Digital Processor Reviews

My first rips were of these LPs: Astaire (English Mercury 9109 702), a 1979 album from the English singer and pianist Peter Skellern in which he imaginatively arranges songs made famous by the elegant dancer for his multitracked voice, piano, rhythm section, and English brass band; "Die Tänzerin," from German singer Ulla Meinecke's Wenn Schon Nicht für Immer, dann Wenigstens für Ewig (German RCA 426124); Sly and the Family Stone's Fresh, from 1973, which features Doris Day reprising "Que Sera, Sera" (Epic ECD 69039); the first album from British blues band Steamhammer (1969, English CBS 63611), which features on second guitar Martin Quittenton, later to achieve fame as the co-writer of Rod Stewart's hit "Maggie May"; that audiophile classic, the late Radka Toneef's reading of Jimmy Webb's "The Moon Is a Harsh Mistress," from Fairytales (Odin LP03); The King's Singers Sing Flanders & Swann and Noël Coward (1977, EMI EMC 3196); and a favorite of John Marks, the 1970 recording by the King's College Choir and the Jacques Orchestra, conducted by Sir David Willcocks with narrator John Westbrook, of Ralph Vaughan Williams's An Oxford Elegy: a setting of two poems by Matthew Arnold, in purported homage to fellow-composer Gustav Holst (HMV ASD 2487). I list these as examples of recordings I used to love but haven't listened to in decades, and which the QA-9 pushed me into digging from the darker recesses of my LP shelves.

I tried a variety of sample rates with these LP rips: 44.1kHz was very good, but didn't capture the essence of the original LPs' sounds; 96kHz was better; but there was no doubt that with a 192kHz sample rate I could not distinguish between the LP and the digital rip. And believe me, I tried. I A/B'd the two versions until blood came out of my ears and I was heartily sick of this music I hadn't heard for, in some cases, decades. When, in An Oxford Elegy, John Westbrook declaimed "Come, let me read the oft-read tale again . . ." for what must have been the tenth time, I felt like screaming "No! Don't read it again!"

With the 192kHz rips, the LP's surface noise floated free of the music in a manner similar to how it does with analog playback, making it easier to ignore it. At 44.1kHz, the surface noise was integrated into the music, increasing its annoyance. While 24/192 rips are profligate with hard-disk space—the 22-minute Vaughan Williams is a 1.5GB AIFF file—this was the only way to go with the QA-9.

Comparisons
I have used three two-channel A/D converters to make the recordings that have been released on the Stereophile and other labels over the past 12 years: two dCS 904s, joined in 2003 by a Metric Halo ULN-2. The dCS is the final version of one of the first 24-bit, high-sample-rate converters, the dCS 900, which was introduced in 1996. The 904 takes balanced line-level signals and uses a 5-bit flash converter embedded in a sigma-delta loop; it can output DSD data via an SDIF-2 connection, as well as 24-bit LPCM up to 192kHz over single and dual-AES links. The Metric Halo is a well-regarded two-channel A/D and D/A converter with low-noise microphone preamps and a headphone amplifier. It uses a 24-bit sigma-delta converter from AKM, their 4393 chip. Although it has AES/EBU and S/PDIF inputs and outputs, the ULN-2 is primarily used to feed 24-bit data at sample rates up to 96kHz to a host Macintosh computer via FireWire.

I'm pleased with the sound of the recordings I've made with these converters; it was to see how a new A/D design from a leading high-end manufacturer compared with them that triggered my doing this review.

Because I have no easy way of storing the dCS 904's dual-AES/EBU 192kHz output as a computer file, and as the Metric Halo is limited to 96kHz, I digitized LPs with these converters at 96kHz. I then redigitized the LPs I'd sampled at 192kHz with the QA-9 at 96kHz. Although the converters were limited to a peak level of between –3 and –6dBFS when the LPs were being ripped, I normalized the peak level of all the files to –0.1dBFS with BIAS Peak, so that there were no loudness differences. Although the dCS offers a choice between brick-wall and slower-rolloff anti-aliasing filters, I used the brick-wall.

First up was the Metric Halo ULN-2 at 96kHz. In "Die Tänzerin," Ulla Meinecke's voice sounded slightly harder at high levels than the QA-9, with a little less palpability to the image. The repeat echoes for which, in the 1980s, this track became renowned as an audio-show demo track were as audible through the ULN-2 as they'd been through the QA-9, but paradoxically, the soundstage overall was a little flatter. Low frequencies were very similar with the two converters, the punchy bass interjections of the electronic keyboard in the song's coda sounding equally weighty—though if I had to swear to it, the Ayre's lows were marginally tighter overall.

With "Puttin' On the Ritz," from the Peter Skellern LP, the Ayre's high frequencies sounded smoother than the Metric Halo's, but without being in any way mellow or rolled off. The small inflections in Skellern's phrasing were slightly better developed than with the Metric Halo; once I'd heard that, going back to the Ulla Meinecke rip revealed the same difference between the QA-9 and the ULN-2.

Turning to the dCS 904, again with 96kHz files, the converters sounded much more similar than had the QA-9 and ULN-2. The low-frequency keyboard jabs in "Die Tänzerin" sounded identical through the dCS and Ayre, as did Meinecke's voice, whereas the ULN-2 had emphasized fricatives a little. After repeated comparisons, especially using headphones, I felt that the reverberation accompanying the finger snaps and occasional handclaps that so effectively punctuate this song had a slightly more integrated relationship with the direct sounds through the QA-9, leading to a slightly more developed sense of image depth than with the dCS 904. With the Skellern, again, the converters were very close, but with a slightly better developed sense of the recorded space through the Ayre. I also felt that the dCS 904 was not so kind when it came to dealing with the inevitable end-of-side distortion.

One peculiarity: In Sly and the Family Stone's "Que Sera, Sera," when Doris Day reaches the end of the first line—"When I was just a little girl, / I asked my mother, what will I be?"—she flattens the final word. With the 96kHz Ayre transfer, it was just a little more easy to distinguish the pitch of that blue note against the accompanying chord on the Rhodes piano and organ than with the dCS. I have no idea what that means, other than to note that listening to the original LP, there was never any ambiguity about the note's pitch.

To put these comparisons into context, the Metric Halo costs $1695 and includes a wealth of DSP functionality; the dCS 904 cost around $7000 when last available. For Ayre's QA-9 to be not only competitive but to excel the expensive dCS converter's performance, even by a small degree, while costing $3950 or $4750, makes it a great value.

Summing Up
Why rip an LP when a CD is available? I'm not going to get into the whole "LPs sound better than digital" matter—pace Mikey Fremer. But CDs from before the mid-1990s were generally recorded and mastered using A/D converters that were only optimistically described as having 16 bits' resolution. It was only at the end of the 1980s, with the advent of the first true high-resolution A/D converter, designed by Robert Adams of dbx and marketed by UltraAnalog, that CDs had a true 16 bits of resolution. (Yes, the word length on a CD has been 16 bits since the crystallization of the format in the famous "Red Book" in 1981, but the least-significant bits of the 16 back then were often quantization artifacts or random noise.) As a well-recorded LP can have resolution greater than CD's 16 bits in the region where the ear is most sensitive, and can have a recorded bandwidth greater than the CD's 22.05kHz, transferring an LP to digital with something like Ayre's QA-9 makes a lot of sense.

On the other hand, and particularly with classical LPs, getting rid of the inevitable ticks is a royal pain. I have always kept my LPs in good condition, and I both clean them with a carbon-fiber brush and get rid of static with a Zerostat before I rip an LP. But there are always some ticks, and classical music, with its wider dynamic range, is less good at masking them. I resolve to eliminate only the worst problems (using the waveform pencil tool in BIAS Peak Pro 7) when I clean up the file (footnote 2), but after I've done so, lower-level ticks that hadn't bothered me before now sound louder, and my obsessive-compulsive tendency urges me to make another pass. And then another.

Given that I believe you should rip LPs with your loudspeakers muted (almost all turntables and tonearms are microphonic, to a greater or lesser degree), and adding the cleaning-up required, ripping an LP becomes a lot like work—even without taking into consideration that I also normalize the 24/192 file after the click removal and prepare a downsampled 24/48 version of the tracks to play on my iPod Classic. It has become something I don't relish having to do.

But when recordings you love have never been issued on a good-sounding CD, it makes sense to rip them with Ayre's QA-9—it's the closest thing to a truly transparent audio component I have encountered. Even though I've purchased many A/D converters over the years—a Manley ADC, two dCS 904s, a Metric Halo ULN-2 and MIO2882, an E-MU 404, an M-Audio Transit—I am buying the review sample. There are many LPs waiting to be ripped.

---

Footnote 2: Pure Vinyl's Rob Robinson points out that if you rip the LP "flat," with no RIAA equalization—which his program can apply when the file is played back—the tick will last only a couple of samples and is much easier to delete. Ripping with RIAA, as I do, results in what was originally a straight up/down pulse being stretched into something that both lasts longer and both under- and overshoots the waveform's timeline.