Stereophile Test CD 3 Tutorial Tracks: Data Compression & Binaural Demonstrations
 Frederic Chopin: Waltz in c-sharp, Op.64 No.2 (ADD) 4:20
(from Concert, Stereophile STPH005-2)
 Frederic Chopin: Waltz in c-sharp, Op.64 No.2 (ADD) 4:20
Truncated to 15 bits (Indexes 1, 3, 5, 10); encoded via PASC (Indexes 4, 6, 8); encoded via DTS Zeta (Indexes 2, 7, 9)
Musician: Robert Silverman
Producer/Engineer/Editor: John Atkinson
Executive Producer: Larry Archibald
Concert Organizer: Ralph Johnson
Road Manager: Danny Sandoval
Piano: Steinway D piano supplied by Riedling Music Company, Albuquerque, NM
Piano Technician: Charles Rempel
Emergency Piano-Stool Repair: Gorm Damborg
Recording Venue: The First United Methodist Church, Albuquerque, NM
Recording Date: November 6, 1992
Microphones: two Brüel & Kjaer 4006 *" omnidirectionals with black (diffuse-field) grids
Microphone Preamplifier: tubed EAR 824
Cables: AudioQuest Lapis balanced fitted with AudioQuest XLRs to mike preamplifier; Beyerdynamic balanced microphone cable to recorder
Recorder: Revox PR99 *", two-track, open-reel recorder at 15ips
Tape: Maxell XL-1
Transfer to Digital: Ampex ATR-100 *", two-track, open-reel recorder restored by Mike Spitz (ATR Service Company) and modified by Robert Harley, with the invaluable help of Steve Hogan (Jensen Transformers), Jerry Ziss (Analog Devices), Steve McCormack (McCormack Audio), and John Curl Digital Transfer: Manley/UltraAnalog 20-bit A/D Converter; edited with 20-bit resolution on the Sonic Solutions hard-disk editing system; mastered and redithered to 16-bit resolution using the Meridian 618 Mastering Converter
Canadian pianist Robert Silverman, who is Director of the School of Music at the University of British Columbia, has made over a dozen recordings—for Stereophile, CBC Records/Les disques SRC, Musica Viva, Orion, and Marquis. His recording of Michael Baker's Piano Concerto (a work written for and dedicated to Silverman) won a 1991 JUNO Award, while his recorded Liszt recital won a Grand Prix du Disque from the Liszt Society of Hungary, resulting in an invitation to give the annual all-Liszt recital in Budapest to honor that composer's birthday. Other than Intermezzo (1991) and Concert (1994) for Stereophile, his most recent recording was of piano music by César Franck (Musica Viva MVCD 1061, 1993).
Robert Silverman's next recording for Stereophile will be the Liszt B-minor Sonata, due for release in the spring of 1996.
Concert, which features works by Schuman, Schubert, J.S. Bach, and Chopin, was recorded, as the name might suggest, live in concert. Robert Silverman played the second of Chopin's Op.64 Waltzes, published in 1847—two years before the composer's death—as an encore. I used the same pair of spaced Brüel & Kjaer omnidirectional mikes as featured in the soundstage-map track to feed a ReVox open-reel tape recorder.
The microphone's-eye view of Robert Silverman
What you should hear: I positioned the microphones so as to place the image of Robert Silverman at the left loudspeaker. This means that the image of the piano should extend from just inside the left speaker all the way to the right loudspeaker position. Although this does mean that the piano is positioned a little right of center, this is what you would have heard in real life had you been sitting on the hall's center line.
After the analog tapes had been converted to 20-bit digital, the editing and master assembly were done preserving the 20-bit resolution—not so much because we were sure that the final four of the Manley's 20 bits were valid, and the analog tape hiss was well above the 16-bit LSB level, but because this meant that any artifacts resulting from the digital manipulation would remain below the CD's 16-bit noise floor.
As the CD only carries 16-bit data, the 20-bit master had to be reduced somehow. Simply dumping the output of the Sonic Solutions hard-disk editor to DAT, thereby truncating each digital word from 20 to 16 bits, audibly reduced the air around the instrument and the solidity of the image. By using the Meridian 618 Mastering Converter, however, we could reduce the 20-bit data to 16-bit in such a way that much of the original signal-resolution is preserved.
The 618 uses a single Motorola DSP56001 digital-signal processing chip to manipulate digital data on the fly. The input and output word lengths can be independently selected; pre-emphasis can be added if the operator so wishes; and, most important, the 618 redithers the data with a choice of noise-shaping curves. By shifting quantizing noise up to the inaudible 20kHz region as it reduces the output word length, it preserves as much as possible of the original's resolution in the midrange.
The interesting thing is that our starting point for the mastering process was an analog original with a noise floor not much more than 60dB down from the peak signal levels. Any changes occurring in the digital domain would be well below the recording's floor of analog tape hiss. Yet using the Meridian's noise-shaping algorithm when we downloaded the master to CD-R gave a noticeable improvement in air, space, and overall palpability.
By contrast with the first cut of this Chopin performance, the second was assembled from "processed" versions so that, by switching between the appropriate sections on tracks 12 and 13, you can directly compare the effects of the various strategies to which 16-bit linear PCM digital can be subjected.
The track is divided into 10 roughly equal sections, each indicated by an index marker. Indices 1, 3, 5, and 10 are the original data truncated to 15-bit accuracy, with a resultant dynamic range of 90dB. (I did this by attenuating track 13 by 6dB, thus moving all the bits in each word one place to the right, then stripping off what was now the 17th bit. Amplifying the data by 6dB brought back the level to what it originally had been, but now the 16th bit carries no meaningful data.)
Indices 2, 7, and 9 are the 20-bit data processed by the DTS Zeta algorithm set to its lowest, 128kb/s/channel (footnote 6) data rate; indices 4, 6, and 8 are the 20-bit data recorded on a DCC cassette with a consumer DCC recorder, and subjected to the PASC data-reduction algorithm with a data rate of 192kb/s/ch. For those who have CD players without Index access, the time information is as follows: 0:00, 0:53, 1:44, 4:04, truncated; 0:26, 2:33, 3:28, DTS; 1:03, 2:11, 2:58, PASC. Note that there were no D/A or A/D conversions; all the dubs were made in the digital domain.
I'll leave it to you to decide how close the data-reduced versions get to the original. I find it intriguing, however, that the truncated version differs the most, particularly in how less easily identifiable audience noises become. Listen, too, to the difference truncation makes to the sound of applause.
 The Binaural Grand Prix (DDD) 3:16
Recording Engineer: John Atkinson
Recording Venue: "The Pits Hairpin," Montréal Formula One Grand Prix Circuit, Montréal, Canada
Recording Date: June 12, 1992 (first day of qualifying)
Microphones: two Sony ECM-150 omnidirectional electrets worn on a headband in front of JA's pinnae
Recorder: Aiwa HD-S1 portable DAT
Due to its extensive use of composite materials, the modern Formula One racing car weighs just 505kg (1111 lbs), yet its 3.5 liter, gasoline-burning engine, capable of revving to more than 14,000rpm, develops more than 750bhp, enabling it to accelerate to 100mph in under 3 seconds and reach a maximum speed of more than 200mph. Unlike the American Indycar series, where many races take place on oval tracks—Indianapolis is the greatest of these—F1 cars race exclusively on what are called "road" courses: ie, dedicated tracks that feature a taxing mix of straightaways and corners.
Shown here at the Montréal Pits Hairpin is the late Ayrton Senna in his McLaren-Honda. Three times a World Driving Champion, Senna, in my opinion, was the greatest driver—in terms of raw talent and the ability to use it to push his car beyond its theoretical limits—of his generation.
As a longtime F1 fan, I try to get to as many races as I can. Sixteen races comprise an F1 season, but the only one currently that takes place comfortably close to the US is the Canadian Grand Prix, which is run on the Isle Notre-Dame in the middle of the St. Lawrence river in Montréal. In 1992, not only did I attend the Canadian race, but I took along a binaural recording system consisting of two small omnidirectional microphones, which I wore just in front of my ear canals.
What you should hear: When played back on headphones, a binaural recording should put the listener at the original event, with all the sounds emanating from outside the head. (This is the opposite of when you play a stereo recording on headphones, when all the soundsources appear inside the head, strung on a line hung between the ears.)
The photograph (fig.3) shows the soundstage captured by the microphones. You should hear spectators chatting to your sides, and a low-level buzz coming from the PA system in front of you. The French-Canadian announcer's voice rises in excitement, and you can hear the cars approaching from the straightaway, braking hard in under 100 yards from almost 200mph in sixth gear to just 40mph in second as they enter the hairpin. (The carbon-fiber brake discs can be seen glowing red-hot through the wheel hubs.) Negotiating the hairpin, the car's unloaded right front tire lifts completely off the ground as the driver pushes the limits of tire/Tarmac adhesion.
Fig.3 The view of the Montréal F1 circuit's Pits Hairpin, as captured by the binaural microphone technique.
Once past the apex of the bend, the driver floors the gas pedal, accelerating hard through the gears and skating across all the available road—mind the outside wall!—to get to maximum speed within a couple of hundred yards. The throaty roar of the V8-engined pack can be easily distinguished from the higher-pitched whines of the higher-revving V10s (Williams-Renault, for example) and V12s (McLaren-Honda, Ferrari), and you can hear one driver running his engine against the rev limiter as he screams away in too low a gear.
The 1992 Canadian race was won by Gerhard Berger in his McLaren-Honda MP4/7A, and the 1992 World Championship was taken by Nigell Mansell; but this cut is dedicated to the memory of Ayrton Senna da Silva (1960-1994), who won pole position in Montréal with a lap time of 1:19.775 (equivalent to an average speed of 124.19mph).
Footnote 6: The DTS proposal for discrete surround-sound laserdiscs is to use the Zeta algorithm set to its best-sounding 240kb/s/channel data rate. See Stereophile, Vol.18 No.3, March 1995, p.33.—John Atkinson