Test reports help, but your ears are the final judge.
CHOOSING SPEAKER systems (a pair for stereo, or two pairs for four-channel sound) can be a demanding job for the serious listener, but it also can be very rewarding. Up to the speaker(s), the music is a "signal"--an electrical replica of real sound. The speaker has the critical job of translating that signal into living sound. How critical this job can be, and how varied its outcome, depending on the speaker, is easily demonstrated by changing the speaker in a typical TV set for a hi-fi model. You'll hear things from that TV set you never heard before: a wider response and a fuller tone--as well as the raspiness and distortion of the TV's audio section.
A speaker system, then, is a sonic window on the entire playback system. The better it is, or the higher its "fi." the more it will reveal both the excellence and the flaws of that system and of the pro gram material played through it. A speaker that makes everything sound the same is, to that extent, not a high fidelity speaker. But the more revealing a speaker is of tonal variations inherent in recordings or even in different FM stations, the more faithfully or accurately that speaker is reproducing the signals fed into it.
How do you judge this quality and then relate it to your own system needs in terms of such considerations as size, personal listening preferences, amplifier power? There are, generally speaking, two approaches. One is to follow the published test re ports on speakers (such as those appearing regularly in HIGH FIDELITY magazine, and reprinted in its TEST REPORT annual): the other is simply to listen for yourself. Ideally, the listening should at tempt to relate what you hear with what you have read about a particular model. But whether or not you can make this relationship in all instances, there are some guidelines, for both understanding test reports and for your own auditioning, that can help you through the plethora of speakers on the market to a wise choice for your own sound system.
Test Reports and How to Read Them
To begin with, it must be emphasized that no set of "numbers," no matter how carefully derived, can tell you fully just how a particular speaker will sound. especially when installed in your room.
What lab tests can do, however, is suggest what you may expect to hear. They also provide a basis for comparing certain aspects of performance among various models since all tested units are subjected to the same treatment.
Impedance
The impedance of a speaker (its voice-coil resistance combined with various reactances) is the actual load presented to the driving amplifier.
Typical manufacturer ratings for impedance are 4, 8, or 16 ohms. These ratings, however, are not constant inasmuch as speaker impedance changes with the frequency of the applied signal. A typical impedance curve is shown in the accompanying graph (Fig. 1). Note that at speaker resonance (about 100 Hz) impedance rises, then it drops. The value in ohms at the "low point" (past the bass rise) is stated, as is the impedance measured over the rest of the audio range. Thus, this particular speaker would be said to have an impedance of about 6 ohms just past its bass rise, and to average about 8 ohms across the range-which is satisfactory and qualifies as an "8-ohm" speaker.
If the impedance is found to dip significantly low, this fact will be mentioned so that you can be assured of your amplifier's safety when used with that speaker. The impedance information also relates to the use of two speaker systems connected in parallel across the same amplifier output terminals.
For instance, while a dip in impedance to, say, 5 ohms is of little importance for a given speaker sys tem, that same dip on the part of two such speaker systems produces a net impedance of only 2.5 ohms, which could be dangerously low for some amplifiers or receivers. Usually the owner's manual (for the amp or receiver) states the minimum safe impedance to use.
Frequency Response
Simply stated, a loudspeaker's frequency response is the sound pressure it produces at some finite distance in front of its surface as a function of the frequency applied. Just what constitutes "good" or even "acceptable" frequency response from a loudspeaker never has been clearly defined. Nor can we discern much general agreement on specific numbers in this area.
Some years ago, Harry F. Olson, the noted acoustician of RCA Laboratories, published a fairly definitive set of criteria for high fidelity loudspeaker performance in which he proposed uniform response from 30 to 15,000 Hz, with a maximum departure from such uniform response of no more than 5 dB at 100 Hz and 8,000 Hz, and a maximum departure of no more than about 8 dB at the extremes of 30 Hz and 15,000 Hz. Interestingly, Dr. Olson allows no provision for peaks above an aver age "zero-dB" level, only for moderate amounts of attenuation. This concurs with the opinion of many experts who have long maintained that narrow peaks of departure from uniform response are much more annoying (and tend to add to so-called speaker coloration) than are slight dips of equal amplitude. While Olson's prescription is properly rigorous, we have found in the course of testing modern loudspeakers that a smooth rise above zero dB by no more than, say, 5 dB does not degrade the sound of a loudspeaker that has very good performance characteristics in important areas other than frequency response.
In judging a loudspeaker's frequency response in conjunction with a HIGH FIDELITY equipment re view, first examine the "on-axis response curve" of the speaker to be evaluated. This is the simplest frequency response curve, since it is concerned only with the ability of the speaker to reproduce all frequencies uniformly directly in front of the speaker radiating surface. If. for example. you note a significant "valley" in the curve in the midrange region (frequencies from about 300 Hz to about 2,000 Hz), listen particularly for definition of individual orchestral instruments and for "presence" in vocal selections. Lack of midrange response often tends to make vocalists sound subdued--almost as if they are singing from somewhere behind the accompanying orchestra. A premature rolling-off of low frequencies appearing in the curve will show up in your listening tests as a distinct lack of "full bodied" bass when such instruments as bass violins, kettledrums, and even the lower octaves of a grand piano are used as auditioning material. To some degree, male speaking voices also will be affected; they will especially show "coloration" when a peak of a few decibels occurs in the frequency response between 100 Hz and 300 Hz. This peak causes the so-called "bass-boom" or "one-note-bass" effect often heard from mass-produced con sole sets. The truth is that many of these sets have a built-in peaked bass response to give the untrained listener an impression of "real bass." If you're in doubt, just listen to any live male voice and com pare it with the sound of a reproduced male voice.
If the on-axis frequency response curve shows high-frequency response poorer than that suggested (in Fig. 2), you have another clue to speaker performance. To verify it, listen to musical selections containing strong high-frequency signals (cymbals, triangles, and the like) as well as to the consonants in speech such as s, f and even the letter t. If their crispness and definition seem wanting in an on-axis test (where highs are more easily projected from a loudspeaker than at any other listening position), you may be fairly certain that degradation will be even more extreme at any other listening position relative to the speaker's axis.
In addition to these obvious performance aspects, a few less apparent but equally important speaker characteristics should be considered. If the pressure waves created by the moving diaphragm of a loud speaker could be directly coupled to the listener's ear, our discussion of frequency response evaluation could end right here. (In fact, properly de signed headphones work exactly that way. When using headphones that form a good seal with your ears, you eliminate all problems of room acoustics and of directivity or "angular dispersion." All re produced frequencies are channeled directly to your hearing system, which probably explains why more and more stereophone enthusiasts maintain that no listening is as faithful to the original or as devoid of coloration as headphone listening. Since we are dealing with loudspeakers, however, we must face up to additional considerations..) Response and the Room Regarding the so-called ideal home listening area, in the last analysis how many readers will actually redecorate their living rooms or dens to conform to that ideal? Many listening rooms, in fact, do not even have the most desirable dimension proportions (3:4:5). Here is one area of speaker evaluation in which intelligent subjective listening can again supplement objectively determined measurements and reports. Try to do your auditioning in a dealer's showroom that is not too unlike your own listening area--both in dimension and in sound-absorption treatment.
Since one of the most often heard complaints from owners of newly acquired speaker systems is that "they didn't sound that way when I listened to them at the dealer's showroom," try to get the dealer to let you "final test" the speakers of your choice by taking them home for a few days.
Faced with the problem of room acoustics and your own listening position (which is most likely to be off-axis in any reasonable stereophonic arrangement), you need to concern yourself with another important characteristic brought to light in the current method of HIGH FIDELITY speaker re viewing-that of directivity. If you examine the other two frequency response curves given in a HIGH FIDELITY speaker report you will see that they are identified as "average front hemisphere response" and "average omnidirectional response." Generally, the three curves are almost identical up to frequencies of about 150 Hz or so because bass tones emitted by a loudspeaker tend to be nondirectional. Mid and high frequencies, however, be come increasingly directional so that as you move off-center of the speaker axis, these frequencies will sound weaker unless certain design precautions are taken. In stereophonic listening this is particularly significant, because the listener is more often than not well off-center of the axis of either speaker (hopefully somewhere in between them), subtending an angle of 45 degrees or more to either speaker.
---
--- Fig 1. Impedance curve.
Fig. 2. Shaded area represents typical response limits of high-quality loudspeakers, approximately plus or minus 6.5 dB from 30 Hz to 15,000 Hz. The smoother the curve (i.e., the fewer dips and peaks) within this area, the better. "Normalized" response refers to the zero-dB reference line's being drawn through the center of the total dB variation, although the actual response measurements are made for a constant 1-watt input, with approximately 80-dB sound-pressure level resulting as the most often encountered nominal average speaker output.
------- Fig. 3 (a and b). Conventional method of presenting loudspeaker
dispersion data (left) does so on an absolute basis for only one frequency
at a time (this graph shows a polar pattern measurement at 5,000 Hz).
In contrast, CBS Labs response curves, which are used in HIGH FIDELITY'S test reports, show complete frequency spectrum on a comparative basis: direct on-axis response, average front hemispheric response taken from five points, and average omnidirectional response taken from fourteen points The most significant curve from the listener's standpoint is the third (omnidirectional) curve, but by comparing the con tours of the three curves, you can form a good idea of the speaker's directivity in terms of actual dB differences all across the total audio band.
--
The two additional frequency response curves shown in the published test reports take cognizance of directionality by including separate measurements taken from various points around the speaker. One of these curves represents average radiated sound-pressure level ("equivalent omnidirectional response") while the other corresponds to average front hemispherical response. They pro vide a complete and easily interpreted picture of both frequency response and directionality characteristics. Previously, directionality of a loudspeaker was usually shown by plotting a polar radiation pattern (Fig. 3a). The problem with this type of graph was that a whole series of them would be needed (one for each frequency tested) to truly de fine a speaker's directional characteristics.
The new presentation (Fig. 3b), which includes the two additional frequency response curves, tells you at a glance just how well the speaker can radiate energy over the entire frequency range and into the entire listening area. As you listen to a speaker which has been reviewed by this new method, you will want to walk to the extreme corners of the room-to every conceivable off-axis location in which you or other listeners in your home may be listening. If the two off-axis-derived curves (particularly the one identified as "average omnidirectional response") exhibit marked attenuation characteristics above 1,000 Hz, pay special attention. Should you plan your home seating arrangement at a considerable off-axis angle from such a loudspeaker you are likely to sense a deficiency of high-frequency response. Try, therefore, to duplicate this condition in the showroom by moving away from "stage-center" and determining whether you can, in fact, notice the loss of highs which the published curves "predict." Again, this is best done by playing music that is rich in high frequencies. Many recently designed "omnidirectional" speaker systems tend to overcome this problem by using multiple drivers for their high-frequency array, often directing them away from the listener, so that the sound is dispersed by reflecting walls or other surfaces. This type of speaker reduces the undesired beaming effect of the highs and tends to average out the high-frequency energy radiated in a room. If you find that such a system pleases you and that the published curves confirm your high opinion, bear in mind that the smoothness of high frequency response depends in part upon where that speaker is located with respect to available reflecting surfaces. Should your home listening situation preclude such positioning, the high-frequency dispersion of these types of speaker systems may well be altered unfavorably.
Another point to remember when considering the high-frequency dispersion characteristics of any speaker system is that many systems are equipped with some sort of high-frequency control--either a continuously variable adjustment or a multiposition switch. Check the setting of such controls as you audition the speaker. As a rule, the curves shown in the published reports are derived with all controls set to their indicated "normal" position. However, the effect of the controls on the response also is described. Thus, you should audition a speaker with the controls set for neither boost nor cut, and then vary the controls while listening for the specific changes indicated in the report.
As a final check, after you are satisfied that the high frequencies are well-dispersed to preferred listening positions, return to the "on-axis" position (directly in front of the speaker). Try now to judge whether the highs sound strident or overly obtrusive, particularly if the on-axis curve of the report shows a substantial increase of amplitude in the frequencies above, say, 4,000 Hz as compared with the other two off-axis curves. This procedure is advisable since in some instances you or others in your listening group may find yourselves in an on-axis, position after all.
Distortion
Distortion in speaker performance has been, and still is to some extent, a confusing subject. The reports in HIGH FIDELITY (based on tests run at CBS Laboratories) include data on harmonic distortion, and a few speaker manufacturers also are beginning to state something about distortion in their descriptive literature. Their reluctance on this point stems from their apprehension that the reader will equate the distortion figures of speakers with those normally encountered in connection with tuners, amplifiers, etc. The latter figures are minute by comparison. For instance, 5 to 10 per cent harmonic distortion at normal operating level would be considered horrendous in an amplifier, but of far less significance in a loud speaker. Of course, by definition, the lower the distortion the better-in any audio device. Still, to understand what 10 per cent distortion at 40 Hz means for a loudspeaker, one must understand the nature of the distortion tests themselves. Suppose a speaker produces 10 per cent distortion when it is fed with enough electrical energy to furnish an average output sound-power level of 100 dB. The same loudspeaker will produce a sound-power level of 90 dB (putting you just a few rows back in the imaginary concert hall) with only 2 percent distortion. In general, the greater the sound output, or the lower the frequency (below the bass resonance of the system, at least), the higher the distortion. To put these figures in perspective, sound levels of 90, 100, or even 110 dB correspond to the loudest sounds normally encountered in a concert hall. For this reason the 100-dB (or in some cases, the 90-dB) level becomes the one we try to get from every loudspeaker--including those that can produce such loudness only with "break-up" or "buzzing" that is, with relatively high distortion. However, the tabulations used in the reports also show what hap pens when the speaker is driven to lower sonic out put. The prospective buyer then can plan accordingly. If, for example, your budget permits purchase of a system that cannot cleanly reproduce those concert hall levels in your living room, it's nice to know that the speaker of your choice will, nevertheless, give clean output at an 80-dB level, even at the difficult low frequencies. The distortion tables thus provide a much better guide in this respect than a single statement such as "5 per cent distortion at 50 Hz" which is absolutely meaningless (though still used in some advertising) when unaccompanied by some indication of sound out put level, or electrical input requirement plus an indication of the system's efficiency.
Using HIGH FIDELITY'S published distortion data, you can judge a speaker system's distortion fairly easily. First, listen to an assortment of re corded material at moderate levels-well below the point where significant distortion figures apply and familiarize yourself with the over-all tonality of the system in question. Then, in separate steps and listening to the same recorded passages, in crease the level of sound with the amplifier's volume control. At some point you will begin to hear a decided change--a decrease in the "cleanness" of the signal, particularly in the bass and lower mid range region. This point will, as a rule, correspond approximately to the dB level at which the tabular report first indicates higher orders of distortion. If that level seems overpowering-and one that you are not likely to require in your home listening environment (either because of personal preferences or considerations of neighborliness)--then the on set of distortion at such a level is of academic inter est only; it need not dissuade you from purchasing the system in question. In conducting these tests, however, you should allow for some margin inasmuch as the dynamic range of some future musical programming may be such that instantaneous crescendos may "push" the speaker into objectionable levels of distortion--if only for short durations. By basing the distortion tables in the reports strictly upon output sound levels (and not amplifier power applied to the loudspeaker) any consideration of speaker efficiency as such is deliberately ruled out for the purposes of this test. But efficiency is important for other reasons too, as discussed below. We mention it here, however, to remind you to be sure that an amplifier of ample, clean power output is used for listening tests--otherwise you may be listening to amplifier distortion or overload rather than the distortion caused by the speaker's limitations. In this regard, at least, most manufacturers of loudspeaker systems are careful to recommend required amplifier power for use with their products.
We suggest allowing a margin of perhaps as much as 50 per cent plus in this area-because the chances that a good high-powered amplifier will "burn up" the loudspeaker are less likely than is the chance that .the opposite will pertain, namely that a low-powered amplifier will be unable to drive an inefficient speaker system to levels at which it is still relatively free from significant amounts of distortion.
Judging Transient Response
Since the early days of speaker evaluation, nearly every reviewer has spoken of good or bad "transient response." This term refers to a loudspeaker's ability to respond with precision to sharply defined percussive sounds of short duration. Such sounds occur often in music and are produced of course by percussive instruments. Moreover, it has been suggested by some experts that all music (whatever the instruments producing it) is actually a series of "percussives" or "transients"-an interesting idea that further points up the importance of this characteristic. Anyway, the best program material for evaluating transient response is music that abounds in obvious percussive effects.
As additional help, HIGH FIDELITY'S published test reports approach the problem in a somewhat more concrete and scientific way. Carefully tailored "tone bursts" are fed into a loudspeaker, while its output is observed on an oscilloscope at various sound levels. The outputs are compared with the inputs for evidence of transient response defects such as "hangover," "ringing," or "muddiness" defects which were often suspected but never easily confirmed quantitatively by experienced listeners in the past.
While the published reports do not, as a rule, include photographs of the actual tone bursts, their relative importance and possible effect on the sound is noted by the reviewer in terms of pinpointing deficiencies in transient response at specific regions in the frequency range. Thus, high-frequency transient response may be adequate, while low-end transient capability leaves something to be desired.
If careful listening tests confirm this, a comment will be included in the published report. When it does, that's your cue to concentrate-during your own listening tests-on such things as the attack of bass drums and timpani, the bowing of bass viols, while simultaneously listening for evidence of "muddiness" or lack of definition. If deficiencies are noted in the high-end transient response, listen for "fuzziness" in the reproduction of sibilant sounds, upper brass tones, and even the upper octaves in a piano solo.
------ Fig. 4. Graph shows amount of acoustic power needed to produce
100-dB sound level in listening rooms of various sizes. Note that for other
dB sound levels, the acoustic power requirements change.
Specifically, for each 10 dB lower sound level de sired, divide the acoustic power shown here by 10.
For example: in a room that is 3,000 cubic feet in volume (length times width times height), the above graph shows that a sound level of 100 dB will be produced by 0.5 acoustic watt. To produce in this room a sound level of 90 dB, only 0.05 acoustic watt would be needed. For 80 dB, the acoustic wattage drops to 0.005, and so on. To estimate amplifier power needed, divide acoustic watts by speaker efficiency: e.g., 0.5 acoustic watt from speaker that is 1 percent efficient requires amplifier capable of supplying 50 watts to speaker.
Efficiency and Amplifiers
The speaker reports in HIGH FIDELITY invariably recommend the amplifier power needed to properly drive the speaker sys tem being evaluated-but this serves as a starting point rather than an absolute rule for your power requirement deliberations. To translate the published (and necessarily generalized) power requirement to the specifics of your own room size and listening preference, a knowledge of the relative efficiency of the system is still needed. Actually, an indication of speaker efficiency is implied in the frequency response curves already discussed. Those curves are derived with a standard electrical input to the speaker of 1 watt, and so the output levels on the dB scale of the graph provide a true indication of efficiency. For instance, an output level of 70 dB (caused by 1 watt of electrical input) represents a fairly inefficient speaker system. A speaker system that produces a sound level of 80 dB from the same 1 watt of electrical input is ten times more efficient than the one that produces only 70 dB of sound level. The acoustical (not electrical) power required to produce a 100-dB level in typical rooms of various sizes is plotted in Fig. 4. The cubic volume is simply the length by width by height of the listening room. If 90 dB of sound level is enough for your needs, you would reduce the readings by a factor of 10. If, on the other hand, you feel that you must have sounds at 110-dB levels, multiply the reading obtained for your room size by a factor of 10. A somewhat simpler approximation is to figure 1 acoustic watt per 1,000 square feet of floor space for a 100-dB sound level capability. Of course, to trans late the acoustic power back into electrical power you would need to know the efficiency of the loud speaker you contemplate purchasing. Here the frequency response curves of the report are used bearing in mind that 10 dB represents a 10 to 1 ratio of power. Thus, if 1 electrical watt produces a sound level of only 80 dB in a room of approximately 1,000 square feet of floor (a large dealer's showroom is likely to have such measurements), it means that the acoustic output is 20 dB lower than our desired "100-dB level," or the efficiency of the system is 1/10 x 1/10 or 1/100th-or 1 per cent (a figure not uncommon with modern, air-suspension bookshelf-type speakers).
Whether you go through all these calculations or not, the final test will still be a listening test involving the use of an amplifier, whose power rating you think is adequate, coupled to the speaker you are evaluating. If the speaker is capable of high sonic levels, as reported in the tests, but there is audible distortion as you approach such levels, you can be reasonably certain that the amplifier is overloading before desired levels of sound are attained-and you will then want to verify this "mismatch" by running the same speaker from a higher-powered amplifier. Most dealer showrooms are equipped to easily switch the speaker under test to any of several amplifiers as a driving source. These few additional A / B tests are well worth your extra time and effort, for there is nothing so frustrating as having selected an excellent speaker system only to find that the amplifier being used to drive it just doesn't have enough power output to do the job.
Power-Handling Capacity
The transient-response tests using tone bursts serve another useful purpose: they enable the reviewer to suggest maximum power limitations for a speaker system. The older method-using continuous tones of ever in creasing power-of determining how much power a speaker system can safely handle is invalid, since musical programming is unlike "continuous tones" and more like the tone bursts used in the transient-response tests.
For judging this performance characteristic we recommend that you be guided solely by the test re ports. In other words, do not keep increasing the sound level in your own listening tests merely to find out how much a speaker system can take be fore its voice coil is destroyed or its cone structure is deformed. Such destructive testing may be worth while in an engineering laboratory (where one or more prototypes are actually driven to destruction in the course of a design program). but it would hardly be welcomed by an audio dealer.
Dynamic Range
An important characteristic of loudspeakers that is not directly. documented in the reports--but which often is inferred by the reviewer from related data-is that of dynamic range, the difference between the softest and the loudest sounds that you will hear when your system is finally put together in your home. The upper (or loudest) end of this dynamic range may well be determined by the highest sound level your selected speaker system can produce with reasonably low distortion, as tabulated in the test report. On the other hand, the lowest extreme (or the softest sound) you will be able to hear without masking caused by residual hum and amplifier noise will be governed more by the performance of your amplifier--not to mention the characteristics of a given phono pickup or tape deck. To gain some idea of a speaker's dynamic range, play a record at the loudest listening level you will want-and then lift the tone arm from the record. With everything (and everyone) else perfectly silent in the listening room, are you then able to hear hum and noise from the system? Is it at a bothersome level? If it is, then your lower limit of dynamic range capability is being diminished by amplifier limitations. Either the gain of the amplifier may be too high in terms of the efficiency of the speaker selected, or the amplifier's signal-to-noise ratio may be inadequate in terms of the speaker's own capabilities with respect to dynamic range. For this reason it is important that be fore purchasing a speaker system you should listen to it actually being driven from your own make and model of amplifier (or receiver). Today's program sources (FM, low-noise tapes, and discs) often contain useful dynamic ranges of 60 and even 70 dB. A residual noise level caused by amplifier problems may well reduce that available dynamic range by as much as 10 dB.
Your Own Listening Tests
While lab test results can help in choosing a speaker system, the lab measurements ideally should be related to listening tests. Conscientious equipment reviewers try to do so, but in the last analysis the "best ears" are your own. What matters most is how the speaker sounds to you, just as how a new garment fits you (regardless of stated size, or color, or style, etc.) really determines how well you will look in it.
Of course, it helps to know what to listen for. To begin with, it must be understood that a loudspeaker is essentially a translator, or-as engineers put it-a "transducer." It changes one form of energy (electrical) into another form (acoustical). In so doing, it invariably adds and/or deletes some thing from the original. Because of the laws of physics there is no such 'king as a "perfect speaker." There probably never will be. But the improvements and "refinements" in speaker design and manufacture over the years do result in systems that perform more accurately as translators or transducers than before.
One way of stating this from the standpoint of the listener would be to say that the better a speaker is, the more it gives you the feeling that you are not listening to a speaker at all, but rather to the program it is reproducing. Audio people refer to this feeling as "listening through the system" rather than to the system. They also refer to this quality as one of "transparency," the opposite being "coloration." Although many persons do not realize it, their own ears are well equipped to discern these qualities. It is surprising, perhaps, that so many who firmly believe that "seeing is believing"-do not ascribe the same infallibility to their hearing.
And yet the ear is one of the most sensitive and reliable "instruments" ever encountered: it can distinguish between subtle differences; it can be selective and hear one sound over another, or mask one sound under another; it encompasses a range of frequency response and dynamics scarcely met by the costliest of man-made test instruments. And it gets better with use: despite the well-known falling-off of sensitivity to extreme high frequencies as one gets older, the over-all response ability of the ear, the total hearing experience-particularly to music and "music-type information"-grows progressively more sophisticated. Listeners, like conductors, improve with age. By virtue of your own ear, you can verify your first impression and, more to the point, you can zero in on the final choice that often has to be made between two or three speakers which have impressed you equally favorably.
Although "transparency" and "coloration" can hardly be defined in clear-cut textbook fashion, they can be described in terms of the specific effects they encompass. Judging a speaker is rather like viewing a painting: from a few feet away you form a general impression; closer, you see how the individual brush strokes contribute to the total.
Sound Coloration: Speaker's Fault or--? Coloration of the sound, caused by severe irregularities of a speaker's frequency response (peaks of, say, 6 to 10 dB or greater anywhere along the spectrum) will show up as a tendency for the speaker to emphasize certain portions of the musical range. The highs may be too prominent with respect to every thing else, or the mid-bass may force itself on your attention while all tones below and above it seem to recede into the speaker box. Coloration also shows up as a speaker's favoring of one group of instruments over the others: the speaker may be prodigious, for instance, at projecting the characteristic guttiness of the string bass while at the same time masking the upper registers, or it may waft bold waves of brass tones into the room while lending the woodwinds and strings a too distant quality. It may favor the male voice (with a characteristic and false heaviness) while slighting female vocalists.
And so on.
Of course, any of these effects might be inherent in a badly recorded disc or tape, and many of them-particularly extreme discrepancies in bass and treble balance-could result from poor room acoustics. For this reason, it is important to listen to the speaker playing material you know well, prefer ably your own fairly new recordings. As for the acoustic setting, you should try to evaluate a demonstration room in terms of how closely it resembles your own listening room. A room that is abnormally small is no place to assess speaker quality; it will tend to suppress or muffle the bass and probably bounce the highs around excessively. A room that departs in shape from the normal rectangle is also bad: if it approaches a cube in shape, it will cause most speakers to take on a hollow sound; if very long and corridor-like, it will introduce false resonances to the mid-bass. Any room that is heavily carpeted and draped will tend to tone down the highs; conversely, any room with prominent hard surfaces, such as glass show windows or shelves loaded with the metal chassis of other equipment, will tend to accentuate the highs.
Unless the demonstration room comes fairly close to simulating the acoustics of your own listening room, you should insist on listening to the same speaker in two or three different locations in that room before making up your mind. Or, ideally, you should try to get the dealer to agree to sell the speaker (or speakers for stereo) on approval-a week's time with them at home should be long enough to let you decide whether you want to keep them or return them for another pair.
What Do You Listen To?
General impressions aside, another tack for the prospective buyer is to try to analyze the speaker's response in terms of its bass, midrange, and high-frequency output. Such analysis is best done with signal generators in addition to musical material. The signal generator can pinpoint specific areas of troubled response, reveal severe peaks and dips, indicate the low and upper reaches of response, reveal the speaker's "doubling frequency" (a bass frequency at which the speaker no longer can respond to the fundamental tone and produces harmonics of it instead), and provide some information as to the speaker's dispersion characteristics (its tendency to beam tones as the frequency is raised). However, a signal generator is not as a rule a normal household item; and even if you owned one few, if any, dealers would permit its use during normal store hours. People come to hear music, not beep tones, goes the argument.
Well, yes. But there is music--and music. To serve as test source material, the music chosen should have some texture and weight. Music lightly scored (pops in particular) is apt to make a "joyful noise" on a much wider quality range of equipment than is symphonic music. It's no great shakes to re produce (or to record, for that matter) simple-textured material which makes relatively little demand of the equipment and thus is of little value in assessing a speaker. A solo guitar, for instance, of ten is used to impress listeners with the transient response of a system, which is all right as far as it goes, but the real test would be a guitar playing against accompaniment: if the twangs are utterly clean along with everything else, then you can conclude something about the system's response. The real test is how well the speaker handles percussives when it also has to reproduce other, different kinds of sounds at the same time. Works scored for large ensembles and encompassing big jumps in dynamic as well as tonal range-operas and most sym phonic works--actually comprise the best speaker test material.
Also useful are passages shared by a deep-voiced instrument, such as an organ or bass viol, with a high-toned voice such as a flute. Note too that instrumental groupings that have their own rich over tones structure-massed strings and woodwinds, for instance--tend to absorb distortion components, while "purer" instruments-such as the solo piano, flute, or horn-will be more revealing of distortion.
The solo piano, in fact, is one of the most difficult instruments to record and reproduce well. The hu man voice is another: one of the touchstones of tonal purity is the male voice that sounds masculine but not "chesty" or with a hint of false mid-bass emphasis; another is the female voice that sounds lilting, "feminine" but not screechy. Thus, good recent opera recordings make excellent speaker test material: a poor system simply will not do justice to a soprano; a good system makes her a joy to hear, actually exhilarating. In general, when you find yourself listening not to the system but to the mu sic, and you sense the goose pimples during certain passages, you can feel secure that you're hearing the real thing.
Popular music is not too valid for speaker test material since many pop records are made with beefed-up midrange tones to make a soloist sound prominent. Similarly, many rock recordings have artificially boosted tones, especially in the mid-bass, that can give a false impression of a speaker's true ability.
Musical test material may not reveal to you the specific low-frequency response limits of a speaker, but you can tell by listening carefully to the music's bass portion whether or not the bass is generally satisfactory. Good bass does not necessarily mean the deepest bass, although ideally the deeper and cleaner the bass, the more natural the sound (assuming the midrange and highs are suitably balanced and full). There are, in any case, limits to how far down a speaker can go-limits governed by the design and cost of the unit. A speaker that can hit rock bottom, tonally speaking, generally will hit sky high in terms of cost. But rock bottom may not be within the scope of your taste, needs, or budget, especially when there are so many less than ultimate speakers that come fairly close and cost a lot less.
Good Bass
Within its design limits (and any responsible engineer will admit that speakers do in deed have design limits), however, you should expect a speaker to sound clean. "Clean" bass is free of boominess, thumps, or a blurry indistinct effect.
The bass notes are clear enough to enable you to distinguish between a string bass and a tuba. The bass drum, when struck hard, ideally should have a deep, somewhat "dull" and tight quality. A bass, ensemble, playing way down and tutti, should pro vide a sensation of energy that you can almost feel as well as hear. The bass viol should growl rather than sound mellow. Timpani should sound some thing like real thunder, and you should be able to distinguish the tonal variation between differently tuned timpani. The low end of the piano keyboard should have a vibrant, almost rugged quality, and with good definition among closely spaced notes what engineers call "being able to read the name on the keyboard." Finally, you want to be aware of possible intermodulation effects in which the powerful bass interacts with higher-pitched tones to cause the latter to blur or waver. A passage in which an organ or string bass plays together with high-pitched woodwinds is useful for checking this effect.
Good Middles and Highs
Listening higher along the response range you still should be alert not only to peaks and resonances (they can occur anywhere along the frequency band to color the sound) but to the unique aspects of midrange and high-frequency sound. The treble portion of a speaker's sound should have an open, full, well-aired quality.
Poor dispersion, or a tendency to beam the highs, means that the highs will be most prominent when heard from somewhere directly in front of the speaker (response on axis will be noticeably brighter than off axis). In addition, the mid-tones may sound honky or "box-y" and/or the extreme highs will take on a "squashed down" tone, as if an invisible hand were compressing them. Outright distortion will be readily discernible as a kazoolike quality or nasality. Lesser amounts of distortion will show up as a "hard" or unnatural sheen over the sound.
In addition to the beaming effect, poor dispersion signifies phase distortion, a kind of tampering with the original time sequence of the elements of the signal that also can be responsible for lending the sound an over-bright or hard feeling. When auditioning the midrange and highs, listen also for good transient response-a speaker's ability to respond crisply to percussives such as drum beats and rapidly plucked strings, especially when played against an instrumental background. Each note should sound distinct, with no blurring from the note before or to the note ahead. Another tip-off to transient response (of a pickup and amplifier as well as of a speaker) is the sound of record surface noise. An occasional surface defect should sound like a quick tick rather than like a prominent or ex tended rasp.
If an FM tuner is handy, tune to a point between two stations where you hear the characteristic rushing noise-white noise, as it's called. Listen for a few moments to the way in which the speaker reproduces this hash. If the sound is fairly subdued, rather like that of a shower running in the next room, the speaker probably has very little audible distortion in the treble and is dispersing the sound well. The brighter or harder this noise sounds, the more high-frequency coloration the speaker is adding to the signal and the more directional its projection of the highs. Many speakers characteristically sound bright on white noise when heard on axis, but more subdued off axis; this pattern is not ideal, but it is acceptable inasmuch as most home listening will be from a position decidedly off axis of the speaker, especially in a stereo setup where the main listening area is somewhere between the two "looking-out" axes of the speakers.
As for instrumental sounds in this frequency area, ideally you should try to recall what instruments (and voices) sound like when you last heard them live. Unfortunately, auditory recall is a very sometime thing, and for practical purposes such "A/B" testing in retrospect is next to futile. Realistically, the best you can do on this count is to listen to the tonal blend of an ensemble even while trying to discern its individual sections and participants: the good speaker will present both sonic experiences equally well. It will have, as they say, its own "internal separation" without at the same time being "clinically analytic." In the long run, listening for a good speaker is fairly akin to listening to music itself: you bring to both experiences a kind of ever-developing sophistication and you get in return a deeply personal revelation. In both, you are listening for the singing strings, the plaintive woodwinds, the brazen brass, the definitive percussion. Eventually, you will recognize an authoritative speaker, even as you applaud an authoritative performance.
Some recommended recordings which we feel provide exceptionally good "speaker test material" follow.
-----------
Full orchestra and chorus; for highs, bass, transients, distortion, middle honk
MANLER Symphony No. 8, Concertgebouw of Amsterdam, Haitink. Philips 6700 049 (two discs).
MOZART Mass in C minor, K. 427.
Berlin Radio Symphony, Fricsay.
Deutsche Grammophon 138124.
These are tremendous recordings for sampling orchestra and chorus. Fortissimo tuttis put heavy strain on the ability of the speaker to maintain differentiation in loud passages. Listen to the fortissimo near the beginning of the Mahler and the soprano solo and choral passage near the end of the opening Kyrie in the Mozart. Do the massed sopranos stay as thrilling voices, or do they become shrieky or scratchy, or hard-the result of peaky highs? Do drums and cymbals have their natural sharp attack (transients)? Do low drums have satisfying weight, and the organ in the Mahler the proper power, indicating good bass response? In the tuttis do you hear male voices, female voices, and the main orchestral instruments all individually? No known sound reproduction system will do this perfectly, but a poor system does it very badly indeed. Try to make sure your standards are not too low in this respect by hearing at least a sampling of your test records on an excellent sys tem.
Also recommended:
BIZET Carmen
Soloists, chorus and orchestra, Bernstein.
Deutsche Grammophon 2709 043.
ANY WAGNER OPERA (Karajan on DG; B6hm on Philips; Solti on London).
STRAVINSKY Rite of Spring L.A. Philharmonic, Mehta.
London CS 6664.
LONDON SYMPHONY, Bernstein.
Columbia MS 31520.
LONDON PHILHARMONIC, Haitink.
Philips 6500 482.
ELGAR Enigma Variations London Philharmonic, Haitink.
Philips 6500 481.
---------- Transients and highs: percussion sounds
VARESE Ameriques.
Utah Symphony, Abravanel.
Vanguard S 274.
WHAT THE WORLD NEEDS Now(The Burt Bacharach /Hal David Songbook).
Boston Pops, Fiedler.
Polydor PD 5019.
XENAKIS Metastasis for Orchestra.
French National Radio Orchestra, Le Roux.
Vanguard C 10030.
Each of these discs contains passages with an abundance of percussion instruments at work. Try the fortissimo about 3/4 inch from the beginning of the Varese or the Stars and Stripes Forever on the Boston Pops record. Wood blocks, bells, cymbals, triangles, and what not all make demands on transient response and on clear, crisp sound at the top of the highs. Cymbals are particularly revealing:
they should "smash," of course, but stay metallic and ringing, not papery or scratchy. It would be most helpful to hear one or more of these records on a very superior system first, because percussion may sound impressively sharp even when it falls short of its true quality.
The Boston Pops record is valuable not just for percussion, but also for general orchestral texture.
Deutsche Grammophon has done a terrific job in Boston's Symphony Hall; the highs are exceptionally clean, wide and smooth.
----
Specials for bass
BACH
Organ music. E. Power Biggs on the Thomaskircher Organ. Columbia KM 30648.
Also recommended:
BACH, J.S.
"Die 6 Orgelkonzerte." Karl Richter, organist Archiv 2533-170
NEW MUSIC FOR ORGAN: Music by Albright and Bolcom.
William Albright, organist.
Nonesuch 71260.
STRAUSS Also Sprach Zarathustra.
Los Angeles Philharmonic, Mehta.
London CS 6609.
The organ is of course the natural instrument for a thorough low-bass test, and there are many good organ recordings. The Bach has the special virtue of using a baroque organ. The pedal notes have the wonderful, slightly snarly quality that gives us a ready criterion for low distortion. The dynamics in the opening D minor Toccata and Fugue are wide indeed-great as test material-and a lot of the bass has tremendous power. In those held chords in the Toccata, when the biggest pipes come in, do you hear great power at the very bottom? Listen most carefully when the pedal notes go up and down the scale, as in the Fugue; does the speaker reproduce them all or does it drop some out of hearing? Listen too for the effect known as doubling, in which the deepest notes seem to lose their fundamental and seem to be dominated by the overtone an octave higher. The truer and more distinct the fundamental tone, the better the speaker. In judging bass, don't listen only to organ recordings, however. You will be listening to massed orchestral instruments drums, cellos, etc.-most of the time, and the lowest organ notes are particularly difficult to record well.
The Strauss is a far-out test that will separate superb bass reproduction from the merely good. The soft opening is played by four instruments: a pedal note at low C (about 33 Hz) on the organ; the C an octave higher (about 65 Hz) on the contrabassoon; the C another octave higher on a kettle drum; and the last two Cs played together on the bass fiddles.
London has managed to get the organ note onto this record with superlative power and clarity. If your speaker is in the big bass class and your room is favorable acoustically, the organ will envelop you, assail you physically with a profound power. If your speaker is not quite in that class, you may know the organ is there but will hear mostly the buzz in the contrabassoon note and the tremolo in the double basses.
Remember, in making this test you are dealing with but a single deep-bass frequency. It does not tell you all you need to know about deep-bass response, and you might evaluate the speaker differently if the musical pitch were altered by as little as a whole tone. (If the speed of your turntable can be "tuned," you can easily check out this possibility.)
---
Some ultimate’s
MOZART Violin Concerto, No. 4, in D, K. 218. Heifetz. RCA Red Seal LSC 2652.
BARTOK Quartet No. 2.
Juilliard Quartet. Columbia D3S 717 (three discs).
LEONTYNE PRICE "Prima Donna," Vol. 3.
RCA LSC 3163.
Individually desirable characteristics should work together to produce a unified impression which is in itself a test. Take Price's marvelous singing. On a moderately good speaker it can be thrilling enough. What a speaker of the utmost refinement in middles and highs does by comparison is to remove some last "support"-coloration, if you will--so that Leontyne stands there alone, free in space, utterly true. All middle honk is gone, the mid-highs are smooth as glass, there is no rough ness to make the sound edgy or hollow. The orchestra sounds absolutely true too.
Similarly, Heifetz' violin comes through wonder fully on fairly good speakers: It is recorded well up front, with plenty of pizzazz. Get good extended, supersmooth highs, and freedom from all honk, and the violin moves a little closer and into sharper focus, but ingratiatingly, sweetly, totally without hardness just what a fiddle is like a few feet away.
And the quartet music can be similarly indicative of the speaker's refinement at the top. It should leave behind all hollowness, all over-sharpness, and stand ultra-clear but sweet.
-----------
(High Fidelity, 1976)
Also see:
EDITORIAL -- Introducing the wonderful world of speakers.