Extraction vs Generation Page 2

After a few hours of listening in ambience-generation mode, I found the Yamaha to be a mixed bag—its coloration is successfully masked when playing large orchestral works, but on much chamber and folk music, that distinctive Yamaha sound could not be escaped, no matter how it was adjusted. In fact, I cannot agree with BS that it is worth an audiophile's time to fine-tune hundreds of potential Yamaha permutations until it closely matches the ambience on an original recording. This strikes me as audiophile masochism, especially when there is an alternative available: ambience extraction.

By the way, even in ambience-generation mode, the Yamaha already provides a degree of ambience extraction by virtue of the initial delay. In other words, artificially generated ambience always gets a free ride on the coattails of extracted natural ambience. How well the two ambiences mix is a function of how well you can adjust the Yamaha.

The DSP-1 as an Ambience Extractor
Experience with the Benchmark and the Phoenix boxes has taught me that if a single initial delay enables ambience extraction, a few properly timed additional delays can increase that ability (footnote 5). So I was excited by the potential of the Yamaha for ambience recovery, provided that a program setting could be found to accomplish it. Not to keep you in suspense, I am pleased to announce that the Yamaha DSP-1 turns out to be a formidable ambience decoder. Its well-engineered, wide-bandwidth digital delay provides purer sound, and freedom from aliasing and low-frequency IM distortion products. The latter problems were occasionally audible in the Benchmark and the Phoenix (footnote 6).

Recommended DSP-1 Set-up for Ambience Extraction
If you follow this recommended set-up, the only active adjustment during use will be the ambience volume! First, for best ambience extraction, use the six-speaker mode (four speakers also work, but you'll never know what you're missing). Next, use a program that has a L–R (L minus R) matrix. This helps to keep center-channel information out of the effects speakers, increasing your ability to raise their volume without hearing an echo effect. The L–R signal also contains a large proportion of ambience. DSP-1 programs that use L–R matrices are Dolby Surround, Surround 1, and Surround 2. I would have liked to try the "Presence" program, which supplies four independent delays to the effects speakers, but that program lacks a L–R matrix. Don't bother with the Dolby Surround Program, which has a 7kHz cutoff in the surrounds as well as a modified Dolby-B chip. In fact, I don't even recommend the Dolby program for use with the movies; you'll discover wonderful natural ambience in your film music if you use the Surround 1 program instead.

Our goal is to regulate the Yamaha's Surround program for minimum artificial ambience effect. To that end, set the hall type to "On Stage" (which primarily contains early reflections), liveness to about 0.4 (anything lower and the rear speakers produce a phasey effect), and room size to minimum. I discovered that the artificial stereophonic effect of Yamaha's left and right effects speakers is superior to the Phoenix's or Benchmark's monophonic output. At the minimum settings, the DSP's pseudo-stereo outputs produce a spatial spread without significant artificial coloration. (The Shure HTS uses similar processing to create its stereo surround outputs.) Because of the Yamaha's pseudo-stereophonic processing, it is no longer useful to connect the rear speakers out of polarity with each other.

The high-pass filter should be set to Thru (flat). Set the low-pass filter to somewhere between 12 and 16kHz, depending upon the treble response, placement, and orientation of your effects speakers (I have great success bouncing my rear-side speakers off the ceiling). Adjust the initial delay to about 30ms. However, if the effects speakers are much farther from you than the main speakers, subtract 1ms for each foot of difference. Be approximate; this is not a critical adjustment (eg, if main speakers are 8 feet away and effects speakers are 12 feet away, use 26ms delay).

Interconnect Method
I agree with Bill Sommerwerck about the recommended interconnect method of the DSP: You may have sonic reservations about passing your preamp's signal through something called "Yamaha" on its way to your Krell amplifier. One way to get around that is to use a preamp with two main outputs, feeding one to the Yamaha and the other to your power amp, thus losing the remote control's ability to control the main speaker level (a small compromise; plus, you won't need the optional four-channel remote unit). Lacking two main outs, you can build a high-quality Y-adapter box, using the most esoteric cables, of course. You should then be able to sleep at night knowing you haven't compromised your sound.

The Listening Experience
With the above settings, the Yamaha DSP-1 decodes the ambience in the original recording and spreads it around the listening room in a very natural manner. You will truly be able to discern the quality of the performing hall. When playing good recordings, I found the effect to be much more pleasing than any of the ambience-generation modes of the unit. In addition, it is possible to switch from chamber music to pop, orchestral, folk, or whatever, without having to reset numerous Yamaha parameters, a distinct advantage of ambience extraction mode. Just adjust volume to taste; one way to determine a proper ambience level is that it is unnoticeable . . . until you turn the unit off or mute its output. This ambience becomes so seductive that muting the DSP will cause withdrawal symptoms. Then you will know you have joined the growing ranks of audiophiles addicted to ambience extraction (footnote 7).

Footnote 5: Recent psychoacoustic research has revealed that the Haas effect can be extended to more than 60–90ms with carefully placed delays at approximately 25ms intervals.

Footnote 6: I found that fast high-frequency transients (castanets, applause) could cause the Benchmark to produce low-frequency IM products. The Phoenix's problem is less evident, and can be audibly eliminated with a graphic equalizer.

Footnote 7: Recommended further reading: "Extractions of Ambiance Information from Ordinary Recordings," by E. Roerback Madsen, JAES October 1970. Explains the Haas effect, and makes a scientific case for ambience extraction using the methods outlined in this article.