circa75 Home | About circa75 | Articles | Links | Contact Us

Posted by aaron at 09:00PM, Sunday, October 31st, 2004

The Linn Numerik Digital-to-Analog Converter: Laying down old-school beats, accurately placed to the picosecond

In which I talk about my experience with jitter in CD replay, and attempt to explain why "bits-is-bits" sounds to me like baloney

I just got a new DAC for my stereo. It's a Linn Numerik -- the DAC that Linn made to go with my transport (a transport is a CD player minus a DAC), the Karik. For the past several years I've been using an old Micromega Duo BS, which is an early-90s French DAC with a nice, up-beat, warm, engaging sound. Since I got the Karik, I've wanted a Numerik (or simply an altogether-better Linn CD player), and they're finally available at prices I can afford.

My new toy is made a little more interesting because I've been telling a few (skeptical?) friends about my experiences with reproduction of digital signals, like those off of CDs, in home audio. One of the things my experience has taught me in the decade or so I've been into audio is that Linn was right: for loudspeaker-based systems, getting a better-quality source component (CD transport/player, record player, tuner, etc.) has a bigger impact on how much you enjoy listening to music than spending money on amplifiers or speakers or anything else farther down the chain ("source-first"). Most normal, sane people, of course, are utterly incredulous that a CD player that originally retailed for something like six grand can really sound any different than an old Technics from a chain store.

Obviously, I hear a difference (some would say, I think I hear a difference): I can hear a pretty big change in the sound when I switch from my old DAC to my new one. I can also hear a change when I use different CD transports to feed a digital signal to the DAC. That's where incredulity generally turns to outright mockery: "That's impossible. It's a stream of 1s and 0s. How can changing the device that reads the stream make any difference in how it sounds, through the same digital-to-analog converter and analog stage, unless at least one of the transports is screwing up, and sending out 0s where it should be sending 1s, or vice versa?"

It's a good question. Clearly, we have the technology to read data off a CD correctly, right? You can stick a CD-ROM in your computer, and double-click an icon for a program on the disc; do it two successive times, on two successive CD-ROM drives, and the same program launches without errors both times, right?

Well, yes, but that's different. A computer is bufferring the bits in memory as it reads them from the CD-ROM, whereas a cd player is constructing a waveform in real-time with (usually) no buffer. Back in the 80s, when the first CD players came out, a lot of audiophiles were more or less offended by the sound from them: bright, brittle, shrill, and metallic; there was detail, but there wasn't as much realism as they felt they heard from the best record players at the time. Linn, whose LP-12 record player ruled the roost back then, sneered at CD and pronounced it a flawed medium. At the time, they didn't see how anyone could make it competitive. Slowly, though, audio companies started making players that sounded less and less awful. They brought out separate transports and DACs, and retro DACs with vacuum tubes in them, to warm up the cold CD sound. Theoretically, a lot of people agreed that any "competent" transport should sound like any other -- but, strangely, they still sounded different. Eventually, Linn started looking at the problem themselves. They came to the conclusion (not exactly original, but also not conventional wisdom at the time) that the biggest issue involving the sound of different CD systems wasn't the quality of the analog output stages of the DACs, but the amount of clock jitter present in the interface between the transport and the DAC.

The idea is that if the samples -- the 16-bit words that make up a 1/44000th of a second snapshot of the audio signal -- that the transport reads off the disc aren't all put together in the DAC with exactly the right amount of time between them, the shape of the analog waveform that the DAC makes will just be wrong. (I'm oversimplifying; check out this Stereophile article for a much more detailed explanation.) The waveform will have artifacts that might result in phase anomalies, or weird harmonics, or ultrasonic ringing. It's weird stuff that seems like it shouldn't make a huge difference, but smart people have posited that the errors that result tend to manifest themselves in the frequency range where the human ear is most sensitive to things like phasing (presumably so that, as monkeys, we could tell where the big eagles were, when they flew near to snatch us off our branch and eat us, and thus avoid them; or, if you prefer intelligent design, because God made us that way; it was pre-ordained that we'd react negatively to jitter.)

The problem was exacerbated by the spec for the S/PDIF (Sony Philips Digital Interconnect Format) interface that's usually used between transports and DACs, which calls for the clock signal, which is what the DAC uses to time the samples as it puts them back together, to be embedded within the rest of the digital data. Extracting this clock signal is error-prone at many steps along the data's path, and it turns out that things like the physical interface -- its conductive properties (resistance, inductance, etc.), if it's an electrical signal; whether it's an electrical or optical signal -- and the hardware used to generate and extract the clock signal in the transport and DAC, respectively, can make pretty big differences in measurable jitter.

Linn's solution at the time was, while they didn't have the technology available to remove the timing errors for the samples from the CDs themselves, to at least fix up the transport-to-DAC interface. They did this by "slaving" the transport's clock to the DAC, so that it sent words down the interface exactly when the DAC expected them. So, the Numerik has a little electrical output on its back panel next to the S/PDIF port, for the clock, and the Karik has a socket on its back panel, too. The Karik can feed an S/PDIF signal to the Numerik without the clock signal being hooked up, so you can actually connect and disconnect the clock link while you're playing a CD.

The effects are shockingly dramatic.

What I hear, without the cable: the CD sounds detailed and clean. It's perhaps a little bit cold -- it's not lush and warm, and I'm not blown away by the realistic timbre of a french horn or an oboe. But the music comes through well enough. The rhythms are engaging and snappy, when the music on the CD is. It's not at all harsh, it's tonally well-balanced, and there's a pleasant sort of aura around the performers, particularly on sparser rock CDs. The stereo image implies musicians burnished in the glow of stage lights shining through cigarette smoke, pretty accurately placed.

What I hear, with the sync cable plugged in -- all the same hardware bits, but supposedly less jitter: that burnished haze goes away, and I suddenly hear three new instruments I didn't know were in the mix, and a bassline that I'd sort of been aware of is now fascinating and played with chops and subtlety. That little tremolo at the end? That's cool. I didn't notice that before. I like how he slides up to that note. Oh, and that thing, that unidentifiable rhythmic sort of sound from before? That's a rhythm guitar slightly to the right of center stage, with a faint tambourine behind it and to the left, and someone else far off on the left shaking some maracas. The sound is darker, and squeaky-clean in a way that sounds almost cold. All that haziness that's disappeared leaves things sounding sparse for a moment, until you realize that all that's missing is some sort of noise. There's so much more bass articulation. Listening to my recording of Lenny Bernstein's Shostakovich 5, what used to be a nasty, typical 70s recording with artificial tonal balance and weird instrumental color becomes a decent recording with good balance and ambience on the winds and horns. The strings are still a little spotlit, but they're obviously (well, obvious now) miked differently than the rest of the orchestra. And the bass! I can clearly hear when the double-basses are doubling the trombones, where I could sort of hear a rumble before. The timpani attacks are astounding. The end of the last movement is gripping and exciting, and the way the performers build tension is great. I find the music more involving, and more interesting; I can hear what instrumentalists are doing, and I'm more aware when they have impressive chops. I feel like I hear a lot farther into the mix, without any harshness or artificial brightness. In fact, some CDs that seemed cursed by a sort of Pro Tools glare that made them a bit spitty on everything I'd heard them on are fine now; I can listen all through the album, and still want to turn it up at the end.

It's remarkable, plugging and unplugging the clock sync repeatedly, how much noise goes away. This isn't something subtle that you only hear with multi-thousand-dollar amps and speakers. This is noise just a little less dramatic than turning on the dolby switch on your 80s tape deck. This is noise like playing a dirty record then playing a clean copy of the same LP; noise like turning on and off some faulty piece of electronics that's creating interference. You plug the sync link in, and the evident noise floor of the CD drops, leaving the sound of instruments playing -- music that was previously buried in the hash. It's astounding, and it's all with the same transport and the same DAC, even. The same bits are coming through; only the timing by which they're reconstructed into an analog waveform is changed.

How does the poor old Micromega compare? I hook it back up, and it still sounds great. It's pleasant and warm (the one fault I hear with the Numerik is that everything becomes a little cold), and it bops along happily. It still tracks those tempo changes, those ritards and accelerandos, that drum player speeding up towards the end of the line. But there's a lot missing. Stuff is obscured a little, and there's a bit of haze layered over, that I didn't recognize as haze before. It sounds a lot like the Numerik without the sync cable, in fact, but warmer and happier, if a little less detailed.

I'm won over. I can't prove that there's not something else going over that clock sync cable, or that Linn didn't perversely cripple the performance of their digital gear when used without the sync cable. I really doubt it, though. The most impressive thing about the Numerik is how exciting and interesting a lot of symphonic music suddenly is. I've never heard an orchestra on a CD, played back over loudspeakers, sound much like a real orchestra before this, and I've certainly never noticed this much drama, tension and relaxation, ease and warmth and timbre and other information, coming out of a CD before. Yeah, rock, pop, electronica, whatever all sound better to varying degrees (that Goldfrapp track? Damn! I didn't realize you could make sound appear to come from outside the room like that with just two channels! It's creepy! Some rock CDs, though, just sound kind of eh.) But all those CDs of symphonies and concerti, masses and suites and tone-poems, that were always tolerable when I was sort of interested in digging down into the music, maybe reading through the score, or listening to in the background, are now gripping and exciting all on their own.

People have a tendency to want to rationalize about things, based on pet theories and their own experience, their opennes to new ideas, or lack thereof. "You must notice this kind of noise in the signal. Look at this graph!" "You can't possibly hear that, because psychoacoustic studies show it makes no difference." I have far too little physics and electronics knowledge to dissect all of the issues in home audio, and as far as I can tell, that's true of everyone on Earth. I can describe what I hear, though; years of listening to music performances have taught me to trust my own judgment and hearing over that claimed by others, who may be sycophants or nepotists or sexists ("That violinist is great! No, of course I'm not won over by her slutty black dress!"), or who may lack the experience to have any basis for comparison ("Of course he's a great violinist; he plays in tune, and has great showmanship. No, I've never heard Midori. Why?"), or who may simply be listening for different things than I do. What I hear in this case is amazing, and I'm thrilled to have bought this twelve-year-old Numerik.

I'll wait to talk about loudspeakers vs. headphones another time.
circa75 Home | About circa75 | Articles | Links | Contact Us

All content copyright © 2001-2009 the owners of