Part I

Whenwe go to the movies, we go for the big theatrical experience of sightand sound. I mean, what are we paying $7 (or more) per ticket for? Wewant sound at least as good as what we get at home from our ownfinely tuned systems. After all, Speaker Builder readers knowwhat good sound should be, or at least have a very strong opinionabout it. And nowhere will you find better big-screen sound than in atheater equipped with the THX Sound System.

    So when I got aninvitation to interview Tom Holman, I figured now’s my chance tofind out what’s behind the THX Sound System I met Tom(Photo1 ) at his office on the campusof USC in Los Angeles. Friendly and soft-spoken, he has ateenager’s enthusiasm for technical details and theself-confidence of someone who truly knows what he’s talkingabout. We spent a few minutes discussing how he got started in audio.

PERSONAL. In high school, Tom gotinvolved with school plays. He worked mainly with lighting until hisjunior year at the University of Illinois (UI) when he shifted tosound, first in theater and television, and then in technical film.

    During this time he alsodid summer production jobs for the college. Graduating in 1968 with aB. S. in communications, he went to work full time for theuniversity; mixing, editing, and doing whatever else was needed.During the ensuing five years Tom says he really learned about soundbecause he had access to the facilities of a great college library.

   “I think it’sbetter than MIT’s or Stanford’s,” says Tom.“I’ve looked at audio at all three of them. And, becausethey (UI) would buy everything I wanted, I was able at that time toread pretty much everything that had been published about audio.Today, that’s impossible because there’s ten times as muchstuff out there. But back then you had to read two shelves full ofAudio Engineering Society material, and over a five-year period Ipretty much did that.”

    Tom spent most of histime working in cinema, eventually concluding that the likelihood ofimproving film sound quality seemed to be pretty slim. So in 1973, atthe age of 26, he went to work for Advent where he felt he could dohigher quality work. While there, he worked as an engineer with AndyPetite, the firm’s chief speaker designer, as well as with HenryKloss, the company’s founding engineer.

    In 1977, Tom left tostart his own company, Apt, because, “Advent wasn’t ingreat shape. Although it was a very successful loudspeaker company,their television line ate up al of the loudspeaker profits. I workedmainly in receiver and radio, and on the loudspeakers, a little onthe television. I started Apt to make the pre-amp and power-amp,which I did for three years.”

    When a chance to work forLucasfilm came along in 1980, it was just too good an opportunity topass up. Film sound was improving. Recently, Dolby Labs had greatlyenhanced the quality of theater audio by adding noise reduction,stereo format and standardization.

   “And it was also afield where things were a little backward. It was pretty easy to makea contribution, because I just used the principles I learned in highfidelity and applied them to film sound,” says Tom.

USC.I readily accepted Tom’s offer of a tour of the USC filmdepartment. A courtyard between buildings was full of sculpturesarrayed around a quiet fountain. Students were walking all about. Twobrass players practiced outside against a backdrop of bird songs anddistant traffic. Tom pointed out that one of his assignments forstudents in his beginning film sound class is to sit in thiscourtyard and write down everything they hear.

    USC has a very completesound facility with Foley stage, mixing stage, and a large scoringstage with 24-track recorder. The screening room also serves as alecture hail. Student productions are kept small enough not to needdialogue replacement.

    When we went through themachine room (Photo2 ) of the two film-mixingstages he stopped to assist a student who had accidentally let thedubber reel run past the end. A few minutes later, the dubber wasreloaded and work resumed. It was almost time for Tom to go back towork, so we made arrangements to continue the interview later atSkywalker Ranch, home of Lucasfilm, Ltd.

    The drive through thewine country just north of San Francisco was mesmerizing. Thickpatches of trees dotted the valleys between grass covered hills, withnarrow, well maintained roads winding their way through it all. Thisis highly recommended territory for anyone who likes wine and countrydrives.

SKYWALKER. Named after Luke Skywalkerfrom the Star Wars  series, Skywalker Ranch occupies asizable chunk of this marvelous landscape. Tom assured me that everybuilding here is brand new, just constructed with different styles,as if over a long span of time.

    Having traded in hisprofessor’s coat and tie, Tom was casually attired when I pulledup the long driveway. Noting my camera, he said managementdoesn’t like pictures to be taken because they are afraid of aphoto being misused by one of the scandal sheets. But he added thathe would let me know when it was safe to take a few shots.

    The centerpiece of theranch is a lake gravity-fed by seven wells in the hills, creating areservoir for fire fighting. That’s important because the ranchis far from town, and those beautiful grassy hills get very dry andbrown in summer. The ranch has its own fire department that is alsopart of the security department. These personnel are also trainedparamedics.

TECH BUILDING. One of the moreobvious structures is the Tech Building, devoted entirely to postproduction. This first segment of the winery-shaped structurecontains two edit/mix suites, back-to-back and mirror-imaged. Eachhas 11 editing rooms, a pre-mix room, a final mix room and machineand control rooms. Thus editors and mixers are brought closetogether. The rest of the building houses support people.

    “The sound studiosare set at the points of a star in order to separate them from oneanother, conceptually,” said Tom. “And the center of thestar is a set of technical rooms with heavier floor loads, morecooling and more power. The idea is that technology today&emdash;orin the future&emdash;is likely to be concentrated in a form where youneed to take care of some high tech equipment with heavy power,heating and cooling requirements. And then your sound editors aregoing to be separated from the equipment, not like the Movieola,where they’re physically working on stuff. So that’s whythere’s a central, heavy duty area.”

MILES OF DUBBERS. “The editingsuites are set up conventionally for two editors” continued Tom.“These are the simplest rooms with just benches and mag readers,squawk-box-type deals. But there are hidden troughs under the floorin order to run any amount of high tech fiber optic or what have you.Each room has its own thermostat and air volume control becauseyou’re likely to have different technologies in different rooms.

    “In the final mixmachine room all the dubbers are centralized in the central core, butthe recorders for a room are located here,” Tom continued.“There can be a lot of debate on which place you want to havethem. It wound up that you want a machine-room operator here and youwant your masters here. The central core basically is miles and milesof dubbers&emdash;about 600; you switch them to whatever room youneed. There are about 600 channels of Dolby SR (Spectral Recording)in the building. I think we’re the largest single customer forDolby.

    “Patching for 24tracks at a time, plus cross patching. You could patch up, basically,your normals with that scheme. It’s Magnotek equipment,thoroughly aligned with this ‘tweak cart’ that we builtwith a Bradford audio noise meter that goes down to 0. 1dBcalibrations. And we use those tenths,” says Tom.Alignment is done for every reel. With 35mm film, a reel is about tenminutes long.

   “And it has a spectrumanalyzer, tones or pink noise. You can choose to azimuth from anypair of channels you like.” The tweak cart is also connected in24-tracks at a time to do everything, including Dolby levels.

   “(In the rack there) is SSL automation. This is dubber andrecorder patching, dubber to inputs of consoles. But the monitorsystem is not a console function. The actual sends and returns fromrecorders and all the things in the path to the loudspeaker are inthis rack we built. We called it the CP-250, a kind oftongue-in-cheek reference to the Dolby CP-200. This actually does thetask of the studio DS-4: encoding and decoding the matrix, plus allthe noise reduction needed for that, [Photo3.] plus sends and returnsto recorders. They’re doing an L-t-R-t [left-total,right-total] here.

    “All the patchconfigurations are on floppies, so if you want to be in the Academymono mode you press the ‘A’ button on the console and loadthis disk. It puts all the processes available into the correctorder. Here are cards from the Dolby CP-200 (Photo4. ) to which we’ve added acarrier card which does differential input and output amplifiershere&emdash; input amps, 100dB common mode rejection input amps andoutput amps. “

    The rack also contains anumber of things for simulation, like optical clash, dirt noise,grain noise. “And this (unit also) simulates the frequency rangeand octave-to-octave balance of the range of different sound systems:A4”s, THX full range, or A4 equalized, or A4unequalized&emdash;pretty much any condition you might encounter in atheater … There’s a background-noise adder, a clipper, aclipper per octave band, and so forth,” he said.

    The signal goes fromthere to the room EQ, the THX crossovers, and the power amps. Theycan also mix for IMAX equipped theaters by switching in the topcenter speaker. The THX and IMAX systems are the same except for theadded channel in IMAX.

    While we are in one ofthe mixing theaters, Tom presses a button on the console and thesound of air conditioning rumble can be heard. “Average theaterbackground noise level,” he explains, “so you know whatyou’re working against. Footage counter and level meters appearout of black under the screen, by virtue of scrim cloth&emdash; theold theatrical trick, scrim.”

    Lucasfilm worked withSolid State Logic on the SSL 5000 series mixing console to turn itinto a film sound console. “These modules, in particular thepanning modules, are the most complicated ones,” adds Tom.

    “We have anotherlittle input console over here called the mix-in-context mixer. Ittakes existing pre-mixes and puts them onto main busses, so you canput up whatever you like. If you’re mixing Foley, you can runthe dialogue tracks and use the whole console for Foley. You canstill add in the dialogue tracks, so you can mix in the context ofthe already existing pre-mixes.”

    The mixing stages arenothing special, just good mixing stages. Tom demonstrated how deadthey are by clapping his hands. The surround speakers are hiddenbehind scrim panels, 14 in all. The rooms are also set up for thepower requirements of digital audio. They have massive air handlersthat take in water, chilled and circulated from an underground plantbelow the parking lot out back.

    Of the two one-manspaces, one is being used mainly for storage; the other is set up forthe sound designer, Ben Burtt. To get the projection to the properlength they use a periscope arrangement. The screen has adjustablemasking for the different film formats and the extra IMAX speaker.

LAWYERS, ACCOUNTANTS, AND ENGINEERS.“The second floor also has editing rooms and smaller workrooms, and also a room that looks down, balcony-like, into the stageso that you could use it as a control room to record in the stage. Ifall the other stages are booked and you just need one effect, just goup there and run a mike line down.”

    Tom continued to pointout things as we walked through the halls. “There are eightoverflow editing rooms here for each floor, which are apparentlybeing used as offices. George’s (Lucas) philosophy about that isif you don’t build a lot of offices you won’t have a lot ofoverhead, because you won’t have any place to put the people. Sodon’t build a lot of offices. He once said to me in a meeting,‘Engineers: I know what they do; lawyers and accountants: Idon’t know what they do.”

    A lot of odd angles inthe Tech Building help break up the work spaces, with no feeling of asterile business office. One section is made to look as if anexterior space were roofed over with a skylight to connect twobuildings. Office windows look out onto this space, providing lots ofnatural light. Very European, and very comfortable, according to Tom.

TRANSFERS AND LIBRARY. The transferroom has just about every kind of tape machine. The centerpiece is amulti-format Magnotek mag film recorder that has head stacks for altrack configurations and film guides for 16mm and 35mm. Sound effectscan be processed by an array of signal processing devices, includingan old Burwen single-ended noise reduction unit. The room also hasits own tweak cart.

    All of Lucasfilm’ssound effects are stored on l5ips tape, using Dolby A or SR-typenoise reduction. This is an unpublished library, but is somethingclients get when they do their films here. An even more restrictedlibrary, not on the shelf, contains the signature sounds for StarWars and Indiana Jones .

    Transfer room monitoringis accomplished with standard LCRS (left, center, right, surround)speakers and encoder so you can really tell what’s happening onfour-channel. “One of the problems I find common inHollywood,” says Tom, “is that unless you’re in adubbing stage, you can’t make any judgements about soundquality, because the transfer rooms are usually badly equipped withold Altec monitors, or something. And operators sit in the transferroom, and they try to figure out what to do about the rumble orsomething. They can’t really do a good job because they’renot hearing it properly.

   “Ourpoint is that every step of the process is made as standardized as itcan possibly be. And at the points where you make judgements, then,you have to have the right monitoring. So we probably do the correctkind of monitoring early on in the chain, like in pre-mix, that otherpeople do when they’re trying to make judgements on a Movieola.That’s OK if you know your effects library and how it’sgoing to wind up in the end. But it’s really hard to makejudgements, say, on production sound recordings. That’s whytransfers are made as routine as possible, so that you’re set upin a prescribed way and you do it every day. So six months later youcan make exactly the same transfer, and you can cut in a word and itdrops in. Levels have to be very accurate, equalization has to bevery accurate in order to do that.”

ADR AND FOLEY. The rooms I mostwanted to see were the ADR (Automated Dialogue Replacement) and Foleystages (all the non-vocal sounds an actor makes, named after an earlysound editor named Jack Foley).

    The Foley stage has abackground noise level of NC 5. They had to extrapolate that valuebecause the official tables only go down to NC 15. The room was soquiet I could hear my ears ringing and blood flowing. When Imentioned this, Tom assured me that the air conditioning was on. Veryquiet. It has to be quiet enough to get quiet clothing sounds withoutbringing up any room noise.

    We stomped on thedifferent “special noise” surfaces such as wood floor,concrete, metal grate, and there’s also a shallow depression forcreating water noises. The room also has an outside door for bringingin cars, and is made as dead as possible with four inches of fuzzeverywhere. The Foley stage is slightly more live than the screeningroom (see below) because of the extra surfaces.

    The ADR stage is similarto the Foley stage. It’s just big enough for a 6 x 14 footscreen. It’s furnished with a stool, script stand, headphonesand microphone. And it has another nice touch: a window to theoutside world. The control room has a very basic one-channel mixerand monitor for the replacement line and the original productionsound.

BIG SPACES. The scoring stage wasbooked, so we only got to see it from an observation deck. This isone of the two largest spaces at the facility, more than large enoughfor a full orchestra. The stage walls and ceiling are made frommassive semi-cylindrical cast-concrete forms. This gives the room avery long reverb time with very diffuse reverberation over a widefrequency range, even at low frequencies. It also provides a highdegree of isolation from outside noise. There are pockets in theceiling and walls for movable panels that allow wide adjustment ofthe reverb time, from 0. 7 seconds to 3. 5 seconds.

    I asked Tom about thetrouble of getting contractors to correctly follow the plans of anacoustic designer. “Well,” said Tom, “Ted Schultzdesigned this. He’s got lots of experience. He’s justretired, in fact. He worked at BB&N (Bolt, Beranek & Newman)forever, and then went off on his own about ten years ago. He didBaltimore, Toronto, Davies Symphony Hal, some in Australia, some inEurope, all over the place.”

    The other large space isthe screening room. “So, this is what we think a screening roomought to be,” Tom noted. “It’s fairly shallow for itswidth. In other words, it’s almost square. It starts with thepicture as the beginning point for determining what the ratio ofdimensions should be. There’s about equal masking all around.That tells you what the height ought to be, and depth for a certainroom volume and listening angle. It’s not really a theaterspace, it really is a cinema space.”

    The room is very deadwith a slight echo off of the screen. It has five main channelspeakers in the new format and hidden split surrounds.

    The projector was made inthe 1950s and came from a theater in San Francisco. Tom pointed outthe flutter idlers for 70mm are very well damped, and very difficultto turn. The film is moving quite fast, 112 fpm or 22. 5 ips.

    “This is an oldprojector,” observed Tom. “And it’s still the bestavailable dual-gauge projector. Well, it’s been repainted, butbasically projection is not a new issue. It was well faced in thepast. Now we do have, for example, much better heads than theyhad in the past: six-track Teccon&emdash;an awfully good headcompared to what they had in the 50s. The booth also has all thestandard sound equipment.”

    As we completed our brieftour of the ranch’s audio installation, I was more than a littleimpressed. It’s quite a place, and it was truly gratifying tosee a facility where no expense was spared to do things right. Butnow it was time to sit down with Tom and get the story on the THXsound system. Here’s what he had to say.

TH:Well, we’ve built a sound system called THX and it’s inabout 350 theaters now. [See Audio, September, 1989, p. 65for THX theater list. ] It’s very strong in some markets andweak in others. It’s in Paris, London, Germany, Canada,Australia, Hong Kong, Singapore, Korea, all over the place. But itreally started in 1980 simply as an experiment to make a better soundsystem for a dubbing stage.

    Is it a refinement forexisting theaters, or is it a completely new system?

    TH: It’sactually both. If you look at the whole chain&emdash;from themicrophone to the listener&emdash;you could say that certain parts ofthat chain were of much higher quality than other parts. If yousimply tune up a Nagra (a portable reel-to-reel) and use it, itexceeds the dynamic range of the sources it’s capturing. If youtune up dubbers, and you do things like adding better azimuthstability to them, we found that mag film was a fairly good medium.

    It has to be what I call“super tuned.” By that I mean tuning the level to within 0.1dB, tuning the equalization at 10k to within a tenth.

    Phase adjustments?

    TH: We make phaseadjustments for azimuth every day. So those mag film generationswhich were, at the time, audible changes from generation togeneration, get to be much smaller changes when you do that.

    In 1980 I felt the filmsound consoles were rather backward in sound quality. In order to getthe features you needed, you had to sacrifice performance, becauseyou needed these customized features. So we chose a music industryconsole in order to start with a basic good quality sound, highisolation and (low) crosstalk, low distortion and all those things.We modified it substantially to turn it into a quad panning, LCRS,console. That was a Neve 8108, our first.

    Later, fortunately, wewere able to work with SSL when they started building modularconsoles. It then became clear that all they needed were a fewdifferent module designs and you could do a real film consolethat’s as good as any today. And that’s what’s in useat the ranch tech building now.

    When it came to thetheater sound system we said, “Well, let’s start fromscratch,” because standardization had arisen, really, as twostandards. In the 1930s when the exhibitors owned the distribution,they owned the theaters. They built a sample theater in Hollywood andthey equipped it with 1. 5-mil slits, ‘xyz’ screen,‘xyz’ loudspeaker, and a certain electrical filter. Theytuned-up the whole thing and made their negatives for that system,and mixed on that system in those theaters. Then they went out andduplicated that theater hundreds of times across the country.

    This was when the“Academy Curve” started?

    TH: Partly. Butthe “Academy Curve” was never really standardized, studioto studio. For example, MGM always put more high frequency boost inthe negative and had more rolloff in projection than other people. Sothey weren’t really interchangeable, but they didn’t needto be.

    Right after World War II,two things happened. First, the justice department stepped in andsaid studios couldn’t own the theaters anymore. Second, thisexplosion of new technology that had been developed during the warbrought about the Altec Lansing A4 “Voice of the Theatre”about 1947. So the standard became the A4 with a certain equalizationahead of it.

   Now the standard was no different from room to room. There wererecommendations, like reverb time. (Tom opens a book. )“ A reasonable summary of currently accepted optimumreverberation times is given for 500Hz . . . it is based on audiencejudgements of acoustic quality of existing rooms and auditoriums.“This is no statement of what ought to exist. It’s a resultof motion picture theaters growing out of vaudeville houses wheresome reverberation was necessary to support the loudness of livespeech. Therefore, they went directly from vaudeville to motionpicture theaters with no stops in between. A lot of those are stillaround: the Castro Theatre in San Francisco, for example.

    There’s one inthe small town I used to live in.

    TH: It probablyhas a lot of bric-a-brac.

    It has anEgyptian-theme interior, bas-relief sphinxes and a tented ceiling.

    TH: Great! Butprobably fairly live. They weren’t very dead spaces.

    It was a theater for awhile, but it’s now a concert hall.

    TH: So no one hadever set out to say what ought to be. “If given a blank slate,here is what you would make. “The A4 was standardized andapplied to many auditoriums and dominated the market. Even to thisday it accounts for about 80% of the installed base of theaterloudspeakers. It has a number of problems that its own designerstried to remedy in the 1960s. They didn’t get anywhere becauseit was so entrenched they couldn’t change things.

    When we came at it in1980, we said, “Well look, the performance of this thing israther poor in a number of known ways. And many people havecontributed directly to making improvements in large-scale soundsystems over the years. Let’s draw on all these experiences andcombine the best of them into one comprehensivesystem.”

    The first battle we hadin designing the system was room acoustics. We started with Beranek.Our first room measured 70, 000 cubic feet, and Beranek says itshould have between 0. 8 and 0. 9 seconds reverb time. Fairly short,but I thought it should be even deader than that for several reasons.Both Acustica (the main European acoustical journal) andJASA (Journal of the Acoustical Society of America) havepublished a lot about the influence of reverberation time on speechintelligibility and background noise, and how they combine to harmdialogue intelligibility. Here, we are our own worst enemies, becausemost of the examples are cases of public address systems in noisyreverberant rooms.

    Now motion picturetheaters are usually quieter and deader than such rooms. But again,we are our own worst competition because we also have sound effectsand music, all going on at the same time, competing for dialogueintelligibility. So I wanted to go for the most transparent channelpossible. The other factor, seldom operative in Beranek’s day,is the widespread use of stereo, where localization of the screenspeakers is so important for giving the kind of wonderfuldirectionality that’s possible on the screen.

   It was easy to determine that we wanted a lower thannormal-reverb-time room, so I went down to a 0. 5 second from his 0.8 second . . . for that size. And I also agreed with therecommendation that the reverb time be flat with frequency.That’s an old idea. It says that music sounds warmer in a roomwhere the reverberation time goes up at low frequencies. If we wantthat in a sound then left-extra is rather close to it, then center,then right-extra and right close together. And those&emdash;track,it’s easy: we put it in a sound track. Yet we maintain the kindof articulation, you might say, of low frequencies. The most obviousexample of low-frequency reverb time problems is in 2010 wherethere was the cut between loud spaceship rumble and the vacuum ofspace. So it’s supposed to go instantaneously from rumble tosilence in abrupt cuts. The auditorium where I saw it had about a 5second reverb time at 31Hz. It just sort of smeared over theedits.

    So we wanted it fiat, andwe wanted a low reverb time. Then, of course, you want it quiet. Wemade our first stage super quiet because we needed to use it for dualpurposes, for a Foley stage as well. So it’s down around NC10,NC12. Which is really outrageously quiet and led to some problemslater on, which I’ll get to.

    Then look at the screenspeaker. You say, “What is it the A4 does right, what is it that it does wrong?” There are some axioms. One of mine isthat sound must be emitted from the same space as the screen. We haveto shoot (the sound) through the screen. When you shoot over the screen you can see people’s heads drift up as they adjusttheir pinae (external ears) to the angle of the sound field, and theynotice this sound coming from above them.

    I always thought theway we perceived direction up and down was by turning our headstoward the sound. You’re saying we perceive vertical positioningby the shape of the ear?

    TH: Right. So,it’s an axiom that we’re going to put left, center, andright loudspeakers on the screen. That was just a given because thesound effects are made so they seem to be on the screen. A lot ofpeople have thought you could have put left and right outside thescreen image and shoot through black transparent masking and getwider stereo. Well, actually that could be kind of nice in the musicand the ambience tracks. But when there’s a synchronous soundeffect of Indiana Jones entering camera-left, moving to the center,when the footsteps come from (elsewhere) suddenly things don’tmake sense anymore. There’s no sense for what we call “theeffect-effect.” It’s off-picture.

    I once rented Silverado and set up a pair of small speakers a couple feet oneach side of the TV. During a scene with a card game, a noise camefrom the right speaker and the card players turned and looked at thespeaker.

    TH: Right! Therethey are, in your living room! And the problem with the typical TVcase is the image simply isn’t big enough. And of courseif the TV image is made big enough, it doesn’t look good enough.That’s another problem.

    We went back to someKodak information from the early 1950s as to grain sharpness, focus,and basically how big a picture can appear. You know, in a homestereo we use a 60⁰-wide field most typically. Well,that’s too wide for films. We say that a 50⁰-wideCinemascope picture with the sound speakers at45⁰&emdash;plus or minus 22. 50, left and right&emdash;isthe kind of optimum seat.

    That’s how wedesigned all the dubbing stages. And in the technical building, everyroom you enter, despite its size, you’re always at a place wherethat same angle is intended.

    In the case of a theaterwhere we have completely adjustable masking, we make the other formatpictures&emdash;2. 2 to 1, 1. 85 to 1, 70mm, 1. 85 to 1, 35mml. l. ~.and change the masking so you get the biggest picture you can fit.You can come as close as possible to that. But the rooms are all laidout for the 2. 35 to I Cinema-scope image. By the way, it’sreally nice to start a room design from a picture because it meansyou can leave, for example, an equal amount of masking&emdash;liketwo feet&emdash;all the way around, and that forms the dimensions ofthe room. Voila! You’ve got the width and height of theroom right there, amid depth is governed by the optimumlistening angle.

    What’s thedifference between 1. 85 to 1, 70mm and 1. 85 to 1, 35mm?

    TH: Everythingshot today&emdash;since Lawrence of Arabia &emdash;originateson 35mm film in one of two formats (Fig.1 ). In the Cinemascopeformat, the picture is squeezed onto the film and expanded onprojection, and the aspect ratio is 2. 35 to 1. When that kind ofCinemascope negative is blown up to 70, it’s de-anamorphized,spread out, and the out-sides are cropped down to 2. 2 to 1, Soit’s not exact. That was in order to provide sound track areainside the perforations. There were also existing standards. Thataccounts for maybe less than a quarter, maybe 20% of the negativesshot in Cinemascope.

    The screening room at theranch has five main channel loudspeakers (Photos5a and 5b ). The old way to do it, the original 70-mil way, was equal spacing,a left, left-center, center, right-center, right (Photos5a and 5b ). That all dropped out of use by the early 1970s, and whenStar Wars came along they said, well, let’s use theseintermediate channels for just bass, and invented the”babyboom” format. And, of course, that’s much less important asto where it is. So what we have is left, right-extra and rightlie just inside the 2. 2, 70-mil format and 1. 85, 35 format. So whatyou do is interchange the outside pairs when you’re going from2. 2, 70-mil or 2. 35 Cinemascope to the 1. 85, 35-mil.

   So it’s a whole speaker at the left- and right-extra, but youjust use the bass for the baby boom.

    I read somewhere thatPanavision now has a 70mm camera quiet enough for use on a set.

    TH: There arecameras, and they’re used in process photography in all kinds ofways, but they’re not usually used for principal photography.There hasn’t been any show shot in it. Maybe someday.

    The other 80% are shotwith a cropped negative. That is, the actual negative area on thefilm is approximately 1. 33 to one.

    They just leave off thetop and the bottom, and there’s a line in the viewfinder to showwhere the 1. 85 frame is. So the cameraman composes for the 1. 85frame. In any case, there’s nothing special done about that. Theprojector in the theater can be framed up and down, and you’llsee the ceiling and the boom and some of the lights, possibly. Somedirectors print a simple black matte in the top and bottom, so if thetheater mis-frames it, it will be very obvious. That’s a bigdebate. And when you blow up 1. 85 negatives, like ET, to70-mil, you can only blow them up to 1. 85. So they don’t fillthe 2. 2-to-one frame. And because you’re blowing them upfurther, the image quality is not as good as the 1. 85 to one.

    With Cinemascope,there’s more glass, there are more lenses, and there’s alittle more difficulty. If you pull focus from foreground tobackground, things seem to change aspect ratios because there’sstuff in there. So there are some drawbacks to the Cinemascope, butit gives so much more negative area that it is in fact the betterprocess.

DIRECTIVITY. So that got us reverbtime, reverb time flatness, background noise, spatialization. Then wecome to the next factor: what should the Q of the loudspeaker be,what should its directivity be? The simple-minded theory is that youwant to basically direct the sound at the audience. You don’twant to put excesses of sound energy on the ceiling, on the backwall, because it’s likely to return to them later and causeechoes or add to reverberation. So basically, you use a Q of aloudspeaker so that the audience lies in the 6dB contours of thespeaker.

    Now that’s onetheory. Another theory goes directly to the problem of amateurloudspeaker design, and that is the “two-wayproblem.” And the “two-way problem” is that if you puta crossover at say, 1, 500Hz, in a two-way loudspeaker, 10” anda 1” dome, you can design that loudspeaker to be flat on axis.The trouble is that in the far field, when you start considering thepower response for the long term, what you will get is a rolloff.

    Including allreflections?

    TH: Including allreflections, you will get a rolloff at the top end of the woofer dueto its directivity collapse. It will cross over to the tweeter, goingwide. So you design it for flat on axis, and you can make it flat onaxis, but it’s got this rolloff and then opening up that causesa hump in its curve. And it is quite audible. It does have a littlecoloration.

    Aha, I’ve heardthat in my own design!

   TH: It’s not an unpleasant coloration, but it’snevertheless there. So that says to me you have to match thedirectivity at crossover. It” s almost impossible to build asystem that’s going to work over ten octaves and still haveconstant radiation with frequency. But one of the mistakes of a lotof amateur&emdash;and especially very high-end designs is that theygo for different radiation patterns in different frequencyranges. They have some theory that it should be forward-facing trebleand omni-directional bass or what have you. And there’s somedramatic directivity change somewhere. I think it should be asconstant and controlled as we can make it. So that’s anotherrequirement of the system. You must have drivers that go over enoughbandwidth with enough uniformity, enough directivity, enough powerhandling. You can add equalization, that’s kind of a minorissue. But they’ve got to cover the bandwidth. They’ve gotto be smooth. They’ve got to cover a uniform directivity. Thoseare the big things by which you choose what technology to use.

THIELE/SMALL. Now, if you look atsome of the ingredient parts, what were these developments we used?Well, one was the Thiele/Small parameters that all speaker builderstoday are using and pretty much understand.

    That’s howI’m learning speaker design.

    TH: Did you usethe LEAP (Loudspeaker Enclosure Analysis Program) program?

    The program I’musing came from a Speaker Builder article. (Thanks, Reid-Thecheck’s in the mail&emdash;Ed.)

    TH: Of course theinfluence they had was not only in understanding how loudspeakers andboxes worked together, but then really going back and redesigning theloudspeakers for particular box configurations. And what we found wasthat you need a certain cone area to reach a certain sound pressurelevel. And that is rather equal to what the A4 is, which is two15” drivers.

    Now you think about whichways you can orient those. Well, if you stack them vertically and youmatch the directivity at crossover, by which I mean you use a hornthat’s wider than it is high in radiation pattern, you can dosomething very nifty, which is , you can beg the difference onthis power response versus axial response. You can make the axialresponse right and, because the radiation patterns are equal atcrossover, they will have smooth constant power. This is a longdebate I had with Peter Snell and others back in the 1970s aboutwhich is more important: the axial sound field or the powerresponse? The system we’re working on kind of begs thedifference. You just make them both the same, and you don’t havethe issue.

    There are some reasons todo coaxial designs, but no coaxial design has really done everythingproperly, I think.

    The high frequencies needto be handled by horns for the directivity control, and they need tobe rather large ones.

    They must have mouthsabout a yard square. Anything smaller, like foot-high types, have aterrible problem since they can’t maintain uniformity down tothe crossover frequency. You choose the crossover frequency based onwhere the directivity patterns match and where the drivers can handlethe power. That gives you one of the constraints of how big the hornmust be to keep the pattern controlled down to that frequency. So allthe horns used in THX are rather big.

    Is it just a two-waysystem?

    TH: It is two-way. I looked at three-way and decided not to, but I’ll cometo that.

    The type of horn used isa new style&emdash;as of the last ten years&emdash;called“constant directivity” because they really are much moreuniform than the older horns. The straight axial horns, used byKlipsch for example, can be more or less constant in one plane,horizontally, but they will collapse vertically across frequencies.So they have a strong change with frequency in the vertical planewhereas they can stay reasonably constant in the horizontal plane.

    That’s the problemwith the old fashioned, pre-1940 horns, the multi-celled onesinvented for the A4. The problem with the multi-cell is that althoughit’s the first attempt to get more uniform coverage, it hasterrible lows. So if you put up pink noise and walk around the room,you get a “wish-woosh-wish-woosh”&emdash;very obviousholes. I was able to measure that in a dubbing stage and confirmedthat what you hear is there. It’s quite clear there are holes inthe pattern.

    What is the name ofthe curve used in the throat of the constant directivity horn?

    TH: Variousmanufacturers use different techniques. JBL callstheirs”bi-radial. “I don’t know what EV calls theirs.But they’re based on several things. They join two hornstogether, in effect. One controls the vertical directivity, and thatfeeds into a slot-type radiator, which then forms what’s calleda diffraction horn to set the horizontal pattern. There’s a kindof joint there. And I don’t know any way to design thosemathematically. You just duke around with ’em till they work.

    You are splicing a coupleof things together. So that’s definitely an ingredient. Someonetried to apply it to film sound before, but it had been done sopoorly, and in such bad demonstrations, that nobody believed it. Theyblew out peoples’ ears in the demos. They tried it at theAcademy. It was not good. Then there’s the compression driver onthe back of the horn. The Altec 288 is a venerable device&emdash;witha phenolic diaphragm which was susceptible to breaking, and waschopped off at about 8kHz (at 24dB per octave). Someresearch&emdash;I think it was done by Howard Durbin atJBL&emdash;showed that the reason it chopped off so abruptly at 8kwas the fact that the surround was going out of phase with the maindome. The surround has a significant area which also feeds right intothe slot. So it was notching by virtue of that feature. So JBL got apatent on a diamond-kind of pyramidal-surround that does, in fact,not go out of phase at high frequencies, and doesn’t suffer theabrupt chop that the others do. Then the diaphragm material becameimportant about 1980. The ability to whip titanium into domes wasnew at that time.

   This was, in classical terms, like a Western Electric driver designedin the 1930s, only it’s designed out of much more exoticmaterials. And today we’ve even taken another step in thatdirection. It’s now a titanium dome embossed with al kinds offunny little patterns to stiffen it up. And now the neodymium magnetmakes it much lighter. Things are getting more exotic and moreexpensive. (But) I don’t know whether it’s actually betterquality.

    It still looks likethe same ol’ speaker.

    TH: It’sdifferent. It’s a very cleaned up version of the WesternElectric, I don’t know the model number. I’ll bet it’sin here (Grabs book ¹). . . 1938!

    This has a lot of greatthings in it. There are whole paragraphs in here you can lift out andput in today’s manuals, and they’re still not done. Likeprojectionists riding the volume control. You know, complete no-no .. . the movie’s already been mixed. (Leafs further throughbook ) There it is. Lansing 285, high-frequency unit, 1938,showing a two-mil Dural, durable aluminum, 2 mils thick, radial slotopenings, a voice coil. Of course they had a field coil theydidn’t have permanent magnets! Oh my God! Look what they wereconcerned with. Directivity. “A certain amount of directivity isrequired since the best illusion is obtained if the ratio of directto reflected sound is as high as possible.” (Snaps bookclosed.) 1938! Still true today. Oh my gosh, I keep findingthings like that, gems of wisdom in this book.

    They keep proving youright.

    TH: That’sright, the ancients have stolen my wisdom! No, the fact is whatwe’re doing here is very much rooted in the history of howthings ought to be. It’s simply been greatly cleaned up.

    You’re provingthe theory of things they came up with over 50 years ago.

    TH: Yeah.Frequency range extended, amplitude capacity extended. But it reallyfalls in the same tradition. These guys would understand itperfectly. So now you’ve got a low-frequency system, ahigh-frequency compression driver and a horn, and you look at it andyou think, maybe a three-way would be a good idea, simply for powerhandling if nothing else. But the titanium, although falling off,does fall off smoothly into the top octave. There’s a problem ofwhere you’re going to line things up. You have a speaker and youput a microphone out in front of it, and if you measure its acousticphase you can find where its acoustic center is. That turns out to bea different place&emdash;electrically, measuredacoustically&emdash;than it is for vertical pattern or for horizontalpattern. All three of those are different points. So where are yougoing to line up the tweeter. . . on which of those three? You wantit to speak at the same instant. You want it to have the same sourcewith all the rays going out from the same point. It gets to betricky.

    Adding a third drivercreates too many variables?

    TH: Well,it’s not impossible, but it’s difficult for a anotherreason. These horns that have to have about a yard-square mouth areabout 40-45” long. Now if you set them where you want to, thatis, on top of the woofer cabinet, the woofer speaks well before thehorn tweeter. And even if we align things with equalization of thetime difference, 1. 9ms, or so, you’ll have noticeable radiationpattern changes, and it won’t pass the 1938 test.

    One of the 1938 tests wastap dancing, with Eleanor Powell. On (reproducing) tap dancing theyheard, “ta-thunk, ta- thunk, ta-thunk,” instead of one hit.So they slid the horn in and out and found they could have less than2ms of delay between the two. Paul Klipsch repeated this in the 1960sand claimed he couldn’t find it. But he’s wrong. He’sjust wrong.

   I mean, his experiment wasn’t any good, apparently. He wasmoving loudspeakers al over space. The frequency response must havebeen changing al over the place&emdash;transfer function changing.About 2ms in the midrange is quite audible as a“ta-thunk.”

    So you say, “Thefirst thing we’re gonna do is delay the woofers.” Now thefirst job of the crossover is to delay the woofers by 1. 9ms. And wedo that with a delay line of all-pass filters. Now if we do that witha delay line of all-pass filters, and we have a tweeter somewhere uptop that is going to speak sooner than the midrange, we’re goingto have to delay the tweeter as well. It turns out you can delay bassa lot more easily than you can delay high treble.

    I did look at the thingbeing a three-way system and decided against it, basically because ofmoving a delay line. I’ve got a delay line in there now thatgoes up to 500Hz and takes up to three op-amp sections to do it. Ifthat delay line had to be extended to 5kHz it would be much morecomplicated and I’d have 15 sections, be less reliable, and allthat stuff. Otherwise, you would have to go with a digital converterand an actual time delay or something to get it. So that’sanother reason not to use a three-way. So we chose not to because,yes the response is falling off, but&emdash;and this is an importantpoint for amateurs&emdash;every real commercial loudspeaker has thecrossover network designed for the specific drivers, in two ways.One, with the driver as its terminal impedance, instead of with aresistor. The most common fault of amateur designs is that theydon’t take into account the impedance of the driver. They justcalculate something out of a book for an 80 resistive load. Andthat’s way too simplified. That’s the first problem, Iwould say, of amateur designs.

    In my own design Iused a stock crossover. That is likely the cause of a lot of thestrange coloring Knowing this, I’m going to tear back into it.

    TH: You have tostart by measuring the terminal voltages with crossover in place andtaking a lot of acoustic measurements. So that’s the secondpart. This is the way THX was designed. You start by measuring thedrivers. You find out their acoustic transfer function. Then youdetermine what target transfer function you want for the wholesystem. Then the crossover makes up the difference between the two.

    So, for example, wewanted a Linkwitz-Riley fourth order alignment for all its known goodproperties. It’s because of the way in which we rank therelative importance of the various things that the crossover doesthat we say the fourth order Linkwitz-Riley is best for our purposes.First-order crossovers are hopeless because they have bad radiationpattern tilts, a very strong positive-going lobe and very badnotches. Which means that as you move up and down with respect tothat loudspeaker&emdash;as you move up and down theauditorium&emdash;you get quite different transfer functions in thecrossover region. That’s not good. Linkwitz-Riley, with its oneprincipal lobe on axis and two very minor lobes off, is the best inlobing behavior if you can’t have a coincident driver.

    There was an extensivearticle in Speaker Builder (1/85 ) on crossovers andlobe patterns. I’ll have to read it again.

    TH: (Laughs ) Good! You measure the woofers and you say you want this kind of aresponse to be Linkwitz-Riley. Acoustical Linkwitz-Riley, notelectrical. ^2,3 But overall, acoustical. Then you determine what poles andzeros you need in the electrical domain to add up with the drivers tomake the final response. Siegfried Linkwitz doesn’t say thatanything that’s called Linkwitz-Riley is Linkwitz-Riley. Becauseelectrically they (may be) Linkwitz-Riley, but unless you knowwhat the acoustic transfer function is, you don’t know.

    And you do the same thingwith the compression driver. The one for the woofer turns out to bepretty simple. It’s a couple-pole high pass (-1dB at 40Hz) toprevent overload at very low frequencies. It has quite a flatpassband and then it’s a four-pole rolloff, as you would expect,because the woofers go quite a bit past the 500Hz crossoverfrequency. On the other hand, the compression driver is kindof close to the 500Hz limit. It’s rolling off itself. So that,more or less, forms one pole of response so that the highpass isthee-pole electrically, and becomes four-pole when you add in theacoustic transfer function. So, you get two four-pole-squaredButterworth response that is Linkwitz-Riley, and it works out for allthe good reasons. Then you come up to a kind of plateau where we findthat the horn is about 10. 5dB more sensitive than the woofers, soyou can set it down by that amount. Then we have a whole bunch ofthings going at high frequencies. For one, the compression driver isrolling off because of air trapped between the diaphragm and thephasing plug, having to squeeze the air in and out. Another reason isthe moving mass. Yet another reason is that the inductance justdoesn’t allow the current to be usable at very high frequencies.So there are a bunch of reasons why it rolls off.

    Number two, we hang amotion picture screen in front of it. And a motion picture screenwith its perforations is a one-pole low-pass, RC and just 6dB peroctave, located between about 5kHz and about 8. 5kHz, depending onthe screen, its thickness and perforations and such. So there’sone factor to account for.

    All of these thingscontribute to some kind of high-frequency rolloff which we compensatefor electrically. And there’s a fourth consideration, which isthat there’s an international standard on what we ought tomeasure in the far field. There’s no standard on what a homeought to be, except people would probably go for flat when theymeasure things. But in a motion picture theater it’s well knownthat if you make things perfectly flat-on pink noise, say&emdash;thenall program material appears to be too bright. So it all has thisstandard curve, ISO 2969, Curve X, which is a Dolby-promulgatedstandard (Fig.3 ). It is flat to 2kHz, andit’s down 1dB per third octave beyond there. It’s astandardized house curve. It’s -6 at 8k. And it tips up and downsome with room volume. There’s a room volume correction in it,based on an average room.

    We took an empiricalapproach. We designed a crossover so the speaker would be flat. Andthen we hung a screen in front of it which we knew would roll off thehighs. Then we corrected the network to get it down to the standard.So we can be, on the average, on the ISO standard.

   So, the audio on a sound track is flat?

    TH: No itisn’t. It’s been listened to and monitored over an X curve.So it has some degree of built-in boost. It won’t be as much as6 or 7dB at 10k. That’s too much. If you play back a CD in amotion picture house, or in a dubbing theater, you have to brightenup the highs to work against the X curve rolloff. But it’s likea standardized de-emphasis like RIAA or NAB. It’s the same kindof idea except that it’s electroacoustic. And if everybody isusing it, then it’s all translatable.

    It’s doneacoustically, with the crossover, instead of e. g. , the playback EQelectronics of a tape machine.

    TH: Right,exactly.

    We get the ingredients bylooking around manufacturers’ catalogs and picking what we thinkare the best available. The ones that have been made to theLinkwitz-Riley network. The ones that will give us actual chambercurves, and such. And we send them up to the

    University of Waterloo inCanada and start taking measurements. We put the speaker on the edgeof the stage, as it would be in a movie theater, and we raise andlower the pit to see the effect of the first reflection off thefloor, and all kinds of things.

    We discover we can’tget anything like the rated frequency response with the woofersystem. There are many phone calls back and forth, and we look at themanufacturer’s data and we sort out how it was made. We discoverthat Linkwitz-Riley all depends on a 2 pi environment. It’snot for 4 pi. It’s for half space. And so we look at thewoofer on its back on the stage floor and hang a microphone up m theair. And sure enough, we get the Roy Allison famous classical dipcaused by the reflection off the boundary behind it.

    How deep was thespeaker?

    TH: About twofeet. It’s down maybe an octave from a home speaker, butit’s still there.

    So we tip it up on theedge of the stage and we put boards all around it, build a wall allaround it and discover, sure enough, the bass comes up quite a lot,as you’d expect. And subsequently, when you install it in adubbing stage here in Hollywood, when you interrupt this walldiaphragm and put two A4 bins on either side and two bass bins oneither side of the woofer section, you lose 15dB at 100Hz.

    You lose. . . ?

    TH: Yup. 15dBworth of loading. You’d think it’d be only six orsomething. But it was a huge notch at a hundred. So the walls are avery important ingredient of the system, because it makes the basssmoother, it eliminates any Allison interactions with the environmentbecause you’ve flush-mounted everything. You have onlybeneficial reflections from the local environment (Fig.2 ). Then it performslike it’s supposed to, as its design standards and Thiele/Smalltell you it’s supposed to be.

PHOTO1: Tom Holman, corporate technicaldirector for Lucasfilm, Ltd. , and assistant professor at USC’sSchool of Cinema-Television.

PHOTO2: Dubbers, USC film lab.

PHOTO3: Dolby decoders.

PHOTO4: Dolby CP-200.

PHOTO5a: First THX installation showingformer placement of speakers behind screen.

FIGURE1: Comparison of different aspectratios and THX speaker placement.

PHOTO5b: Behind-the-screen view of firstTHX installation.

FIGURE2: Side view of the THX baffle.Re-drawn from THX… Instruction Manual… .⁴

FIGURE3: Graph of ISO 2969, curve X.

REFERENCES

1. Academy ofMotion Picture Arts and Sciences, Motion Picture SoundEngineering , New York: D. van Nostrand, 1938.

2. Linkwitz,Siegfried H. , “A Three-Enclosure Loudspeaker System: Part III,“ SB 4/80, pp. 14-19, 22-24, 30.

3. Linkwitz,Siegfried H. , “Active Crossover Networks for Non-CoincidentDrivers, “ JAES, Vol. 24, January 1976, p. 2.

4. Holman,Tomlinson, THX Sound System Instruction Manual: Architect’sand Engineer’s Edition , Fourth Edition, Theatre Operations,a division of Lucasfilm Ltd., October 1987, p. 15.