Make your own free website on
Sonata LP
By John Atkinson, Wes Phillips, Michael Fremer, March 1997

There has been much argument in audiophile circles about whether an LP or a CD is a more faithful representation of a master tape. Although we recorded Robert Silverman's thrilling performance of the Liszt B-Minor Piano Sonata for CD release, we also had in mind to issue an LP. As the source for both would be the same, the question we can answer is: Will an LP cut straight from a 20-bit master tape via a Class A 20-bit DAC sound closer than a CD noise-shaped to 16 bits from the same 20-bi original? 

Although we did record backup analog tapes at the 1993 Sonata sessions, the edited four-track master existed only in digital form, as 18 gigabytes of 20-bit data on the Sonic Solutions digital audio workstation hard drives. The sound quality of these 20-bit data was actually analoglike in that it was free from high-frequency glare and other digital artifacts. It also offered a view into the acoustic of the Albuquerque church that the 16-bit noise-shaped CD, as good as we felt it to sound, only hinted at 

The CD had been mastered from a CD-R, which in turn had been created from a two-channel mixdown from the four individual 20-bit microphone tracks, noise-shaped to 16-bit resolution with the Meridian 518 processor. For the LP, an identical two-channel master tape but with 20-bit resolution would be played back on a Nagra-D digital tape recorder with the data decoded to analog by a Mark Levinson No.30.5 D/A processor. (The Levinson has a 20-bit signal path and true 20-bit resolution.) The analog signal would feed the Neumann cutting lathe at AcousTech Mastering, the facility shared by RTI and Acoustic Sounds. 

However, as we were reminded by veteran cutting engineer Stan Ricker, the lathe's computer needs enough advance warning of changes in the signal's envelope to optimize the track pitch to accommodate the music's dynamic range. When an LP master is cut from an analog tape, a special preview playback head provides this advance warning, but this is not possible with a digital tape. The answer was simple. As only two of the Nagra's four tracks are used for the stereo master data, I duplicated the data on the other two tracks, but advanced in time by about two seconds. These second two tracks could be used to feed the lathe's preview computer with the exact preview time adjusted with a Lexicon Model 300 digital delay unit. (This unit output the analog preview signal but was not in the signal path for the cutting.) 

To capture the sound of Bob Silverman's 9' Steinway D, Robert Harley and I had used a central Schoeps "Sphere" stereo microphone, with outrigger B&K omni mikes (see sidebar). Because there's a significant difference signal between the channels, a result of the "bloom" added to the sound by the spaced-omni mikes, I asked Stan Ricker if cutting this LP would present any problems. He laughed, and reminded me that he had cut some of the early Delos albums in the late '70s, which he had recorded with spaced omnis. "The engineer's art is to cut what's on the tape," he told me, "not to compromise it to make the cutting or the playback easier." 

It was with some trepidation, therefore, that I lowered my Linn Arkiv phono cartridge into the lead-in groove of the test pressing. Although the groove's vertical velocity is large, the Arkiv could track it set to 1.9gm downforce. I breathed a sigh of relief. Wes Phillips's Transfiguration also had no problem staying in contact with the groove walls---I'll leave it to him to give a blow-by-blow account of the cutting and to analog maven Michael Fremer to decide whether we were successful.--- John Atkinson

Cutting at Camarillo:

Wes Phillips John Atkinson and I flew to Camarillo in November 1996 to test a pet theory: Given the audible superiority of the 20-bit data coming off the Nagra-D to the 16-bit CD standard, would an LP cut from the 20-bit digital master tape preserve enough of that quality to justify pressing a vinyl disc? Our partners in this project were RTI's Don MacInnis and Analogue Productions' Chad Kassem, who had recently purchased David Wilson's disc-cutting facility and had installed it within RTI as AcousTech (footnote 1), under the direction of legendary mastering engineer Stan Ricker. Stan's done it all: an accomplished double bass player, he's also run record presses for Keysor, mastered the original Mobile Fidelity half-speed-mastered LPs at the now defunct JVC Cutting Center in Los Angeles, engineered some of the earliest digital recordings for Delos and (with Keith Johnson) the legendary 1987 Moscow Sessions CDs for Sheffield Lab, and won countless Grammys.) 

At AcousTech, stuffed into what used to be a high-speed tape-duplicating room, was the Neumann VMS 70 cutting lathe (fitted with a Neumann SX 74 cutter head) and control desk, a Studer tape deck, a pair of WATT/Puppies that looked mighty like the Ur-pair, and an extra table shoehorned in to support Stereophile's gear. We promptly set up the Mark Levinson No.30.5 on the table and used the Studer as a stand for the mighty Nagra, in order to have it close to Stan's console. The Nagra fed the Levinson processor via the Meridian 518 (this time used solely as a jitter-reduction box) and two lengths of Madrigal AES/EBU cable. The Levinson in turn was connected to Stan's console with MIT balanced cable. After the requisite -20dBFS reference tone from the Nagra tape and we promptly ran into our first problem of the day. 

"Where's the gain?" asked Stan. 

"What do you mean?" 

"If I bypass the board---which I'd really like to do, since it's got about a mile of wiring in it---I come up 6dB short of the lowest level I'd feel comfortable cutting at. I could even use another 8dB or so to feel really good about it." 

"The No.30.5 puts out the CD reference level of 2V RMS," John reassured Stan. 

"That may be. But my concern is that the quiet passages will put the pressing at a disadvantage. The average level will be -20 or -30dB, while the noise level of an LP is usually about -50dB. That's not much of a spread." 

Even using the board's active stages, we could only boost the input to Stan's minimum requirements, and he clearly wasn't happy about that course. We discussed our options and rethought the chain for a while. Finally, Stan suggested we use a Mackie 1202, a small mixing board that had been used for tape-duping. "Real audiophiles might be too snooty to use this,
Stan declared, "but it's real clean. I'm amazed by how good it sounds." 

This entailed yet another search for cables, this time terminated with ring-tip-sleeve phone jacks. "Oh yeah! Now I've got gain out the ass!" Stan burbled happily. 

Time for a reference cut. As Stan opened a box of lacquer blanks, John---remembering all the furor over lacquer quality that accompanied the rise of CD---asked what Stan thought of the current crop of lacquers. "I really like these Apollo lacquers, but the Transcos aren't worth doodley-squat! The Apollos are as good as any I've ever cut. 

"I'm ready for this session. I even put a new cutting stylus in the head this morning." We ran into problems with the first lacquer, however. Stan: "I've dropped a chip! First time that's ever happened here." When grooves are cut into a lacquer blank, the material carved out of the lacquer is lifted in a single continuous "curl" that must be sucked out of the way by the cutting lathe's vacuum system, or it will mar the lacquer's surface. The "chip" was that curl of lacquer, "dropped" by the vacuum system.

Examining the cutter-head assembly, Stan was stunned to find that the new stylus was, in fact, defective. We all breathed a sigh of relief. Everyone knew that something would go wrong; we'd gotten our disaster out of the way early, we figured, and it was relatively minor. 

"Whew," sighed Don MacInnis. "I thought we'd blow the whole day trying to link up our gear with yours. This is nothing compared to that." 

Stan set the reference levels and the preview time, made sure the cutter-head was functioning smoothly with the new new stylus, and we were ready to cut our first real lacquer of the day at 11:28am---not too shabby. "Believe it or not, if you talk while we're cutting the lead-in and early grooves, you can actually modulate the lacquer to the extent that you can hear it on playback," Stan cautioned. We fell into respectful silence---even Chad. After about 60 measures, Stan quietly said, "Don't be startled. I'll be switching between the preview and the master from time to time." 

JA responded, "I'm assuming that's why I didn't hear the first note of the piece." 

"Actually, I haven't switched anything yet," Stan replied. 

We scrubbed that cut and synched the master carefully. John had recorded a 1kHz reference tone 20 seconds before the music began, so we went back to the tone and counted off 15 seconds of silence. We cued another lacquer and started over. The first side
Doesn't even boast any band separations, so we all lost ourselves in the music---at this point, there's just not much to do. John was on his knees in front of the table behind Stan's console. 

"You want a chair?" Stan asked. 

"No, this is the proper position for prayer." 

At 12:11pm, with six minutes left to the side, Stan looked up from the console and exclaimed, "This sounds a hell of a lot better than that CD you sent me." 

"That's the difference between 20-bit resolution and 16-bit," I interjected. 

"It doesn't sound like digital at all." 

The side ended and we all examined the lacquer---without touching it. It looked perfect. Stan had spaced the grooves so that there seemed to be plenty of "land" between the dynamic grooves, and the 21-minute side spread over the blank evenly. This is tricky; you want to use as much of the disc as possible, but without bunching up too much music---especially dynamic passages---in those inner grooves where the groove velocity is so much faster. 
Just to be on the safe side, we cut another side A. This, too, was uneventful. John remained on his knees, alternately chewing on his cuticles and the ends of his mustache. At side end, Chad, Stan, and I high-fived one another and started discussing lunch---it was going on 1pm. John interrupted our chatter. 

"Stan, can you dig out the CD we sent you? I think I need to check something." We'd routed so many wires through different paths during the morning that Don had to dig up a pair of headphones in order for John to hear the CD player. He listened and glowered.
"Can I hear the beginning of the master tape, please?" he requested. 

As we heard the beginning of the sonata for the fourth time that morning, John groaned and smacked his forehead. "We don't have the first note on the master!" he wailed. It seemed that when John had mixed the data to the Nagra tape the night before we flew out to AcousTech, he had decided he needed to add some more ambience before the first piano note to allow for the LP lead-in---but under the time pressure he'd inserted the ambience before the second note of the Sonata, not the first. Murphy's Law wins again! "Damn, damn, damn!" 

Subdued, we went to lunch. Stan, Don, Chad, and I traded stories of our biggest blunders, thinking that John might derive some measure of comfort from them. Fat chance. The man was inconsolable.

When we went back to AcousTech, we cut threelacquers of side B with no delay. Why three? Stan examined the first cut of side B and determined that the spread could be wider. If the two sides of a disc have widely differing grooved areas, the record tends to dish. Not everybody pays attention to this kind of detail---but Stan Ricker does. He spread the groove more and evened up the sides. 

Another benefit of this is the increased amount of land between the grooves, which helps to cut down on the groove malformation that produces pre-echo. Think of a lacquer as a plastic substance just barely removed from a liquid---it tends to flow a bit. When the grooves are tightly spaced, the more dynamic side of a single groove can dominate the opposite side---the lacquer will try to achieve an equilibrium, resulting in the ghostly intimation of the music yet to come, or that just passed. 

This tendency for lacquer to "flow" is one reason why AcousTech located the cutting room at the same location as RTI's plating facility. Lacquers can begin to be plated within minutes of having been cut. 

Gary Salstrom, who runs RTI's plating operation, is part chemist, part technician, and part alchemist. We watched, awed, as he processed our lacquers. The first step is a thorough cleaning. The cutting process itself raises oils out of the lacquer, and it is vital to remove them: The electroplating process replicates the lacquer's surface on a molecular level, so it is important that nothing---not even an oil---be present upon it. 

The lacquer is soaked in a mild detergent solution for 30 minutes, then rinsed in filtered tap water, rinsed again in de-ionized water, and yet again in stannic chloride solution, for 45 seconds. Finally it is ready to be plated with silver, to give the surface a conductive medium for the nickel to adhere to in the plating baths.

The cleaned lacquer is fixed to a spinning table in the spraying booth, where a multinozzle head, mounted on a pendulum, sprays it with a mixture of dextrose, sodium hydroxide, and silver nitrate. The dextrose acts as a reducing agent, precipitating the silver nitrate into a layer of silver that adheres to the surface of the lacquer. (Common wisdom has it that the layer is about one molecule thick, but Gary tells me he's never actually measured it.) 

Watching the black lacquer turn into a reflective silver disc after just a few rotations is a magical experience, rather like watching a photographic image's rush toward reality in the developing tray. No matter how many times I witness it, I always feel some degree of awe. 

Why silver? Salstrom explains, "We're trying to duplicate the groove cut into the lacquer, and silver is the best conductor that's affordable. It has a lot of desirable qualities. First, we get an even spread of amperage from the center to the outside edge. You can also get silver that is very pure---impurities will show up sonically---and it can be applied in an incredibly thin, grainless coating. When I was at Wakefield [a 1980s-era plating and pressing
 facility in Arizona---WP], a chemist and I developed our own silver spray. It was so hot there we had problems with the ammonium hydroxide escaping from the bottle of silver spray, which meant that metal started dropping out and forming solids at the bottom of the bottle. So I started mixing my own. 

"Now I get silver nitrate crystals from a company in Chicago because dissolving my own silver 'coins' in nitric acid can be hazardous, but that's the only thing I have made up for me. And even though that results in very pure silver, I use a micropore filter before it goes into the lines feeding the sprayhead."

The silvered lacquer can now be put in a plating bath, there, over the next 14 hours, it will grow a master---a metallic negative that is then used to produce the mother, which is a positive, a direct copy of the lacquer (footnote 2). Stampers---negative images of the mother and, of course, of the final records produced from them---are struck from the mother in a similar manner. This is known as "three-step" plating, which is used when many thousands of pressings are required, or if it is important to preserve the master. (Dance records and other "disposable" projects frequently employ a two-step process, which can put the master at risk.) 

For the actual plating, RTI has one four-position rotary tank (which uses a rotating cathode as opposed to a fixed one) and a nine-position pre-plating tank (which uses a fixed cathode putting out no more than 20A, but circulates the plating bath more vigorously). The terms are confusing, because the pre-plating tank is actually used for fully plating most projects. Traditionally, pre-plating tanks, which use lower amperage and a lower temperature in the bath, are used to build a thin layer of metal onto the lacquer before the disc is moved to the rotary tank, where higher amperage and bath temperatures complete the job quickly. Quality plating facilities prefer to use the pre-plating tank for the entire process of forming a master, however, to prevent exposing the fragile lacquer to high temperatures. The plating process thus takes about 14 hours, as opposed to less than half that. Stampers are formed in the rotating tank, since the lacquer is not involved. 

Where did RTI get their plating equipment? "Our plating tanks were built by Keysor Century, but they stopped pressing records and plating years ago," Salstrom explains. "They put the tanks in a storage locker at the back of the plant and essentially forgot about them. We were having a hard time finding equipment and, in the middle of a conversation with Keysor, this stash of gear came up. We went out to look at it, and it was all knocked over and jumbled up from the earthquake. But we were able to rebuild it with Rick Hashimoto's help." 

John and I witnessed the spraying of our lacquers and watched as they were lowered into the pre-plating tanks. Then we were taken on a tour of RTI's pressing facility. RTI has seven presses, all going full-bore. It's an impressive sight, to be sure, but I think I'll leave a complete description of a record press for another time and another project. Ponder this, though: Making a record is an artisanal affair---it requires a great deal of hands-on intervention from the press operator. If records have become a premium-priced product as digital has taken over the marketplace, that is a reflection of their true nature: They are just one step removed from being handmade. As we watched the press operators baby their charges and the QC people run into the listening room to check pressings, I was overwhelmed by the realization that the technology worked at all---and even more that it works so well. 

In fact, proof of this was delivered to us by way of our test pressings. Quite apart from the missing first note and our desire to recut without using the Mackie unit, we had severe nonfill (footnote 3) and pre-echo problems throughout both sides of the disc---and these were the cuts that RTI had determined to be the best! I QC'd several discs one evening---my notes of audible events averaged one page per side. 

Test pressings are usually noisier than the product run anyway. You stop the press to change stampers and the vinyl in the hopper warms up, then cools off, then warms up again---all of which affects consistency. In a short run of, say, 20-50 pressings, more than enough to produce multiple copies for the client plus several file copies for the pressing facility, the vinyl never achieves optimum flow---as it would 100 pressings into a production run. When I called RTI to discuss the tests, they astounded me: the TPs I'd been listening to were the product of an extended run---they'd actually pressed a few hundred copies hoping to eliminate the nonfill by simulating real production circumstances! 

The nonfill was annoying enough, but the almost constant pre-echo was extremely distressing. We all conferred and concluded that substituting a Mark Levinson No.38S for the Mackie 1202 would only take us so far. We needed to monitor cutting gain, watch the groove spread, and pray. 

Footnote 1: The deposition of metallic materials upon a mold  (ie, the silvered lacquer) is accomplished electrochemically. That is, the lacquer is connected to the negative pole of a DC circuit, making it a cathode; while the positive side of the circuit uses the metal to be deposited (in this case, nickel) as an anode, which is consumed as it migrates to the mold. The whole assembly is submerged in an electrolytic bath in the plating tank, which at RTI consists of nickel sulfamate.

Footnote 2: When a record is pressed, the vinyl biscuit is deposited in the center of the stampers, occupying the label area. When the stampers are forced together, the vinyl spreads from the center out toward the edges of the disc. This forces it into the inner groove wall (left channel) and, in most cases, the outer groove wall (right channel) as well. Sometimes, the outer groove wall doesn't form a perfect negative of the stamper---this is called nonfill---either because the vinyl hasn't achieved optimum consistency or because the outer groove is heavily modulated, making it too complex a form to force the vinyl into.

Almost a month after our first attempt, JA and I made our way back to Camarillo to try again. This time it took us only a few minutes to set up. We arrived at 10am and were cutting side A before 11. This time, the first note was on the tape. "It took me four hours to restore the 20-bit master from the Exabyte archive tapes," John mused, "just five seconds to redo the problem edit, and another two hours to remake each master. With an analog master tape, we could have fixed the problem in five minutes total!" Stan was dubious about our first attempt---A3---and insisted on cutting an extra back-up, so we wound up with five A sides. As we listened to one of Silverman's thunderous climaxes, John asked Stan if people would have any problems tracking the disc. "Hey, man, that's your problem---all I do is cut what you give me!" 

By 4pm we had another pair of B sides cut. The day had seemed uneventful, but we were too cautious to celebrate yet. I should point out that though Stan seemed---even to such eagle-eyed observers as JA and I---to be doing very little between takes, underneath his casual-seeming exterior he was calculating constantly. He made subtle adjustments with each pass and, when we got the next batch of test pressings, it became obvious that he was picking up on cues that the rest of us had missed. We wound up approving the last cuts we attempted: A5/B4. 

Awaiting those new TPs, we were tense. What if we had bad lacquers? What if we had to start over again? It fell to me to listen first---nobody else had the guts. Truthfully, neither did I---my fingers were shaking as I cued the lead-in groove. Side A4 was much better. Nonfill was nonexistent, and, while there was pre-echo, it was much quieter and far less prevalent than in the last batch. Before I went on to listen to the B3 cut on the opposite side, I went back and listened to A4 again. It was definitely better, but not, I felt, quite there. I cracked open A5 and listened. Yes! You will still hear some ghostly pre-echo, but this side has all the magic. I went back and compared B3 and B4, settling at last on B4. 

Before we approved the sides, however, we held a group listening at my house. Executive Producer Gretchen Grogan, JA, and I agreed that, mild pre-echo aside, A5/B4 was a recording we could all be proud of. Just to be certain, we listened to my LP of Bolet playing the B-Minor Sonata. London had managed to fit the entire work on one side of the disc, mostly by cutting the level way back---and even then, they were plagued by pre-echo far worse than we had to settle for. And the dynamic range that results from our cut is so seductive that I'd be loath to surrender any of it---not that we could guarantee that a reduction would eliminate any problems. 

In a sense, the very quality of RTI's 180gm pressing works against it. They've managed to produce some of the quietest vinyl I've ever heard, and that means that you can hear even further into the recording and all of the mechanical parts involved in playback. 

I'm glad we decided to release Sonata on LP, even though it was a lot of hard---and frequently frustrating---work. For one thing, I believe it to be a spectacular performance. We must have listened to it hundreds of times in the process of releasing the LP, and it holds up. Oh my yes, it definitely holds up. But the whole process also taught us some lessons about making black discs. I love LPs, but now I understand why the major labels converted to CDs so precipitously. In digital production, if you follow the rules meticulously, you will get consistent results. Not so in the exciting world of analog. There you can follow all the rules and still be at the mercy of your lacquer, plating facility, or vinyl pellet producer---to name just a few. Every day is an adventure. Let's fac it---the major record labels just aren't all that adventurous. From the evidence, we are. And I sure hope you are, too. After all, this Sonata takes you on a trip to both heaven and hell. As for JA and me---we've been there already.---

Wes Phillips

Original text can be found on the site

back to the index