SIGNALS & NOISE (Letters to Editor) (Jul. 1990)

Horns vs. Direct Radiators

Dear Editor:

In the September 1989 Issue of Audio, we noted the article entitled, "THX Sound System: Certified Hi-Fi for the Movies." One particular portion of the article was intriguing to us for its contention that horns produce more harmonic distortion than direct-radiator woofers and that this single criterion should condemn them. As designers and manufacturers of both types of transducers, we feel obligated to address this issue. Although we consider frequency modulation distortion to be a more offensive contributor to total distortion, we recently conducted additional tests on harmonic distortion.

The cabinets tested were a dual 15 inch horn-loaded enclosure (HORN) with a designed lower cutoff of 45 Hz (a Klipsch MWM) and a dual 15-inch vented enclosure (VENT) with a designed lower cutoff of 40 Hz (a Klipsch KP-450). The identical 15-inch drivers were used in both cases. Our measurement equipment included B & K 4133 microphones, Hewlett-Packard 3325A function generators, and a Techron TEF System 12 Plus. Tables and graphs were generated containing harmonic distortion data with fundamentals at 40, 50, 60, 70, 80, and 200 Hz; FM data with f1 and f2 at 50 and 230 Hz, and frequency response of the tested HORN and VENT (reflecting the 6-dB difference in sensitivity). The data for 40 Hz reveals that the harmonic distortion levels are essentially equal for the HORN and VENT. In both cases, the woofers are in (or nearing) their nonlinear excursion range at the upper input voltages.

At 50 and 60 Hz, the harmonic components of the VENT are large when compared to the HORN. At these frequencies, we are well into the designed operating range of both systems, but the HORN is more efficient, requiring less voltage to obtain the same SPL. The HORN is not nearing its nonlinear region like the VENT. The data for 70 Hz represents a region of apparent transition. Neither woofer is driven into a nonlinear region, and harmonic components are again at about the same levels. The data for 80 and 200 Hz confirm this.

The data for FM distortion clearly demonstrates that the HORN produces far less of this far more objectionable type of distortion. These components are not harmonically related and are therefore more irritating to the listener.

The spectrum of a musical instrument is constructed of harmonics along with the fundamental and, in general, does not contain FM components. Some might point out that FM distortion goes to zero at 90° off axis, but this is true only in the theoretical case of a plane wave. In actual tests of real speakers, the reduction generally falls in the range from 4 to 10 dB, as does harmonic distortion.

Our tests indicate that the HORN produces somewhat less harmonic distortion than the VENT. We feel, however, that an evaluation of horns versus direct radiators should not be made solely on the basis of harmonic distortion data. In the case of FM distortion, the HORN is clearly superior. Regardless of what parameters are measured, the result of proper horn loading is demonstrably cleaner sound over any range. All types of distortion, as well as many other characteristics, should be considered when choosing between transducer types.

We acknowledge that speaker evaluation is a highly subjective practice, yet it is our experience that proper y applied horn design offers the greatest reduction in total distortion and compression while improving power output, transient response, and control of coverage angles. We submit that direct-radiator woofers are basically a compromise of distortion characteristics in order to obtain a better ratio of bandwidth to cubic volume. The complete test data is available from Klipsch upon request, for the price of a large envelope and 50-cent postage.

To quote the late Dick Heyser from the pages of Audio, "... in my personal opinion, accurate percussive bass is a specialty which a properly set-up horn seems to have to itself."

-Jim Hunter; Design Engineering Manager; Klipsch and Associates Hope, Ark. 71801

Author's Reply:

Regarding my article in the September 1989 issue, I regret that space considerations did not permit a more comprehensive explanation of the experiment I performed regarding the distortion performance of two low-frequency loudspeakers.

The experiment was performed in a 70,000-cubic-foot dubbing stage belonging to one of the most prominent post-production sound companies in Hollywood. The acoustical design called for a relatively short reverberation time (0.5 S), flat with frequency.

Two sets of three loudspeakers were installed side-by-side, one set being a conventional one, with horn-loaded woofer and tweeter sections, and the other being the THX system components, with direct-radiator woofers and horn tweeters. Both were flush-mount ed in a large, flat baffle, for optimum low-frequency loading, and were side by side. I spent three days after the installation adding 1/3-octave equalization to make both systems as smooth on a spatially and temporally averaged basis as possible. Then I conducted several experiments on the systems, one of which, the distortion series, was reported in Audio.

My first objective was to measure distortion over a much wider frequency range than is typical for loudspeaker measurements. Most loudspeaker manufacturers measure the amount of the second- and third-harmonic distortion versus frequency with the parameter of sound pressure level. Since second harmonic typifies asymmetrical distortion sources, and third harmonic symmetrical sources, users have often believed that the job of reporting was done when these were measured, expecting that higher harmonics would fall away in progression. I found this not to be true, due to a variety of distortion mechanisms which add various amounts of different harmonics.

A precursor for a good experiment is to make sure that what is being measured is not an artifact of the testing process; ensuring this was my second objective. In order to ignore power amplifier distortion, for example, it is necessary to measure the amplifier distortion and see that it is in fact several orders of magnitude lower than the measured distortion. The generator used was a Krohn-Hite Model 4200A driving a Crown DC-300A power amplifier through the installed ADM con sole and Dolby cinema processor.

Measurements at the output of this chain at every frequency and level used in the experiment showed the driving system's distortion to always be negligible as compared to the loudspeaker distortion (except where the loudspeakers were shown as having harmonics more than 70 dB down-for these, noise in the receiving room competed with the measurement of distortion, particularly at the higher harmonics). To prevent amplifier distortion in the sound pressure pickup, a dynamic microphone with no internal electronics (Electro-Voice Model 668) was connected directly to a Hewlett Packard 3580A spectrum analyzer.

Since the diaphragm motion of the microphone is so small at the levels em cloyed, we expect virtually zero distortion from it, and with low distortion in the driving and receiving electronics, I believe the distortion that was reported is accurate within the resolution of the analyzer, which is about ± 1 dB. I then sampled the three available loudspeaker systems (left, center, and right) to confirm that there were no large differences among them. This is a relatively small sample (these loud speakers are as big as a car, so it's hard to assemble a large number of them), but I believe the sample was a good one, because the post-production sound company had maintained a long-term relationship with the manufacturer and felt these to be prime examples of the model.

The next question to be raised concerns the model selected and the levels and frequency range over which it was driven. At that time, some years ago, this model represented about 80% of the installed base of theater loudspeakers in the U.S. The frequency range was chosen to be that over which the loudspeaker is routinely driven by the program material of the movies. It is true that the loudspeaker is being driven below the frequency of horn loading for this model in my lowest frequency test, but it is equally true that this is routine use of this loudspeaker, and the distortion that I measured is thus valid in practice. The sound pressure level which I chose to measure at is actually a very conservative one compared to the highest levels to be found in motion picture mixes.

Measured in the dubbing stage, Return of the Jedi produces a maximum of 105 dB SPL in the 63-Hz octave band, obtained with all channels operating and a fast, peak-analyzer time constant. (Film sound has one great advantage over other areas of sound reproduction in that standards exist for the playback SPL of mixes, including loudness calibration, so that the experience designed by the filmmaker in the dubbing stage is transferred to the listening situation.) On the other hand, the level of measured distortion was so high in the horn loudspeaker that I felt driving it to higher levels might be destructive.

A remaining question is: Why do I rate harmonic distortion as more important than intermodulation distortion? Since "everybody knows" IM to be more audible than harmonics, which are, after all, "musical," I wanted to sort out this audibility question. If we are interested in human observers and their perception, then I believe high order harmonic distortion to be more damaging than intermodulation distortion. The reason is frequency masking.

The presence of multiple tones gives rise to intermodulation distortion in the ear itself, so intermodulation distortion arising from loudspeakers is probably often masked by the nonlinearities of the hearing mechanism. New experimental work on masking curves for human observers, arising from single sine waves, was performed by Lewis Fielder in his study on subwoofers and published in the Journal of the Audio Engineering Society. His sets of masking curves form, for the first time, a method of weighting harmonics which makes sense to me. Using his data, I evaluated loudspeaker distortion for speakers in the THX program the following way: The harmonic spectrum arising from sine wave excitation was measured at octave-band centers and at 80, 100, and 110 dB SPL, corresponding to Fielder's masking curve data. The masking curve was then plotted on top of the data. The next step was to discard as insignificant any harmonics which lay below the masking curve.

For harmonics which were above masking, I added up the number of decibels by which each harmonic exceeded the masking level. This ascribed the correct weight to, for example, the very audible 15th harmonic while perhaps discarding altogether the perceptually unimportant second harmonic. In measuring a number of speakers over the last two years, I have found frequencies and levels where the "audible distortion products" range from 0 (audibly distortion less!) to 200 dB! Perhaps this test is a little too sensitive, since it gives values as great as 200 dB, which certainly seems exaggerated, but it is nevertheless well derived from the known psychoacoustics of the system.

The point of all this is really as simple as this: I believe the choice of the "best" qualified loudspeaker for a particular job depends not on the technology of the loudspeaker but on its performance. As I found, one example of a horn system may be worse than one direct radiator. Jim Hunter found the opposite. This is precisely my point: It's not the technology, it's the actual performance of individual models that matters.

The THX System is a little heretical to those who believe in a technology based solution, since it uses different technologies for the woofer system (direct radiator) and the tweeter system (compression driver). This combination of components offered the best available method to meet goals of frequency range, smoothness, uniformity of coverage, dynamic range capability, and low distortion. It was not an easy task to combine these two technologies, and it depended on recent developments on many fronts (e.g., low frequency drivers built with Thiele Small parameters in mind, constant directivity horns, Frater-Linkwitz-Riley crossover) to stitch together an optimum system. In fact, four parameters have to be matched at crossover to make a seamless transition: Complementary amplitude responses, matched time delay, matched phase throughout the crossover region, and matched directivity at crossover.

-Tomlinson Holman

(Source: Audio magazine, Jul. 1990)

= = = =