Plangent Processes offers a unique playback system that corrects for the wow and flutter present in the master tape of a recording, to provide speed stabilization. It’s a combination hardware and software suite that, as their website explains, “combines state-of-the-art analog electronics with unique digital signal processing.” The system employs a custom tape playback head and preamp, followed by proprietary DSP that provides speed stabilization and wow and flutter correction.
Jamie Howarth, President of Plangent Processes (photo above) gave a demo and a seminar about the Plangent system at AXPONA 2024. I was highly impressed and wanted to learn more, and talked with Jamie recently.
Frank Doris: Can you explain the Plangent process? What gave you the idea for it?
Jamie Howarth: Well, the germ of the idea was somewhat accidental, but pretty obvious in retrospect. I’m fortunate to have solid absolute pitch, so, records with wow, and warped or off-center records, drove me batsh*t since I first began listening to 45s in the Pleistocene Era.
I got hired in New York 1980 by a great legend named Howie Schwartz, a highly successful jingle studio owner. He had opened a brand-new top quality facility I had helped install and he wanted to compare the tapes of the day. He had the resources: the studio maintenance shop had some very early extremely sophisticated spectrum analyzers/ FFT (Fast-Fourier Transform) analyzers, and digital frequency oscillators. This stuff cost $50,000, $60,000 bucks, and now they’re free plug-ins. But back then this kind of analysis had never been done before at the studio level. Howie was jazzed to prove he had the best taste in tape formulations.
So, I'm testing a variety of tape stocks, and ironically, I found what that I thought was the best – and not his favorite 3M 250 – or the most popular Ampex 456. Probably not the smartest move but I was too young to know he wouldn’t be thrilled to hear that the most linear tape was Agfa 468. I learned a lot about tape compression and the effect it had on the sound of rock recordings, and why 456 was favored – in the nature of its overload characteristics, it de-essed a lot. But another more hidden and surprising aspect of the sound was the distortion caused by the tape transport. One of the newer MCI machines sounded rough off tape, and it turned out to be a bad bearing causing the distortion – causing fast flutter that acted as intermodulation distortion. IM is the result of sum and difference beat frequencies with no harmonic relation to the music, and it sounded nasty with this brand-new, highly-advanced-design tape recorder. When the mechanical bearing was replaced the machine sounded good. The mechanics of the transport influenced the audio quality.
With these test playbacks I could always tell that it was not the same as the original, even on an extremely good machine – in this case a top-notch brand-new MCI JH110B with extremely tight digital servo control. It had amazing long-term speed stability, but still something wasn't quite right. When we looked at a frequency synthesis input we saw a perfect spike in the FFT, a perfect vertical line at one frequency. And when it came back off the tape it looked like a shimmering pine tree sliding sharp and flat – varying frequencies… and that was the machine’s speed sliding back and forth, AKA wow and flutter. Rather than a simple percentage metric the speed performance was starkly rendered graphically by the new digital test tools that made their debut back then.
As good as the servo of a modern tape machine (or many turntables) is, it’s always trying to catch up with itself, and it often is too slow to correct the faster mechanical variations in the machinery. It's never perfect. There's a phase lock loop in there that takes a moment to deduce and correct the speed. So, it's always trying to catch its own tail – some measurable transport flutter even in a servo controlled machine like the JH110 is still rather high – the servo itself jitters and under/overshoots, which causes a very fast flutter >50 Hz in rate. We didn’t think of flutter [as being] that fast; it was defined at lower frequencies as an obvious warble, much slower. Instead, via the faster flutters that were too high in repetition rate to be perceived as that familiar “gargle,” the transport was causing distortion, intermodulation distortion to be exact. IM was the province of electronic circuit performance, not considered as a problem in the tape transport. And yet there it was.
This was in 1980, about the same time that digital audio was just beginning to become possible. The digital tools had only recently become available to carefully analyze a piece of analog like a tape recorder. But if you can do that, then of course, why bother, just record the audio itself digitally. So, these advanced techniques for analyzing the machines were relevant, but only for a very short period of time. Most of the machines had been designed at least 10 to 15 years earlier, even the best ones back then in 1980. And still, to this day, one of my gripes is that the tape machines that are still revered – the Ampex ATR102 and Studer A80 and 820 – and exclusively employed in archive and reissue remastering - are great in and of themselves, but they're also 1975 designs. There’s resistance to the idea they can be improved, even 50+ years later, because to the older engineers there’s simply no concept that they could be faulted in any way.
I was triggered by the revelation that there was so much pitch and speed deviation going on, and it was visible, and I could adjust the servo of the MCI to minimize it, but I could never get rid of all of it. And I saw some patterns in it that were kind of odd. You'd see not only the big “pine tree,” but also some little tiny “pine trees” on either side of the main fundamental tone, symmetrical little deviations. I didn't know what that was until much later, but I saw it. Those are actually the beat frequencies that are caused by the flutter modulating the fundamental. They are the intermodulation products, sidebands straddling the fundamental signal. Imagine the motor cogging at 60 Hz. Put in an A 440 tone and you’ll get back 380, 440, and 500 Hz. That’s not gonna sound good. And it’s not that specific – there are a myriad of provoking causes and it’s not on A 440, it’s on everything. So what’s the fix?
And [one day], I was working with 30 ips (inches per second) tape, and I was scrubbing (manually rocking) the tape back and forth just to make edits and stuff. And I was hearing this whistle as I scrubbed the tape, and asked one of the older guys what it was and he said, “that's the bias.” That's the signal that is recorded on the tape in order to be able to make the magnetic materials respond pretty much linearly. Basically, the tape is blasted with a very-high-frequency signal way above the audio, which serves to pre-condition the magnetic domains of the tape so that the oxide will respond linearly.
AC tape biasing has been around since the very late 1930s. One of the perhaps apocryphal, but perhaps true stories was that although the nonlinearity problem was known, no one knew what to do about it. As legend has it an experimental German Magnetophon design accidentally picked up a radio station from a nearby ship-to-shore radio transmitter broadcast leaking through a cold solder joint, and it ended up bleeding at a high level onto the recording, which suddenly sounded fabulous, and they couldn't figure out why. They went back and corrected everything they could see that was wrong with the circuit, and they lost it. They went back and realized the trail went cold when they repaired this solder joint: a cold solder joint was acting like a crystal detector, like a rectifier, and the result was the carrier of the radio station being heavily amplified and blasted onto the tape. Well, there was a good illustration of what was among the first uses of AC bias.
FD: I wonder how long it would've taken somebody to discover that if that accident hadn't happened.
JH: There's a great book by a guy named Thomas Kuhn called The Structure of Scientific Revolutions. It speaks to the fact that simultaneous invention is a thing, and stuff tends to come along in waves of consciousness. We’re pre-conscious about the art of science, if you will, and that's going on in various places at the same time. So in the late 1930s or early 1940s a US inventor, Marvin Camras – an early designer of wire recorders – figured this would work. Another guy in England came up with a similar idea.
Back to scrubbing the 30 ips tape at low speed and the bias oscillator in this MCI recorder: I was hearing the ultra-high-frequency signal and realized that it's got an actual wavering pitch to it. Then I looked at the schematic; It was very clear that the source of this high-frequency tone in the MCI was a quartz oscillator, with a frequency of parts per million, and yet the machine couldn't [reproduce] anything like parts per million mechanically. So, what I realized was that the recording of the fixed bias oscillator after it was recorded reflected a perfect pattern of the speed variations on the tape.
In other words, the playback variations in what was originally a fixed oscillator are frequency deviations – created solely by the mechanical errors of the machine – and that turns out to be exactly proportional to the tempo and pitch variations in the music – the wow and flutter in the audio is imprinted in the frequency modulations on the bias. For example, if the tape speed varied by a semitone then so did the frequency of the recorded bias – exactly.
I asked my mentor, the late Dave Smith, Phil Ramone’s chief engineer, a much beloved genius whom I was fortunate to have worked for at the Hit Factory and elsewhere: “Why don't they use the bias playback as the servo source, rather than trying to measure the machine's speed indirectly with an optical sensor or something? Why don't we just use the actual recording?” He thought this was a distinct possibility, but others scoffed that it was too high a frequency: it clocks in at anywhere from 50 kHz to 432 kHz. “You'll never be able to play it back,” they said. Well, I was hearing it at low speed, so I dunno fellas, it's playing back!
The idea sat in the shadows for a while. I worked for about 15 years in the TV industry, not in the record biz, but I always had this idea in the back of my head… David and I talked about this again when I left the ABC network in 2000. He said, “don't try to control the tape machine. You're going to be back to chasing your tail because you're at the mercy of the servo system.” He said, “just do this in DSP (software).” So I took a chance and hired smart guys and built it.
The Plangent Processes system is an FM detector chasing the speed of the original recording, followed by a DSP pitch shifter; a very, very, high-precision pitch shifter. It operates internally at a minimum high sampling rate of 24 MHz/32 bit sampling rate and it's rendered, not real-time. This minimizes any potential artifacts coming from the fact that the musical material is constantly being resampled; the patented system’s pitch shifting is done with an irregularly-shaped resampling algorithm, aka weird math. Don’t ask.
That technology was barely available in 2003, when I went to the AES convention in Amsterdam and I ran into a young fellow, Patrick J. Wolfe, who was studying for his PhD with Simon Godsill – who among other things had developed the CEDAR noise-reduction system. I asked him to write up a software routine that looks at a varying ultrasonic frequency signal coming in and translates that into a pitch map or a drive mechanism, if you will, a servo drive mechanism that's done digitally, followed by a sophisticated time warping routine to reverse the wow and flutter in the recording. And he said, yeah, I can do that. And it worked on the first try. I disrupted the tape path with a pencil eraser and squiggled the audio horribly, and what came out of Patrick’s laptop 30 minutes later was perfect. Next try I couldn’t get the eraser in, so some “unwowed” audio was rendered. THAT sounded way better than the original source – no more bees and sidetones in the clarinet. “Why? “Patrick said, “look at the frequency modulation before the processing. It has all these sidebands and we’re removing them when we take out the FM.” That’s the IM, the “little pine trees” – beat products caused by the mechanism.
Sample of "Born in the USA," FFT spectrum of drift, wow and flutter with sidebands on a 240 kHz bias signal, Studer A80 tape deck.
The same sample after being corrected with Plangent Processes: all speed-related issues are resolved.
FD: So, basically you had the idea a long time ago, but the technology wasn't available to implement.
JH: Correct; it was conceptualized long before it could be achieved. One challenge is that tape heads are designed such that they'll give you a lot of audio output, but they'll roll off at something like 30 kilohertz. That's the obstacle – the recording [of the high-frequency signal] is definitely there, but the playback of it was impossible. Only NASA-style instrumentation recorders could record and play back such very high-frequency material and they sucked for audio. In fact, when we were first doing this, John French, the unsung hero of the audio business who’s designed or installed practically every replacement magnetic tape head that's ever been sold in the United States, loaned us some rare instrumentation heads that would actually play back the bias, and still work in the audio spectrum.
Then we had to do a fair amount of work on the preamp in order to be able to retrieve a signal that's minus 70 dB [down] off of the tape. The technical challenges were [solved] by two guys. One was Dave Smith, and another unsung hero – John K. Chester, who's been by my side with all this stuff all the way through after David tragically was lost in 2005. John started as a front of house guy at the Fillmore East and was on the audio team at Woodstock. They were able to develop an extraordinarily low noise, wideband tape system with not only this trick capability but also a substantial update to the analog electronics of tape playback, where, as stated earlier, not much had been done since 1975. So it’s current-day electronics, with this additional purpose. As a piece of analog gear, it’s superb, and modern.
FD: When you digitize the audio signal, aside from the processing for the pitch correction, what is the sampling rate and bit rate?
JH: 384 kHz/32-bit, currently is tops. We've got an RME analog to digital converter that'll do 768 kHz, but we don't prefer it for audio, at least not yet. We employ both Mytek and Prism ADCs.
FD: What is the result of the Plangent Processes system? What are some of the things to listen for?
JH: What I would suggest is, don't get too hung up on the fact that it’s “digital.” Don’t even necessarily think “audio”…pitch and tempo are musical performance criteria, and we find that while the sonic quality is certainly advanced, there’s definitely a positive effect on recapitulating the original performance, which is what high fidelity is all about. There’s an authority that comes through – and more personality. More realness. What they played.
We recently worked on Phil Ramone’s original A&R Recording masters of Getz/Gilberto for an IMPEX SACD, and the you-are-there factor is stunning. “The Girl From Ipanema” players and singers are in the room with you. There’s absolutely no timing/pitch-based abstraction layer messing with your hearing and perception of the performance and dimensionality.
Our interview with Jamie Howarth will be continued in Issue 213.
