Posted July 2011
by William R. Dudleston
I recall a very interesting conversation with Dr. Belle Julesz, Research Director of Psychoacoustics at Bell Laboratories for 35 years. This remarkable man came to understand more about the human hearing mechanism and the way the brain localizes sounds than anyone on this planet.
Somehow, we got into a lengthy discussion about how a barn owl rotates its head to triangulate the location of mice in the dark. We then related this to human hearing and localization of sources of sound. It seems that we, like the barn owl, tend to turn our heads a bit when we hear an interesting sound. Our brain wants to “view” things at a different angle to try to assign positional coordinates to the sound. Sometimes this rotation of our heads is obvious, most of the time we don’t even notice. Yet, these subtle rotations allow us to “lock-on” the apparent source of a sound.
The Physiology of Localization
Understanding the localization process is fundamental to making good recordings and to understanding how the stereo process works (or doesn’t work). A quick study of the human anatomy reveals that we have not one, but two ears separated by about six inches of gray matter. How convenient for the brain to be able to compare and analyze inputs from these two antennae at separate locations. Even more fascinating is the fact that the six-inch separation of these receptors corresponds to one-half of a wavelength around 1000 Hz, the very frequencies where your hearing is most phase sensitive! This means that sounds coming from your extreme left or right with 1kHz content will arrive approximately 180 degrees out of phase at each ear, thus demonstrating the maximum amount of phase difference.
Have you ever reached for your keys in a dark parking lot, only to hear a quarter hit the pavement? Without fail, that quarter will land on edge, circle around you a couple of times and come to rest just out of your reach under the car. But how can you discern this when you can’t see the quarter? The clues are found in amplitude, frequency, and phase differences.
As the quarter circles around you, it obviously appears louder to the closer ear. This allows you to perceive a lateral path. Phase or timing differences present evidence that the quarter is moving in an arc, not a straight line. A frequency-shifting phenomenon (Doppler Shift) lets you know if the quarter is moving toward or away from you. But how can you tell if the quarter is in front of you or behind you?
The answer is frequency masking. It turns out that those fatty ear lobes have a purpose besides flaunting jewelry. As the quarter rolls behind you, high frequencies are attenuated somewhat. When it rolls to the front of you, the sound is efficiently collected by the pinnae of your ears. The asymmetry of the pinnae even facilitates height differentiation.
The home stereo system was assigned the challenge of recreating a three-dimensional soundfield while utilizing only two speakers. Think about it. Height, width, and depth from two loudspeakers in the same plane. Sometimes I am amazed that the stereo effect works as well as it does.
In the days of John, Paul, George, and Ringo recording engineers didn’t quite know what to do with the 2nd channel. We had drums coming from only the left channel, guitar from the right. This recording style, while amusing, was actually less realistic than the earlier mono recordings. After about 10 years of this, recording engineers began emphasizing more classical recording techniques; coincident mikes mixed with an ambient pickup, and far less artificial reverb on voices. Finally, theoretical stereo had chance to show its stuff. Then along came quadraphonic stereo; which quite simply didn’t work.
You see, STEREO is intended to be a simple comb filter effect. When a recording is reproduced through a pair of loudspeakers, the left and right channels must interfere and crosstalk at each ear. This is the very concept on which stereo is based.
The key to good stereo effect is to preserve the interchannel differences all the way to the listener’s ears. Any foreign phase or amplitude differences that intervene will be detrimental to the true stereo image. This places an absolute premium on carefully matched loudspeakers and careful room positioning.
Why Most Loudspeakers aren’t up to the Task
Only a handful of loudspeaker manufacturers take the time to carefully hand-match loudspeaker pairs. Each pair of LEGACY speakers goes through a lengthy, cumbersome process known as nulling. We will place a microphone equi-distant from each speaker, then feed identical pink noise to both speakers. One of the loudspeakers is intentionally connected out of phase. The goal here is to create the strongest cancellation possible. We then adjust the individual crossover components; turns are wound off inductors and resistors are trimmed. We will often improve the null by as much as 3 dB!
I’ve seen many well meaning “experts” recommend the purchase of dual mono blocks to improve channel separation, when a conventional stereo amplifier may have 90 dB of separation already. Since the channel separation at the listener’s ears rarely exceeds 6 dB, the best way to guarantee good imaging is to simply use a matched pair of loudspeakers which are designed to function properly in a reverberant field listening room.
Loudspeakers and The Listening Room
A quick trip to your local high-end dealer can often leave one in bewilderment. If all these speakers are supposed to be accurate, why do they sound so different?
The answer lies primarily in the way the loudspeaker couples with the listening room. How it couples is a function of the power response and the physical properties of the listening room. The power response is related to the dispersion pattern of the loudspeaker and its amplitude response. Room properties include such factors as geometry, speaker, and listener placements, reflectivity of surfaces, and even ambient noise levels.
Today, much attention is given to the way a speaker measures in an anechoic (reflection free) environment. While an on-axis pulse, FFT analyzed, can tell a speaker designer a great deal about how his speaker will behave at a distance of 2 meters in an anechoic chamber, what about the real world? Clearly these anechoic measurements weigh only the DIRECT ARRIVAL path to the listener.
A real listening room with walls, floor, and ceiling will have an infinite number of paths from the speaker to the listener. Most of the reflections from these surfaces have long path lengths, are diffuse and exhibit random phase, amplitude, and directionality cues. These reflections are usually not detrimental, and in most cases add to the “air” or “ambiance” in the recording. Reflections of this type are termed LATE REFLECTIONS.
But what about the short path reflections such as the inevitable floor reflection? These EARLY REFLECTIONS also tend to be the strongest of the reflections. Their single bounce pattern leaves very little opportunity for absorption or randomization, particularly problematic is the frequency range from 250 Hz to 1500 Hz. Such frequencies possess wavelengths whose dimension is similar to that of the reflected path- length, thus causing strong response anomalies. These frequencies also fall into the range where the human ear is most sensitive to the phase anomalies. (Current research indicates that human phase acuity diminishes sharply above 2 kHz. It appears that the hair-like transmitters (cilia) within the ear begin to scramble phase information when forced to change direction thousands of times per second. Amplitude information ultimately reaches the brain at these frequencies, but with unintelligible phase relationships.)
Early reflections can effect tonal balance, clarity, and image localization. The brain has a tendency to fuse these EARLY REFLECTIONS with DIRECT ARRIVALS into one smeared arrival. Unfortunately these troubling reflections are only about 6 dB weaker than the direct sound arrivals. In fact, most listening rooms will have average level differences of early to late sound of only 6 to 8 dB. Even more surprising is data demonstrating that only 15% of the reflections typically reaching the listener are from the sidewalls. The remainder consists primarily of the floor and ceiling reflections.
How do we deal with the dreadful floor reflection? Most of you have already done something by carpeting the floor of the listening rooms. Unfortunately, carpet is of little benefit below 1500 Hz.
We have found, as did Roy Allison a long time ago, that by elevating the lower midrange driver the proper distance off the floor, one can reduce mid-bass anomalies significantly. Then by establishing a crossover point to the woofer in the mid bass range, the responsive dip is virtually eliminated! This is owing to the averaging effect caused by the unique path lengths from the woofer and mid-woofer to the listener. This technique is effective in the Signature III, Focus and Whisper speakers, where multiple mid-bass drivers share the load.
What about the sidewall reflections? We have found that the biggest reason sidewalls cause a lack of clarity is that they are typically reflecting a signal that is inaccurate. Simply put, most loudspeakers exhibit horizontal dispersion that is not uniform.
How Does a Loudspeaker’s Dispersion Pattern Influence what we Hear?
Consider the classic 2-way speaker system utilizing a 1” dome and 8” woofer with a 2800 Hz crossover. While this speaker could appear near ideal on axis, when measurements are taken 30 degrees off axis laterally, it will in fact exhibit an 8 dB suckout near 2800 Hz, and an abrupt roll-off above 13.5 kHz. This undesirable beaming effect applies to all drivers and becomes a limitation when the wavelength of the radiated sound is smaller than the width of the diaphragm itself.
Such traditional design neglects the audible effects on power response and thus results in strong colorations in the reverbant field. In contrast each model of the Legacy speaker line is designed for specific application. Whether a corner sub, an on-wall rear speaker, a downward directed center channel or a forward firing tower, Legacy has taken great care in addressing directivity pattern and power response.