Feature Archives | Feedback

The late Nineteen Sixties was a wonderful time for audiophiles. The quality of vinyl records and Dolby noise reduction tape systems made the source no longer the weakest link in the chain. Amplifiers also were excellent. Tube amplifiers, using the "Williamson" circuit, and transistor amplifiers, with many dB of feedback sounded wonderful. Speakers became the weakest link, and soon there was a proliferation of speaker companies offering new design concepts. The leading manufacturers offered a variety of designs. JBL, Altec Lansing and Klipsh all used folded horn designs to provide a strong bass. Acoustic Research and KLH offered an "Acoustic Suspension" design, which sealed the speaker enclosure, reducing efficiency but allowing a flat bass without any resonance. Quad from the UK built a massive electrostatic speaker that covered the entire audio range.

Then came "Bose." The Bose Model 901 speaker system departed significantly from the rest. The other speakers used a traditional, forward facing, woofer, midrange, and tweeter with crossover networks. Bose used nine, 4 -1/2" speakers, with eight of them directed out from the rear of the speaker enclosure, and one in front. Emir Bose the inventor said that by reflecting the speakers off the wall behind them, (they came with stands and instructions that they be mounted two feet or so from the wall) the sound produced simulated that of the concert hall, where, he estimated, 90% of the sound heard was reflected. To correct for the deficiency in the bass of the smaller speakers, Bose included a frequency equalizer that boosted the amplifiers gain at each end of the spectrum. It was connected through the pre-amplifier's tape monitor input.

I had a pair of Acoustic Research, AR3A speakers and my friend Gerry had just bought a pair of Bose speakers. We brought them together and had a sound off. A number of us would listen to first one speaker and then the other using different music sources. Blood Sweat and Tears, Spinning Wheel, Beatles, "Day In the Life," Holtz, "The Planets," Cat Stevens, "Baby itýs a Wild World." These were our standard audiophile sources. For each source we would compare how the speakers sounded on the drums or the violins or the horse drawn cart. Both speakers sounded great, but different. Not different enough though. We noticed that the speaker whose volume was set the highest generally sounded better.

What caught my attention, was that nine little speakers could sound very good, when mounted in an acoustic suspension enclosure and the frequency response corrected electronically.

The Light Bulb goes on!

Wow, it hit me. What a great step in speaker design this is. Then I thought, the next step should be to integrate the speaker with the amplifier, an amplifier that had no transformers!

Amplifiers had been, and still are, large and heavy. As acoustic suspension demanded higher and higher power, the amplifiers began to outweigh the speakers. A stereo tube amplifier, delivering 100 watts RMS per channel, would require a 600-watt power transformer. That was because you have two hundred watts output and about a 30% efficient class A, audio amplifier. Tube amplifiers also required two audio output transformers, each capable of 100 watts of audio, covering the 20 - 20kHz spectrum. The output transformers were required to match the high impedance of the tube amplifier to the 4, 8, or 16-ohm impedance of the speakers being driven.

Transistor amplifiers only required one four hundred watt transformer as the outputs came directly from the transistors and the amplifiers operated class AB or B. The power transformer was required to provide isolation and the DC voltage (typically +/-40 volts) for the amplifier. This voltage matched the maximum output to the speakers.

The reason for the transformers actually was to provide the minimum DC voltage consistent with the output power of the amplifier when driving a low (4-8 ohms) impedance. Bose kept this tradition by connecting his nine eight ohm speakers in a series parallel (3 and 3) arrangement, maintaining eight ohms.

I thought, " What if eight of these speakers were connected in series. That would be a total impedance of sixty-four ohms?" If the amplifier could produce 100 volts RMS that would calculate out to about 160 Watts. Why 100 volts RMS? Because that is what you would get if you used a voltage doubler rectifier off a 115-volt AC line. Two diodes and two electrolytic capacitors and we replace all the transformers. Of course you required an audio amplifier that could swing three hundred volts with some pretty good dv/dt if you were to accomplish this. Quite a challenge, they weren't available on the market. The only high voltage power transistors made were used for auto ignition systems. Mosfets were not yet invented or at least not available. This is my next project I thought.

I started working on the amplifier design and also researching speakers. CTS was one of the manufacturers that Bose used, and so I ordered a couple of dozen of the same speakers. They had four and half-inch cones, ten ounce magnets, a long excursion, and a power rating of ten watts continuous, and twenty watts peak.

An old garage in the back of the house was converted to a wood workshop. I bought a radial saw and a drill press. Working with wood was not my strong point but soon I learned how to build a basic enclosure. First I tried porting the speakers with narrow channels in the cabinet. It was not very different from the Bose "Wave Radio" of today. The Q of the port was too high resulting in peaked bass. Next try was an acoustic suspension sealed enclosure, which was filled with dampening material. The results were excellent, as they should have been. Essentially, I had built a Bose 901 speaker with all the speakers facing forward. "Not very creative" I thought, "but hey I'm a power supply designer not a speaker designer.

Using a borrowed HP spectrum analyzer in combination with a calibrated Neumann microphone, I made many frequency response tests. The tests were made outdoors to eliminate room effects. My neighbors spent a summer listening to either pink noise or the shrill sweep up and down the audio frequencies.

The final speaker design consisted of eight 41/2"speakers, and one tweeter set flush in the front of a two-foot long by one foot high by, one foot deep, box. Measurements showed that the speakers were flat from 120 Hz to about 12 kHz. The low end dropped very evenly at twelve dB per octave below 120 Hz , and the high end fell off completely by about 15kHz. The tweeter was added to shore up the high end. It was placed in the center of the speaker face with four speakers (two over two)on each side. The tweeter crossed over at 10kHz and had an L pad to match its output to the rest. This fixed the high end, the amplifier would correct for the low end.

My day job was designing pulse width modulated switching power supplies (the speaker design was my night project). The amplifier design was similar in many ways to a half bridge PWM. The difference was that it had to be linear and have a high enough bandwidth to accurately produce output at 20 kHz. It took some doing but after a few starts the amplifier was finished. The frequency response had a 12 dB per octave rise starting down from 120 Hz, and flattening at about 40 Hz. The output sine wave could reach 300 volts peak to peak (100 volts RMS) driving a 64 ohm load, with a 1 volt RMS input. The input audio was connected through a small audio transformer. It was the only isolation from the AC line and the only transformer used. Mechanically the design was simple. A 19" x 10" standard aluminum rack panel served as the chassis and heat sink. A PC board containing the amplifier was mounted on standoffs to the panel. Power transistors and diodes were mounted to the panel/ heat sink. The panel was mounted on the rear of the speaker enclosure with standoffs providing the two-inch spacing required for the input electrolytic capacitors.

It was with bated breath that I hooked the amplifier to the speaker, and fed in a high quality audio source. It sounded great. I was ecstatic. Frequency response tests verified that the amp-speaker combination was flat from 50 Hz to 18 kHz. Wow. I did it.

The amplifier only weighed six pounds and had a bill of materials of about $35. The sound was better than I had expected. My AR3A reference speakers did not sound as good. Soon I had two prototype speaker/amplifiers with handles on them. I took them to various audiophiles listening rooms to compare them with other speakers. In all cases they sounded as good or better. Although my original plan was to develop a loud speaker that had a built in lightweight amplifier, I never thought that it would be of premium sound quality. It turned out that a few things I never considered made the unit superior.

One was the low crossover distortion. When the amplifier goes from a negative to a positive output, and back, you switch output transistors. This typically produced a dead band voltage of about 4 volts. To compensate for this dead band, designers generally used a DC bias generated by about three diodes. This circuit helps but due to the different temperature coefficient of the diodes and the transistors, total compensation is impossible. If both transistors were on due to the voltage being too high, you would suffer a runaway failure. The fact the my offset voltage was relative to a 150 volt swing and not a 40 volt swing meant that the crossover distortion became very low.

The second feature was that the resultant speaker had great imaging. Imaging is one of the most important criteria for a true audiophile. It is the ability of the system, to create a presence, by spreading the sound across a "Stereo Sound Stage". The sound of each instrument actually appears where it was recorded. The ear does this not by the loudness of the instrument but by the phase difference of the sound reaching the left and right ears. Sort of like an active GPS system. This phase difference is most pronounced in the mid range area (5,000Hz). Of course only recordings that are true stereo recordings can produce a high level of imaging.

The "Greenhorns" (the name I gave the speakers) covered the entire frequency range from 50 to 10 kHz using the same speaker. There would be no phase difference due to phase shifts caused by crossover networks or changing speaker parameters. The Greenhorns had superior imaging, no doubt about it.

After completion of the design, testing and comparing it with the best speakers I could, I was convinced that I had a winner. The low cost of the power amplifier, the excellent sound, and the imaging, should make this the desire of every audio buff and manufacturer. "I'm gonna be rich" I thought. All I have to do is sell this design to a rich audio manufacturer and I can retire on the royalties.

TO BE CONTINUED

Frank Greenhalgh
Sept. 14th, 1999