In the October 1954 issue of Audio the writer described an experimental speaker system (Edgar M. Villchur, "Revolutionary loudspeaker and enclosure," Audio, Oct. 1954.) in which the bulk of elastic restoring force was supplied by the pneumatic spring of the enclosurs' air rather than by the cone's mechanical suspensions.

The speaker mechanism itself had a subsonic resonant frequency, but when mounted in its acoustically sealed, Fiberglas-filled enclosure the final resonant frequency of the system was raised to a predetermined value of about 45cps.

The present article is a report on a commercial unit built to the above design. It is considered that there is general interest in performance measurements of a device constructed on a new principle; in addition quantitative data on loudspeaker performance under carefully defined test conditions is relatively rare, and it is hoped that such a report may stimulate the publication of similar reports on other speakers.

The commercial unit is made as a two-way system, using a 12" acoustic suspension woofer in combination with a conventional cone-type high-frequency speaker, and is also made as a woofer system for use with other high-frequency units.

The system is illustrated in Fig. 1. It was decided to report on the model with the woofer alone for the following reasons:

• (1) The point of the article is to make a quantitative report on the capabilities of the acoustic suspension system, and
• (2) test procedures are simplified and made more reliable, thus better subject to accurate duplication by others. Variables such as radiation angle, microphone calibration at higher frequencies, and interference effects between the two speakers which may make microphone positioning critical, are largely eliminated. The performance of the high-frequency section will be only briefly summarized.

.

The two basic criteria of measurement techniques are validity and reliability. Validity refers to the degree to which the tests measure what they are supposed to measure, and are uninfluenced by other factors. Reliability refers to the accuracy of the measurements: it is an index of the extent to which the measurements can be duplicated at other times and places.

A valuable paper on loudspeaker frequency response measurements (Jensen Mfg. Co., "Loud Speaker Frequency-Response Measurements," Technical Monograph No. 1, 1944.) has pointed out the dangers involved in interpreting speaker frequency response curves when the exact test conditions are not known. Three such curves, made by three different acoustic laboratories using the identical speaker, were compared with each other and shown to differ by as much as 10 db at different portions of the frequency spectrum.

In the present case it was decided that unless test conditions could be established that would make it possible for results to be readily duplicated by anyone with the necessary facilities and skill, the measured data, while it might have a limited usefulness, would be unsuitable for publication. The frequency response and distortion measurements published here have been duplicated without significant variation. It is believed that any tests conducted under the same controlled conditions will achieve results which are within 1 db of the frequency response curve and will add or subtract less than one per cent to the values of the distortion curve.

The enclosure was placed in a hole in the middle of a 2-acre field, its face flush with the surface of the ground (see Fig. 2). The speaker was fed from an amplifier with a controllable source impedance, and a microphone was suspended at a distance of 5 feet above the enclosure, on axis with the cone. This means that the speaker was radiating into a controlled 180-deg. solid angle, into essentially free space, driven by an amplifier with a controllable damping factor.

The reliability gained by such test conditions involves a certain sacrifice in validity if what we are measuring is musical fidelity. One does not listen to a loudspeaker when one is suspended from a boom in the middle of a field, and in addition the effect of refraction from the cabinet edges, which would be present under any conditions .except a bookshelf installation, is not taken into account.

The alternatives, however, are worse. If we test the speaker in a normally live listening room we will get a different frequency response curve for each microphone position and for each room we use, and interpretation becomes difficult. Room resonances may accentuate or suppress various harmonic distortion products. Similarly, diffraction from the cabinet edges will create interference effects that will change the response curve with even small changes of microphone position.

Thus RETMA Standard SE-103, in describing methods of measuring speaker pressure-frequency response, states - that if there are no manufacturer specifications on mounting for a direct-radiator speaker:

• ". . . it shall be mounted on a rigid, non-absorbing, effectively infinite baffle. (This implies a radiation solid angle of 180 deg.)
• "Substantially free-space conditions shall exist in the acoustical environment of measurement.
• "The microphone shall be placed on the axis of the speaker as specified by the manufacturer and at a distance at least three times the maximum transverse dimension of the radiating area. (About 30" minimum for a 12" speaker.) "The values of Eg (signal voltage) and Rsg (driving source impedance) used in the measurement, and the value of Rsr (rated speaker impedance) shall be specified."

The conditions of measurement also comply with the recommendations of Standard C 16.4-1942 "American Recommended Practice for Loudspeaker Testing," published by the American Standards Association, for both pressure-frequency response and distortion measurements. The ASA specifies 5 feet on-axis for microphone position. Under the above conditions the measured data was found to be reliable, and results could be repeated at will.
.

.

1. Krohn-Hite audio oscillator, model 430-AB.
2. Altec 21-BR-150 capacitor microphone, with cathode follower output.
3. Freed a.c. vacuum-tube voltmeter model 1040.
4. (Low-frequency pressure response was re-checked in repeat tests using General Radio Sound-Level Meter Type 1551-A.)
5. Hewlett-Packard Distortion Analyzer 330B.
6. Dumont Oscilloscope Type 208.

.

In a narrow sense the validity of the tests is assured by the use of established and recognized techniques. The readings did measure r.m.s harmonic distortion, pressure-frequency response, and bass transient response.

In a more general sense, however, it is difficult to state that "flat" frequency response under free-space conditions represents musical fidelity under normal listening conditions. We know, for example, that the perception of relative bass content in reproduced program material varies with the volume level of the sound, a phenomenon called the Fletcher-Munson effect. Room acoustics, speaker placement, and other listening conditions can strongly influence the actual perception of sound in the final listening.

Validation must then be made by correlating objective data with subjective judgments, and by deductions that cannot, by their nature, be as rigorously tested as can the data itself. It is universally accepted, however, that the avoidance of dips and peaks in the response curve (ignoring slope) is a good thing. The writer then suggests that, considering the sensitivity of control that can be exerted over electronic as opposed to mechanical devices at the present state of the art, any equalization be assigned to amplifier circuitry. *3

This leaves us with a response curve for the 180 deg. solid angle which, if flat, will be transformed into a response that includes boost of the low bass when the speaker is mounted in a corner or at the junction of the floor and the wall. In the case of the unit tested the effective amplifier source impedance can be changed externally by switching from the 8-ohm to the 4-ohm connection, providing flexibility of bass response for different mounting positions. A test run was made with the speaker placed in the corner of a normal listening room, the driving source impedance being low - that is, using an amplifier with a relatively high damping factor.

The results of this test are indicative of general performance under such conditions but not as rigorous as the free field tests. Optimum use of the acoustic suspension system, as of any other speaker, is considered to be with an amplifier that has a variable damping factor, which is adjusted to optimum (not maximum) bass response under the conditions of operation. It is agreed among acoustics authorities that there is an optium source impedance from which to drive a given speaker mounted in a given way for the most uniform and extended bass, but this fact is only recently acquiring general recognition.

A brief listing of characteristics of the unit tested appears below:

that the main factors which determine the final resonant frequency are the mass of the moving system, the cubic volume of the cabinet, and the amount of Fiberglas filling, all of which are readily subject to accurate control. The elastic stiffness of the suspensions, a factor which is not so easily controlled with accuracy, contributes only about 10% of the total elastic stiffness, and the resonant frequencies of different production units can therefore be kept within small tolerances.
.

Distortion data is listed before frequency response data in order to validate the former. We are not interested in the total amount of sound put out by the speaker when stimulated at particular frequencies, including spurious harmonics, noise, etc., but in the output of reasonably undistorted sound at different input frequencies. A frequency response rating that extends down to 32 cps has little significance if the harmonic distortion at this frequency is 40% at moderate power, other than to indicate that the designer has attempted to extend the bass response further than he should have for that particular speaker. Attenuated output is by far preferable to distorted output.

Similarly, a low distortion reading at 32 cps has little positive significance if speaker response is down 20db at that point. The distortion graph of Fig. 3 should be read in conjunction with the frequency-response graph of Fig. 6, so that the distortion curve refers to the reproduced frequency range, and the frequency response curve refers to the range of low-distortion reproduction. The necessity for validation of the frequency-response curve by distortion data is clearly indicated in ASA Standard C 16.4-1942, which in listing essential practices for plotting pressure-frequency response characteristics states:
.

• "Unless otherwise stated, the values of pressure plotted shall be those corresponding to the fundamental frequencies."

.

The graphs of Fig. 3 plot the r.m.s distortion-frequency characteristic of the AR-1W at 10 and 20 watts input to the rated impedance (4 ohms). Amplifier gain was adjusted to 6.3 and 9 volts output across the speaker at 175cps, and left that way for the run. The radiation angle was 180 deg.

Figure 4 plots the distortion-frequency characteristic of the AR-1W in the corner of a normally live living room, and Fig. 5 shows oscillograms of the acoustic output of the speaker at 32, 60, and 100cps, with 10 watts into the rated impedance under the latter conditions.

A word of caution about the interpretation of the distortion figures must be inserted at this point. The reference power levels are electrical, and the inefficiency of the speaker (referring to the electrical power required for a given acoustical power) therefore favors the readings. The sound level is also quoted, but will be meaningless to many readers. The only correct way to compare the data with corresponding data from another speaker would be at the same sound pressure level at a given frequency, not at the same electrical power input. The range of difference in commercial speaker efficiencies is probably at least 25 to 1.

The efficiency of the AR-1W, on the other hand, is close in value to that of several other commercial units for which the writer has a high regard. It should also be noted that the efficiency of the AR-1W remains essentially constant down to very low frequencies (see Fig. 5), and the absolute efficiency in the 30-60-cps octave may be greater than that of another speaker with a much higher over-all efficiency rating. For example, if pressure response at 40 cps is down 9 db (which still allows for an excellent low-frequency reproducer) the efficiency at that frequency is reduced by a factor of 8, and it will require 8 times the amplifier power to create the same sound pressure level as at the reference level. The AR-1W used as an organ pedal tone generator could not be considered an inefficient speaker. When the 8-ohm connection is used for 180 deg. radiation conditions (mid-wall shelf mounting) the efficiency is halved. The conditions of the 8-ohm connection can be achieved without such loss of speaker efficiency by using an amplifier with a damping factor of 1.

The graph of Fig. 6 plots the frequency response of the AR-1W under the open-field conditions described above. The ratio of horizontal-to-vertical scale follows RETMA Standard SE-103, which states:

• "The length of a 10 to 1 frequency interval shall be the length of 30 db on the ordinate scale."

The graph of Fig. 7 plots the frequency response of the speaker in the corner of a normally live living room, fed by an amplifier with a damping factor of 4.

Good transient response is associated with a uniform steady-state frequency-response curve. The transient response of the AR-1W can be predicted from the frequency-pressure curve of Fig. 6.

The bass transient response was also checked visually with square waves and an oscilloscope. Figure 8 illustrates the response of various speakers to a square wave of subsonic fundamental frequency. In (A) the cone of an ideal speaker moves forward to the top of the square wave and remains completely motionless over the horizontal portion, while air pressure at the microphone decays smoothly. There is no hangover whatsoever. A poorly damped system is represented by the acoustic output in (B), which exhibits definite ringing after the initial stimulus. (C) is the acoustic output of the AR-1YV speaker system as recorded with microphone and oscilloscope, showing slight overshoot.

Efficiency has no direct relation to quality, but it does have an indirect one in that the power demanded from the amplifier by an inefficient speaker may exceed the amplifier rating. If the available voltage driving the amplifier is great enough the amplifier may then overload and distort. It is also true that in A-B tests the louder system tends to sound better automatically, and an efficient speaker has the edge in audio salesrooms if the electrical levels are not adjusted for equal volume from each speaker.

Interpreted in general terms the sound pressure levels indicated in Fig. 6 mean that a good 10-watt amplifier is adequate for the AR-1W or AR-1 speaker for moderate listening levels in typical living rooms. For larger rooms and for those who like very high levels of reproduced sound, at least 30 "clean" watts are required. The RETMA efficiency rating of the AR-1W at 100cps is 21.5db.

In constructing a figure of merit for a loudspeaker system designed for home reproduction the question would arise as to what place, if any, efficiency would receive. If manufacturers were canvassed (sie bewerben etwas) as to the significant factors in such a figure of merit a strange correlation could undoubtedly be made between the features of a particular manufacturer's speaker and the qualities emphasized as most important to the figure of merit. Manufacturers of low-efficiency speakers would tend to deny the relevancy of efficiency, while manufacturers of high-efficiency speakers would probably take an opposite stand.

At the risk of the writer's seeming to have an ax behind his back in need of grinding, it is submitted that efficiency should not appear in the main term of a figure of merit for loudspeakers, but in a second term connected by a plus sign. The second term would include other factors such as price and size.

Low efficiency simply means that for given performance results more money, weight, and space must be invested in amplifying equipment. The relative cost, in terms of these three factors, of the added electronic capacity can be calculated, but should not be reflected in the index of quality.

In the case of the Acoustic Research woofer, efficiency has been deliberately traded for extended and uniform bass response and low distortion. It is obvious that the magnetic circuit used in the AR woofer is sufficient for a motor of very high efficiency. The sacrifice of efficiency is justified, in the mind of the writer, by the performance data reported in this article.

The high-frequency speaker used in the model AR-1 system is an 8-inch direct radiator. Its performance characteristics, as used in the system with a 12 db/octave bass-droop network and pad, and as measured by Acoustic Research, are: frequency response 800 - 13,000 cps ± 5 db, and distortion over above range with 10 watts input, 1 per cent maximum.

hier sind die geparkten Bildunterschriften 


Fig. 1. Acoustic Research AR-1W low-frequency speaker system. The two-way system uses the same cabinet.

Fig. 2. Field measurement of the AR speaker, using General Radio Sound Level Meter (the microphone is detachable).

Fig. 3. Harmonic distortion-frequency characteristic of the AR-1W, under conditions noted.
Fig. 4. Harmonic distortion of the AR-1W in the corner of a room.

Fig. 5. Oscillograms of acoustic output of AR-1W (conditions listed in Fig. 4) at 10 watts to rated impedance, at 32, 60, and 100 cps, respectively, from left to right. The extreme uniformity of output is accidental, as can be seen from the frequency-response graph of Fig. 7.

Fig. 6. Frequency response of the AR-1W under conditions noted.

Fig. 7. Frequency response of low range of AR-1W in the corner of a normally live room.

Fig. 8. (A) Acoustic output of a speaker with unlimited and perfectly uniform frequency response, when stimulated by first half-cycle of square wave. (B) Acoustic output of poorly damped speaker system, same stimulus. (C) Oscilloscope photograph of acoustic output of AR-1W to first half-cycle of a square wave of 8 cps fundamental frequency. The rounding of the initial impulse is due to the lack of woofer high-frequency response.