Lab Notes: Shepard Functions By: John S. Simonton, Jr. The first audio ittusion 1 ever heard was the Shepard Tone. Maybe you know it by the more dee eriptive corm "barber pole” tone. It got that name because, Like che stripes on a barberpole, ir seeas to defy the ald sew "what goes up must come down". the effece is that of a contindously rising (or falling) tone which never re= solves. How the Shepard Tone works. ‘there is nothing very mysterious ut the Shepars Tone, #4 discon Certing #2 it can be at First, and EP you've worked with synthesizers for a vhile you can figure out pretty quickly what's going on. The spectrum of the tone consists Of m large number of octavely Telated components, all stepping upscale together. The harmonics st the high and low ends of the Spectrum have relatively low am~ plitudes, while harmonics in the middle of the tone are at maxinua jeplitude. Inegine for @ moaent chat you are following the lowest harmonic thee makes up the Tone, At First the amplitude of this component is so low that it is, for all practi- cal purposes, inaudible. ‘But as Ie steps upacate its level in- creases, peaking when (ta frequen cy corresponds tos paint midway between ehe high and Tow limits. After peaking, the saplitude de~ creases a2 the frequency continues to step higher until finally, at the upper frequency limit, ‘the harmonic is again inaudible. When the harmonic reaches the high frequency liaie ie dia pears, only to be replaced by a new harmonic at the lover limit, Since the eight or ten har~ monies which ake up the Tone are AIL vising in» "ataggerea™ pro~ pression, cach in turn starting over agein ay it reaches ice upper frequency limit, the overall ef= fect is that of tone which is constantly increasing in pitch while noe actually getting aay “higher” (or lower if che tones ace all falling). It's an in teresting illusion. a2 Pophory Shopara's original work waed 4 computer program written by Max Matthews, bur the came type of effect can be accomplished vsing analog synthesis equipment con Erolled by a gadget which, for lock of @ better nome, we may ae well call a "Shepard Function Gen ‘The Shepard Fonction Geners tor, Thinking about whae happens vith the Frequency and amplitude 9€ each harmonic of 9 Shepard Tone fakes it eagier to understand the composite sound. The frequency {nereases conatently and Linearly from low to highy uotil the higher Limi is reached. At this potat, it begine again at che lower limit. This is a ramp function, The amplitude of any harmonic increases from the lover liait until ie reaches the middle fre~ quency, and then decreases aa ft Approaches the upper lisit. This can be a triangle. function $0, we ozed a gizmo which will produce a bunch of ramp wave~ Foras. (sight is a convenient num ber), and an equal number of tri- angle waves. The vanps and trian sles born must have fairly precize Phase relationships to one anc ther, as sunmarized in the diagram in figoce I. Tn the interest of conserving Araving time and apacny T have shown only four of che eighe fune= tions pairs that our device will generate, I'm sure that you can See the patcern, and thac the Bissing odd funcelons fit between the even functions thet are shown, Notice thet! each function gate a the complete serids ia 45 degrees (pi/e-radtans) ove of phase vith each of its neighbors. The even pairs shown are 90 degrees out of Phase with one another. Now, there are almost cer~ tainly lots of possible analog ways €o generate these function, But the simplest circuits Lean think of to do chis are too dinple (for example, they woulda’e be able to generate the functions over a very wide frequency range), and whet complicated approaches cone to wind are very complicated. but 1 know of sone workable digi~ tal approaches and 1'd like to show you one, a discrete logic sachine. The day will soon de here when we woulda'e even discuss a Logic machine approach to this problea. We would just truck Durselvea dovn to our local elec- teenies store and pick up a D1 ter packes Single Chip Date Pro- cessor to be progranmed on our trusty Home Dota Syaten. No doubt, but let's Look at a way t0 do esmentially the sam Thing with councers, DACs, MUKesy and suchauthings we can get to" day. "Figure 2 shows some compon- ents which we'll use in the Shepard Function Generator (as you probably realize, « counter con fected directly to DAC generates Fazpe). TE the counter iy count ing up, upward sloping ranps com out, Having the counter count down, or inverting the counter oucput before it gets £0 the DAC, Produces downvard rape (ace fig bre 3 Consider thin: 4 triangle aay be thought of a2 a reap which changes its mind halfway up. If ve replace the inverters ia the Figure above vith txclusive-o8 gates, we can produce a single Togie ‘inpue ‘that when high, causes the DAC to produce an upward ramp and when low, causes @ downward Tamp. By using che gost signifi cant bir of the counter ae well as the control signal to the EX-ORs, the digital input co the DAC will count up until the MSB goes high, then it will count dove -~ 19 other words, @ triangle function Geoe figure 4). If'we're inteceated in gen- ecating ‘only a single function pair, See a Simple setter t9/ pick Up 'new Least Significant Bie on the Counter and une it to effec tively aviten back and forth be Eween the cireuitey of figures 3 and &, producing first a small fection of the ramp, aod then a iaail section of the triangle, This gew LSB also switches between two sample-and=hold circuits to de-aultiptex the coapasite output of the DAC. Figure 5 shows how & Tieele more Logie gives FeO) and Peto). ‘itm sure that we're together so far, and to make sure that ve Poiyphony stay together I should mention a useful way Co think of the eight Most Sigoificant Bits of the coun- tere. Think of thes as phase, sunearized in Table I below. TABLE 1 Counter output ary hex 0000000 - $00 0 degrees 00100000 - $2045 degrees ‘01000000 - $40 90 degrees 1109090 - $60 135 degrees 19000000 = $80 180, 0100000 - $40 225 11000000 - §¢o 270 degrees 11100000 = $20 315 degreesIf you're more confortable with « graphic representation, see Figure 6. The benefit of thinking of the counter data in thia way is that phase shifte are produced by simple additions, For example, £0 Shift the phase of the waveform produced by the counter and DAC by 45 degrece, aimply aad $20 to the output of the counter. ‘This in pretty hendy ching co know, parti cularly when it just happens thet we are looking for a way to gen erate eignt sete of functions which are 45 degrece aparts Figure 7 shows a block dia- grem of the comploce Eight Phase Sheperd Function Generator vhich reeuite when ve include an added IC to calculate digital phase offsets and devmultiplex the out pur with 8/1 analog ewiten Ci Figure 8 shows the schematic for the complete shepard Function Gon- erator. Inthe same way that the circuit of figure 5 alternately generated pieces of Fz(0) and Ft(O), the Shepard Function Gen- erator sequentially puts out pieces of Fr(0), FCO), Fr(1), FEC), Fe(2)evseensFEC6), PECs Fe). Details: Starting from the Most Significant end of the coun ter, the first eight bite of the counter serve the same functions that they did in the vara-ups. and welve decided to think of that function as phase. Unlike the previous sketches, these phase Bite are broken down into two groups! the three Mont Signifi- Eant and the nexe five. If you don't cee the significance of this grouping, review che binary repre Sentation in Table 1, To produce 45 degree phage shifts, che three Nose Significant Bite are the only ones which change. Below the eight ghase bits, you'll see another grouping of Eneee bits, Think of these as “offset” bits and notice thar they ate what's added to the three Most Significant Bice by the adder. And note that the offset bite also Serve as address bits for the De~ Muxes so chac any given phase offset alvays gets strobed into the same sanple-and-hold The next "Lesa Significant bit can be thought of as avitching back and forth between ramps and triangles, as io figure 5, And since figure 7 is @ block diagram of our working Shepard Function Generator, and not just theoreti cal like the previous figures, the Least Significant sit of the coun ter serves as strobe which al- lows time for the DAC to settle “4 Pobyphony: an, 7 . ES a trianates zy Wo" and A'S contd by ‘may be sbstiate Tor OOSbefore selecting a sample: pola. ‘So, the Shepard Function Gen~ erator which velve developed isn't Simple (though I would like to think it hae a cercain elegance) and vhen you consider that ve'll leo need eight VCOs and eight Vas to produce the Tone (see Figure 9), you might question if ivft really worth the hasale. But Shepard Tone generation in not the oaly application for this cireuity recently the same principles have been applied to Other areas af electronic ausic. Por exauple, the Barberpole Phaser Snvenced by Harald Bode in a sig~ pal processing device which subs Citute pave change componente ep- plied to an external signal source Pobphony for the frequency components of the Tone. The characteristic of multiple phase shifters are con trolled by Shepard Functions that the phasing effect doesn't simply owing back and forth, like Getre used to hearing, but rather Sweeps up and down eternally. it's really a most unuoval effect and if it hag occurred co you that the same principle might also work with other processing elements (ouch as filters, maybe) you're on the right track, Figure 10 is the configura tion such approaches would cus: tomarily take, wich ehe camp func~ tlone controlling the parameter doing modified (phase, corner fre~ quency, time delay, ete.) and the Grlangles controlling VcAs to fade the output of each modifier in and You may have noticed chat we're using gobs of equip Bentwulote of phavers or flangere Or whatever, Chances are that you don't have eight flangers laying around, Even if you ue the least expensive modifiers available (pata's Ekx module series, f: example) you will #eill have bucks tied up in repetitive eli snte. For thowe who Tament the fact that there don't seem to be any new effects, this one quali- fies, Tele unique all right, but vorth the cont.n! Naic. | We're being prejudiced by what velve seen ao far (always adunger). We're thiaking of the 1rd Function Generator only 45to generate monophonic, noa= eyelic illucions by always using ll 16 output funceions to control \6 corresponding processing ele~ uence, But that's where we're getting of the track: You don't Rave to use all the outputs ell the time, and the results don't have to be monophonic. Now, there's ao doubt that eight phase Shepard Functions are the absolute minimum number of components which will still pre- aerve the "barberpole” illusion, but there are other timer whea sets of phase synchronized func~ tions are useful, 2 ic obvious that any pair of triangles 180 degrees apart ~~ Ft(0) and Fe(4), for instance ~- may be used with a pair of VCAs to give automatic Stereo panning? Or that four triangles 90 degrees apart. provide quad panning? with the arrenge- Bent shown in figure 11, the ap- parent "revolution" of che sound Source is clockwise, To reverse ent direction, reverse Various combinations of tri- angles with unequal phase rela: tionships may be used to produce effects which don't just swing Found and round, but rush out of fone of the "cornecs", sving around in froar of (or behind) you to disappear into the other corner. When you start adding effects into 46 Pubphary: Setup (such as phase shife- ers) under control of the ramps, as shown in figure 12, the sound really begins to move around you in some strange way A nice thing about this is that the effects devices don't all have to be the same to produce interesting resulta, In facts ome of the most interesting re sults come from using completely differenc effects (such as phaser and echo) in opposite corners with only VCA processing on the other corners. White you might be hesi- tant to rush out and buy eight YCOs just to get a tone, you pro- bably have enough modules or ef- fects to get sterted. Voltage control is obviously preferable, but even effects which have only ‘manual control are useful. Among. Other things, be sure to try syn~ chronizing the frequency of the affects orcillator to the frequen cy of the Shepard Function Gen~ Y think you get che idea: Play. Try different effects and different functions applied to different effects. Try control- Ling the VOAs with the ramps and the effecte with the triangl: try leaving out the VOAs alto- gether, Not all of the results Will be particularly pleasant, but you will surely also find some that are unique beyond vord While many of these effects are somewhat less spectacular when done in stereo, they are still very effective. This is getting long jure when I could go on forever; but it has to end ae soon as T dray your attention to the hard sye (npute A positive pulse applied to this input resets the counter chain and causes all functions to stert from the same knows point. Thiv few ture will,be particularly useful to us as Craig Anderton introduce tus to Synehro-sonie techniques in future issues of Polyphony. No new effects indeed! The following is available from PATA ELECTRONICS, INC., P.O. Box 14359, Oklahosa’ City, Ok 73116, (605) 843-9675: Experimenter's kit of circuit’ board and electronic couponents (does aot include hardvare or jacks. specify Now BK-9 Shepard Fuaction Generator Experimenter's postpaid. Etched, drilled and legended Rit, $24.95 circuit boerd alone for the Shepard Function Generator Specify Nd. ER-9pc, $4.95, postpaid, Vina & Master Card accepted ($10 winisun charge) or include check or money order vith order. eanuany 1909Practical Circuitry Build A Quadrature Function Generator By: Tom Henry wnat exactly in a quadratore function generator? To answer thae, we must first understand the concept of a quadrature oscilla for, An oscillator is said £0 have quadrature outputs if 16 produces simultancous eine, co Hine, segative sine, and nogative fosine ovtputs. Such an oscillo~ tor can be created quite easily by setting @ four pole loupaxe filter into oscillation, and tapping the required outputs’ from consecutive Stagesl, Te erigonometry alarms you, just consider the outputs to be four sine waves, each one nine ty degrees out of phase with the previous output Despite the interesting patches possible with a quadrature Srcillator, it does have one main Grawback: | Sine wave oscillators are fairly touchy, aod you Will Often tind that. the amplitude will Change aa the frequency is swept bver a wide range. In addition Gnder some conditions the oscilla tor say fail to oscillate: or, at the other extreme, hard clipping ay bring about some undesirable distortion. That's where the quadrature function generator coset in, since Se wilt chow out the oselilator Eonptecely and replace it witha foneelon generator. Gacillacors ire reactive; they sepend on Tesonat ing RE network, Punecion teneratore are mow-reactive Gn tho sense that they don't reso nate); thei timing depends solely Sn the chaeging and discharging of Tne ieputs are very stable fa amplitude and pucity over a very Wide ranges Finally, one more distinction between eset liacors Grd funetion generacore. La ehae ihe former generates. sinusoidal cueputsy while the latter gen erates triangle waves (or come~ Eimes ‘ramp waves). How it works, To fully us derstand the workings of a quadre~ ture Cuaction generator, ve aust recort to some mathematics There's hardly room to do that here, 49 if you're interested in the math behind che circuit, please refer to another article of Mine which gives the complete Tnalysia of a quadrature function Generator2. (actually, the cir Cait presented here fa more com pact and uses fever parts chan the barlier version, but the eireuit ction in very similar.) But evea though we can't go through the fathenacics here, wo cam still get an intuitive fool for how the circuit works, Referring to the Schematic, op amps IC1A and ICIS Form a Schmitt erigger/ integrator function generator, am old friend Trom way back. C3 sets the basic frequency range, with @21 allowing for aa adjustable rate. Te is important co aote that the output of ICIA is a triangle wave, and the autput of L078 ix = square Save. The triangle wave goes divectly to the "Primary Triangle Output” via R2, 1p addition, various Line segnents of the tri~ ingle wave are used in conjunction With Quadrature Function Gencra~ Yor's other circuitry to construct a new triangle wave ninety Jogrees out of phase with the firat and that's where the square wave ou pet from ICIB comes in This output tells the cireuit when to grab the various segments needed th the construction of the new triangle wave. 103 is configured as an BX- cLUSIVE~OR gare; it seemed more cost-offective to use the dirt cheap 4001 quad gate for this, father than using one gate out of 4 0070 EX-OR gate. sual power supply hookup for thie chip. Since thie part of the Circuit drives the gate of an N= channel FET, ve need a swing fro= negative to’ ground. Nore the unu- IC1D, Ql, and their asro- ciated cireuitcy comprise a sign changer. This cireait will invert or noe invert the input, depending upon che control voltage at the gate of Ql. Once again, see (2) For more details on the function of this sub-eireuit. Note that, so far, all of ehe 2p amps have cach been 1/4 of @ TLO84 quad bi-fee op amp package. Thie particular chip muse be used Cinstead of 7418, for example), Since this circuit coquires an ex” tremely high slew rate =~ we want all of the switching to be Clean as possible. fut now we're going to reverse this philosophy and specify as slow an op amp as posible for 102A ead 1C281 1C2A Suse together the various line Segments to fore a new triangle Wave hich is ainety degrees out Of phase with the originel. By specifying a low slew rate for this aap, eny of the discontinu Hes in the derived triangle will be masked by the op aap's inabili~ fy to slew fast enough. Pretey sneaky? In addition, TO2B is see Up ao lowpass filter with a cutoff Frequency of 1.5 kllz, vhich Biso helps smooth out the new triangle output. Adjusting the trimpote is fairly easy. To simplify the process, temporarily replace C3 wich a 0.05 uF capacitor. This Mill increase the frequency to a more easily observable ranges How monitor the "Derived Triangle Out- put" on an oscilloscope. While Watching the vaveform, go back and 26 Popphory: February, 1984‘QUADRATURE FUNCTION GENERATOR Ail Resistors in Ohms. il Capacitors In Mi, Tex: Pin 4~+15V Pin fai8¥ c2:Pin 4= 15 roazey a) >=} ote Unstal Power Supp Schnee forth between trimaer: Gntil the waveform gether” and connects to form a Smooth criangle wave. This pro“ ceee (4 quite magical; I think the Sigh on the scope vill really amaze you! Tf you have a dual Channel scope, Compare the two waveforms (*Prisary" versus "De~ Fived") and confirm that they are indeed ninety degrees out of phase vith respeet co each other. AL and 12, To round out the circuit you probably want to provide versions of the "Primary" and "Derived" triangle outputs. ‘This will give you a total of four outputs, each one ninety degreex oat ‘of phave with the previous. Applications. Well, vnac shall ve use ie for} Por a’ start, how about automatic quadraphonic panning: Gang the Inpucs of four Yeas together, and control each one by a different triangle wave. Send each output of the VCAs to a Popphony ‘Connect Tem Tom) a 50K $3 80K Re OOK different chasnel in your quad system, and che result is circular location modulation. what?!? You don't have 2 qued systea? Then you can still use the quadrature Funetion generator for sone neat stereo effects. For a really Bizarre sound, try this patch: Gang the inputs of VcAs 1 and 3 and apply a dry signal to these Vode. VCA I should feed the left channel and VCA3 the right chan~ nel. Now gang the inputs of VCA 2 and VCA 4, and epply an echoed Signal to these inputs. VCA 2 Should ix into the Left channel, while VA 4 mixes into the right. Now really crank up the delay tine and feedback and hit some etaccats notes =~ but don't get seanick! what?! You say you don't have @ stereo rig either? That makes it harder to think up Patches, but here's « good one to try. Apply an audio signal to four different flangers, vith exch one set for a slightly different initial delay time, Thea send each flanger outpue to a Vea, and Hinaily sum the VOAs together into a monsural output, See the func Derived wot Sato tion generator goings you should hear an incredibly dense aad lush sound, Thie ie especially good for full-bodied instruments, such a6 rhyth guiter. You may think ['m getting ridiculous, but what do you do if you don't have any synthesizer at allt (Ta serious nou!) Well, you can create some great Lis jour figures on an oscilloscope Sereen, or better yet, hook the unit up to your Taser art show for afar out display. (What?! Wo laser? Well...) There's quite a lot this Little black box can do, I've had a real blast dei ing and building the qued Fanetion generator. The © has a real "that's neat!” aspect to it, and is lots of fun to play with. “TE you come up with sone interesting applications be sure te jot we a Line c/o Polyphony. (editor's note: Splitting a signal into four filters, whose Sutputs feed four VoAs controll by the Quadrature Function Gener: 20» February, 1984 a ‘continued on pagePractical Circuitr: continued from page 27 tor, can make for fascinating timbral changes. Also, for tho ef you who are into synchro-sonic recording techniques, adding @ CMOS switch in parallel with C3, and feeding the switch control terminal with an appropriate trig~ ger pulse, will reset the oscil! cor at the rate of the trigger. All in all, for chose experimen- tere who don't quite need the sophistication or versatility of John Simonton's "Shepard Function Generator" presented in the Feb- ruary 1982 issue of Polyphony, the Quedvatore Function Generator’ pro= vides a low-cost way to experiment wich voltage controlled panning, cross-fading, channel-splitting, and the like.) NOTES. (1) For example, see J. Patchell, “Build a Voltage-Controlled Quad~ rature Oscillator", Polyphony, Nov/Dec 1980, pp. 26-27. (2) 1. Henry, "A Function Genera- tor With ‘Quadrature’ Triangle Wave Outputs", Electronotes #122, pp. 13-20. PARTS LIST, QUADRATURE FUNCTION GENERATOR Resistors (1/4 Watt, 5% tolerance preferred) RL 470 R2R30 1K Ra. 10K RS 12k R6,R7 15K R8,R9 027K R10 33K RLL,R12 50K trim pot R13-R20 100K R21 100K potentiometer R22 150K Capacitors (15 or greater working Volts) cl 0.005 uF, aylar pre ferred 2 0.01 uF, mylar preferred 3 0.33 uF, mylar preferred Semiconductors Tcl TLO84 quad bi-fet op amp rez 1458 dual op amp 103 4001 CMOS quad NOR gate a MPF-102 N-channel FET DI,D2-IN4148 or equivalent switching diode Pobphony February, 1984