system Q, as explained in a previous section. The frequency at which response drops by a given amount is a function of the low-frequency resonance of the woofer in its cabinet. This, in turn, is set by the compliance and moving mass. There is a limit to the attainable compliance, so the only way to lower resonance further is to increase the mass of the moving system. This lowers the efficiency. Increasing the flux density to raise the efficiency decreases the Q, which makes the bass response roll off. Electrical compensation is feasible, and is used in at least one commercially available speaker.
Horn loading can increase efficiency greatly but results in increased bulk and high cost.
Efficiency is of greater theoretical than practical interest. At a given frequency it is not very easy to calculate or measure how much power is being fed to the speaker. A more useful rating is the amplifier power needed to produce a given Sound Pressure Level at some fixed distance from the speaker, or in an average living room. See "Sensitivity."
Impedance--Determines how much current flows when a given a. c. voltage is applied. Impedance can consist of any combination of resistance, capacitance, (of a "condenser") and inductance (of a coil). The voice coil has resistance and inductance, but if we measure the current flow for a given applied voltage we find that less current flows than these two elements would dictate. This implies that there must be another impedance in series with the circuit. It is called "motional impedance." The motion of the voice coil in the magnetic field generates an a. c. voltage that opposes the signal voltage, reducing the flow of current. Since this is a current-determining element, it can be represented by impedance. At resonance it can be quite large because of the large excursion of the voice coil, and can be represented by a resistance.
The combination of many resistances, inductances, and capacitances causes speaker impedance to vary widely over the frequency range.
With a single speaker, there is a broad region, usually around 400 Hz,. over which the impedance is low and nearly constant. Speakers are rated at this value. Multi-speaker systems which include additional capacitors, inductors, and resistors in the dividing network usually also have such a minimum but may have additional minima at higher frequencies.
Variations in impedance result in changes of current flow, so the power input to the speaker also varies over its range. The speaker designer must see to it that frequency response remains uniform by tailoring the efficiency to compensate for this effect.
Impedance is of importance because the maximum power available from a solid-state amplifier increases as the load impedance decreases. Below a certain impedance value, the amplifier output current becomes excessive and fuses may blow or damage can result. Even where special protective devices prevent this, the distortion goes up. The distortion increase can take place at low power levels as well.
So it is a good idea to keep speaker impedances within bounds. One speaker manufacturer-probably because he was originally an amplifier manufacturer-pays special attention to maintaining speaker impedance as uniform as possible over the frequency range. These systems are called Controlled Impedance speakers.
Phasing--Phase is the time relationship between two sine waves of the same frequency. In-phase waves are in synchronism. If one is delayed with respect to the other, the signals are out of phase. The amount by which they differ in phase can be expressed as a fraction of a wavelength or in degrees, where 360 deg. equals one wave-length. It can also be measured in radians; one radian equals 180 deg./Tr. One can specify the difference in fractions of a second, but in this case the figure is called delay instead of phase difference.
If out-of-phase speakers are close to each other, they interact unfavorably. One cone moves inward when the other moves out. At low frequencies the air moves back and forth between the speakers rather than out into the surroundings. At higher frequencies the speakers operate more independently because the sound from each tends to form a beam on the speaker axis. The result is that bass response is decreased. To facilitate correct phasing, speakers usually have one terminal coded. According to the EIA standard, one terminal shall be coded by means of a + mark or a color dot, preferably green.
In multi-speaker systems, there are frequency ranges where the outputs of a woofer and a tweeter (for example) overlap. If they are out of phase their outputs cancel and a dip occurs in the combined frequency response.
Observing the polarities marked on the woofer and tweeter does not guarantee correct phasing because the dividing network introduces additional electrical phase shifts, and relative time delays are caused by the different path lengths between the two speakers and the listener. Smoothness of the frequency-response curve is the criterion.
Phasing is also important between the two speakers of a stereo system. If identical signals are fed to two separated speaker systems, a listener located equidistant from the two speakers hears the sound coming from a point exactly centered between them.
If the speakers are out of phase, there is no definable apparent source of sound, and the correct spatial effects of stereo program material are lost.
Phase (Delay) Distortion--Program material consists almost entirely of waveforms that are not sine waves.
Any waveform can be analyzed into a series of sine waves of different frequencies: the fundamental and its harmonics, which are multiples of the fundamental frequencies. The harmonic structure of a wave plays a large part in determining the nature of its sound to the ear, which uses it to identify the different musical instruments.
Now, if a complex wave is formed by the addition of a fundamental and a number of harmonics, its reproduced waveform is altered if any of the components are displaced in phase, or delayed. Under some circumstances the ear can detect this change in waveform, which is caused by phase or delay distortion.
Polar (Directional) Response(angular distribution, dispersion). At low frequencies the sound waves produced by one side of a speaker diaphragm spread out uniformly in all directions. The sound becomes more and more concentrated into a beam with increasing frequency. Polar response is seldom specified for hi-fi speakers, but it is important. Verbal descriptions tend to be even less definite than those of frequency response.
"Wide-angle" and "60-degree dispersion" are examples. Since the sound level decreases fairly gradually as one moves off the axis of a speaker, the amounts of decrease should be specified for a series of angles and at several frequencies.
Complete information can be represented graphically. A series of frequency-response curves, on-axis and at various angles off-axis (Fig. 2A), are difficult to interpret, because the off axis responses tend to be rather irregular and may even overlap each other.
Greater clarity results from a series of response plots at different angles for a number of representative frequencies. Fig. 2B.
Power Rating--The concept currently in use, while rather inexact, is the power rating of the most powerful amplifier that can be used with a speaker without damaging it when reproducing average program material at top power. Ratings for electronic guitars and the like are different and must be specified.
A continuous signal of the same power produces far more heating of the voice coil, which gets no chance to cool off as it does in the softer passages and pauses of program material. At low frequencies, the large motions of the voice coil help the air to cool it.
Here the principal danger is from mechanical damage due to the coil striking something, tearing of suspension material, breakage due to fatigue of the flexible voice-coil leads, and the like. Consequently, the rating does not apply to continuous-wave power. Note that this kind of power rating is quite unlike that of an amplifier--it is not based on distortion.
The power rating is sometimes mistakenly interpreted as the power required to drive the speaker to a reasonable level (see Sensitivity). Presence-Some speakers seem to give sound a projected or "forward" quality, noticeable particularly on voice. It is caused by a region of somewhat elevated response around 3000 Hz, where the ear has its maximum sensitivity. It tends to bring a voice out so that it is not "inside the box," but it also makes music overly brilliant and "hard" in sound.
Sensitivity--This term is coming into greater use to indicate how much sound is produced by a speaker for a given power input.
The sound pressure level may be given at a point a stated distance on axis for a given power input. A typical figure for an efficient speaker is 90 dB SPL* at 4 feet on-axis, for 1 watt. The frequency should be specified. It is preferable to make the measurement over a band of frequencies such as 800-1200 Hz. Otherwise one may be tempted to specify sensitivity at a peak in the response curve, which could easily be 5 dB or so. Incidentally, for purposes of comparison, the peak intensity level of a 75-piece orchestra might be 116 dB, and the long-term average at 30 feet distance 96 dB. Normal speech 3 feet from a talker is 65-74 dB. The EIA (Electronic Industries Association) Pressure Efficiency Rating, used mostly in public address and similar work, has been applied to hi-fi speakers. It is the SPL obtained at 30 feet on-axis for 1 milliwatt input. The EIA figure can be converted to SPL at 4 feet for 1 watt by adding 47.5 dB to it.
Fig. 2A-Response curve of a speaker on axis and at a number of angles. Fig.
2B-Polar response taken at 500 Hz and 5000 Hz.
Fig. 3-Power requirements for various room sizes and conditions.
*Sound Pressure Level
The latest IEC (International Electrotechnical Commission) definition of characteristic sensitivity is the ratio of the average sound pressure, over a stated frequency range, and referred to a distance of 1 meter (from the speaker) to the square root of the nominal power (input). These rating systems provide means for comparing the sensitivity of loudspeakers but they do not permit the user to determine directly how much amplifier power he needs. A far less technically involved rating is used by one of our leading consumer organizations. This is the amount of power required by the speaker to provide somewhat louder than normal reproduction of a variety of recorded program material in an "average" living room of 3000 cubic feet, as judged by a group of listeners. A chart is provided that shows the relative power required to obtain the same sound level in rooms of different size as well as those that are more "live" or "dead" than the average. See Fig. 3. "Sensitivity" sounds desirable, and it is, other things being equal, but it is not related to the quality of a speaker.
Transient--Although this is not a speaker characteristic or measurement, it requires explanation before the following terms are discussed. A waveform that repeats itself forever is called a continuous or steady-state signal. A waveform that occurs but once (in a while) is termed a transient. Music and speech waveforms are constantly changing and hence are transient in nature.
Transient Distortion--The degree to which a speaker fails to reproduce a transient perfectly. A reproduced transient can differ from the original signal in so many different ways that there is no accepted number that expresses transient distortion; it is more of a qualitative term. It can be represented graphically, however. See below.
Transient Response--The response of a speaker when a transient is applied to it. This is of the greatest importance, since the function of a speaker is to reproduce transients. In general, a speaker with a flat, smooth, frequency response has excellent ability to reproduce transients. However, very small, sharp peaks and valleys in the response curve can cause disproportionate amounts of transient distortion, which is revealed by transient response testing. See "Tone-Burst Test" below.
It is widely believed that transient response correlates closely with listening quality. However, at least one series of experiments has failed to establish such a connection.
Tone-Burst Test--Transient testing on amplifiers uses a square-wave input.
With loudspeakers a better method is to use a "carrier frequency" modulated by a square wave. To put it more simply, the signal is an audio frequency signal that is started and stopped periodically, forming a "tone burst." The speaker output is picked up by a microphone and observed on an oscilloscope screen. The burst builds up gradually to full amplitude instead of instantaneously, and takes time to die out after the signal has been turned off. Good speakers have fast build-up time without overshoot and little hangover. With a really bad speaker, it may be hard to tell when the signal has stopped, the "decay transient" almost completely filling the interval between tone bursts. The frequency range of the speaker under test is traversed by varying the frequency of the sine wave in the tone burst. The bad spots are selected for photography of the oscilloscope trace. Equipment has been built to record the average value of the sound that persists after the burst, providing a curve of this type of transient distortion vs. frequency. The set-up is complicated and difficult to use, and the curve does not show the nature of the hangover. At the present time there is no accepted quantitative measure of tone-burst response.
Conclusions
Experts differ greatly in their opinions regarding the relationship between measurements and listening quality. Indeed, there is not even general agreement as to what listening quality should be. Measurements and specifications are chiefly of benefit to the speaker designer. They are of value in providing a means to ensure that speakers produced according to a given design all perform within predetermined limits with respect to the design standard. There is little question about the desirability of wide range frequency response, and especially smoothness of the response curve. Good angular distribution at all frequencies is also important. It is not entirely a matter of how a speaker sounds off-axis. The polar response determines the ratio of direct sound to that received after reflection from the boundaries of the listening space. The latter, which is called the reverberant sound, must be sufficiently great to provide an adequate feeling of ambiance--that of being immersed in the sound-as in a concert hall. Low distortion is a contributing factor, but perhaps not quite as much so as most people think. Transient response provides a reasonably good guide to overall quality. But there is no accepted quantitative relationship. It is safer to assume that bad test results connote poor sound quality than that good specifications necessarily indicate fine audible performance.
As to listening quality, it is generally assumed that the objective is the most faithful reproduction of the original sound that can be attained. There are obvious exceptions: users of guitar amplifiers, for example, want artificial sound effects. They have been known to ask for "good, clean distortion." But even for those who want the best possible reproduction, it must be realized that, at best, one can only achieve a good illusion of the original sound.
Without extremely elaborate systems that call for special recording techniques, the acoustics of the listening room simply prohibit the exact duplication of concert-hall sound. This is one of the reasons why microphone pickup and recording techniques are so varied and complicated. They are intended to create the desired illusion, and they succeed to a truly remarkable extent.
All this means that the final evaluation of speaker system performance is subjective. One must listen to a variety of program material, and decide on the basis of one's own preferences.
(adapted from Audio magazine, Jan. 1970)
Also see:
The Loudspeaker as a Spherical Sound Source (Mar. 1973)
Some Loudspeakers Past and Present (Apr. 1970)
= = = =