1) From Steel Wire to CD album to Grammy Nomination

Last December, the Recording Academy released a list of nominees for the 2008 Grammy Awards. One of the more obscure categories covers released albums of historical significance. Among the nominees this year was a release of a recently discovered live concert recording of Woody Guthrie from Fuld Hall in Newark, NJ in 1949. This album, The Live Wire, carried significant historical importance, because of Woody Guthrie’s role as one of the key figures in folk music, and since he served as an inspiration for many modern performers, often cited to include Bob Dylan, Bruce Springsteen, Billy Bragg and Wilco. The album’s release also included great technical challenges, since it came from a wire recording. Wire recordings were made on spools of steel wire that resemble the sorts of things that one used to find in old-fashioned hardware stores. Many people do not even realize that wire recordings were ever used, but they enjoyed a very brief period of popularity just after World War II, ending pretty soon after the introduction of magnetic tape around 1950. This article will discuss the technical efforts that were made to restore the recording to a quality level that would be acceptable for today’s digital technology, and will focus specifically on the challenges imposed by the wire medium, the speed variations in the original recording device, and the problems caused by the wire itself.

2) Quick synopsis of the re-discovery and restoration

Before getting into the technical details, here is a brief history of how this wire recording became a CD. More information can be found at http://www.woodyguthrie.org.

The story begins when Paul Braverman, a student at Rutgers University heard about a concert Woody Guthrie was giving at Fuld Hall in Newark, NJ. He used his new wire recorder to record the concert – it was an intimate gathering moderated by Woody’s wife, Marjorie Mazia, and so there were numerous family stories recorded along with the songs. Braverman knew that Woody was an important figure and musician, so he saved the wire recording - in a box in a closet for 50 years.

Later in life he was preparing to move from his home, and while he and his daughter were clearing out the house, they found the box and decided to send it to the Woody Guthrie Foundation. This was in 2001, right after the anthrax scare, and the arrival of a heavy box in a brown paper wrapper with wires inside made for some initial consternation at the Guthrie Foundation. It took a while to realize that the wires held actual recordings.

Nora Guthrie and Jorge Arevalo Mateus (curator of the Woody Guthrie Archives) assembled a team of audio folks to work on the restoration project, including David Glasser and Charlie Pilzer at Airshow Mastering, Steve Rosenthal and Warren Russell-Smith of The Magic Shop, Art Shifrin (who converted a tape machine to be able to read the wire recordings), and Jamie Howarth and I for Plangent Processes. These other folks were primarily responsible for the transfer of the audio from the wire, the digitization, and a great deal of audio work to remove noise, bring out the voice, and restore levels. The algorithmic work discussed below came into play only after the transfer process was complete – and it was a torturous transfer process, since the wires break and tangle, tensioning was difficult, and the wire can twist away from the recorded surface.  The descriptions of the transfer session include Jamie Howarth of Plangent Processes holding his thumb against the wire as a human tensioning device and one of the Magic Shop guys working tirelessly to untangle the snarls that occurred when the wire snapped. Once the digital files were created, the algorithmic work could begin to detect and correct deficiencies in the original recording equipment.

The net result of the efforts of this dedicated team was that the “old recording crackle” was minimized, unintelligible segments became clear, pitch was correct, and the entire concert seemed brought back to life. It puts the listener almost in the room with the audience, despite the imperfections that remain.

3) Key technical challenge – fixing speed variations that lead to “wow” and “flutter”

One of the biggest problems associated with moving old recordings to CDs is that the motors, tensioning devices, ball bearings and other equipment that was used on older recording devices did not have the consistency of behavior that is expected for recordings in the digital age. Jamie Howarth of Plangent Processes spent many years working in recording studios and knew first hand that these effects were present on recordings. For many recordings, this resulted in slight speed variations of the tape (or steel wire!) used in the recording process. To understand how these inconsistencies can affect the recording, consider a couple of examples that are fairly extreme, but can be understood from experience.

Example 1 – Doppler shifted train whistle

The first example is the sound of a train whistle as it is going off in the distance. As the train moves away, the pitch of the whistle seems to drop for a listener at the station. This is the well-known Doppler shift effect, and it can be understood by thinking of how the train’s motion affects the sound waves as they travel to the listener’s ear. If the train is idle on the tracks and the whistle is blown, there is no pitch shift, but when the train is moving, the Doppler effect kicks in. The sound waves produced by the whistle are compression waves, but that means that the pressure oscillates up and down. This can be graphed as a "sine wave"; see the first diagram below. The more quickly the wave moves up and down -- i.e., the more peaks per second (frequency) -- the higher the pitch. When you and the train are motionless, then the waves encounter your ear (or your recorder) at a normal rate. If the sound of the whistle is the musical note called "standard A" the peaks will arrive at the rate of 440 each second. However, if you stay motionless, but the train approaches you, the sound wave will move toward you at a greater speed, so the peaks will arrive more often, say at 466 per second. (Picture this sine way moving leftward pretty fast.) This will cause the pitch to seem higher, in this case an A# (A sharp). In most discussions of the Doppler shift, the focus is on the speed of the train, but one can just as easily adopt the viewpoint of the listener and say that the effect is to change the rate that the sound waves get to the eardrum – hence any audio event that delays (or speeds up) the arrival of sound will change the pitch for the listener.

Now consider an example from a tape recording process that can produce a similar effect. The simplest example is when a particularly long song is recorded. The tape reel is pretty heavy, and on older equipment, the motor speeds can be impacted by the weight of the tape or the relative distribution of the tape between the take-up reel and the supply reel. This means that over the course of recording a long song, the speed of the tape can vary by slowing down or speeding up. Now, imagine what would happen if that tape were digitized on modern equipment that pulls the tape through at a more exactly consistent speed. If the original tape sped up during the recording, then the sound waves it recorded would have been “stretched out” over a greater length of tape. When digitized, though, these sound waves would be sampled consistently, and hence the pitch would appear to drop over the course of the recording. Of course, an opposite effect would occur if the original tape slowed down during the recording. You will find that there are musicians and audio engineers with “golden ears” that can detect these pitch changes in audio recordings, even on some classic tracks.

Example 2 – Cassette tape gets “eaten”

A second example that is a pretty fair analogue for the sorts of effects that were present in the Woody Guthrie wire recording, is the classic problem of having a car tape player “eat the cassette.” When this happens, the only way to save the tape is to carefully extract it, untangle and unkink it, and then rewind it onto the cassette, using something like a pencil or a pinky finger to wind the reel. It often occurs that in this process the tape gets stretched in a number of places, or it gets wound on with different tension in different places. When such a tape is played back in a cassette player, the sound often exhibits sporadic drops in pitch that can be very destructive to the listening experience. These sorts of errors in the original recording media must be corrected before the material can be used for a CD, and with material such as the wire recording of the Guthrie concert, it is a one-of-a-kind record of the event and so one can only hope to detect the recording problems and devise a means to correct them.

4) Wow and Flutter Correction

The previous examples just give everyday examples of the pitch changes that can happen. In practice, it is generally mechanical problems with the recording equipment that create the difficulties. Jamie Howarth, founder of Plangent Processes, spent many years working in recording studios and knew the ins and outs of the recording equipment at a high level of detail. What is more, he could associate noise signals he could hear with the likely sources in the mechanical apparatus of the recording device (details can be found at http://www.plangentprocesses.com/ ). Howarth postulated and later showed that it would be possible to find trace signals on the recordings that would serve as markers for the speed variations that occurred during the original recording process. These speed variations are generally called “wow” and “flutter”, and when they can be measured from the original recording, it is possible to restore the audio to a corrected state. For the Woody Guthrie work, it was possible to find a faint signal caused by the electrical supply line. This 60Hz hum proved to be the timing marker that allowed the wire recording to be stabilized. Plangent uses a patented process to do this speed correction, but in the next section some figures will be given to illustrate the process.

5) Sine waves distorted by timing errors

The simplest example of wowing can be illustrated by looking at a simple sine wave, as shown in the following figure. As should be apparent from the picture, we can consider the sine wave to be a continuous signal with no breaks if we imagine a line connecting the blue dots. In the digitization process, the sine wave is sampled and only the sampled values are stored – this leads to the PCM, or pulse-code modulated format that is the basis for CD audio. In this example, we assume that the correct, desired values from the sampling are the ones highlighted as red dots. So, on modern digital equipment, only the red dots would need to be stored for the audio signal to be properly recreated in a CD player (for more information on digital sampling, look up sampling theory associated with Shannon and Nyquist and others in Wikipedia or another source).

When the recording process has speed errors during the recording, the net result is that the recording does not sample the red samples in the figure above, rather it captures other nearby values, as depicted by the green diamonds in the next figure (and note that in this example the digital values are oscillating faster than the original sine wave, so the pitch will be higher). If these timing errors are severe enough, the listener will get the distinctive “stretched cassette tape” effect described earlier.

After detecting the speed variation errors and applying a correction factor, the restoration process can put the digital samples back where they belong, as shown by the black stars in the following figure. These black stars lie on top of the red dots that were the desired and correct digital samples.

As a final example, look at a slightly more complicated case where there are multiple oscillations occurring at once, and where the oscillations can be thought of as interfering with each other.

The next figure shows that the wowed samples that appear as green diamonds can be returned to almost exactly their original positions, where the black circles overlap the correct samples indicated by the red dots.

Summary

In summary, speed variations in recording equipment can have a strong effect on the quality and intelligibility of recorded audio. There were instances in the Woody Guthrie wire recording where the meanings of the stories or lyrics could be completely misinterpreted before the timing corrections were applied, so the improvements in the audio can sometimes be dramatic. The Plangent Processes algorithms have also played a role in improving the audio in recent releases of several movies, most recently the BluRay release of “Close Encounters of the Third Kind.” And, for those readers who are fans of the Grateful Dead, the Plangent techniques were applied to the “Cow Palace” release from last year. Look for it on more releases of old videos and classic rock albums in years to come.

In this brief article, the focus has been on the wow and flutter corrections, but mathematical algorithms are underlying all of the digital audio processing techniques that are involved in the production of an album. The noise reduction, soundstage shaping, and vocal emphasis effects that were applied by Steve Rosenthal and his team at The Magic Shop are also implemented through mathematical transforms and filters, of course guided by their expert ears and knowledge. But at its heart, the digital age of audio is built on a basis of mathematics.