Everything I Need to Know I Learned From the Hubble Telescope

Should a recording document reality as faithfully as possible, or should it improve upon or somehow transcend the music it records? Both, actually.

[Image Credit NASA, ESA, and the Hubble Heritage Team STScI/AURA)-ESA/Hubble Collaboration. Acknowledgement: B. Whitmore ( Space Telescope Science Institute) and James Long (ESA/Hubble).]

The topic is very much in the news. Organizations like Harvard University and public radio station WDET-FM Detroit have publicized plans to digitize their vast archives of interviews and live performances, both for preservation and for cloud accessibility. The Harvard archives include rarities like the first appearances of Joan Baez at Club 47 in Boston, so it is important to do right by the material. But what exactly constitutes a high-quality digital audio preservation process?

Consider the image from the Hubble Space Telescope above, showing the slow collision of two galaxies producing super star clusters. It's spellbinding, dramatic, but it isn't real. Space doesn't look like this.

Hubble images are triumphs of sensor fusion. They are made, not born. Images must be woven together using the incoming data from cameras in many different spectral bands, cleaned up and given colors that bring out features that eyes would otherwise miss. The images illustrate the wonders of the universe rather than documenting them.

[I wonder if the sound subjectivists who believe that lossy compressed audio is “just fine” also dismiss space exploration with sophisticated instruments, because they can see stars “just fine” with the naked eye.]

It may seem quite a leap, but a strong analogy exists between the production process for a Hubble image and a high-quality digital restoration process for analog source material. There are three fundamental steps in both:

• Data Collection. The dramatic Hubble images have their source in high-resolution instrument data; do no less with your source audio.
• Sensor Fusion. Use software tools to analyse the audio from multiple perpectives to execute operations to improve audio quality.
• Presentation. Render the results in formats accessible to the intended audience(s).

When the process is complete, two products have been created. First, there is an archival recording that faithfully transfers the original analog recording, in all its flaws, to digital form. Also, there is a restored image of that recording as processed sound files that transcend the moment to reveal what may not have been originally apparent to the naked ear.

But how do you design these steps to achieve a high-quality result? There are five fundamentals:

• Accuracy, with regard to the source material
• Appropriateness and Adequacy, with regard to your processes
• Consistency and Explicitness, with regard to your execution

In an upcoming three-part series, we’ll examine how to stay true to the five fundamentals in your project. Stay tuned.

[Originally published 2-February-2012. Revised and expanded.]

            Vinyl-to-Digital Restoration #15            



Artist: Pink Floyd
Title: The Dark Side of the Moon
Genre: Rock
Year: 1973



What other title could be more appropriate to discuss? DSotM presents a presentation challenge due to the extensive use of crossfades between tracks, making it hard to identify the instant where one track ends and the next one begins. They slowly collide with each other in passing. As with live albums, you'll be glad to have accurate high resolution sound data when confronted with a crossfade to zoom way in to make a precise cut at your decision point.

© 2012 Thomas G. Dennehy. All rights reserved.


Automated Tools Can't Fully Replace Listener Insight

Which has more impact: the power of automated software tools, or the power of human insight?

Automated software tools are the backbone of the sensor fusion step in the digital transfer process, the step that mines through your source recording to recreate the best possible audio image. However, they cannot replicate your knowledge of what your music is supposed to sound like. And only you can recognize when they have not gone far enough to recreate a full-fidelity image.


Never is this more apparent than in cases where special repairs for old damage are needed.

As you execute the final presentation step of isolating and labeling your tracks, don't speed from track start to expected track end. Examine the wave form as you proceed, looking for blemishes software tools could not correct algorithmically. These include: pops and clicks that somehow did not fit the pattern that cleanup algorithms sought, larger skips, special surface problems. All must be corrected manually.

Do you get more satisfaction out of the bang-for-your-buck from software tools and processes, or from the small hand-crafted victories?


            Vinyl-to-Digital Restoration #19          

Title: Chase the Clouds Away

Artist: Chuck Mangione
Genre: Jazz
Year: 1975

When reviewing this recording, I encountered an odd repeated sound pattern around the 13:00 mark of side one. The "comb" pattern is the hallmark of a "broken record," a permanent defect in the groove wall of an LP that causes a turntable stylus to skip back to the previous groove, endlessly repeating the same 1.8 seconds of audio. In this case, the phenomenon went on for over five minutes before the stylus somehow got jostled enough to get past the defect (otherwise it would still be recording).

You can listen to a copy of the affected passage here.

Because automated tools see nothing unusual about oft-repeated audio patterns, one would think the recording is ruined. But a hands-on matching technique was employed to save it. By matching, I mean the sound wave just before the throwaway audio repeats is matched to the wave just after, cutting out the sound in between to effect a smooth transition.

Judge for yourself the quality of the audio match here. If you are interested in the details of the technique, a step-by-step guide is available.

© 2012 Thomas G. Dennehy. All rights reserved.