Vintage Technologies: The Eltro and the Voice of HAL

The Eltro Mark II  "Information Rate Changer"  As Monty Python might say: "Now for something completely different." In the course of human history what I'm about to describe amounts, at most, to a bit of amusing trivia. But it's a trivia I was around to witness by improbable connections, mentioned below. If you would, please join with me as we set the way-back machine to over four decades ago, sometime in the mid 60s. The story begins with an idea then making the audio rounds, one which was an offshoot of the video tape recorder. Most VTRs at the time used something called a rotating head drum. It was a novel way to obtain a very fast head-to-tape speed, as the head gaps whirled along narrow tracks angled on the tape. Meanwhile, the tape itself chugged along at a regular modest speed. That trick gave us the high frequencies video requires, without running out of tape every 5-10 minutes (pretty clever). Home VCRs yet to come would use the very same principle.   But what if the heads followed in the same path as the tape motion? Say you mounted four of them on a small upright cylinder, and could rotate this "drum" in either direction at any speed. Before the tape begins to move those heads could repeatedly "scan" a short section of audio track and unendingly play the sound at that one spot. Start the tape moving, keep spining the heads against the direction of the tape's motion. The relative head-to-tape speed would now be faster than normal, shifting the pitch of the audio upwards. But the length of tape would take the same amount of time to play. Which some audio people pointed out would give us a nice way to increase the pitch of a recording, without changing its duration!   Instead rotate the head drum in the opposite direction, so the gaps move with the flow of the tape, as if trying to catch up with it, like cars passing on a highway move apart rather slowly. The relative head to tape speed in this case would be slower than normal. Yet again, the same length of tape would take the usual amount of time to playback. Result: you'd decrease the pitch of a recording, without changing its duration. Simple and clever, no?   Finally, let's install a small differential gearing to couple the capstan (which drives the tape) to the rotation of the head drum. If you did it correctly, you could maintain a constant head-to-tape speed, no matter how slow or fast the tape itself moved. You'd have a practical way to alter the duration of a recording, without altering its pitch (i.e., time compression / expansion). And that, in a proverbial nutshell, is exactly what the Eltro device we're about to talk about, did! (BTW: I'd like especially to thank writer Dave Tompkins for his speech processing feedback that lead to this new page, for reasons about to become clearer below.) Actor Douglas Rain, 1968  So how does this concern us? Well, if you've ever watched Kubrick's classic SF film, "2001 -- A Space Odyssey", you've heard the Eltro in action(!). It's a story I don't think has ever been told before now -- here's how it happened. The sentient computer in Arthur C. Clarke's story is called HAL (and no, Clarke did not plan the name to be the three letters of IBM shifted one place to the left -- that myth overlooks an even lovelier example of serendipitous coincidence!). In the film the role was voiced by Canadian actor, Douglas Rain, who was able to give a cool, detached -- yet feelingful duality to the character. Here's a publicity photo of Rain taken from around that time.   During the scene in which Dave (Keir Dullea) "lobotomizes" HAL, you'll easily hear how the tempo of Rain's voice becomes slowly expanded and pitch-shifted gradually downwards. Actually, his entire performance as HAL has a mild amount of time stretching (no alteration of pitch) going on, as Stanley confided to me. I told him I hadn't noticed it before, and he smiled: "it was about 10-20%, rather subtle." But that was enough to enhance Rain's performance with a slightly more measured quality. It's in the final HAL scene that the Eltro effect is cranked way up. "We did that in two passes", Kubrick quietly explained. One pass gradually dropped HALs pitch down to almost zero, remaining at a constant speed. The other pass gradually stretched it out in time, but not as extreme, as HAL sang "Daisy, Daisy" (Bicycle Built For Two by Harry Dacre). And indeed, you couldn't do this simply by slowing down a regular tape recording, as many pundits have since wrongly guessed (to reach the final low pitch, the tempo would crawl to a near-stop). Music From Math LP (click for larger view) Trivial pursuit: this is the same song HAL sings that Max Matthews (with Joan Miller) originally programmed several years earlier (using Music IV) on an IBM 7094, probably the first example of a "singing computer." Clarke heard it during a visit to Bell Labs, Murray Hill, as he was friends with James Pierce, another member of the digital synthesis team. So that particular song in 2001 became also a sly reference to one track of the 1961 demonstration disk (LP cover and liner note seen above, yes, in fake stereo...), which is where I first heard it. (Impressive in '61, you'd find it sounds "cheesy" today, like an early MacinTalk.) Previous knowledge of that track in turn made the HAL "death" scene simultaneously appalling / amusing... Hey, I also laughed out loud in the theater when the too-familiar strains of Strauss' "Zarathustra" began the film. I thought it was meant as a wink-nudge over-the-top joke, as I'd used it for such comic effect in a student radio parody in grad school. But no, I discovered (with a discomforting blush) that Kubrick didn't share my insider's POV, and took it completely seriously, as now do millions of fans who were first exposed to the fanfare by the film.

(Top of the Page)

Studio Tales  Top Panel (click for larger view)  When I first experienced 2001 (in the huge Cinerama theater on Broadway), I guessed that the effect of HAL dying simply had to have been done on an Eltro machine, or a close copy of one. By absurd coincidence, I was an engineer in NYC who may have had the most experience with an original Mark II, at Herb Moss's Gotham Recording Studios, now long gone. (As is Gotham Audio, the separate company which sold the Eltro to Moss's studio.) When I finally met him in August '71, Kubrick admitted they'd used one on HAL. Stanley had inquired around the recording and film mixing studios there, until someone told him about the Eltro Information Rate Changer. (Yes, that's what they called it -- look at the pdf file down below.) So he rented one of only a couple in England, just for this purpose. You can see an Eltro Mark II in the photo above, a surprisingly handsome device for a piece of audio gear, with both matte and polished chrome finishes, two stylishly fluted massive knobs, down to the sturdy real oak wooden case/cabinet. You may note the two narrow brownish ribbons of tape at the very top, which navigate between the Eltro and a host tape recorder. A clear plexiglas top cover kept out dust, and was hinged, to be lifted to one side for easy access. You could also unclip it completely and set it aside if you were working in haste on a complicated job (the device didn't make much noise).   Inside the cabinet the mechanism was well machined (wish I had a photo). A small Nortronics/Viking of Minneapolis tape playback electronics box sat in the lower left corner and provided the main mono output. The two small knobs jutting out beside the output XLR provided Volume and Tone adjustments (the latter being rather an anomaly on pro audio equipment, even back then). I laughed when I first saw it, since I'd been using Viking equipment for about a dozen years by then, for my early home electronic music projects. A small "muffin" type fan inside kept everything cool. You can see the air exhaust slots to the far left of the case, above. The right top knob is rotated to select the function: Off, Pitch (using the rotating head drum alone -- the host tape transport just behind provided tape drive and rewind / fast forward), or Tempo (the Eltro's own capstan and pressure roller then moved the tape). That third setting, equivalent to Time Compression or Expansion, is what 95% of the people who came to the studio used it for.   Gotham Recording initially bought their Eltro to process audio/visual learning tapes for a project they were contracted for by a Long Island company: EDL (Educational Development Labs). As a novice recording engineer hired to edit tapes for this project (long story -- another time), and a bit of a tech-head myself, I became quickly involved with their new Eltro. Gotham's Helena Sterling (what a wonderful boss) inaugurated a series of EDL "Speeded Listening" tapes, for children in primary school. The "speeded" part is where time compression comes in. We clocked the narrators who recorded the lessons to maintain a constant normal word per minute rate. Then for the speeded listening sections, we ran the masters through the Eltro at precise settings to achieve graded, accurate, incremental faster speeds. And the results, synchronized to the EDL film strip projector, was used to train young people to listen more astutely, maintaining comprehension (that was the idea, anyway...). Control Knob CU (click for larger view)  The large knob on the left side of the Eltro controlled the motor drive speed, using calibrated dial positions. If you chose only pitch change, this knob merely varied the rotating head drum speed, and the tape was pulled through the unit by the large "host" Ampex 300 tape machine, not shown (someone constructed a small wooden table on which to set the Eltro so that its top plate was exactly the same height as the Ampex 300 top plate). If you went the full nine yards, you'd set this same knob to the percentage of time compression or expansion you desired. And now the tape would move either faster or slower than normal, while the head drum maintained the same effective head-to-tape speed. I don't have a high res photo, but have tried to sharpen this CU so you can make out most of both the settings -- percentages for duration changes, and musical intervals (major/minor second, third, and so forth), for pitch changes.   There was something convenient about being able to twist the speed/pitch control any time at will, nudging it quickly or slowly. The newer devices which replaced the Eltro, mentioned below, can usually be set only when they're not operating. With a popup menu, say. All well and good and accurate, but like most software emulators, you feel like you're disconnected from immediate control. Bummer. And the Eltro, as I discovered while playing around, could be used in ways not intended, again something you probably couldn't do within the circumscribed parameters of new digital implementations. As an example, I fed a blank tape through the host Ampex, looped it through the Eltro in pitch mode, and back to the takeup reel, and pressed record. Then I sent a small sound into the Ampex's input, while mixing the output of the Eltro along with it. Sure enough, in a stately instance of good-old tape-feedback, the signal returned and looped around, gradually getting lower OR higher in pitch. Kewl. A nice way to create the effect of a flying saucer taking off or landing, perhaps, or some gorgeous background whooshes, whirrings and musical twirlings? (Old synthesists never die, they just whirr away...) The Rotating Heads (click for larger view)  Beneath the 2-1-2" diameter head drum cap (which acts as magnetic and dust shield) the heart of the system, the head drum, is exposed. How the hell did they manufacture such a complicated, tiny physical structure? The cover could be lifted and swiveled on its support rod, to access the heads for threading tape, and for cleaning the drum (it tended to get more covered by tape oxide than standard heads, due to the path and tension). I was curious enough to bring a magnifying glass to work, so I could examine the head gaps. Yup, there were four of them there all right, not easily seen. You'll notice the two tape guides on either side. Those were actually part of a critical adjustment. They were drilled off-centered to their support screws. So you could loosen the guides with an Allen's wrench, and swivel them slightly. That way the nominal 90 degree angle the tape wrapped around the head drum could be increased or decreased, for best sounding results. A greater angle gave more of an overlap as one gap pulled away and the next one came in contact, and vice versa. You eventually learned where the optimum settings were for different material: speech, music, sound effects.   While most of the time I made the speeding listening transfers for the EDL project, every week or so we'd have other clients come in to use our magic box. Most often it was to change a program that ran too long (or too short) so it would fit the alloted time duration. Radio and TV commercials were often brought in for "time tweaking", to fit the :10, :20, :30, or :60 second standard spot durations exactly. I never mentioned it, but I found it ironic that actually they were repeatedly throwing out many-many itty-bitty bits of their recordings, cross-fading the remaining ones together, which is how we could effectively shorten the duration without changing the pitch on the tape itself. Such time constraint problems obviously still occur today, but we generally now solve them on a computer. And yes, you still end up tossing -- or repeating -- a multitude of micro snippets of the original -- that's just how time compression and expansion are done!   Anyway, a decade or so after the Eltro most of these jobs were done using digital boxes like what my friend, Ricky Factor (of Eventide Clockworks) designed and dubbed: a "Harmonizer." You may have heard of (or even used) one of them. Later on, for my own time compression and expansion and pitch changing tasks, I got a (British) AMS dmx-15-80s, which had perhaps the least audible artifacts to pitch shifting available at that time (Ricky's newest "Ultra" Harmonizers are even better). Such rack mounts made tempo changing a bit more complicated than using an Eltro. You'd have to vary the tape recorder's speed with one device (a VSO, or variable speed oscillator), and then bring the pitch back to normal with the processor box. But it worked nicely, indeed, and sounded better than the Eltro (no head-crossover constant-rate flutter). Plus it could do it in Stereo! In another dozen years Sound Designer was released, and the familiar DAW (digital audio workstation) paradigm became standard for high quality audio. That's still the platform on which such audio processing tasks are performed. Nevertheless (*sniff*), long live the Eltro -- the device which started it all!   Postscript: as I tried to describe the events above, more memories of working at Gotham Recording Studios (located near the corner of West 46th Street and Fifth Avenue) flooded back, first time in many, many years. So it was probably worth it to create this unimportant new page. And what may not be clear above is a perception that required the more mature perspective of time. You see, the PEOPLE in that studio were really a special, supportive group, the kind you find in a large city. They were men and women of all ages and sizes, a kaleidoscope of nationalities, personalities, races, and skill sets. I've mentioned Herb Moss and Helena Sterling above. I also think fondly of patient Joe Smith, generous Dio (his last name escapes me, but he played a pretty mean pipe organ at a large church in Harlem), endearing Martha Glass, all of whom worked in the marketing and shipping area. I met pros like Arnie Rosen and chief engineer, Jack Wallace, his assistants: Jack Franz, good friend Bob Schwarz, and the man who wired my recording console, Donald Longmore. Jay Windwer was Gotham's soft-spoken manager who "tried to herd cats" with our eclectic group of recording engineers (I was a novice member of the team). He was always fair, diplomatic and knowledgeable. A few years ago (through this website) I heard again from Jay, and also from Eric Toline, another capable, amiable former-Gotham engineer. I miss you all -- thanks for the memories! (I'm not generally good with names, but if I remember a few more, I'll add them here.) And now, to read, save, or print the original Eltro descriptive brochure, which was created by my late friend, Steve Temmer, of Gotham Audio (for his ISI division: "Infotronic Systems Inc."), you may access the PDF file HERE. (It's under a megabyte.) There's also a zipped version of the file HERE, choose whichever works better on your computer, with my compliments. It's a perfect way to remind ourselves that excellent, innovative technology didn't just explode into reality in the past five or ten years. There were pioneering examples of great ideas that arrived on the scene much too early to hit the serendipity of "an idea whose time has come." Even the Mark II we're discussing here had its own precursors: there existed a few related rotational playback head prototypes in Germany from before WWII and again in the 50s and early 60s (I assume leading to a Mark I...).   It's true, the newest pitch/tempo tools that anyone can run on a modern computer can generally run rings around those venerable electromechanical devices like the Eltro. But that all being said, there really was something quite special and fun about working with, handling, adjusting, these early physical tools, especially when so beautifully crafted. That's something that obviously doesn't exist with software emulations -- different paradigms. Such thoughts inevitably take me back to a time when mature versions of exactly the same principles embodied by these brilliant, intricate hardware tools, could be found only in a daydream, something that wouldn't exist for decades to come... --Wendy Carlos, New York City, July 2008 New images, scans, pdfs and texts Copyright 2008 Wendy Carlos and Serendip LLC.

(Top of the Page)

Back to the Wendy Carlos Home Page