Behind The Scenes (Apr. 1975)

I'VE BEEN in the audio field a long time and I've also been a member of the audio press for a long time. As you might expect, like most of my colleagues, I have developed a pretty thick skin and a highly developed attitude of skepticism. God knows how many wild claims have been bounced off my head. If I had a nickel for every time I was shown the "ultimate" loudspeaker/phono cartridge/amplifier, etc., I could retire! Of course, out of all the welter of high-flown claims and representations have come the hi-fi products of genuine merit which have been the bulwark of our industry.

And as our technological base has broadened over the past few years, some very bright, young engineers have come up with clever new products, formed companies, and become part of the mainstream of audio.

Well and good. Nevertheless, as many of the "old boys" will admit, we find it wise to keep that "grain of salt" handy, just in case some of the hip, new PR types try to give us the "con." You know, things go in cycles in the hi-fi business. For example, after years of relative quiet, electronic crossovers are becoming popular again, which means another round of bi- and tri-speaker amplification. The use of "raw" speakers, full range, woofers, mid-range, and tweeters, in "do-it-yourself" enclosures, is on the rise (as witness a new booklet on the subject by Altec). After years of concentration on receivers, there is a new trend to separate power amplifiers and pre-amplifiers.

And therein lies my tale... I had heard rumors about a certain preamplifier with highly unusual signal processing facilities, and such were the claims that I frankly said "Hoo boy! I've been this way before," and made sure I took a full dose of my Skeptic® pills. Well, the rumors persisted, and I began to get some input on this unit from people whose opinion I had to respect. No one is more curious than an old audio hound, so I finally arranged to get one of these new-fangled pre-amps to see what all the fussin' was about.

I've had this unit for some months now, and I have tried every deviltry I know to fool it. cheat it, lie to it, sneak up on it, confuse it ... everything but adjust it with a ball-peen hammer. Damn thing just keeps on working, doing what it's supposed to do. "Okay," you say, "a good preamplifier should do its signal-routing and amplifying of program material without adding significant noise or distortion, or adding coloration to the sound. In fact, the unit should be completely neutral and have no sound of its own. So what is so special about this pre-amp?" Well, this Phase Linear 4000 pre-amplifier has noise reduction and dynamic-range expansion circuits that are extremely sophisticated and, in fact, are partially a "spin-off" from the space program.

The combination of these circuits, plus several other processing aids, finally enables us to cope with some of the restrictions of dynamic range and signal-to-noise ratio of phonograph discs and, to a certain extent, several formats of magnetic tape recording.

To anticipate a question, the Phase Linear noise-reduction system is not in the same category as Dolby noise reduction. As is by now well-known, the Dolby is a two-pass or closed system which, through special multi band low-level encoding and decoding, prevents additional noise from being added to the tape during recording. In contrast, Phase Linear's noise reduction system is of the single-pass or open-ended type, in which no encoding of tapes, discs, or FM is necessary for its operation. In other words, this is an "after the fact" noise-reduction device and it affords about 10-12 dB hiss reduction from any program source in which hiss exists. Most importantly, this noise reduction is accomplished without any degradation of the signal, such as loss of high frequencies. This seeming impossibility will make sense shortly.

The full title of this pre-amp is the Phase Linear 4000 Autocorrelator/Pre-Amplifier. The autocorrelator is the noise-reduction circuit and, as noted previously, is a "spin-off" from the space program. When our Mariner spacecraft transmits photos back from Mars, for example, the initially received resolution is quite poor. Highly advanced autocorrelation techniques gradually remove large percentage of the detail-obscuring noise "glitches," and the results are the excellent photos you see in your newspapers.

How does the autocorrelator work? First, a few preliminary ideas. Noise energy (in our case, mainly hiss) is considered random in nature, non coherent or uncorrelated. Unfortunately for us, noise energy exists continuously and uniformly within the audio passband. In contrast, music (or any information carrying signal) has an energy spectrum which is neither continuous nor random.

Now here are some words from Bob Carver, President of Phase Linear and inventor of this autocorrelator noise-reduction system:

"Music energy appears in discrete energy 'bundles' throughout the audio band and is therefore not continuous. In addition, if some musical energy appears, for example, at a particular frequency, we know for certain that even and odd harmonics will exist simultaneously throughout the passband, and that energy will not exist between these harmonics. In other words, with music we are able to 'predict' where energy is likely to occur, if we know where the fundamental is, or even if we know where only one of the harmonics is. Also, and importantly, we know where the energy will not appear. In other words, music is coherent or correlated." Since pure hiss is totally uncorrelated, it is assigned a correlation co-efficient, which is zero. Highly correlated signals, such as a sine wave or a linear sum of sine waves, have a correlation co-efficient of one. As noted, music is a correlated signal, but the correlation value of music varies continuously from moment to moment.

The Autocorrelator may be considered a real-time Fourier analyzer with correlation co-efficient estimator. It is a very complex circuit, and the following is a simplified explanation of how it works: A series of electronic gates or "windows," each window controlling a certain frequency range and just "overlapping" its neighbor, is placed in the audio pass band. These windows can be either closed or open. If closed, energy, be it correlated (music) or uncorrelated (noise), cannot be transmitted at that point in the frequency spectrum assigned to the particular window. If the window is open, energy may be transmitted at that frequency. An operational description from Phase Linear says that each window is controlled by two sets of control circuits coupled together by diode logic "OR" gates. The primary circuit is located in the control band, between 200 Hz and 2 kHz. The secondary circuits are located throughout the harmonic band from 2 kHz to 20 kHz. The presence of energy in the control band, if above a level of-65 dB, will cause each window associated with the appropriate harmonic to open. If energy appears in the harmonic band at a particular frequency, each window associated with each upper harmonic of that frequency will be opened. All lower windows will remained closed. Notice that as the frequency increases, fewer and fewer windows are required to be opened.

Notice also that for the autocorrelator to recover a harmonic "buried" gates are opened, the incoming signal passed through the gates is almost always of such amplitude, frequency, and correlation that the well-known "masking" phenomenon occurs, and the noise is subjectively "covered."

Along with the autocorrelator circuit, but not actually part of it, is a signal-controlled, low-frequency dynamic filter which is designed to reduce rumble, hum, and other extraneous low-frequency "garbage." Low frequency noise reduction begins at 200 Hz, ultimately reaching -20 dB at 20 Hz. This low-frequency filter is activated when the correlator switch is put into its On position. There is also a screwdriver adjustment for calibrating the circuit to a particular phono cartridge. In spite of its relation to a phono cartridge, this circuit works with tape as well.

There is a correlator threshold-adjustment control on the lower right of the front panel. It is very simple to use. The correlator switch is placed in the In position, and the correlator control is rotated to the extreme beneath the noise, it is necessary that the harmonic have associated with it a lower frequency fundamental whose energy level is above the noise. Fortunately, this condition is common in musical energy.

=============

"I've resurrected ancient mono LP discs, run them through the 4000....You wouldn't believe how alive and open the sound was."

=============

In similar fashion to most noise reduction systems, the autocorrelator commences with a reduction of 3 dB at 2 kHz, and reaches 10 to 12 dB from 4 kHz to 20 kHz. As noted, the device has a circuit which estimates the degree of correlation of incoming signals and generates a control signal.

This correlation function signal is combined with another signal which indicates the harmonic content of any incoming music energy. This controls a threshold level designed to detect musical information within each of the bandpass frequencies. It must be appreciated that in the music output of a phono disc or tape, the noise (hiss or uncorrelated sound) is always present "outside" the gates, and is always ready to "rush in" along with the music when a gate is opened. By automatic adjustment of the various threshold levels at which the bandpass clockwise position. In this position there is no attenuation of hiss. Turning the control slowly counterclockwise, you will perceive a dramatic reduction of the hiss. If you continue to rotate the control toward its full counterclockwise position, there will be a sudden loss of high frequencies. It is not all that critical and works in the broad area of roughly 10 to 2 o'clock.

Once set, you can flip the correlator switch to the Out position, and 10-12 dB of hiss will come rushing into your ears.

The other processing circuit on the Phase Linear 4000 pre-amp is the "peak unlimit/downward expander" dynamic-range-enhancement system. Of equal sophistication and very nearly as complex a circuit as the autocorrelator, this system is designed to restore the dynamic range lost in the use of compressors, peak limiters, and similar devices in the disc-cutting process, as well as the manual "gain-riding" so often used to raise soft passages above the noise.

This is a four-band expander system, controlled by changes in high- and low-level signals. The high-level peak-unlimit band has a gain variation of +1.5 dB; there is 0 dB gain in what is termed the "don't care" band (the music is neither too loud nor too soft, so nothing happens); at -30 dB is the linear-expand band with a gain variation of ±1.5 dB, and the downward expand band has a gain variation of-3 dB. Since the linear-expand and the downward-expand bands are in essence "hurrying up" the rate of decay with decreasing amplitude of signal, there is ambience truncation. To offset this, ambience is obtained by internally generated L-R and R-L signals, which are specially processed by frequency weighting and phase shifting, and then injected into the output during the linear and downward expand operations. With carefully controlled attack and decay rates and with such small values of individual expansion in each of the bands, the overall operation of the system is undetectable. Yet the increments of expansion available from the four bands add up to 7.5 dB, and this combined with the 10 dB from the auto-correlator noise reduction give us 17.5 dB, which is a very considerable increase in the overall effective S/N ratio and dynamic range.

On the lower left of the front panel is a peak-unlimit threshold control and to its right, a red LED. In use, the control is rotated clockwise until the LED just flashes on the highest program peaks. This is important, for if the control is advanced too far and the LED is continuously on, distortion will be generated. The peak-unlimit lever switch activates the circuit.

Now on to the use of the special circuits. The autocorrelator is just plain amazing in what it can do for signal-to-noise ratio with any recorded medium. Most noise reduction circuits usually perform best on material that is fairly quiet, and this is true of the Phase Linear, but the degree of improvement is far greater, and in most cases noise becomes virtually inaudible. With noisier, older material, there is residual hiss, but the overall improvement is even more dramatic than with the quieter material. I've taken tapes that were 25 years old ... and loaded with hiss ... and put them through the autocorrelator and the drop in hiss and the gain in clarity were astonishing. I've resurrected ancient mono LP discs, that sounded like a steam calliope in heat, ran them through the autocorrelator and the peak unlimiter. You just wouldn't believe how alive and open the sound was, with a large proportion of the hiss gone, and the heavy hand of the compressor removed from the dynamics. Early pre-recorded tapes, primitive stereo discs, all benefitted from the combination of autocorrelator and peak unlimiter. On the other end of the scale, brand-new discs, cassettes, and cartridges were run through both circuits and here again the results were dramatic. Hiss ceased to be a problem and one is a bit shaken to hear how much compression is being used today. This is true with symphonic material too, and if you want to flip, play a pianissimo section with the peak unlimiter/downward expander in operation. You'll hear a lovely true pianissimo. Now take the expander out of the circuit and you will immediately hear the passage jump up 3 or 4 dB in level. The same holds true for triple forte sections in symphonic music, when you run them through the expander. Man, the music becomes really LOUD, just like 15 ips master tapes. In fact, potential users of the pre-amp would do well to check fusing on their speakers, particularly if they are using high output amplifiers.

Another thing the expander reveals is that commercial recorded tapes have been compressed, especially so in the case of 8-track cartridges. The autocorrelator is a nice augment to the Dolby NR system. You may be aware that although cassettes and open reel tapes may have been Dolby processed, it is possible they were not made from Dolby masters. Hence, there is a residual hiss on the tape, which is ever so easily removed by the autocorrelator. Speaking for a moment of the low-frequency dynamic filter, there is one company (who shall remain nameless) who is notorious for the amount of low-frequency garbage on their discs. The circuit cleans this up beautifully, without disturbing any of the low-frequency musical content.

Now, all this wonderful action of the autocorrelator and the peak un limiter circuits would be totally useless if we could hear the systems "working." By this I mean the unfortunate swishes, pumping, and breathing sounds that plagued many of the less sophisticated approaches to single-pass noise reduction systems in the past. Nor would the autocorrelator be very desirable if it in any way attenuated high frequencies. As related earlier, I tried screwing up the system every which way. I fed the circuits with high- and low-frequency, high- and low-intensity sine waves, fed it noise pulses, played heavy, sustained organ pedals, sustained piano chords, high levels of tinkly harpsichords, and wildly assorted high and low percussion. You name it, I tried it.

Nothing happened. Theoretically, the system is capable of being fooled by some sounds, among them wire brushes, which while "musical," could be treated as uncorrelated sound by the processor. In any case, I just didn't encounter any problems.

I've had some of the most prominent "golden ears" in the audio business over to the house for intensive listening sessions, with special attention being paid to the possibility of high frequency attenuation, and we could not detect any such aberration.

Many uses of these autocorrelator circuits come to mind, including some ideas in the professional area. For example, many record companies, especially in England, have extensive reissue programs, and here would be the ideal way to reduce the noise from older tapes and tape copies of old discs before the music signals are fed to the cutting amplifier.

In summation, let me say that once one has become accustomed to listening to music through the autocorrelator, and when appropriate the peak unlimiter (you wouldn't use it with master tapes!), it is difficult to do without their aid. It would be logical to assume that some time in the future, the autocorrelator and peak unlimiter/downward expander would become available in a separate add-on or "outboard" configuration. In the meanwhile, congratulations to Phase Linear for a technical tour de force and a job well done.

(Audio magazine, Apr. 1975; Bert Whyte)

= = = =