The Art and Aesthetics of Audio Restoration

posted in: Essays | 0

Before I deliberately chose to specialize in the narrow field of audio restoration, I read books about noise. Some audio engineers devour gear manuals, debating the merits of particular capacitors and tubes, or they read music biographies that detail the tools, processes and partnerships that led to brilliant music making. I, however, have a soft spot for a good philosophical yarn about how noise can symbolize social and political order/disorder, such as Jacques Attali’s book Noise: The Political Economy of Music, a far-reaching theory with Marxist overtones about of the role of music in society throughout human history. So I’ll start by saying that my job, which requires me to sit under headphones for hours at a time identifying, evaluating, contemplating and then delicately etching out noise, is a good fit.

True, I haven’t cracked that Attali book since grad school, but it made enough of an impression on me that I do think about it when I’m restoring a damaged recording or remastering an album that has never been heard in the digital world. My job is to identify that division between the music and the noise inherent in any physical recording. It’s not a line; it’s a gray space in which all judgements are necessarily subjective. That’s why I consider my skills to be as much about taste and cultural/contextual understanding as knowing how to use digital audio restoration tools like declickers, denoisers and spectral repairs. The noise imprint on historic recording media – I’m talking about things like tape hiss, disc crackles and pops, cassette lo-fi-ness – locates those recordings in time, place and musical moments. Those noises tell us a lot about the music. Good audio restoration lets the music shine through without artificially erasing or altering this sonic landscape.

I always liked these two quotes from Attali’s book:

“Music gives structure to noise.” (Attali, 9)

“What is noise to the old order is harmony to the new.” (Attali, 35)

Without getting too bogged down in academic concerns, these theoretical snapshots act as signposts for making aesthetic decisions about what is music and what is noise. Attali’s statements remind us that noise is a subjective and shifting concept, that noise is not simply bad stuff (as we audioworkers often assume when obsessing over signal-to-noise ratios, etc…). Instead, after a century of humans accumulating and continually re-experiencing millions upon millions of noisy recordings, noise has become an integral part of our sonic culture and, therefore, of our consciousness. The recent rise in noise-adding algorithms in many pro-audio plugins (however sonically dubious such digital overlays may be) stands as testament to the integral role of noise in today’s sonic culture; the big difference is that noise was an unintended byproduct of recording in the past, while today we pick and choose how much and what kind of noise to add or, in the case of a restorationist, to subtract. The art of audio restoration lies in subtracting the unwanted noise without damaging the wanted noise, and the aesthetics of audio restoration lie in deciding, quite subjectively, what is wanted and unwanted. Restorationists do not just remove noise; we sculpt it.

Let’s break this down with an example:

I recently finished restoring and remastering a very rare cassette, Obaa Sima, by Ghanaian highlife artist Ata Kak, for the label Awesome Tapes From Africa. Recorded, surprisingly, in Toronto in the 1990s, it’s an infectious, danceable, bizarre mix of clubby techno beats, synthesizers and chant-singing. This is a holy grail project for blog/DJ/label founder Brian Shimkovitz. He had been searching for Ata Kak or his family to obtain the rights to reissue this album for nearly a decade and finally found the man last year. This remaster will be the first time this recording has been properly released in the digital realm and reissued on vinyl.

My work on Ata Kak began last February when I got an email from Brian Shimkovitz that read, “Ata Kak found his DAT master!” Incredible! I could remaster from an actual master recording rather than a worn cassette copy. A package arrived from Ghana containing the DAT master and a commercially manufactured cassette of the album Obaa Sima.

DATs are a notoriously finicky medium, and this one appeared to be in bad shape, so I used a DAT Xtractor, which reads the DAT like a digital linear tape, rather than playing back the digital audio and recording that signal. This method tends to result in fewer audible dropouts and glitches and has the added benefit of producing a report of all errors. None of this mattered when the Ata Kak DAT master snapped six minutes into playback! After two decades in Ghanaian heat and humidity, mold had grown on the inside spool and glued the tape to itself. I disassembled the DAT and tried baking, lubricating, cleaning and splicing, but it was a goner. I even called up Specs Bros, the leading experts on media restoration — the people the Pentagon calls when a DAT of military secrets has been burned in a helicopter crash — but they told me there was almost no chance of salvaging this DAT.

I turned to the cassette, not in the best shape but playable, which was a big improvement over the now ruined DAT. I did two high resolution transfers of the cassette at 192k / 24 bit, thinking a second full play-through might loosen up kinks in the tape and give a more consistent speed. (Cassettes, like most analog tapes, will lose temporal linearity over time, but can sometimes regain linearity through “exercising” the tape). I finally had a digital transfer of this recording. Then I played the cassette transfer against the first few minutes of the DAT to assess their relative sound quality. This is where the story gets interesting. The cassette played at a slower speed and was riddled with dropouts and anomalies. The DAT sounded ‘Chipmunk-y’ and frenetic, which arguably added to the excitement and danceability of this recording and, importantly, matched the mp3s that have circulated on the internet among Ata Kak fans. Nevertheless, this presented a dilemma, the first of many: which version was correct?

This was not a judgement call I could make alone. Brian and I had long conversations about what Ata Kak intended during recording and about what the Awesome Tapes label fans expected. Who was the target audience for this release? Which version were we music archivists beholden to preserve?

Because I’m a people pleaser by nature, I argued for both versions. I wanted to please the fans who would want to dance to the hyped-up, sped-up version, and I wanted to please the purists who would want to hear Ata Kak as he was recorded in the studio in the 1990s. And I wanted to please the music journalists who would be really into the crazy story of how this album came to exist in two speeds, and would launch debates about which version was more authentic. Meanwhile, Brian was trying to track down a better source, and on the off-chance his own cassette copy sounded better – the very one that started this odyssey when he bought in a Ghanaian record shop – he decided to FedEx it to me from a DJ gig in Australia. This is the life of a restorationist.

I transferred Brian’s cassette. Its speed matched the DAT – the sped up version – but, even after a decade of being played at DJ gigs, it had a far more consistent and full sound than the slower cassette. I decided to use Brian’s cassette as my source, but I planned to slow it down to match the first cassette, which appeared to be closest to what happened at the moment of recording, and thus – here’s that word again – most authentic.

But how did I come to make the call that the slower cassette was more “accurate” to the moment of recording? The songs were synthesizer-heavy, so, assuming the synths were tuned to 440Hz, I nudged the speed in that direction using Celemony’s speed correction software, Capstan (very expensive, but, fortunately, rentable). I also used Capstan to correct the warbles and wow and flutter that are inherent in a well-played twenty-year-old cassette. I played this new slowed down version against the slow cassette for comparison. I relied on my ears, trying to pinpoint the moment when Ata Kak’s vocals sounded most natural. Even as we rely on steady electronic signposts like tuning to 440Hz or locating the 60-cycle hum existent in every (US/Canadian) electrical recording (in Ghana it’d be 50-cycles), deciding what’s “right” is necessarily a subjective art and decidedly not a forensic science.

Slowed down, it sounded accurate but aesthetically wrong — and here’s where this recording’s noise became integral to rendering it in a way that located it properly in historical time while remaining satisfying and listenable.  Varispeeding the recording to a slower speed warped the cassette’s lo-fi-ness further, highlighted sonic artifacts and anomalies (noises), and effectively transformed its noise from inherently analog artifacts that locate the recording in time into obviously digitally modified sounds that pointed to the hand of the digital restoration work itself. (For example, a single channel dropout, seamlessly corrected using the interpolation tool in iZotope’s Spectral Repair, sounded artifacty and choppy when slowed down.) I had corrected the speed of the music but lost the vibe of the recording.

I thought back to the Attali quotes, and about how history changes the relationship between music and noise. Unwanted noise almost always sounds and feels better than artificial artifacts, even if it’s barely perceptible. The noise that was in the artificially sped up recording had transformed, and in its new context, the sped up version sounded and felt right.

Meanwhile Brian had tracked down a third cassette copy a few blocks away in NYC, but the owner seemed reluctant to lend it. He transferred it himself and sent me the files, but they were of no better quality than Brian’s copy. A fourth cassette copy surfaced, but its owner was out of the country and couldn’t retrieve it until spring of 2015, which would postpone release of this record more than a year. At this point, I had been working on Ata Kak on and off for months. Sometimes in audio restoration, you have to use what you have.

Finally, I returned to Brian’s cassette copy, the one I had digitized eight months ago, to remaster the sped up version for release. Brian had consulted with Ata Kak, and together, we decided the sped up DAT version (safely archived on Brian’s cassette), while most likely different from the original recording, was the version the fans would want, and, more importantly, was of superior sound quality.

Not having to varispeed the cassette made the job much easier, though I still used Capstan to correct wow and flutter issues. Then, I attempted to adjust severe stereo imaging problems. These could have emerged at any number of points during this recording’s lifespan – unusual panning choices during mixing, poor alignment during cassette manufacturing, damage to the cassette during its 20 year lifespan. The lead vocals leaned far to the right, and the high end of the sides was out of phase. When I listened in mono, the sides were barely audible. I pulled out every tool in my arsenal to realign this image: mid-side processing, using the Waves S1 plugin to draw the lead vocals to the center, simply lowering the volume of the right channel and panning it more toward center. After hours of trial and error, I finally cracked it. The simplest solution gave the best results (though, to be fair, this is not always the case). I separated the left and right channels into two mono tracks, zoomed in as far as I could, and manually nudged the right channel of each song until the phase aligned with the left. The stereo image snapped into focus. When I listened in mono, the high end snapped to the center instead of disappearing. The lesson here: when there’s a sonic problem, deal with the fundamentals before wielding powerful digital audio restoration or imaging tools.

The Chipmunk-y lead vocals, however, were terribly distorted during the recording process. I wanted them to come through without the piercing crunch of distortion – again, an aesthetic decision made within the context of the noise inherent in this recording, and here I wanted less noise. I used a simple EQ technique I call stabbing, which is pretty much what it sounds like. Using a digital EQ plugin, I set the narrowest Q possible and swept the frequencies for the spots where there is a build up of audible noise, then I stabbed them out. It’s like popping a balloon. I used this technique to find balance in the high end too. Cassettes are hissy, but a dehisser or a broad EQ shelf can pull the liveliness and energy out of a recording. On the Ata Kak recording, I used a push and pull technique – adding a gentle, broad swath of high end to capture energy and minimize wooliness, then stabbing out the frequencies that hissed or sizzled too much until I reached that happy medium. There’s an analogy in painting or sculpting or architecture – the process of selectively adding and subtracting until balance is achieved.

After correcting these sonic anomalies, the rest of the remaster/restoration fell into place. It was sent to manufacturing, only eight months after I received that package from Ghana.

The process of remastering Ata Kak was largely a matter of determining which version to release and correcting sonic anomalies related to a combination of deteriorated source material and problems during initial recording. In drawing out the division between noise and music, I thought a lot about the moment of creation, Ata Kak in the studio, and the decades that followed, his recording’s slow burn toward cult popularity. Fans seemed drawn toward the energy of the sped-up version. Obaa Sima always pulled crowds to the dance floor when Brian played it at clubs. Whether the sped-up version was a deliberate aesthetic choice during engineering or a manufacturing error (and it is still not clear how it happened), it is now Obaa Sima’s historical and cultural context. Noise, distortion and anomalies born of medium degradation contribute to the recording in unexpected, and unavoidable, ways that we restorationists have to grapple with on an aesthetic and historical level.  There is no “simple clean up job,” no matter how elementary the technical clean up itself might be, because noise and anomaly have come to matter too much.

With any remastering project, I consider the condition of the physical media, the anticipated audience, and the cultural moment of the original recording. I think about how these factors have imprinted noise onto the music, what aspects of that noise inform and add to the context and vibe of the recording, and what elements distract from the the heart of the music. Yes, it’s a lot to think about when I’m sitting in the studio with a stack of dusty tapes that I have to turn into commercial digital audio products. But I believe it’s important to approach restoration projects with conscious deliberation in order to hear what exists now and be able to imagine what existed at the moment of recording; to extrapolate what could or should have existed when someone hit record; to take what I have from the existing source and transform and translate it into a digital audio file that communicates the energy and creativity of a recording across decades, continents and physical and digital media.

I use a lot of brilliantly designed digital audio restoration tools in my studio: Celemony’s Capstan, iZotope RX, CEDAR, methods of equalization, compression and limiting. But all the technological wizardry that allows audio engineers to correct speed fluctuations and extract clicks and pops – these are all still just tools. The art of audio restoration lies in perceiving beyond what is audible and imagining what the sound could and should be. Only then can I engage those tools to excavate the music, etch and dust away the noise — or at least some of it. I know when I’m done when I realize I’ve stopped thinking about the tools, and about the division of noise and music, and I’m lost in the beauty of the recording.


Author: Jessica Thompson
Publish Date: 11-15-14
First Publication: Pink Noise

Ata Kak’s Obaa Sima will be reissued for the first time on Awesome Tapes From Africa in January of 2015.

Jessica Thompson is the Audio Restoration & Mastering Engineer at The Magic Shop in New York City. She has restored and remastered historic live recordings from the Newport Jazz Festival, Bottom Line and Caffe Lena Archives, and has digitized, restored and remastered rare vinyl, cassette and analog tape recordings for Awesome Tapes From Africa, Jeff Buckley and Woody Guthrie. Her remaster of Mickey Newbury’s seminal albums, An American Trilogy, was Mojo Magazine’s Reissue of the Year, 2011. Jessica holds a B.A. in Anthropology from Wesleyan University and an M.A. in Media Studies from New School University.