TIf,E E'ATRTTGIIgEXPEATSUED

When the Fairlight CMI was released seven years dgo, both the machineand its software represented a significant step forward in the applicationof computer technology to music. Today, the Fairlight is used in thernaking of popular music the world over, as well as performing an importantrole in the field of musical and technological education. Despite this, veryfew people are fully aware of what the CI}II does and how it does it. JimGrant, who's been working with one for a number of years at the LondonCollege of Furniture, has decided to rectify matters by writing a seriesdedicated to explaining the Fairlight's workings. Part one appears below.

E&MM AUGUST 1984

efore the Fairl ight's appearance,most computer music systemswere the perogative of mainframes

and their contr ibut ion to the wor ld ofeveryday music was slight. Kim Ryrie(the Fairl ight's father) and his Australiancolleagues soon changed all that, how-ever and their invention is now used inalmost area of music production, to theextent that many people appreciate itssound without realising that they'rel istening to 'music by numbers' .

However, despite its widespread use,there are relatively few Fairlights in generalcirculation - less than .100 in the UK -and to see one in action at close quartersis a real treat. Herein l ies the rationale forthis series of articles. What does theFairl ight do? How does it do it? Andwhat can the average musician do withi t?

HardwareTo take delivery of a Fairl ight leaves

your bank balance empty and your l ivingroom full. The hardware consists of aCentral Processor Unit (CPU), one - oroptionally two - six-octave keyboards, atypewriter-style QWERTY keyboard, anda VDU with added lightpen. ln addition,there are some long connecting leads, aSystems floppy disk drive and a box ofdisks containing l ibrary sounds.

SoftwareA foolproof system of connectors -

and a quick glance at the manual on thepart of the user - ensures that the Fairlightcan be powered up in no more than fiveminutes. The VDU displays the expectantmessage 'CMl READY', whi le the CPUhums quietly: there are three fans pull ingair through the innards, keeping 500 wattsof power dissipation down to an accept-abletemperature. . . .

lnserting the Systems disk in the left-hand drive (Drive 0) results in a faint clickas the stepper motors engage. Theoperating software is loaded as a seriesof 'fetches' - each section of programloaded pulls in the next section. Whenthis process has been completed, theuser is faced with the Index page. SeeFigure 1.

Here l ies one of the Fairl ight's mostpowerful features. The whole system ismenu-driven and the different optionscorrespond to different VDU displaysand sets of commands which are enteredfrom the alphanumeric keyboard. Eachoption is referred to as a Page. A Pagehas one or more fi les resident 9n theSystems disk which are loaded when thePage is selected. Page 1 is the IndexMenu itself (Figure 1), while Page 2manages the fi les stored on the disk inthe right-hand drive (Drive 1). These fi lesare user-created, and there are sevendifferent types, as indicated by the suffixafter the fi le name. These are as follows:NAME.VC is a voice fi le occupying about2OkBytes. lt holds waveform data (1610and extra information regarding loopingand so on.NAME.CO holds control informationsuch as portamento, vibrato frequencyand depth.28

0r,1i l8Nn

I |1IRLI[HT

INDEXDISK DONTROLKEYBDHRD CONTRclLI I I IRI '1ON I C ENUELDPESI, IEUEFORIl EENERBT I ONI, IHUEFI]RI '1 I IRf i I , I I NC

cc|NTR||L PFRRIlETERSS|JUND SRIlPL I t I | ]SEUUENDERffNffLOG INTERFHIECOI' IPOSERl, IHUEFORH D I SPLFIYD I SK L I BRFRYRERL-T I I lE CD14P|]SERSCREEN PR I NT

, JI l l DRRNT

I l,l. Iu3 , t5, R l , l l

PIIDE I

rnbL <

rNbL J

PRGE 4PI i | jE 5f n l - 'L b

PBriE 7PRIJE DPFDE SPH6E FIPBl]E [ ,PFEE DPBriE LPrir lE RTNbL 5

I .JSER NFI14E

l@t

011f l f f t iD'

DISK, EXEIlPLEUSER. JIT'1 GRflNT

FREE SPHCE. 536LIBRffRY NO. 8

1 INUENT . IN I6 STIDK C|]2l l i t r I lErm t7 I t iuENT .S03 PRELUDE , IN I8 PRELUDE .SO4 UIUFLDI , I I I I9 UIURLDII .SO5 Bf iSSGT .UC ?B LOCUST .RS6 BDD2 UC 2I PBGEI .PX7 CBELLO(][} UC E2 PFGEz .PX8 GRNGSR UC 23 PFGES ,PXS STICK .UC 24 SYHTH .PC

1B STRI ' I I D UC 25 SYNTH . PT11 I f rNN UC 28 BFSSD .SST2 ERSSGT CO 27 CNTRL .SS13 BBDz.. .C0 28 SN3 .SS14 CBELLOOO.CO 29 TUHEl ,SSI5 GRNGSR .CO 36 TUNEz ,SS

TRBNSFER DELETE

IlULT I CI iNDEL

clt r EII t lcLEUERY

! .qgH$I . , - l | : | . tvee:- ln$IRU|]1ENi . , , ,ye| l , . . : i , , r ,e , 24 . .voices'BDD2 STICK BFSSGT STR|4 ID

T

l@

Com'irterrcthsb@rd

Addressbus

Cmlfolcrrcur lry

wovelorm RAM

dio signoloulpui

AUGUST 1984 E&MM

sheet, 233.

3

4c

f

6b

76

I18,i t

IfIIiIIEIIIIE and typesFf im SPRCE on disklfnfilfli$ numberEf i f f i l f l In t i les { t4ULTI & CBNCEL)f l l l f f iE disk, usen or f i ]e nenret

EII f I I commandEN:I I I | IE instnument on voire f i lesml! f i f l In instrument f i lesI [ l l | i f f lE voice f i les]m:f i l f ln sEquence f i les|lJifllfll somnrendIf#IIEiIfE cor'mahdnFffiff:dmftfmlfi pnotect i onImmLEf0 orr mtsster k eybosrdmfi i l : Im shortcuts

sny l0X r^r i th LIGHTPEN or type, n(set)uherE: n = sneel

toucn

touch l I f f i or tvPe' H I ( netur n )

EE!tr4

NAME.IN configures the CMI to a parti-cular instrument state. Voices are auto-matically loaded and spread across thekeyboard.NAME.SQ holds polyphonic keyboardsequencer information.NAME.RS is a real t ime sequencer (PageR) fi le.NAME.PX corresponds to a screen dump(Page S) to disk. This can be spooledlater to a dot-matrix printer for hard copy.NAME.PC, PT, SS are Music CompositionLanguage (MCL) files. These are gener-ated on Page C and hold text files thatdescribe notes with duration, dynamicsand so on.

A voice file can be loaded in a numberof ways. Probably the easiest is to pointthe l ightpen at the volce name and thenat the command LOAD at the bottom ofthe display (Figure 2). Drive 1 springsinto action immediately, and after asecond or two the selected voice appearson the keyboard.

So far so good. But where does thevoice information go, and how does itresult in a sound when the kevboard isplayed?

E&MM AUGUST1984

Ghannel Cards

Inside the Fairlight, there are usually atleast 16 circuit cards, the exact numberdepending on various options such asan analogue interface and sync card,and eight of these are known as voice orchannel cards.

The Fairlight produces sound by aprocess called Waveform Synthesis. Eachcommand that deals directly with soundgeneration must involve at least a sectionof a waveform. The waveform itself isheld in 16K of RAM on each channelcard as a direct digital representation, sothat increasing amplitudes give largerbinary numbers. Therefore, when aneight-note polyphonic sound is presenton the keyboard, each channel holds thesame voice data.

Put simply, the channel cards can beregarded as digital oscillators whosewaveform is determined by the contentsof 16K of RAM (Figure 3). Different pitches- as played on the keyboard - conespondto the MM"information being read andconverted by a DAC at different rates; the

channel cards perform this functionautonomously. The computer section ofthe Fairl ight passes parameters such aspitch, vibrato, portamento rate and loop-ing points along its data bus, and oncethese have been received, the channelcard outputs the sound unti l the para-meters are updated.

Overall pitching of the CMI is deter-mined by a system clock resident on asoecial card known as the Master card.We'l l be referring to this on numerousoccasions over the next few months,since it holds the circuitry for a goodmany of the CMI's functions. A 34MHzoscil lator is onboard, and this is dividedand fed to the individual channel cards.It's from this clock that the RAM clockingrates - and thus keyboard pitches - aregenerated. The whole instrument can betuned by scaling the master clock.

Page 2 CommandsLooking again at Figure 2, there are

several commands at the bottom of thedisplay. TRANSFER allows fi les to becopied from one disk to another, usingDrive 0 as the destination drive. This isessential for creating backup copies ofimportant music and/or sounds. DELETEerases unwanted files to make room onthe disk. Invoking this command promptsa confirmation message to prevent ac-cidental erasure of important files. Whena fi le is deleted, FREE SPACE increasesby the deleted file size.

At the very bottom of the displdy is anexample of the QUERY command. Thistells us that LOCUST.IN fi le wil l auto-matically load eight voices, whose namesare shown.

Help PagesBy this time, you're probably wondering

how anybody using a Fairl ight evermanages to remember all the commands,especially since we've only consideredPage 2 and there are another 13 still togo.

The answer is simple: Help Pages.Figures 4 and 5 show examples of

Page 2 Help Pages. In fact, the entireuser's manual is held on the Systemsdisk. and sections relevant to the currentdisplay Page can be inspected at anytime by typing HELP (what else?).

Init ially, an index sheet is displayed(Figure 4). Touching any of the highlightedoptions with the'l ightpen results in theHelp sheet specific to the selected optionbeing loaded and displayed. The usercan flick backwards (BWD), forwards(FWD)'; or recall a previous place (PRE):commands can be entered from the Helpsheets while viewing their correct format.The CMI then automatically reloads thedisplay page that called the Help sheetin the first place, and executes thecommand. And yes, there are even Helpsheets that explain the use of the Helpsheets. . . .

That about wraps up the first part ofwhat will doubtless become a saga ofsome duration. Next month, we'll takea look at Page 3 - the keyboard map -and the waveform display page, Page D.

E&MM29

TRfiNSFER FILES TO BNOTHYPE . L I GHTPEN ,

T, f i le(retunn) (select f i les)or <TRfiNSFER)

T, f i le, f i le, l i le( , . .etc)(return)r , rhene, f i le = FiLENEIIE SF or f* or **-* l or I

(ssme as DELETE; see above)

EXRI '4PLES, T. CH0RUS, IN(retunn)T, 4-18, FBLE. UC, 2S(neturn)

Fi Ies ui I I be copied FR0l '1 disk in RH dnive ( DISK f i ) T0 anothendisk in LH dnive (DISK B). Give TRRNSFER cotnmand ui th svstemdisk in LH dnive and DISK R in RH dr ive. l , lhen the messaEe'

PLRCE FILE DISK IN LH DRIUEEpp€tsFS, place DISK B in LH dr ive. l , lhen the tnansfer iscornpleted a f inal message rr i I I nequest the replacement of thesystem disk. I f a f i le alneEdy exists on DISK E (has semEname and suf f ix as f i le on DISK R), i t u i l l N0T be ovenuni t tenwithout youn consent. See also PROTECTION.

TYPE, T, t (neturn)l , lhen the messege eppeErs place a EEII t r f i 1e disk in LH dr ive.Replace system disk r , rhen comPleted. Ner*r d isk ui I I usual lyshor.r an incnease in FREE SPFCE avai lable

@

Jim Grant

TIIE T'AIRLIGHTE)KPEATTUtrD

The second part of our insight into one of the world's most popularcomputer instruments lool<s at display pages and what they tell CMIowners. Jr* Grant

ast month we described the gen-eral software conceot of the Fair-l ight and how much of its power l ies

in its abil ity to present its functions to theuser as a group of related files calledPages. We saw that the CMI powered upwith the Index Page (Page 1) and thatsound and music fi les were managed byPage 2.

One of the oroblems associated withmost sound-generating equipment isthe way in which the sense sight isexcluded from the orocess of soundformation. Most instruments operate on

=--*.u+'\\S..S*\

\\R'\\- N \. \._>

l , , . l l t i l ; f , 1T t : t_ l l , tF, t Tl'1 t-t It E ,tJ i i t , lEFr:rFt t1 I r I : iF Lt lT

inLrL; , {r l l l |1Hf i t r

. IUII -L. :RUIlFET

l l tsuEFrJPF Ir I r : ,FLt i i ' f l l I rE 4

Fr:r t ' l ' '1HT,H E t l i i l litlj--ffEl E;4 :i i !l-El- L E- + ir I

B6 SEPTEMBER 1984 E&MM

the basis of the user twiddling the con-trols and stabbing the keyboard. Fairenough: of all our senses, the ears areby far the most acute. Yet if we considerall the electronic music equipment cur-rently available, the most user-friendlyinstruments have graphic displays, per-haps in the form of panel legends andLEDs or liquid crystal display. Despitethe fact that sight is a very poor quali-tative sense, it can be of enormouspsychological help in our field. Basicallyit boils'down to: ' l f I can see it. I canunderstand it. '

D^ ^^ r-\r d.Bu, u

There's no doubt that the Fairlight'svisual presentation is founded on thispremise. Each display Page is graphicwithout being ostentatious, and Page D,the voice waveform display, is a primeexample of this. Typing PD followed by aRETURN on the alphanumeric keyboardwill result in a display of the type shownin Figure 1. To appreciate the signifi-cance of the display, we must delve alitt le deeper into the workings of theCMI.

Remember that voice information isheld in 16K of RAM on each channelcard. To simplify matters, the CMI div-ides the memory (and thus the wave-form) into 128 sections called segments.Each segment consists of 128 bytes, sothe waveform comprises 128 segmentsmultiplied by 128 bytes to give 16384bytes, ie. 16K. This saves the musicianhandling unwieldy computer numberswhen dealing with the waveform RAM.

The display shown in Figure 1 is apsuedo-3D representation of the voicecalled TRUMPET. Each line from left toright is a segment, and the foremostsegment represents the beginning of thesound. When a keyboard note ispressed, the CMI reads out the RAMinformation segment by segment fromthe front to the rear of the display.

There are two display formats, A andB, and a number of options within eachtype. Figure 1 is in format A, and seg-ments 1 to 128 are shown in steps of 4.Figure 2 is again format A, with the endsegment number 32 and steps of 8:therefore only five segments are shown.Format B gives an oscilloscope{ypedisplay, but with each segment slightlyabove the preceding one. Again, thereare a number of display options. Figure3 shows TRUMPET segments 1 to 128in steps ot 2, and Figure 4 segments 1 to64 in steps of 8.

Although Page D is purely for displaypurposes and does not support anysound creation commands, it 's sti l l aninvaluable aid. At its most basic level. itanswers the ouestions 'where has thesound gone?' and ' is the waveformzero?'.

l \^rage 5Page 3 is another util ity-type display

Page. lt deals with voice tunings, Chan-

nel allocation and keyboard maps.Figure 5 shows a typical display with

eight separate voices loaded into theCMl. Registers A to H are groups ol oneor more of the eight Ch.annels and'NPHONY' is the number of notes thatcan be played with the sound held in aRegister. A quick look at Figure 5 revealsthat there are eight active Registers,each holding the voice indicated. In thiscase, all eight Channel cards hold a

unique sound, and therefore the maxi-rhum number of notes that can be play-ed on the keyboard with any singlesound is one. This is indicated by thecorresponding 'NPHONY'. lf a singleeight-note polyphonic vorce is required,only Register A wil l be active and Chan-nels 1 to 8 wil l be allocated to A. Theactive Registers are mirrored on Page 2so that they can be loaded with soundsfrom disk. The Fairlight will flag an error

TEUHPET

I , IFUTFDRI'1 TI I i ;PLRYf4ulrE, 4

:J.r STEP.l E 4

INI lEX0l ' l l , l f fNIr ,

fi5t'

DEFGH

I I l I r !a2 i Sr I Cl l3 ) BH5:,r . iT{ i STRI,I I I I 15I 'STF;14ID26 r c8ELL00q7 ) Tf fNNP I r lHNr:SF

:l , l ' ,1i{: :rTEF:E . !JLf fUE

FITr: :H, l?Usi :FLE, t ( lT a ' j

OHHfiND'

SHXYSNfiREBDRU14ORGtiN

Lg25886?,48BAVJ

3 . I , lFSTER6 .SLFUE

P I ICH' I28SCFLE, l?r- : ;

V <.UU

SEPTEMBER 1984 E&MM

OMPUTER MUSICIANmessage if you try to open anotherRegister or increase the NPHONY bey-ond eight. A s imple rule appl ies: the sumof the active Registers times theirNPHONY must be less than or equal toeight.

Figure 6 shows another example ofthe Register allocations. Here, anotherset of voices has been loaded. so theNPHONY and Registers are configureddifferently. Although the Register andNPHONY settings may seem a l itt leconfusing and limited, the exact con-figuration is determined entirely by themusician, and changes can be effectedvery quickly for evaluation.

Keyboards and TuningAny voice can be tuned in increments

of plus or minus one hundredth of asemitone up to t6 octaves with crystalaccuracy. Scale allows the Westerntempered tuning of 12th root of 2.00 tobe changed to any other macro/microtuning, eg. for quarter tones, you simplychange Scale to 24th root of 2.00. Pitchis a master tuning control which canvary tuning of all the loaded voices by aquarter of a tone in 256 discrete steps,to bring the CMI in tune with otherinstruments if necessary.

The Keyboard Maps each consist of akeyboard number (1 to 6) followed by sixletters indicating the Register assigned

to each octave. As the C[/l is a music-ian's instrument, it supports two six-octave keyboards called the Master andthe Slave. Using the maps, it 's possibleto create eight different keyboard con-figurations by choosing which soundswill play on each octave within a key-board. The Master and Slave can belinked (as shown in Figures 5 and 6) toany map by changing the SelectionnumDers.

The information presented in Page 3 isknown as an Instrument fi le. The fi le canbe saved on the user disk and is giventhe suffix NAME.IN. When this fi le isloaded it wil l pull the specified voicesinto the CMl, allocate the Registersautomatically, adjust the tuning andspread the sounds across the key-boards. Instrument fi les are a usefullyquick way of bringing the CMI up to aplayable state with 'preset' voices andtuning.r t I11arcware

At the time the Fairl ight was designed,the microprocessor was considered tobe a medium- to slow-speed device. Toincrease the power of any computingsystem, designers have two basicchoices.

One of these (the Synclavier approach)is to base the instrument around a dis-crete logic minicomputer, thus uti l isingthe raw speed of logic chips. This is

qui te an elegant solut ion s ince'music bynumbers' requires lots of number crun-ching, but the other choice, and onewhich is becoming increasingly popular,rs computrng concurrency. In a basicCMI system, there are four micropro-cessors of the 6800 family. Two of theseare in per ipherals, ie. one each in themusic keyboard and the alphanumerickeyboard, and this means there can beseveral independent processes beingexecuted concurrently.

The microprocessor in the alpha-numeric scans the keys and passes thedata to the music keyboard whenrequested. At the same time, the musickeyboards are scanned for pressednotes, key velocities are calculated, andthe control sliders and switches read.

At the right-hand end of the Masterkeyboard is a calculator-style keypadand alphanumeric display used for rapidloading of voices in a live situation.Music keyboard information has thehighest priority of all data in the CMl,which responds instantly to the packetsof data sent flying down the cables at9600 Baud.

Well that about finishes off our des-cription ol the utility-type display Pages-In case you're wondering what the Modesetting on Page 3 is for, don't worry: allwill be explained.

Next month, the controls on Page 7and sampling on Page 8. r

OMPUTER MUSICIA

TIIE T'ATn.LTGIITEIKPEATTUED

Part three, and a discussion of how the CMI samples a sound and why itsown particular sampling techniques are employed. Ji* Grant

results in a short sound when the samole isplayed on the keyboard, and this becomesshorter as we ascend the octaves. To over-come this, the Fairlight allows sections of thewaveform RAM to be read out repeatedly (orlooped) as the key is held down, thus sustain-ing the sound. The smallest section of RAM

0t ' l l ' l l iNIr ,

: l r - r l_ lHD tHl ' lFL I HrJ

SRI ' IPLE FTI iTEFILTER LI: I I , IFILTER HIGHIi t ' lFFrEigr:rFi

Sf i I ' IPLE

II ISPLFIY

Although sampling i tself is very simple, andimpressive results can be obtained very | I ll IrE :i -quickly, i t 's well worth the trouble spenrj ing l t

r-rt l t1Hfl I lsome time adjusting the Sample Rate to a Ivalue that suits the pitch of the input signal. l - t I iFtf i i FIFJ'1Fi T +

Zv I EB HiIE

I 'FF

EHI, IPLE LEUELTR I GI]EF; LEUELTFII I - ; i jEFI TIELB\.

c:: . :E t l i

A t last we've covered enough of the CMI/.\ basics to concentrate on the more

, \ interesting sound creation Pages. Now,the single feature that characterises theFairlight in many people's minds is its abilitytosample natural sounds: this aspect is dealtwith by Page B, and is surprisingly simple toUSE.

At the rear of the CMI lies a selection of lineand mic inputs to suit most applications. Thatabout takes care of the hardware. becauseeverything else is dealt with by software.Typing 'S' or touching 'Sample' with thelightpen initiates the sampling process. After asecond or so, the Display box shows thesound envelope for quick monitoring of inputlevels (see Figure 1). lf all is well with theKeyboard Maps on Page 3, the sampledsound will be playable on the music keyboard.

So far so good. But what are the otherfunctions for? Well, some of them are self-explanalory. Sample Level is a software-based volume control and can be used toattenuate signals that exceed the input rangeof the Fairlight. Unwanted frequencies can bereiected by using digitally-controlled high-pass and low-pass filters: their cutoff pointsare set by Filter High and Filter Low. The actualcircuitry lives on the ubiquitous Master Card,and takes the form of switched resistornetworks using the much-loved CMOS 4051chio.

Whenever a signal is converted to a streamof digital numbers, it's necessary to bandlimitit to one half, or less, of the Sample Rate.Stated simply, this means that we must have atleast two sample values of the input signal'samplitude for the highest frequency present. lfthis condition is not met. the information thatthe sarnpling process has captured is notsufficient lo reconstruct the original signalwithout frequency distortion. This type oldistortion is known as 'aliasing' and is bothextremely noticeable and rather unpleasant.The CMI guards against 'al iasing' by incor-porating tracking filters controlled by theSample Rate.

Sample Rate

than can be looped is known as a SegrpeitHere lies the crux of choosing a suitable

Sample Rate. lf we attempt to loop a Segmentor group of segments that doesn't contain awhole number of cycles, the 'ends' of the loopwon't join up without causing a suddenlump inamplitude. Choosing an inappropriate sample

UDIr- 'E, II'r H Fi I l'1 :1

I i l t t : ,

Figure 1.

Ul '1 l ( l l i { I t ,

Figure 2.

RTUNEl40DE, 4I , ]FUEF|]RI ' I ! I 9PLHY

l , lHr l lEF r-1Fr l ' l t r MLx' i .

imporrant aspect of the CMr that obtiges us ro ffili,iii?Frpay attentron to the correcr^:T?1"^ r.u^t:^- ,^ [l:liili.#,The number of original samples taken is llii',:;ttlj,,fixed and equals the length of the waveform [,1]lll'*ni:L1:----.=

:-. r' --rli./ .,'

l - - ts

RAM, ie 16334. The faster these samples aretaken, the shorter the duration of the sound L''Figure3.becomes (although the fidelity increases). This

UiJIUE. 1LTt- t t {El lDIrE, . {

11 Ir ::i E t:; : : r? 5TEF, I

86 OCTOBER 1984 E&MI

I1

lrate results in a dreadful glitch which increasesat a rate proportionalto the pitch played on thekeyboard, lf the sample rate is almost right, aone-segment loop produces a sudden slightpitch shift, and waveform crests and troughs'drift' laterally through a Page D display. This isshown in Figures 2 and 3, where a drift to theright (sharp) is caused by the sample being settoo high and a drift to the teft (ftat) by it beingtoo low.

Figure 4 shows a sound which is in tune withthe system and therefore loops perfectly. Atthe other end of the scale, if the sample rate istotally wrong the display becomes a hopelessjumble (Figure 5). The relationship between awhole number of cycles and each segment ofwaveform RAM is also the relationshiprequired for a visually coherent display. Thussamples that look good will inevitably soundgood, too.

Now, if all this sounds rather complicatedand you're beginning to wonder how anyonegets anywhere near choosing the corectsample rate, then take heart. lt's all in the helppages for Page 8 - see Figure 6. A usefulsample rate table is included, and with a littlepractice it becomes quite easy to arrive at theconect setting within the space of a few trialsamples.

ADCThe actual analogue-to.digitral conversion is

accomplished by a 10-bit converter, eventhough the CMI is an eight-bit machine. Onlythe top eight bits of the sample values arestored, while the two LSBs (Least SignificantBits) are ignored. This improves the linearity olthe conversion, which means that the signalstep size required to cause a conversion valueto change by one LSB is fairly constant overthe range of the ADG. /

The relationship between the amplitude ofthe input signal and the sample valuesgenerated is linear. When the signal levelchanges by a given amount inespective of theabsolute value the conversion code alwayschanges by the same amount. This is wherethe Fairlight differs from most other samplingmachines such as the Emulator. That uses anon-linear conversion method (called'companding') which allows more codes to begenerated for small signal values than for largeones. Whenever a signal is represented by afinite range of numbers - in this case f255(eight bits) - two things suffen noise anddynamic range. The noise is only heard when asampled sound is actually being playedthrough a DAC, ie. the ADC and DAC are not inthemselves inherently noisy. Making sure thatthe peak ol the input signal causes themaximum ADC code to be generated ensuresthat most of the noise is masked by the volumeof the signal on playback.

The Display box in Figure 1 is an invaluableaid in this respect.

Dynamic range, on the other hand, is ameasure of the range of different amplitudevalues that the ADC can handle. In a linearsystem, this is directly related to the number ofbits used in the process and, roughty speaking,the dynamic range of the sampled signal is6dB times the number of conversion bits.Since the Fairlight uses eight bits, this gives 8x 6dB = 48dB dynamic range. Compandingtechniques result in a larger dynamic range(about 70dB)forthe same numberof bilsused,but at the expense of greater noise at lowsignal amplitudes. The reasons for Fairlight'schoice of a linear converter will become moreapparent when we look at the functions onPage 6.

The actual sample rate is very cleverlygenerated on Channel card 1. Normally, theonboard circuitry is used to generate the

OMPUTER MUSICIANconect clocking rates required for digital{o-analogue conversion when a keyboard note ispressed. However, for the duration of thesampling period the CPU grabs Channel 1,and forces it to produce a stream of pulses at afrequency of 128 times the sample rate shownon Page 8. This is supplied to the ADC, whichresides on the Master card, and once thesampling process is finished the CPU restoresChannel 1 to its original task.

Trigger Level is the amplitude threshold atwhich the sampling process is triggered tobegin. When the Sample command is given,the system waits until this level is reachedbefore proceeding. Once the threshold hasbeen exceeded, it's possible to delay theconversion by using the Trigger Delay, whichhas a range of 0-65533 milliseconds. This canbe especially useful when sampling from tape,for example, as a tone burst can be recorded

shortly before the signal to be sampled andused insteadof thesignal i tself to tr igger thesampling process. Trigger Delay can then beused lo define the precise point at whichsampling will actually begin. This is extremelyuseful for sounds with a gentle attack such asslow strings.

Lastly, the Compressor is a software switchwhich controls a hardware option. Basically,this turns the conversion process into a non-l inear system, thus enhancing the dynamicrange. The electronics use the same type ofcircuitry as that in many analogue compandingsystems. However, very few Fairlights arefitted with this option as it can have a strangeeffect on the commands on Page 6.

Well, that about wraps it up for Page B.There isn't room this month for a discussion onPage 7, so we'll have to leave that for nextmonth. r

ITUIIL

I

.l

I iLEi ;t r lJ8i l I l

Figure4.

I't0l ' l l ' lFNIr '

l , lSuEFrtPi l I r I : :FLfr i ,

I , I ITUEFcJRI ' I D I SFLRY

I

i

JTUNEt'100E'

F|: lRI '18T.F 4 32 STEF'12

Sff I , lPLE RI lTE9noie +gvts -Evr - tE: , " ' i

14817 29S35 74SS 3?t3E 15884c 18744' ,

17748rr sTsstsst2

E 2I BSBF 22351

?3EBB

Ti iEr! : jB5 IEr:r7e 4 I FE8r:r lB 44:{53337 4EtSi ig5E 4S7' i l

1Erg49 5t7 ' {I I I I i , i i , , : , ,1

l l t r4El Sr l lFI U 544 Etr?t3t38 EE45

E&MM OCTOBER 1984 6l

u..:=;

fit

ETHE EATn'LTGHTtr)KPEATTUtrD

How the CMI provides for special effects and looping - in a language just

about everyone can understand. Ji* Gra,nl

I ast month we dealt with one of lhe

I Fairl ight's main sound creationL methods: sampling. The simple act

of pointing a microphone at a soundsource and typing 'S'on the alphanumerickeyboard transforms the Fairl ight from anexpensive computer into a powerfulmusical instrument. And the keywordhere is 'musical'. The abil ity lo create newand interesting sounds (or indeed samplethem) is not in itself enough. What isrequired is controlover that sound and, tocoin a popular phrase, the control mustbe real-time. Musicians, of course, callthis control 'expression', and it 's a par-ticularly diff icult feature to build into acomputer-based musical instrument.

Consider a typical case in which aFairl ight user might be playing the musickeyboard while l istening to a sequencepre-recorded on Page 9. Everything isrunning smoothly: the CMI is readingsequence information from the disk, sort-ing it out, and sending the data to thevoice channels to be played. At the sametime, the music keyboard is being scannedfor pressed notes and more data sent tothe channel cards: notes are stolen ifnecessary. Next the user may decide toswell a parlicular voice by moving theappropriate footpedal. Here the CMI isforced to deal with an asynchronousevent in the normal proceedings, so thepedal value has to be updated constantlyand the values obtained used to scale theamplitude of the voice throughout itsduration. And if this were not enough, theFairl ight has 17 parameters capable ofbeing controlled in real-time.

I t 's the unusual mult i -orocessor archi-tecture of the CMI that enables it tohandle so many asynchronous taskssimultaneously, but let 's move on to lhepresentation of the controls and their use.

D^ -^ - lrdtc /Al l the controls are handled by Page7,

and a typical display is shown in Figure 1.The page features all the usual controlsassociated with processing sound: eachof the eight voices loaded can have itsown unique control set t ing and can besaved to disk wi th a chosen f i lename andthe suffix CO. When a voice is loaded intothe CMl, i t wi l l pul l rn a speci f ied controlf i le if i t was previously Linked to the voice,usrng the command LNK.

At the bottom of the display is a boxwhich indicates currenlly-loaded voices.The gontrol f i le may have a di f lerenl namefrom its intended voice so the two aredrfferentiated visually. The active controlf i le is the name hight ighted whi le the

active voice is shown in the top right-hand corner. Other control f i les can beinspected by pointing the l ightpen at thenames in the display box or by typing'V.n ' . where 'n ' is the voice number.

There are six real-time faders and fiveswitches patchable to most parameters.Threie of the faders and two of theswitches are on the left-hand side of themusic keyboard, while the other faders(or footpedals) are accessible vra Cannon-type connectors on the rear of the key-board. In addition, the music keyboard

passes key velocity information to theCMl, and this can be patched to Leveland Attack as KEWEL.

Below the voice l ist in Figure 1 is acomolete l ist of the controls and switchesthat are available. This is used in conjun-ction with the l ightpen and provides aquick way of patching the controls tovai'ious paramelers: the lightpen is pointedfirst at the parameter and then at thecontrol l ist. A patch can also be estab-lished by tabbing a cursor around thedisplay using the QWERTY keyboard andtyplng in the appropriate name or numericvalue. Figure 2 shows one of the 'Help'sheets for Page 7 which provide a quickreference for the range and possiblepatches available.

Some of the control parameters aresel f -explanatory, such as 'Level ' , 'VibSpeed' and 'Vib Depth' . Again, we comeacross the enigmat ic 'Mode' switch, whichis best left unti l Pages 4 and 5 arcdrscussed (fhe suspense is kil l ing rne -Ed). 'EXP' is the other half of the com-panding process thal was an option onPage 8 discussed last month. As youmay remember, i t 's a hardware opt ionand is very rarely fitted to the CMI dueto the non- l inear sampl ing data that

results from its use. The Filter is a lor'.,-pass tracking fi l ter resident on eachChannel card, used to at tenuate anvunwanted high-f requency conten:present in the voice: the cutoff f re-quency is raised by simply increasingthe value. l t 's a l l real ly a case of svr ingsand roundabouts - a high f i l ter set t inogives a bright realistic sound but oftenwith digi ta l b i rd ies warol ing in thebackground, while low fi l ter valuessuppress any funnies but reduce thesound to a dull noise.

When Portamento is on, each Chan-nel allocated to the voice produces acontinuous glide between each nei., 'pitch it is to play and the last pitchplayed, the rate of note glide being setby the Speed control. 'Glissando' drf-fers from Portamento in that the glideis not continuous but chromatic, andall the notes on the keyboard betweenthe stad and end notes are played. lfboth Portamento and Glissando areselected, Portamento takes preced-ence. 'Constant Time' is a switchwhich selects between two types ofglide: when it 's turned on, the sametime is taken to travel any musicalintervgl and the rate of change altersac_cording to that interval, hence thename 'Constant Time'. This results inpolyphonic portamento or glissando.in which the notes arrive at theirdestinations at the same time prod-ucing a coherent chord. With theswitch off, the rate of change remainsfixed (determined by Speed) and thetime taken to glide varies with the sizeof the interval.

Attacl< and DampingThe Attack parameter has a range of

zero lo 16,384 mi l l iseconds, and may bepatched to 'KEYVEL' for touch-sensitivecontrol of the at tack t ime. l t 's act ive onlyfor Mode 4 sounds, and is extremelyuseful for imposing a degree of artif icial

' enveloping upon sampled sounds.'Damp-ing' has a range of zero to 65,536mil l iseconds, reduced to 16,384 mi l l i -seconds in Mode 4. The value determinesthe f inal decay t ime of the voice, ie. f romkey release to s i lence. l f a loop is act iveand one or more segments are repeatedcont inuously, the voice plays the loopunt i l the damping t ime expires (when thekey is released), o lherwise the voicecont inues through lhe remaining seg-ments. Should the end segmenl bereached before the damping l ime expir:sthe voice stops abrupt ly.

The 'Slur ' switch is useful f or o l issando

OMPUTER MUSICIANIj 'and portamento ef fects, as i t causes

sustain indefinitely in a loop that may be

[' i'tt i, i-:::

l i r r r l r ' lFr i I r

Figure 1. I

lChannel cards allocated to a voice to

ive unti l a new note is played. Newnotes are started at the beginning of theloop without playing any of the precedingsegments. 'Sustain' determines the be-haviour of the voice once the key isreleased. Normally, a voice fades outeither playing its loop or unti l i t hits theend of the segments, but when Sustain ison, Damping is ignored and the voiceloops for the duration of key depression:upon key release, the voice continues toplay its remaining segments with nodecay of amplitude.

I [ iHF I l lE = : ' :1r : it lTTBt l l , , = Et

UiI I :ET

r l f lTFrL, I : l

f r l l i rFF iFpf l

I t i lEt ir:r tl l'1 F tt Il

r : r l l 'TFt i ,L PFPRtlE T EF:3

r:r : r l lTFtr : rL FILE ' lEIr?

l l r r l rE=4 r : iL l ! : j t i r { I rB=r lFFE' lF = r : rFF FL:LFrTFrjEftT0:0FF

LEr.rEL : tE\ , r rEL l_ iF,EEIr =,_lFILTEF: = l ! r_:Uf l : : iT TI I ' lE = t t l t

V, /17'//l

t. l tt I t- 'E , IX! IJ2

t: tl

t , l I I r ITEPTH = Bl l lB iFFFn = ! : !

t i

Lr l r lF r : f iTF:LLI] I ]F ' :THFITLr: i l : rP Ll i r lTH

' I I nE I i 'Lrr

Lrr r

IIB:T

. :htr_Fl- :

r - t iTF,Lt-r ,I ' lE\ r , rEL

: lL lJFl = l : lFFSUETBIH = DFF

nnnil ET tCt( lH:l ! ;r lT : lTFtl l I Ir IrTF: l '110e l i lELLDc|D T8l t f l r lRf lGsB

:1 l , lTtH, l . 4 5

h-EYUELI ooninq

r" 'bChoosing the correct looping point of a

voice wavelorm can make or break agood sound on the Fairlight. Nasty glitchescan occur if an inappropriate sample rateis chosen or if the section to be loopedspans a natural change of amplitude.

lmagine trying to loop a percussivesound such as a drum. The three loopcontrols on Page 7 provide a quick way offinding the best loop, and a typical settingmight be as shown in Figure 3. HereControl 1 is used to define start point ofthe loop while Control 2 sets the length.Switch 1 freezes the effect of Control 1and Control 2 when off, preventing acci-dental movement of the looping pointsonce these have been decided. A usefulfeature is that loop parameters are svedwith the voice information as well as anyLinked control file, so that the sound isplayable even though no performancecontrols are required. The actual looppoints are displayed graphically on Page4, as shown in Figure 4, where thehorizontal axis represents the segmentnumber and therefore time. The loop isindicated by the row of highlighted boxesunder the Harmonic Profi les graph, andsince the voice shown was sampled,there are none of lhese present. ThisPage offers a convenient method ofselecting looping points using the l ight-pen.

'Start Seg' is a powerful expressioncontrol. lt allows the starting segment ofthe voice to be chosen according to acontrol value as a new key is played. Toexplain: suppose we had sampled theclassic synthesiser fi l ter sweep and play-ing the keyboard resulted in a 'fruity'decay. Using a control fader to set thestart segment would then enable us toplay the synth sound from ditferent partsof the fi l ter sweep. As the control wasmoved from segment 1 to 128, the soundwould begin with plenty of f i l ter sizzle atlow control values but become shorterand more mellow as we started thesoundfurther down the sweep (by increasingthe Start Segment number). This tech-nique can also be used to control-theamount of 'breath' on sampled windsounds or the amount of bowing onstringed instruments.

Next month, page 5 and some revel-ations concerning the mysterious'Mode' . r

E@m@: :h, : , r ,1! : r3n.-r : : : ;n i i Ft , t ! iL ' l * p: t , :h*:- i ' : , r

! th t ,3n F 3r 3f i r* tPr

[ l , rntr , : I P. : r . inreter Ranq*, F:1, :hl lL l t rE 1, ILI,F

I I L I LFJ

NHNP I TID Lt- t t : ,J5BTTTiDK 8-16: l :13EL I S9 NIIO ON, OFFPgRTIi I ' lENTIJ OT.I , OFF, ]TLL! 8- l !7I : IJHSTRNT T I TIE OH, BFFUIBRIITI I IEFTH 8-I?,7UIBPIITD 9FEED A-I?7LI]DP I ]ONTR|]L ON, DFFLDDP STf,RTL00P tElrETH S-128STBRT SEDMENT I- I28SLUR9U5TFI IN

HFIRt l0NIC

Figure 2.

Figure

LE(IEL rr l - : '99 | l . :r - l l l . L- IFF ! ,

f

ON, OFF0n,0FF

t:

LI

:i

c.f

cID

cDci

lI\J

IIII\I

IItt

II

\

i

II

\I

UUILL:gTEI ' I IDI

CONTROL PRRF}lETERS

TONTROL FILE' LOOP CD

HoI 'E = 4 r :L lggFt iDoEXF,= I ]FF PNRTFII IENTO

LEUEL = KEYUEL SPEEIJFILIER=8 r lDt iSTTlHE

5l, lTcH3 Lr i0P Ci l rRL = : :HTr-H IOFF LO|JP STI iRT = IT{TF;L 13{ LIJI ]P LN.GTH = DNTRLZOTI :JTffRT sEG = I

CNTRL3 SLIIR = Sl , lTr-HaS SU9TFIN = t lFF

DFI ' lPI l lG = SERTTf,CK = IF

UIB I IEPTH =UI8 SPEED =

UOIEEs,

INTRL.

EII I I? STICK BB9SGT EI:$T|I IT9TFTNINS CBELLOBD TRHN GRNGSR

SI, ITCH.

STRtl l I r lH0I |E, q

L00p l [ l |n PLDT

.1

THE T'AIn'LTGIITETKPEATTYED

Sampling may be the CMI's most talked-about feature, but as this articleshows, defining sounds using harmonic information can be just as dramatic.

o far we have discussed only onemethod of actually creating soundswith the Fairl ight sampling. This

aspect of sound formation is probably thesingle most important feature of the CMl,and was cerlainly the focus of publicattention when the machine was an-nounced. Horvever, the ability to specifysound by means of harmonic informationcan not only result in some very interestingsounds, it 's also rather useful in aneducational environment.

Two display Pages, 4 and 5, allow theconstruction of waveforms by harmonicdata. They deal with exactly the sameinformation but present it to the musicianin different ways.

First of all, though, let's clear up a l itt lemystery that's been evading us for somemonths - the Mode switch. Actually, thisis very simple and is therefore somethingof an anti-climax. When a voice operatesin Mode 1, only the first 32 segments ofwaveform RAM (4k bytes) are used torepresent the sound. An unlooped Mode1 sound wil l stop at the 32nd segment,even though another 96 segments ofRAM exist. ln order to compensate forshorter note event time as played on thekeyboard, each of the 32 segments islooped several times before moving on tothe next segment: this maintains a fairlyconstant net event length for any pitch.Mode 4 uses the entire waveform RAM(all 128 segments of it) and is always usedfor sampling since long, high bandwidth

Frgue 1.

f ' a\Izm GrantJ

sounds need lots of numbers to representthem.

So what's the use of Mode 1? Well,calculating a time waVeform from har-monic dala can be quite time-consuming,especially if the supplied data is detailedand enables subtle nuances of sound tobe generated. However, more often thatnot, only a simple waveform is required,and to calculatethe RAM waveform for all128 segments when a short loop is allthat's needed is rather wasteful, to saythe least. There's no hard and fast ruleabout which Mode a sound should be in:the choice is entirely the musician's.However, using a voice as the destinatlonfor sampling dalaalwaysresults in all 128segments being ovenvritten, even if thevoice selected is Mode 1.

h?rage )Figure 1 shows a typical Page 5 display.

This page displays the harmonic overtoneseries as a set of 32 'faders' similar tothose on a graphic equaliser. Each faderis logarithmic in nature and has a range ofzero to 255, allowing a good degree ofcontrol over harmonic amplitudes andthus enabling the application of a Fouriertype harmonic series.

As an example, Figure 2 shows asquare wave generated by the CMl, com-puted from the values of Fourier com-ponents shown on the faders. The result-

ant waveform is visually very similar to thereal thing, and perhaps more importantly,sounds indist inguishable.

However, the power of the CMI l ies in itsabil ity not only to compute a complexwaveform from a set of Fourier com-ponents so that it can be played on thekeyboard, but also to compute a differentwavelorm for each segment. Every seg-men\ has a unique Page 5 display, sowhile a Mode 1 sound has 32 sets offaders, a Mode 4 one wil l have 1281 Thecurrent segment number is indicated onthe display, and this allows a synthesisedsound to change drastically throughout

FBue2

i ts duration, simply by the user fi l l ing eachsegment with a different waveform calcu-lated from its own Page 5 fader settings.

ln fact, the technique of using differentwaveform segments as the sound pro-gresses is very much the domain of PPGsynthesis. Generally speaking, these pro-gressions are known as wavetables,and in the PPG, a sound consists of a setof 64 waveforms that reside init ially inEPROM (they are transferred to RAM onpower-up) which are read out sequentiallywhen a key is pressed. The idea behindthis system was to circumvent the needfor f i l ters by constructing wavetables thatheld a set of reoresentative waveformsof, say, the classic fi l ter sweep. Unfodu-nately, this results in a very hard, metall icsound, as the sound changes abruptlyfrom one waveform to another, slightlydifferent one. lt 's sti l l a good sound, but inthe interests of f lexibil i ty, PPG havechosen to incorporate the usual VCFsand ADSRs as well as exiensive wavetablemodulatlon.

The CMI is also capable of this forrn i, lsynthesis to a l imited degree, using theloop controls on Page 7. For example,

HRRNON I C Ff iDERS

UOICT' IS11UBRE11 ODE' 1

XHD,

NT

4

0

r lF

KL

J

Lt{

Nfl

K

c

z

0

Uc

I

I

1 r 1 t l l I I I I 22????2??23330 r 234se7 890I234567890l2I

E

SEGI' lENT

5678

c0t lPlJ

ntratrrratra 10Al t r tMl '

suppose we had fi l led all 128 segmentswith waveforms that change very slightlyas we progress through the waveform(see Figure 3). Now, if the loop controlswere set up as shown in Figure 4 (this is aPage 7 display), moving CNTRL1 on themusic keyboard would result in a differenttimbre when the note was played. Usingthis technique allows for some expressiveplaying, since the principle is rather akinto varying the filter frequency control on asynthesiser, the only difference beingthat the actual timbres can be radicallydifferent from one segment to the next.

To increase the timbral movementwithin a sound, CNTRL2 can be patchedto 'LOOP LENGTH' on Page 7, resultingin sections of differgnt waveforms beingread out repeatedly. Since the waveformdata is computed by the CMl, it 's alwaysconstructed so that the waveform fitsexactly into one segment, thereby over-coming looping problems.

Fourier SeriesWell, with all this talk of 'Fourier com-

ponents' and the like, some of you mayreasonably be thinking 'what's it got to dowith music?' The answer, of course, isnot much. Only scientists and engineersdelight in quantifying the world which oursenses seem to handle perfectly ade-quately. However, in order to exprbssourselves explicitly and unambiguouslyabout a wide variety of concepts (some ofwhich may be abstract) we need to usethe language of mathematics. Fourieranalysis and synthesis are mathematicalstatements about something which is notintuitively obvious: the fact that any trulyperiodic waveform can be decomposedinto an infinite sum of sinewaves, usuallycalled harmonics. Similarly, any periodicwaveform can be constructed from thesum of an infinite number of sinewaves.The sinewaves have frequencies that arerelated to the fundamental of thewaveformin such a way that the second harmoniclies at twice the fundamental frequency,the third harmonic lies at three times thefundamental, and so on. The fundamentalitself is often referred to as the firstharmonic.

Of course, obtaining a reasonable rep-resentation of a desired waveform doesnot require an infinite number of sine-waves: more than 16 is enough to give agood approximation. The Fairlight uses amaximum of 32 harmonics, which enablesmost waveforms to be synthesised with afair degree of accuracy. Figures 5 to 8show the development of a square waveby successively adding further harmonicsand using Page 5 to compute the resultantwaveform. The square wave doesn't con-tain any even harmonics (2, 4, 6 and soon) and we can see that the lowerharmonic numbers set the basic squareshape while the higher ones fill in thebumps and sharpen the edges. Even withall the odd harmonics the Fairlight cancompute, the square wave is still notvisually perfect, but it sounds OK. I

0t ' tH8NIt ,

Ftguo4

I, IFIUEF| jPI l N I SPLRY

CONTROL PSRRI' lETERS

CONTROL FILE. I , IR

HODE = 4 I ILISSFNI0EXP = 0FF P0RTFHENT0

LEUEL = KEYUEL SPEEDFILTER = 1l t l0NST TIt4E

DFHPING = lAg UIB DEPTHfiTTffCK = 0 UIB SPEEI]

ut I| ]FFaOFF

L|] | ]P CNTRLLOOP 9THRTLLlOP LNDTH

J t lK I CLb

SI, IEEP

ONC}I TRLlI

CNTRLT

=&=g

SLUR = 0FFSUgTFIN = 0FF

Flgnr,o5.

FilNTiiI

1:- t45

FI

Fgtre8.

UOIIES,

ili'i"i"iI t lITTI

, l

E&MM DEGEMBER 1984

TIIE T'ATRLTGIITE)KP&ATNED

Waveforms, lightpens and interpolation all come under examination inthis instalment of our Fairlight CMI Grand Tour. Ji* Grant

ust when you thought it was safe toopen up a copy of E&MM withoutreading anything about the world's

most influential comouter musical instru-ment, your intrepid reporter returns froma New Year hangover with another action-packed episode. This month we look atthe information presented by Page 5 in aslightly different l ight. You'l l rememberthat Page 5 held the values lor 32harmonic faders and computed the result-ant waveform for the current segment.You should recall also that the only way tocreate a complete sound of 32 segmentswas to define the fader levels for eachsegment and compute over the wholewaveform; or define a few segmentsand Fill the harmonic data to the rest ofthe segments before computing. lt's nothard to see that this method of creatingsounds may be very precise but can alsobe extremely tedious. In a lot of cases, allwe need is a way of tailoring the harmonicsas the sound progresses: harmonicenvelopes, in other words. Enter Page 4.

Figure 1 is a typical Page 4 display, andshows that it 's one of the two Fairl ightdisplay Pages to be almost exclusivelylightpen-driven. The large dark area is infact a reverse video image, and pointingthe l ightpen in this region results in anarrow cursor appearing on the screen atthe current l ightpen position. For thoseunfamiliar with the term, the l ightpen isnow a fairly common computer add-on,mainly because of its simplicity of opera-tion. Contained within every l ightpen is afast photoelectric diode or transistorwhich produces a voltage pulse as the Wline passes beneath it. Usually, the pulseis squared up and passed to the videocontroller chip, which stores the W linenumber and position along the l ine in acouple of registers. This information canthen be used by the programmer toinitiate predefined evenls such as plottinga point or executing a command.

The Fairlight system is no different,except that the video controller is con-structed from discrete logic chips andresides on a single eight-inch boardwithin the CPU. In addition to latching theTV co-ordinates when the l ightpen isused, it generates an interrupt to theprocessors to execute the selected task.

At f irst glance, the graph area in Figure1 looks a bi t confusing but i t 's real ly qui testraightforward: the vertical axis repre-sents amplitude while the horizontal showstime and hence the seqment number.

NIIEi ;rl tl |1 ri ll tl

I .LEHF, [ I ILETE

OMPUTER MUSICIAN

HfiFi I ' ION I ROFILES

Along the bottom of the display are theharmonic numbers 1 to 32. A smalltriangle under the number indicates thatthe time profi le of that harmonic is beingdisplayed on the graph, while a crossshows that the profl le has a non-zerovatue.

So what does all this mean? Have alook at Figure 2. Two profi les are shown,one of which is the First harmonic (left toright downwards) and the other the Third(left to right upwards). On receipt of aCompute command, the CMI wil l f i l l thewaveform segments with sound whichinit ially at least, has a strong fundamental

HFR145l ' l r l l rE, I

mrlfl F Lr:r T

t1

n

r l l l rI [ ! : i

l l l t t ll ! i45r

tr

!2:rrJ: j l! : tEt1c"

trFF' I-I 11 Eff

| r : r l r :E, IH B FI I,1 ::;f l n lE' I

i tFt- t

I l l I rEi :r: r:r 11 f1 H fl Il

HFiF]11[ iH I I : FF] i IF I LEi

Ir irf l l FLrrT

. ! i : i : j i : ii ts j r l i t !

trl: f1 li'fti

L r:r r:r F

MARCH 1985 E&MM

' tI

l .

:)ut degenerates into dominant thirdrarmonlc.

,. lank profi le graph.

i"lore About ModeOne of the previously mentioned

features of a Mode 1 voice is the way inwhich the first 32 segments are loopedseveral times to maintain the net eventtime of the sound across the keyboard. Infact, we have some control over how longa segment lasts before everything moveson to the next one, and this is accom-plished vra the profile. Figure 4 shows aharmonic profi le graph with the durationprofi le indicated by a double l ine: thedefault value is approximately 50mS persegment and increases as the profile isdrawn higher up the graph. This is parti-cularly useful for creating sounds with ashort click at the beginning of each note,such as that of a Hammond organ. Averyshort duration value can be drawn for thefirst one or two segments, and then alonger profi le for the remainder of thesound. lf the duration profi le is madezero, the sound degenerates to a Mode 4condition, except that it only lasts for 32segments.

Another interesting aspect of Mode 1sounds is their ENG profi le. This is anartif icial envelope that's superimposedon the waveform in much the same waYas the more usual ADSR principle. Butthis one's a lot more flexible. When theCompute command is given, the CMIcalculates the waVeform segment bysegment and scales the amplitude sothat it f i ts exactly into the dynamic rangeof eight bits. The ENG profi le is alsogenerated (and its shape implied) by theharmonic data, but can be altered by thelightpen to control the amplitude of thesound on playbacltr

A' ' I 'FAuxrlary FunctronsAlong the very bottom of the display

Page are a number of useful commands.Clear deletes all the displayed profilesfrom the graph, but they remain activeand can be brought back to l ife simply bythe programmer selecting the harmonicnumbers with the l ightpen. Delete, on theother hand, merely removes the currently-selected profile. The same soft of structureappfies to Reset and Zero. Invoking

l [ r f1 l '1Hf1I l

l l i r rEFI: 'F l1 I r I :FL H,

t r iF :

r l : I t

t11 !_._r{r , _ l : j_: , t_!EB_.__:TEF [ | . c :

EE

HirF; l l i r l l I r - : PF:r tF I LEg

l l l_r I l - :

l lFFr l . ' :l'1 [r Ir I

mfnl rL,: jLUI: IP

el45

C

Reset causes a confirmation message tobe printed and also results in Page 4being restored to a complete defaultcondition. Zero isn't quite so drastic, andresults only in the current profi le being setto zero. Every time you give a Computecommand, a new energy profi le isgenerated. Scale is the opposite: it re-drawsthe harmonic profiles from a modi-fied energy profile. This is not without itsdangers, however, as it can result insome harmonic profi les being scaledbeyond their maximum amplitude, whichleads to clipping. The Fairl ight w7l informyou of the situation when it occurs (bydisplaying 'Overflow') but is powerless toprevent it happening if the Scale commandis issued. The only way to recover thesound then is to reload the voice.

Time now to introduce another conceotwith which some of you may not be overlyfamiliar. lnterpolation is the skill of guess-ing an unknown value that l ies betweentwo known ones, and is commonly usedto predict values of points on graphs thataren't the actual ones originally plotted.When the Interp switch is On, eachwaveform segment is computed from amix between the harmonic nrofiles of thatsegment and those of the next one. Thedifference between the two is subtle, andis only really noticeable when the profiles

I t l t: 'a: '

l r i ir r : lEr I

' - - t " ;* , . l * u . - , - . t . - , . - .

i i l i FI

r l iFi I i; l

oI : :LEFF, I IELETE F.E'1ET l ' IFr: i r : iJ t l FUTE ' : r - t l LE I t1 TEF: F r- ' . Ef , t i

contain rapid changes throughout theduration of the sound.

Incidentally, becoming proficient atusing the lightpen for drawing can take alot of practice, so the CMI helps out byproviding a Join Plot selector: when Joinis active, any two points struck on thegraph are immediately connected bya straight line, while fine detail can bedrawn by selecting Plot. I imagine thatmost of you wil l be familiar with theFairl ight's Loop function by now, so Iwon't go through it all again. Suffice tosay then that Page 4 offers a quick way oldrawing the loop start and length. Mode 1sounds are always calculated so that thewaveform fits perfectly into a segment:the first harmonic does one cycle, thesecond harmonic two cycles, and so on-Gone are the Bad Old Days of trying tosample a sound to make it fit segmentsevenly. All you have to do now is use anyold loop to span the sections of a wave'form that are of interest, and Bob Moog'syour uncle.

Next month (yes, there's still more tocome), we'll take a look at Page 6 which,arnong many other weird and wonderfulthings, allows you to splice a sound downto no more than 16,384th of its length.Ahd you thought a razor blade waspowerful . .. r

E&MM MARCH 1985

TIIET'AIRLTGIITE)KPEATTUED

When a CMI page is as helpful a source of music graphics as Page 5, you canbet your l i fe i t needs a lot of explaining-so this month's episode is

dedicated entirely to ()ru,rttit.J i,ntI f you're trying to create a new sound

I on the Fairl ight, whether it 's fromI existing sampled data using Page 8 orharmonic profi les generated on Pages 4and 5, you can be sure that Page 6 wil l bereferred to over and over again.

At its simplest level, Page 6 is a siatic'oscil loscope' containing one segment ofthe waveform RAM. Like Page 4, thelarge black area is a reverse video imageand represents the region that can be hitby the now-infamous lightpen. But unlikean oscil loscope, which acts only l ike awindow on the information, Page 6, inconjunction with its commands and light-pen, behaves more l ike a door throughwhich we can directly access the wave-form data.

The Manual ApproachRegular readers of this series wil l pro-

bably recall that each segment consistsof 1 28 bytes, and that each byte can holda range of values from -128 to 127, givinga grand total of 256 possible values. Now,we can highlight any particular byte bypositioning a narrow vertical window overthe display area. The byte in question,shown in Figure 1, has its position indi-cated by P0INT and i ts value byL E V E L: these can be changed by typingin different values from the alphanumerickeyboard. So, theoretically at least, youcould type in 128 levels for each byteshown in Page 6, and repeat the processfor all 128 Page 6 displays to cover theentire sound, giving 16,384 levels in all. Inpractice though, such a feat would takean unbelievable amount of manual labour(exactly what much of the Fairl ight'ssoftware was designed to eliminate) toachieve, and this has led me to wonderwhether this feature wil l be droppedwhen the Series l l l Fairl ight, completewith waveform RAM of megabyte dimen-sions, makes its appearance later thisyear.

A more practical approach is, ofcourse, to make use of the l ightpen. Thisis as simple as drawing on the back of abus ticket, but just in case even that isbeyond the user's artistic capabil it ies, theJ 0 I N and P L0T funct ions also presenton Page 4 are available here, too, to helpout with the drawing of geometricsnapes.

In fact, trying out a few sample sket-ches soon reveals that waveforms whichlook drastically different may sound sur-prisingly similar. Scientif ically speaking,this is probably due to the fact that as a

I i l [ r t t , , :r:r l'1 ll H ll Il t l

l r i r T r- C

5lrE * f

l t H | | E F r : rF 11 IrF: B! i I t t r j

llntllEfliJilIl I::FiBFHtillf,G[m r:: E [i t{ [ f{ T !:::: i .-r r, r, E r, tIiIEf,I[E';,'11i1n -3 ;1';{FIEf,ffiEIID :-t:F'p i r' -r il i !F 1 :?[ I f tMEm rnjr ]Ff ' -1rni i1 , : , i l i t i ' : : t i ' : r r

: rr ,:lT ' iFE

r , rh ' : t 'E

T iFE

ul ' lNr i f l I l

: : ;HI ' lPLE RI iTEF I LTEFI LUI, IFILTEE HILiHtr-rHF'FrEgSuR

iHt '1FLt

TII : lFLH\.

t5Et5u 1,1!

: : 'TEF F:BTE,

iTBFiT

r j l r_rF

Figure 1,

I?fjltrro{E[ r.I.3L.'e f 'rr rr nri r:i i I i i 3t i,:'r-lfiirilUfiflfifi :-rr i tf hEffil] ,: ,:, n, n, 3 r;ilm[film L-: i ft, Iir .-rr'ilnt[ i l Imfi f f i l ' . , i l r t t , : ,r LllililI] ,: ,:, ri rlr 3 i'riltifffn ': ,:, ri I' 3 n iltlf{iliti ,: i, r,r L, 3 r, '|:lIIIIH i ii tri tr 3 r-r ilmlffi[i ':,:,rir ti' :ti,:lnEfitrFEm ,: i,L,r,i 3trilR!roI{#Tl i,-l r,r i, i r-, ilIff,illEnnB ,:i, r,,n -JritllifiE : ':,r,ir,, i,r',:]t [ i l l f l r i l i t t i , -1 r ,l![AlffilfffE I r''| :-tl'r'rin+r'iI r f : : : : , , , r r t l ' , L I r .HTiI i i , r r -

[[IG 'rr

r : i i r Uf l I r i t i l ' IFLINC

I 4Ej i jBI.1

r : rFF

5FI ' IF.LE LEUELTFIIT:GEE LEI, IELTR I I I ]ER I IELHT

.JPl ' r*r 3t l , : , r l

l ' l r iRI t l t r l

APRIL 1985 E&MM

species, we have litt le prior experiencegained from our senses that relates visualt ime domain waveforms to what weperceive in terms of harmonics or par-t ia ls.

But Page 6 waveform-drawing is ofenormous help, f irst because it gives theuser extensive and direct control olwaveform RAM with one sweep of thehand, and second because the abil ity tobring about such drastic sonrc change sosimply and visually (almost artistically,some might say) is a practicality l ightyears away from the sophistication andcomplexity of the underlying machineand its software. Or to put it another way,it makes the CMI seem less of a beast andmore of a pet.

The Practical ApproachAlteration of the display image is not

reflected in the waveform RAM until theF I L L command is used. This allows thereverse video field to be used as ascratchpad area in which waveshapescan be developed before they are finallycommitted to RAM.

Looking again at Figure 1, notice thereare two 'sl ider' controls similar to thoseon Page 5. These control the segmentnumber to be displayed and the steppingrate through the entire memory instigatedwhen the command Step is issued. Alsoshown are three classic waveforms -triangle, sawtooth, and square - whichdeposit a perfectly fitting single cyclewave into the segment. The pulse widthof the square can be varied from 17o to99% by changing the value held next tothe square wave symbol.

Auxiliary CommandsIn addition to the graphic capabil it ies of

Page 6, there are a number of commandsthat can only be entered from the Fair-l ight's alphanumeric keyboard. Figure 2shows the range of functions available,as presented by the H E LP sheet menu.

First off are a f6w uti l i ty-type com-mands. GA I N scales the displayed databy a specified percentage: if the rescaledwaveform is about to exceed the ampli-tude range of the system, the CMI wil l askyou whether or not you want to proceedand thereby induce clipping. lf you replyin the affirmative, the command wil l beduly executed. Meanwhi le, the I NV E RTcommand inverts the phase of the wave-form: this is useful as a prelude to some ofthe other funct ions such as MIX,MERGE and ADD.

A parlicularly neat l i tt le corirmand isZERO, which al lows us to create a nul lvoice in preparation for the A D D com-mand. The entire waveform RAM can be

i r t : I t : f Ii '18F,I 11I B

:59r5

I l '1':,

1.1i : 'J

r_r F i-

i55t5

E I ' l l

F] : .HI]F 'LETF: l r . r : IF,TF' I I : iL,EF'

LEI, IELLEIIELII ELBI

Figure 4.

I lJ I r t f lulr l ' lFf l t l

r:i r:r l_l lj Ir ::i H l\l F, L l il [

: iHI IPLE EFITEF I LTEF Lr_r i lFILTER HIIJHr: Lr l ' lFFiEi t rJR

i lil'lP L E

II ISFLItY

I qi l : lU HiIiJ

r- lFF

iBI{PLE LEI. IELTFt l r : r : ;EF LEr.rELTE I i r : ;EF: t rELhi '

Figure 5.

turned end on end by using R EVE RS E,and this results in the backwards soundsthat have become familiar to pop musicfollowers the world over.

REFLECT is a less commonly-usedbut if anything more interesting variationon this theme. lt allows us to olace animaginary mirror in the sound and reflectevery part of the waveform in front of thismirror to waveform RAM behind it. Take alook at Figures 3 and 4 for before andafter views: the 'mirror' is at segment 64.

Now on to another of the seeminglyinsurmountable problems brought on bythe onset of new musical technology.Very often, when a sound has beensampled, the beginning of the captureddata does not occur at exactly the start ofthe RAM, perhaps due to some extra-neous noise pretriggering the ADC. How-ever, if the effect isn't too severe, aconvenient method of correction is toR 0 T A T E the sound within the waveformRAM to bring the start in l ine with byte 1.This has the (often desirable) side-effectof shifting the first part of RAM to the end.See Figures 5 and 3 for another revealinqbefore and after

l imagine N0ISE wi l l be fa i r ly sel f -explanatory to even the least clued-up ofthis column's readers: it f i l ls sections ofthe RAM with the output of a randomnumber generator hidden inside the sott-ware. Funni ly enough, the G A I Ncommand is also used to tailor theamplitude of the noise to that of the restof the sound...

Now, BLEND is a strangely out ofplace command. Personally, I think itshould be on Page 4, since its role is tohelp smooth out glitches caused byimperfect looping points. lt may be that tofind a good loop it makes extensive use ofextrapolation techniques, and that see-ing as this is a central feature of the M I Xand MERGE commands, Fair l ight deci-ded to bung it on Page 6 alongside them.

However, rather than discuss the re-maining Page 6 commands in any detailr ight now, I think it 's probably best for allconcerned if I demonstrate their power inthe context of a typical edit session inwhich a sound is created from scratch, ie.without any sampled data. So, now youknow what l ' l l be talkinq about nextmonth.

I f i i i

- ' ' r l l l r i l r i

: i i I 'FLI FHIL

FIt . I I i Lr : r t ]I - ILI tF 'HII .H'r-r iJFF[: ' :11F

_ f l t r I t

I t ] :FI t i , '

OMPUTER MUSICIAN

H^{M/I /1I !A, . .JHINK I 'LL STARTWITH A +, '+ RHYTHM,..

'AHIGETTING A BIT OVERDRAWN'AT THE BANK, TIME To Co/YlPoSEANoTHER MEGA-SMASH Ht7A{oTHER MECASIN6LE. I THINK. '

TERL''TE CHI

5AiINUTES LATER

.. .ANOfHER A ASTERPIEC EFINISHEDJTHINK I 'LL JUST DO

.FouR REA^IXES OF TntS ONEJ,1

S /r{E A HEAD-: JUST TR/ING

E&MM APRIL 1985

ost people involved in the modernmusic industry are already awareof the Fairlight's incredible poten-

tial as a music prodrction tool. Thesedays, you switch on the television, radioor record player in the almost certainknowledge that a CMI wil l make itspresence felt somewhere along the line,and when you consider just how useful itsspecification is to studio Ongineers andproducers, that's hardly surprising. Evenin 1985, there aren't many machinescapable of spreading six octaves ofsampled sound acyoss the keyboard, andmanipulating that sound within usersequences to the nth degree of preciSion.

But if you're fortunate enough to sit infront of a CMI for any length of timewithout any production deadlines tomeet, you'll soon discover that its creat-ive power lies as much with soundsynthesis as it does with music produc-tion per se. Pages 4 and 5 are goodexamples of this in that they offer thefairly standard synthesis tools of har-monic sliders and profiles, but Page 6,which we introduced last month, issomething of a software oddity, since itallows control over the whole waveform -from a single byte to macro type com-mands such as GAIN, I i1 IX andIIERGE.

Now, for any command or process tobe really useful in the field of soundsynthesis, it must be responsible forsome radical change in the sound struc-ture that's both intuit ive and easilyunderstandable. For example, the VCF ofan analogue synth changes the sound agreat deal, and can be simply explainedand understood in terms of the attenu-ation of harmonics. FM synthesis, on theother hand, also results in vast t imbraldifferences, but comprehension of theprocesses involved (and their possibleresults) is a lot more difficult. That's whyactually arriving at a pre-specified soundon sonrething l ike a Yamaha DX7 requiresso much in the way of practice andpatience - and why so many musiciansprefer programming analogue synths,even if the ultimate sonic potentialisn't asgreat.

Fortunately, the Fairl ight's internalconfiguration side-steps most of theseoperational problems, and a good ex-ample of how this is done is the A D Dcommand. This takes a choice of seg-ments from one loaded voice and adds

THEE'ATRLIGIITE)KPEATTUED

In which we take a look at the difference between linear and logarithmicconversion, and incidentally end up creating a sound using the Fairlight's

synthesis faci I iti es. J im Grant

I IRUTFORN D I SPLSY

RDDCOIlnoDE, 4

nautFoRn DIspLny l loDE ' 4

0it18t lD, BDDC0n

f.BUtFoRn DIsptBy l loDE' 4

them directly into the same segments ofthe currently selected voice, scaling theamplitude to avoid clipping if necessary.lf you were to A D D all the segments intoanother, playing the keyboard wouldresult in both soundd being heardtogether but only using one voice. Figure1 shows a square wave and Figure 2 a

sinewave both resident in different chan-nels of the CMI: the resul t of ADDingthem together is shown in Figure 3. Theaddition's proportion can be varied byusing the G A I N command prior to theact ion, or by repeatedly ADDing onevoice to another to increase its amplituderelative to the comDosite sound.

l rav r oa< trn MM

l "

OMPUTER MUSICIAN

onrAi lD, f l I xcon

mtntmum

MSB=1

1111 Voice Whveform(Triongle)

Figure 4. Linear conversion.

Fi lD, SYNT0f4

Sl]UND SRNPLING

S8I1PIE RRTE I 4O8O HZ SAi lPLE LEUEL 255FILTER to ' l I TRIGCER LEUEL 15FILTER HIGH 8 TRICGER IEL8Y O NSCOI1PRESSOR OFF

SSIlPLE

DISPLRY

Figure 8.

0111

Figure 5. Non-linear (logarithmic) convercion.

LinearityNow seems as good a time as any to

explain sornething we've mentionedmany times in the past but haven't reallydiscussed in any detail, namely thedifference between linear and non-linearvoice data. As you'may remember, a

waveform is stored in 16K of RAM, inwhich each byte has a binary value thatc\oresponds directly to the amplitude ofthe waveform at that point. A byteconsists of eight bits, and considering allthe possible combinations of these res-ults in the amplitude of the sound at anypoint being limited to one of 256 levels.

The term 'linear' refers to the relationshipbetween the actual amplitude and thevalue of the binary number used torepresent it. lf all this sounds a bit on thetechnical side (and it ought to), have aquick glance at Figure 4. This shows thatzero amplitude corresponds to binary 0,while maximum negatiVe excursion' is

0t1l1RND, SYi lTor ' l

SOUND SBIlPL I NG

SRIIPLE RSTE I4O8A HZ SBI1PLE LEUEL 255F I LTER LOII 1 TR I GGER LEUEL I SFILTER HIGH 8 TRIGGER DELRY 9 NSCOI1PRESSOR OFF

SRIlPLE

D I SPLf iY

Figure 9.

VUIUL:11t xc0t! FERGCOI'1

noDE, 4

E&MM MAY 1985 85

|-\

I IRN! PFt l

I . IBUEFORI, I DISPLf,Y i lODE, 4

l l i f lND, TEST

lr f ,uEFoRrl DIsPLBY noDE' 4

l1l48ND, TEST

rf ,uEFoRi DISPLBY I ' lo lE ' 4

X I I I I ;T UUILL. It lAND. TESTI

r, lnuEFoRr, i DIsPLBY l ' loDE ' 4

represented by 1111 1111, or 255, andthe maximum positive value is held as0111 1111, or 127. St i l l confused? Wel l ,the value of the Most Significant Bit holdsthe key. lf that value is 0, the waveform ispositive, while a 1 gives negative excur-sions. Anything else in between is insimple proportion. This form of represen-tation results in a ratio between thesmallest and largest signal that can behandled (or in other words, dynamicrange) of about 48d8. You might con-sider that to be not a particularly impres-sive figure, since it means that at lowsignal amplitudes, the sound is more orless surrounded by hiss.

The fact is, storage (in one form oranother) of low-amplitude sounds is aperennial engineering problem, to whichthe most common solution is some sortof noise reduction system such as Dolby.In the digital world, and as a direct resultof research into digital telephony, aditferent solution is to use more binarybits of the byte to represent low-signallevels than you use for the high ones. Thisis shown in Figure 5, in which the lowervalues of the triangle wave use up more ofthe binary bits than a correspondingincrease at large triangle amplitudes. Thebinary data is now no longer l inear - it 'slogarithmic. For the waveform to berecovered, the data must be passedthrough a DAC that has a curve bent theopposite way to 'straighten out' thesound. I know all this sounds more than alitt le involved, but it does bring the magicdynamic range ratio up to about 72d8,which is at least respectable. Only catchis, the process only achieves this with acorresponding increase of quantisationnoise at larger signal levels, though this ismasked by the volume of the signal itself.

You might be familiar with this conver-

sion process under its commonly usedname of companding, and it's a systemused by many hardware manufacturersincluding Linn and E-mu.

So if it's so good, why doesn't the CMIuse it? Simple. Remember your schooldays when you added log numbers tomultiply? Well, this is what would happeni f theeADD command was used withsounds held in the form generated byFigure 5: instead of adding the soundstogether to produce a mix, we'd get theproduct, and end up with VCA-typeeffects at low frequencies and strangesidebands at higher ones. Which isn't, allthings considered, a particularly desir-able state of affairs.

l .a 'l - l lx lng

Anyway, enough of the lecture andback to the Fairl ight. The M I X commandcan also drastically alter the waveformRAM. Essentially, it generates a cross-fade between two specified segmentswhich must not be adjacent, ie. theremust be at least one segment in between.The waveform memory of each segmentbetween the stad and end points con-tains a proportion of the existing wave-form in that segment and that of the endsegment: this is best i l lustrated byexamining Figure 6 and Figure 7 forbefore and after views. Remember thatthe new contents of each segment is amix between where you are in thewaveform and the destination segment.Thus from Segment 2 onwards, thewaveform simply fades up to a squarewave. M I X is most commonly used toadd a clean fadeout to a sound thatdecays to noise or doesn't decay pro-perly in 128 segments.

Have a look at the percussion sound

sampled using Page 8 and shown inFigure 8. lt's pretty clear that the sampleends in a dither of noise. Now, supposewe needed nothing more than a shortpercussive strike, and that only thebeginning of the sound was of anyinterest to us. A quick solution would beto turn to Page 6 and ZERO, say,Segments 64 to 128, halve the sound,and then M I X from Segment 45 to 64.Looking at Figure 9 shows the result - asound that dies away evenly to a noise-free end, much to the relief of allconcerned.

l . lERGE is fa i r ly s imi lar to MIX - wi thone fundamental difference. Again, aform of crossfade is generated betweenstart and end segments, but this time, theprevious contents of intermediate seg-ments don't figure in the result. Quitesimply, the segments in between containa decreasing proportion of the startsegment and an increasing proporlion ofthe end segment. Figures 6 and 10 (ohyes, very logical- Edl reveal all. The M I Xand MERGE commands are t remen-dously powerful for splicing togethersounds of differing origins and producingan even fade from, say, a violin bowattack to a sung 'ahh'.

Creating a SoundSo, now that we've discussed most of

the commands available, let's try tocreate a sound using everything exceptthe Fairl ight's Page 8 sampling facil i ty.The question is: am I allowed to use Page2 and pull a sound off-disk to work with?Well. I 've decided l ' l l have to cheat a bitbecause I already have a thoroughlymarvellous sound called PAN.VC. whichattacks with the characteristic breathchiff of pan-pipes.

B6 MAY 198s I E&MM

OMPUTER MUSICIANFirst off ,First off, we configure Page 3 to

generate two voices, one with anN P H 0NY of 7 in Register A to be playedon the keyboard, the other monophonicin Register B as a scratchpad voice.Using Page 2, we load Register B withPAN.VC, which can be sben in Figure 11.The breath chiff is clearly visible at thebeginning of the sound, but unfortu-nately, the sampling staded a fraction toosoon, and the waveform has a few initialsegments of low-level rubbish - nothingto do with Electronic Soundmaker, youunderstand. The cure is to rotate thevoice left to bring the start of the soundproper coincident with the start of theRAM.

The next step is to flick to Voice 1 inRegister A and ZERO i t . Using a newcommand, TRANSFER, the f i rst fewsegments from PAN.VC can be copied tothe blank voice currently selected. Stab-bing the keyboard at this luncture revealsthat all is well, so the next thing to do is to

work on the body of the sound itself.Before any sound can really make the

grade as far as aesthetics are concerned,it must have plenty of t imbral andamplitude movement within it. A goodway of producing harmonically-richwaveforms is to use Page 5 and create afew segments spread across those un-used by the chiff. Figure 12 gives thegeneral idea - note that the createdwaveforms are all ditferent. So whatabout the segments in between? Well,th is is where MERGE comes in handy,f i l l ing in the ZEROed segments, andusing the segments created on Page 5 asthe start and end points.

OK, so far it sounds quite interestingtimbrally (and looks it too, as Figure 13shows) but it 's sti l l in need of someamplitude variation. An easy way toachieve this is to invert a couple ofsegments (numbers 32 and 96, say) andl.l I X from segments 1 to 32, 64 to 32, 64to 96 and 128 to 96, using Page 6. lf you

look closely at the differences betweenFigures 13 and 14, you shouldn' t havemuch diff iculty identifying the variation inampl i tude, especial ly in the sound's f i rstquaner.

All that remains is to inserl loop pointson Page 7 or Page 4, and adjust theattack and damping on Page 7. I

t've tusr aou6l{T THts NErrrr \'meon-Ruro-col PoSE R'- SoFTWAREONLY COST

'I\E T5OOI

IT CONTAINS ALLOF THE TECHNIQUESANO STYLES OF IHEWONLDs GREAT.EST COMPOSERS!.?AN TUi}| THE/l osr BoRDtGAxD,r{tDtocRr solt6.,f lo.l$""r!L.os1 E r r ADL t c nstp

i low I CAN COI PoSE LIK! A{oZABTAltD BEEtHoylr frL Dl lll6^-f^/{outIN NO TI'',IE! lETs SEE WKAT THEPRO6RAM DOES WITH TITATPIECEoF nAtBlENT rraUSlC t DtD IAST rltoHT.t

i+/.€oD