Eµ Modular OpManual

7/29/2019 E Modular OpManual

1/88

MODULAR SYNTHESIZER

OPERATION MANUAL

RETROSPECTIVE

By Dave Rossum


2/88

2

The E Systems ModularSynthesizer Owners ManualRETROSPECTIVE

By Dave Rossum

Edited by Riley Smith

A Retropective

The operation manual for the E modular systemwas never officially published by E Systems forreasons unknown. This retrospective, compiledalmost 30 years after it was originally written, isfar more than just a time capsule to the goldenage of analog synthesizers. It is an advancedcourse in the art of sound synthesis.

This manual should be of interest to anyoneinterested in building or playing analog modularsynthesizers, or those wishing to expand theirknowledge of synthesizer programming ingeneral. Although electronic music machines havechanged dramatically over the years, many of theunderlying concepts of the modular system

remain.

This retrospective manual was compiled in thesame spirit that created the E modular systemit was a labor of love. We sincerely hope youlove reading about it.

Thanks Dave, for permission to finally release it.

R. S.


3/88

3Modular Synthesizer Owners Manual

Table of Contents

Voltage Controlled Oscillator ............................................................................................................ 4

Sawtooth/Pulse Voltage Controlled Oscillator ................................................................................. 9

Voltage Controlled Amplifier .......................................................................................................... 10

Quad Voltage Controlled Amplifier ................................................................................................ 14

Voltage Controlled Lowpass Filter .................................................................................................. 16

Voltage Controlled Highpass Filter ................................................................................................. 20

Universal Active Filter ...................................................................................................................... 24

Resonant Filter ................................................................................................................................. 28

Lag Processor ................................................................................................................................... 29

Dual Transient Generator ................................................................................................................ 31

Voltage Controlled Transient Generator Input Unit ....................................................................... 36Noise Generator .............................................................................................................................. 38

Sample and Hold ............................................................................................................................. 40

Envelope Follower ........................................................................................................................... 43

Ring Modulator ............................................................................................................................... 46

Quad Inverter .................................................................................................................................. 48

Dual Preamp .................................................................................................................................... 50

Dual Reverb ..................................................................................................................................... 51

Voltage Controlled Clock ................................................................................................................ 52

Hex Digital Inverter ......................................................................................................................... 54

Triple Or Gate .................................................................................................................................. 55

Triple Latch ...................................................................................................................................... 56

Dual One-Shot ................................................................................................................................. 58

Analog Switch .................................................................................................................................. 59

Eight Position Address Generator ................................................................................................... 60

Voltage Source Output Unit ............................................................................................................ 62

Digital Sequencer ............................................................................................................................ 65

Memory Address Generator ........................................................................................................ 65Programmer................................................................................................................................. 66

Memory Module .......................................................................................................................... 68

Tape Interface .............................................................................................................................. 74

4000 Monophonic Keyboard........................................................................................................... 76

4050 Polyphonic Keyboard ............................................................................................................. 80

Applications - Speech Synthesis ...................................................................................................... 85


4/88

4

Voltage Controlled Oscillator

Function

The VCO is, of course, the vital organ of asynthesizer, being the signal source fromwhich most tones are derived. A few words onvoltage control of pitch would be helpful forstarters. To say an oscillator is voltage con-trolled is to say that its frequency is somehowdetermined by a control voltageit doesntsay what this function is.

Some synthesizer VCOs have linear frequencycontrol, which means if the control voltageincreases by 1 volt, the oscillator frequency

increases by a certain number of Hz, say, 1V/1000 Hz. In other words, the pitch increaseslinearly with voltage. Most synthesizers,including Es, have exponential frequencycontrol, which means that a 1 volt increase incontrol voltage multiplies the originalfrequency by some factor. If this multiplicativefactor happens to be 2, the exponentialcontrol is said to be 1V/octave (since adoubling of frequency is equivalent to raising

the pitch one octave). The distinction between

linear and exponential voltage control iscritical. Consider two VCOs harmonizing anoctave apart at 1000 and 2000 Hz.

With linear VCOs (at 1V/1000 Hz) theaddition of 1V to each control input increasesthe frequencies to 2000 and 3000 Hz. Theseare no longer an octave apart; linear oscillatorswont track with each other.

On the other hand, two exponential VCOs(1V/octave) at 1000 and 2000 Hz respond to a1V increase by shifting to 2000 and 4000 Hz,

keeping the harmony the same.

The E 1200 VCO submodule, when properlytrimmed by the Volts/octave trimmer on themodule board, responds at precisely 1V/octaveto its incoming control voltage. The controlvoltage which the submodule sees is the sumof several voltages coming from differentplaces on the front panel. The range switchand the coarse and fine initial frequency pots


5/88


contribute to the summing node, as it iscalled, and so do the three frequency controlinputs at the upper left. The lowest of thesethree has no attenuator, so any voltage appliedto that input has a precise 1V/octave effect onthe frequency. The upper two inputs are bothattenuable; 1 volt applied at one of these

inputs raises the frequency roughly one octaveif the attenuator is completely clockwise.When the attenuator is turned down some-what, the same 1V input may raise the pitchby only a fifth. Its control function wouldthen be 1V per 7/12 octave, or about 1.71 V/octave. In other words, the sensitivity of pitchto control voltage can be one octave or less pervolt. Frequency control voltages can comefrom an infinitude of different sources. Onesuch voltage source is a keyboard (q.v. 4000Keyboard), which gives a different voltage foreach key depressed.

Our keyboard, and most other synthesizerkeyboards, puts our zero volts for the lowestkey, 1/12 V for the second key, 2/12V for thethird, up to 1V for the first octave. An octaveabove that is 2V the next octave is 3V, up to,5V for the highest key. Patched into the 1V/octave input on a VCO, the pitch of the VCOthus follows the keyboard in tempered tuning.

This is such a commonly used patch that it is

preset on our VCOs through the KYBD switchin KYBD 1. The VCO is pre-patched throughthe KYBD 1 bus in the ribbon cable to theKYBD 1 voltage of a keyboard, if present.Oscillators on KYBD 2 are similarly pre-patched to the KYBD 2 bus, which may alsohave an associated keyboard. Assuming akeyboard is present on the KYBD 1 bus, itscontrol voltage also appears on the jackslabeled Control Voltage 1 on the powersupply front panel. It is important to under-stand that these jacks, as well as the gate and

trigger jacks, merely tap into the bussesrunning throughout the machine and canserve either as outputs from or inputs to thesebusses. Thus, if a keyboard is in use on theKYBD 1 bus, a control voltage from anothersource can be sent to VCOs not needed for thekeyboard by using the KYBD 2 bus; this isaccomplished by patching the DTG into thepower supply control voltage 2 jack, and

switching the desired VCOs to the KYBD 2position. If the DTG were patched into theControl Voltage 1 jacks, it would becompeting with the keyboard for control ofthe bus. Since the keyboard has a very lowoutput impedance compared to the 1 Ohmoutput impedance of the DTG, the keyboard

would completely overpower the DTG and isthe only voltage source which would appearon the bus.

Since the effective control voltage to a VCOsubmodule is the sum of initial frequencysettings, keyboard control voltage, and thethree F.M. inputs, several controls can behappening simultaneously. For example, youcould play an arpeggio on the keyboard, havea low frequency sine wave on one attenuableF.M. input to give a shallow vibrato, and a

clocked VSOU on another F.M. input causingthe arpeggio to be transposed in any desiredmelody and rhythm. This still leaves one F.M.input unused, which, if you patch to the slowrandom noise output, turns your masterpieceinto total garbage. (Dont feel too bad if you likeit.)

The 2200 VCO simultaneously outputs sine,triangle, sawtooth and pulse waveforms. Theseare available from the full-level outputs at 10Vpeak-to-peak for sine, triangle and pulse, and

zero to 5V for sawtooth. A mix of the fourwaveforms is available at the mix output, toprovide greater selection of timbres. The mixoutput is inverted, making it possible to obtainone of the basic waveforms and its inversefrom the same VCO.

This allows two VCAs to be used as a voltage-controlled panner, or a VCA control to go upwhile a VCF control goes down, or anythingelse you can think of.

As can be seen from the panel drawings, the

phase relation among the waveforms is suchthat the sine trough, the triangle peak, thesawtooth fall and the pulse rising edge aresimultaneous. The pulse falling edge can occuranywhere from 0% to 100% of the waythrough the cycle. This is determined byanother summing node which includes theinitial pulse width pot and the attenuablepulse width input. With the attenuator fullyclockwise, the input increases the pulse width


6/88

6

10% per volt. If the effective pulse widthcontrol voltage (pot plus attenuated input) issufficiently negative, the pulse width goes tozero turning the pulse wave into a -5V DClevel. Similarly, a large positive voltage gives a+5V level. These DC levels are useful forvoltage sources, and since they are not audible,

they provide a good way to shut off the slightcapacitive pulse-wave buzz which leaksthrough the output mixer even when all thewaveforms are supposedly off. (Simply turnthe initial pulse width pot full left or right). Ifyou find the pulse output of a VCO dead, orcutting out intermittently, check the pulsewidth modulation to determine whether thepulse width is pinning at one of theseextremes. Pulse width modulation is a veryuseful tool for achieving smooth and interest-ing timbral changes. Try it with a transientgenerator, or as a source of vibrato. Notice thatP.W.M. has no effect on the other waveforms.Actually, the VCO submodule generatessawtooth first, then the wave convertersubmodule inverts half the cycle to givetriangle and distorts the triangle to give sine(as such, sine is not completely pure, but canbe trimmed to be nearly so and filtered to beperfectly so. Pulse is obtained in the WCsubmodule by watching the sawtooth output,and giving a +5V level when the ramp is below

a critical voltage, and -5V when above. Thepulse width summing node simply determinesthe value of the critical voltage.

The sync switch allows the VCO to beconnected to one of two sync busses runningthroughout the machine. Oscillators which areconnected to the same sync bus and allsomewhat less than a semitone apart in pitchwill all be pulled, or synced to precisely thesame frequency (equal to the highest fre-quency among them). In addition, an oscilla-

tor within a semitone of any harmonic ofanother oscillator will be pulled up if itsfrequency is lower than the harmonic, or willpull up the other if higher than the harmonic.This is essential for keeping oscillators in tunewith one another or in constant phaserelation. If the so called chorale effect isdesired, the sync can be turned off.

VCO 1

No Sync

VCO 2

VCO 1

Hard Sync

VCO 2

VCO 1

Soft Sync

VCO 2

Syncing is accomplished by the use of syncpulses; when the sawtooth of an oscillatorfalls, a sync pulse is issued on any engagedsync bus. Any other oscillator an that sync busalready close to discharge will discharge itssawtooth when a sync pulse is seen on thebus. Thus, with all the oscillators both issuing

and observing each others sync pulses, all arepulled to the rate of the fastest. This is knownas soft sync, since the slower oscillator willonly discharge its sawtooth if it was alreadyclose to doing so anyway. Hard sync causesthe sawtooth of a slave oscillator to dischargewhenever the master oscillator does. This isequivalent to soft sync when the slave is lowerin frequency than the master, but results indeep harmonic sidebands and consequent-bizarre timbral changes when the slave ishigher than the master.

In any case, hard sync is incapable of syncingto harmonics, as can be seen in the drawingbelow, which describes the operation of bothtypes of sync:


7/88


The gate input is a digital input (q.v.) sensitiveto two states: high (>2.5V) and low (


8/88

8

Another way to say this is that all non-harmonic sidebands are eliminated, leavingonly those sidebands which are harmonicallyrelated. With the syncing on, sweep themodulation and carrier frequencies indepen-dently as before. Notice that their respectiveeffects on the tone are now much more clearly

defined; the apparent pitch (the fundamental)of the F.M. output is actually the modulationfrequency, since a full cycle of modulation isrequired before the F.M. waveform repeatsitself.

The carrier frequency, on the other hand, has aprofound effect on the timbre of the output,introducing more high harmonics as itsfrequency goes up. If both VCOs are switchedso as to track a keyboard, the F.M. output willkeep the same waveform regardless of pitch.

Some readjustment of frequencies may benecessary if the sync is operating close to thelimit of its pulling range. This is evidenced byfluttering on certain keys, and sections of thekeyboard being out of tune with respect toother sections. It is always correctable with thefine tuning on the VCOs. If you need toduplicate the same F.M. timbre in two voices,the tuning can be difficult. Tune the modula-tion VCOs together, then tune the carrierVCOs together (with the modulation off, ofcourse), then turn up the modulation indices

until the desired timbre is achieved in eachvoice. Notice that you only have two syncbusses, so youll have to use sync 1 on onevoice and sync 2 on the other. More than twovoices are not possible with syncing unlessspecial sync busses are installed.

An interesting special case of F.M. is somethingthat could be called auto-F.M., or feed-backF.M. Patch the full-level sine output of a VCOback into an attenuable F.M. input of the sameVCO. Listen to the mix output (any waveform

will demonstrate the effect, though sine mayshow more extreme timbral changes). As theattenuator is turned up, the fundamental goesdown and the timbre gets more razzy. It isdifficult to explain these changes preciselywithout plenty of math, except to say that theoutput at the full-level sine jack is the one andonly waveform which will give itself at thatjack as a result of the functions performed onit by the VCO and WC submodules.

The output waveforms which appear at theother jacks will be other functions of the sameinput waveform. In this example, where thesine jack (no longer a sine wave) does theauto-F.M., it may be fairly obvious that withinone cycle of the resultant waveform, the VCOwill sometimes be running faster than the

original frequency and sometimes slower. Itwill, in fact, spend more time running slowerthan it will running faster since the output-voltage feeds back to control the oscillatorfrequency. The result is a lowering of pitchwith increased modulation index. (This can becompensated by adjusting the initial fre-quency setting). Note that when sawtooth isthe modulator, the pitch goes up withmodulation index this is because sawtooth isalways positive. As for the timbral effects ofauto-F.M., they can probably be best under-stood by seeing the output waveforms on ascope, if available. Since any of the fourwaveforms can be used as modulator, any canbe used in the output mix, and the modula-tion index is variable, an infinite number ofnew timbres is made available.

An effect somewhat related to audio-rate F.M.is audio-rate synchronous gating. This hasalready been discussed for sub-audio rates; thegate input performs the same function ataudio rate. The figure below shows the result

of gating an audio rate VCO with the pulsewave from another VCO running at a consid-erably lower audio rate. Though any waveformwith portions over 2.5V will operate the gate,pulse is used here for simplicity.

+10

-10

+10

-10

+10

-10

Synchronous Gating


9/88


Since the gating is synchronous, the samenumber of complete cycles is always present ineach burst of the higher frequency, and theyalways begin and end synchronously. Com-pletely asynchronous gating could be achievedby turning a VCA in the signal path on and offwith a pulse wave, but since there would be no

phase-locking between the two frequencies,the output would contain non-harmonicsidebands much like asynchronous F.M. Thegating and gated VCOs can be made to trackeach other with the keyboard, thus givingconstant timbre, as in the case of F.M.

Since gated oscillators maintain the phaserelation automatically, syncing the twooscillators is not necessary, and any number ofvoices can be patched with identical timbres.(Recall that synced F.M. needs a separate sync

bus for each voice, so only two voices arepossible).

Though the similarities between F.M. andgating are many, you can hear that each givesunique timbres unavailable with the other.Check out the effects of gating frequency,gated frequency and pulse width on theoutput pitch and timbre.

The sawtooth/pulse VCO is a simplified VCOcontaining a 1200 VCO submodule but nowave converter. It thus outputs only full-levelsawtooth and pulse waveforms. It differs inother respects from the 2200 VCO only in thefollowing ways: it has only one attenuableF.M. input, no syncing switch, no hard syncinput and no gate. For some purposes, it is

fully as useful as a 2200 VCO, particularlysince the full-level waveforms can be filteredto give reasonable approximations of the otherwaveforms, and attenuated at most inputs toother modules.

For control purposes, of course, it is ofteninadequate. Generally, if a system has several2200 VCOs, it can have a number of SPVCOsto be used in simpler voices, and much moneyis saved.

An ApologyThis description of the VCO has been prettylong-winded, mainly because it is such animportant module which must be understoodwell before you can program a synthesizer.Almost every patch discussed in later chapterswill include one or more VCOs, and you maybe referring back to this chapter often fordetails of its operation. Thats why we put itfirst.

Sawtooth/Pulse VCO


10/88

10

Voltage Controlled Amplifier

The 2000 VCA is a three-into-one mixer/amplifier whose gain can be controlled, eitherlinearly, exponentially, or both, by control

voltage inputs. It has three signal inputs whichare mixed in a summing node before amplifi-cation; one of these is full-level, one isattenuable, and one is attenuable and invert-ing. (The inverting feature will be discussedlater in this chapter.) The mixed signal is thenamplified by a factor determined by the modeswitch, the initial gain pot and the controlinputs, and the output appears on the outputjack.

The function which determines the multiplica-tive factor, or gain (gain = Vout /Vin). isselected by the mode switch. The simplest

mode to understand is linear. In linear mode,the three control inputs (two attenuable, onefull-level) mix in a summing node with avoltage from the initial gain pot. This sum-ming node enters a linear control input on the1000 VCA submodule. The gain of the VCAincreases linearly with respect to a voltageapplied on this input, as shown in the graph:


11/88


Note that the gain will not exceed 2, or about6 dB up, regardless of further voltage increasebeyond 10V. Since the initial gain pot puts out-5V full left and +5V full right, zero volts isobtained approximately at midrange.

In this position, with no voltage on thecontrol inputs, the VCA has zero gain; in otherwords, it is turned off. With the pot full right(+5 V), the gain is trimmed to be 1 (Vout=Vin)using the gain trimmer on the module circuitboard. Study the linear mode by patchingvarious control sources into the control inputs.A sub-audio sine or triangle wave applied to anattenuable input can give sub-audio amplitudemodulation, or tremolo. When an audio-ratesignal is applied to the control input, audio-rate A.M. is obtained, which will be discussedin detail later in this chapter. Notice with thesignal input entirely off, the control input, ifaudio, can be heard leaking through to theoutput of the VCA to a small extent. Thereason for this is that the circuitry in thesubmodule sees a slight offset in the signalinput even though the voltage applied at thesignal input is zero. The VCA tries to amplifythis offset by a gain which is varying at theaudio rate applied to the control inputs. Theresult is a signal with the same frequency

appearing at the output of the VCA. Theability of a VCA to prevent this occurrence iscalled its control rejection; the controlrejection of the 1000 VCA submodule can betrimmed on the module circuit board to about40 dB.

To better understand sub-audio gain control(actually a much more common application ofthe VCA), consider the following patch:

This is a simple voice (q.v.) When a key isdepressed,, and its gate is used to initiate atransient from a DTG. This transient, patchedinto a control input of the VCA, controls theenvelope, or loudness contour, of the signalfrom the VCO. For this reason, transientgenerators are sometimes called envelopegenerators. (We chose our name to indicate

that the transients produced by a DTG can beused for many purposes other than controllinga loudness contour.) In linear mode, the initialgain pot can be adjusted to squelch theoutput. When the transient is then gated, theamplitude of the VCA output follows thegraph of the transient proportionately. Noticethe independent effects of the signal attenua-tor, gain control input attenuator, and initialgain control. When the initial gain control isturned further left than the zero gain point,more and more positive voltage must be

applied to the control input to turn on theVCA. It is thereby possible to cut off the lowerportions of the transient, which is a differenteffect from attenuating the transient. Attenu-ating the signal simply reduces the peak-to-peak voltage of Vin seen by the submodule,resulting in a lower amplitude at the outputwithout affecting the gain.

Apply a sub-audio triangle wave to anattenuable control adjust its rate and theattenuator level to give a good tremolo. Now

switch the mode switch to exponential andreadjust the settings to give a similar tremolo.Notice the difference between the twotremolos. In linear mode, the VCA sounds as ifits either on or offyour ear is not very ableto hear the linear increase in gain once it hearsany output at all. This is because the earresponds to volume exponentially, notlinearly. In exponential mode, the amplitudeincreases faster as it goes up, and decreases

VCO VCA

KybdControl

Voltage

KybdGate

Out

gain gain =

5

5

2

1

10 volts

V in


12/88

12

slower as it comes down. The effect is abouncier tremolo with more apparentvariation in its amplitude, even though theactual peak-to-peak variation in the envelopeof the output may be the same as in linearmode.

Electronically, exponential mode switches thecontrol input summing node to an exponen-tial control input on the submodule, giving again function like this:

input is zero, the response of the VCA to thepot and the other inputs is identical to thelinear mode.

When a positive voltage is present from eitherthe pot or the inputs at the linear input to thesubmodule, the gain responds exponentiallyto any changes on the right-most controlinput. In addition, when the effective linearcontrol voltage is zero, the VCA is off. If atransient is patched into the linear attenuableinput, and a tremolo voltage in the exponen-tial input, the VCA will turn off when thetransient goes to zero (the advantage of linearmode), and the response to the tremolovoltage will be exponential (the advantage ofexponential mode). Notice that in linearmode, the same patch gives a very strangeeffect, in that the gain sweep of the tremolo is

the same at the loud portions of the envelopeas at the soft portions. Since the ear does notrespond linearly to gain, the tremolo effectsounds shallower when the output is loud, anddeeper as the envelope fades away. In mixedmode, the tremolo depth is scaled by thetransient, giving a more even-soundingtremolo. In a sense, mixed mode gives asecond VCA in the control path; without it,another VCA would be required to perform themultiplication of the linear and exponentialinputs. VCAs can be used in control paths as

well as signal paths, and the possibilities areimpressive. A simple example of using onecontrol input to control another is thefollowing patch for voltage-controlled vibrato:

1

1

2

2 3 4 5 6 volts

gain

VCO 1 VCO 2VCA

control

f.m.slow

sine AudioOut

The maximum gain available is still limited to6dB (gain - 2). Notice that the gain can betrimmed (exponential zero trimmer) to beunity at 5V input; but, being exponential, itnever goes to zero, even at negative controlvoltages. It can become essentially inaudible,

however. Patch a transient generator into acontrol input in exponential mode. Since thetransient already has exponential rises andfalls, the gain responds in a doublyexponentiated way. This effect can be undesir-able, particularly since the VCA wont go offwhen the transient falls to zero.

Clearly, then, there are advantages andproblems with both linear and exponentialamplification. It is possible to have the best ofboth in the middle switch positionexponen-

tial times linear, or mixed mode. In thisposition, the full-level input, the left-mostattenuable input, and the initial gain pot allsum into the linear input on the submodule-the right-most attenuable input goes into theexponential input. The gain function in thismode is proportional to the product of thefunctions in the linear and exponentialmodes. Specifically, gain = Vlin/5 x 10 v exp/2.Thus, when the voltage on the exponential

VCO 2 oscillates with a steady tone as long asthe VCA is off. As the gain of the VCAincreases, the depth of VCO 2s vibratoincreases. The control voltage applied to theVCA could come from say, a foot pedal, atransient generator, another slow VCO, etc. Itis when the VCA is used in a control path thatthe inverting input is useful. At audio ratesthere is no difference between a signal and its


13/88


inverse, except that they will cancel each otherif mixed 1:1 (try this with the two attenuableinputs of a VCA). At sub-audio rates, however,there is a big difference. Compare an ADSRtransient and its inverse, or an upward glideand a downward glide.

(A basic note on our general philosophy is thatif something is possibly useful and easy to do,we might as well include it. It is true that aquad inverter (q.v.) will perform the samefunction as the inverting input of a VCA, butyou might be able to save the cost of one ifyou could use a VCA instead, and youd havethe option of voltage controlled gain inaddition).

Audio-rate A.M.This effect, produced by applying audio-rate

signals to both the signal and control inputs ofa VCA, is still another way to achieve newtimbres. It is closely related to ring modulation(q.v.), except that a ring modulator givesnegative gain (gain with inversion) fornegative control voltages, while a VCA onlygoes down to zero gain. Another way to saythis is that a VCA is a two quadrant ampli-fier, while a ring modulator is a four quad-rant amplifier. As in the case of all audio-ratemodulations, the carrier and the modulatorcan be synched to eliminate non-harmonic

sidebands. Unlike exponential F.M., however,the pitch of the output is not affected by themodulation index, so voltage control ofmodulation index, and, hence, of timbre, is amusically sensible patch.

Consider the following:

Since both VCOs track the keyboard, and bothtransient generators are gated by the keyboard,the evolution of the timbre of a note is thesame for all keys. TG 1 is controlling theenvelope of the output; TG 2 controls the gainof VCA 2, and hence, the modulation index.As the modulation index increases through the

attack, the spectral bandwidth of the outputbroadens, changing the timbre considerably.As the modulation index falls, the bandwidthis constricted once again, and the output willsettle into the unmodulated tone of VCO 1 (ifTG 1 has not turned off yet). We just discov-ered this patch specifically for this chapter, sothat may explain the excitement over it!

VCO 1

TG 1

VCA 1Kybd

Control

Voltage

VCO 2 VCA 2Kybd

Control

Voltage

Kybd

Gate

TG 2Kybd

Gate

Out


14/88

14

Quad Voltage Controlled Amplifier

The 2010 Quad VCA contains four simplifiedVCAs; each channel contains only one signalinput and one control input, both non-attenuable. Each channel has two features notfound in the 2000 VCA: a control polarityswitch, and a lag time pot. With the controlpolarity switch in the + position, a positiveincrease in control voltage on the input jackwill produce a linear increase in the gain ofthat channel. In the - position, the gain woulddecrease linearly. (The response to the pot isalways positive). The lag time pot allows the

digital outputs (q.v.) of other modules to beused as control inputs without producing apop in the QVCA output. The attack andcut-off do become somewhat mushed by theintroduced lag time, so the pot is provided toadjust the lag time for the proper compromisebetween popping and mushing.When continuous signals are used for controlinputs, the lag time can be turned completelyoff (full left) if you want.

In addition to these special features on eachchannel, the four channels are interconnectedin a way that makes the QVCA very useful forvoltage controlled switching, mixing, andpanning. The tip of each signal input jack iswired on the circuit board to the shunt of thejack below itthus, a signal applied tochannel 1 is automatically put through all thechannels below it, unless one of them has apatch cord plugged into its signal input. Inthat case, the signal on that patch cord is fedto its channel and any open channels below it,

etc. The same firm-wiring is done with thecontrol inputs. (Notice that a patch cordplugged into a jack breaks the firm-wiringchain even if it is not used to input a newsignal. A patch cord so used is called adummy patch cord: not too useful in thecase of the QVCA, but may be in otherapplications). In addition to the firm-wirefeature, the four channels are also mixed; theinverted mix is available at the -sum jack.


15/88


A few diagrams of patches on the QVCA willclarify its uses considerably. In these diagrams,a filled-in jack indicates the presence of apatch cord, and its signal is identified with aletter. Pre-patched inputs to other channels areshown in parentheses:

A

(A)

(A)

(A)

B

(B)

(B)

(B)

A

(A)

B

(B)

C

D

(D)

(D)

A

B

C

D

E

F

G

H

Since a positive voltage on each control inputcan turn that channel either up or down,depending on the polarity switch, you canimagine a myriad of uses for the QVCA. Thedrawings at right show a few possibilities:

That last one is a pretty obscure patch, but itillustrates a point. When D is at a positivevoltage, the mix output contains A and B only.When D is zero, all channels are shut off(provided, of course, the initial gain pots areset appropriately). As D goes negative, A and Bremain off, and C comes up in amplitude.Notice that the effect is different if channel 4spolarity switch is +; in that case. C would beheard at either positive or negative values ofD. This could be corrected by patching adummy patch cord into channel 4s signal orcontrol jack, but, of course, switching itspolarity to - is easier.

As you can see, the QVCA can do some

complicated things; as such, it can be confus-ing. Watch your polarities and initial gains.Since there are no attenuators, even turningthe initial gain pot full left wont shut off achannel if a sufficiently large (>5V) signal isapplied to the control input. Switchingpolarity will often solve this, provided thecontrol input then never goes below -5V.

Remember it can also be used for exactly whatit isfour separate VCAs.

A B

A

B

C

channel 1

channel 2

mix out

voltage controlled panner

(A and B) or (C) selector

D

A

B

C

D

E

F

G

H

A

BC

D

mix out

mix out

voltage controlled, 4-into-1 mixer

knob controlled 4-into-1 mixer

D

A

B

mix out

voltage controller crossfader

E


16/88

16

Voltage Controlled Lowpass Filter

The 2100 VCLPF (VCF for short) is, a lowpassfilter whose cut-off frequency is determined by an

applied control voltage. Signals enter the 1100VCLPF submodule via a summing node whichincludes a full-level input and two attenuableinputs. The total control voltage is anothersumming node comprised of the initial cut-offfrequency pots, the KYBD switch, one full-levelinput (1V/octave) and two attenuable inputs. Anydiscussion about filters will necessarily refer toaudio spectra, so you should review yourmusical physics if you dont already feel

comfortable in frequency space. Speakingideally, a lowpass filter is a black-box whose

gain is unity for all inputs with a frequencyless than some cut-off frequency fc, andzero for all input frequencies greater thanfc. In actuality, the cut-off is not so sharp;the 2100 VCLPF has a cut-off slope of 24 dBper octave. That is, if the cut-off frequencyis 1000 Hz, a 2000 Hz input would beattenuated 24 dB. Since the cut-off corner isnot sharp, fc itself experiences someattenuation, equal to 12 dB.


17/88


Thus, the attenuation curve for a VCF whose fcis 1000 Hz is as follows:

-10

-30

500 1000 2000250

Frequency

slope = 24 db/octave

fc

-20

attenuationin

decibels

Sine

f

Sawtooth

f 2f 3f 4f 5f 6f 7f

1

21

31

41

51

61

71

Square

f 3f 5f 7f

1

31

51

71

Triangle

f 3f 5f 7f

1

91

251

391

and notice the effect of the initial fc pots. As fcis lowered (counter-clockwise), the timbre ofthe output gets less razzy, approaching andfinally reaching the pure sine wave timbre.

At the same time, the overall volume goesdown. A look at a few spectra will show whythese effects happen. The diagrams belowshow spectra for several waveforms withfundamental frequency of 500 Hz:

Suppose the 500 Hz sawtooth is passedthrough a VCF whose fc is 1000 Hz. Thedrawing on the following page shows whatoccurs; the attenuation curve appears indotted black, the input sawtooth spectrum indotted red, and the spectrum of the filteredoutput in solid red.

Notice that all harmonics above the 3rd are

heavily attenuated, and the third is attenuatedby about 15 dB. Such cutting-out of the highharmonics results in a mellowing of thetimbre, and, since high harmonics contribute

First off, what is this good for? For one thing,

filtering of razzy tones will give mellower,more spectrally pure tones, and this can bedone to any extent desired. Run a sawtoothwave through a VCF with the Q off (full left),


18/88

18

on the VCF they control the value of fc. Thus,if the VCO and VCF in our example were bothtracking a keyboard on one of the KYBDbusses, we could use the initial fc pots to tunethe cut-off of the VCF to 1000 Hz when theoscillator is at 250 Hz (on the same key, ofcourse), and they would track together up and

down the keyboard, keeping the output timbrethe same. (Just how fc is tuned to a particularvalue will be discussed in the section on Q.)If a keyboard is not used, the VCF can be madeto track with the VCO by patching a controlinput to the 1V/octave inputs of bothmodules.

Notice that the effect of having a VCF track aVCO may also be desirable. Try filteringwaveforms other than sawtooth; rememberthat for any one of the infinite timbres you

can input to a VCF, an infinite number of newtimbres is available at the output.

Probably a more useful application of the VCFthan merely generating new timbres is theability to voltage control a timbre. Considerthe following patch:

500 1000 1500 2000 2500

frequency

fc

VCO VCA

TG 2

VCF

TG 1

Out

a certain amount of power to the input signal,the overall volume goes down when they arefiltered out.

This kind of timbre generation is calledsubtractive synthesis, since an original signalwith many harmonics is filtered to give a newsignal with few harmonics. In the case of thefiltered sawtooth above, the output is spec-trally so simple that it could be duplicatedalmost exactly by an additive synthesis.

To do this, you would mix three sine wavesfrom three synched oscillators at 500, 1000,and 1500 Hz in the appropriate ratio. Noticethat if the cut-off frequency is brought low

enough (fc = 500 Hz will do), an almost puresine wave results. This is how sine waves canbe obtained using only sawtooth/pulse VCOs.In fact, pulse at 50% duty cycle can be filteredto give an even cleaner sine wave with lessattenuation, since it has no 2nd harmonic.

Suppose we run a 250 Hz sawtooth throughour VCF with its fc = 1000 Hz. You can see thatthe output now will contain the first fourharmonics in almost unchanged proportions,and significant levels of the 5th, 6th, and 7thharmonics as well. Clearly this output willhave a razzier timbre than the output at 500Hz. In order to maintain the same timbre overthe whole range of desired pitches, we needsome way that the value of fc will track withthe signal input frequency. If the filter isvoltage controlled, this is easy. Notice the 1V/octave control input and the KYBD switch.These serve much the same function on theVCF as the analogous controls on the VCO;

It is assumed in this patch that the VCO, VCF,and TGs are all on one KYBD bus, that theVCA is in linear mode with the initial gain setto squelch the output when the transient isoff, and that the TGs are patched intoattenuable control inputs. (These assumptionswill be made without comment in the future.

Exceptions, of course, will be discussed).Notice that the timbre of every newly-attackednote goes through a reproducible evolution asa result of the transient an the VCF. Experi-ment to see the effects of initial fc, attenuationof the transient, and transient parameters.A fast attack and decay on the VCF transientgive a sound suggestive of a plucked string; aslow attack on the VCA and the VCF soundslike the attack of a brass instrument (brass


19/88


instruments are characterized by their spectralbandwidth increasing as the volume in-creases). Try inverting the VCF transient (ifyouve played in the same orchestras I have, itshouldnt sound familiar). Sub-audio sine ortriangle control of a VCF gives an interestingvariety of vibrato. Try this on a sine wave

input for starters; you might expect the effectof a VCF on a sine wave to be similar to aVCAyou will notice, however, an apparentpitch vibrato as well as the expected tremolo.This is due to the fact that a filter introducesphase changes in frequencies close to fc (phasechange = 180 at fc for the 2100).

As such, the VCF vibrato is an effect unlikeeither a true vibrato or a true tremolo. Trysynchronous audio-rate fc -modulation. This isa powerful effect, capable of even more

versatile timbral changes than audio-rate A.M.If a transient is applied to one attenuablecontrol input while the modulation frequencyis applied at the other, very complex time-evolutions of timbre can be obtained. You canalso control the modulation index with a VCA.

The Q pot allows a resonant peak at fc to besuperimposed on the attenuation curve of theVCF. The numerical value of Q is proportionalto the increase of gain of the VCF at fc as aresult of this resonant peak. As the Q is

increased, the height of the resonant peak isincreased, and the rest of the attenuationcurve is depressed (the overall volume goesdown). The attenuation curves at right withincreasing values of Q show the effect.

In frequency space, a filter can be viewed as amodule with a frequency- dependent gain.This is the view we have taken so far. Atextremely low values of fc, it is more intelli-gible to look at the VCF in the time domainas a module which limits how fast its output

voltage will change. Turn the initial fc pots fullleft. This value of fc is around 2-5 Hz. Nowapply the control voltage from a keyboard orother source to the signal input and patch theVCF output to a VCO F.M. input. Notice theexponential portamento (q.v.) which results.By applying a control voltage to the fc controlinputs, the rate of the portamento can bevoltage controlled. In fact, fc will go wellbelow 0.1 Hz with -10V on the 1V/octave

input. Higher values of fc (around 1 Hz andup) may find many applications wheneversharp voltage transitions need to be smoothedout. See the chapter on the lag processor formore ideas.

In the 2100 VCF, the Q can only be increasedto about 20. Past this point, the filter breaksinto oscillation. If the Q is Set barely into theoscillation region, the oscillation is purelysinusoidal, and at precisely fc (as the Q is

increased, the distortion increases and thefrequency goes down). The fact that thisoscillation occurs at fc makes it possible to setfc at a particular value; it also makes it possibleto trim the full-level control input to precisely1V/octave. Thus, to assure that several VCFsare set to the same initial fc, simply turn offthe signal inputs, turn up the Q until oscilla-tion first occurs, and zero-beat the VCFs at thedesired frequency. Then turn the Q back

Q

No Q

Mid Q

Hi Q


20/88

20

down. Similarly, to set up the patch usedearlier, which cut out all harmonics above thethird from a sawtooth input, just tune the VCFto oscillate one octave above the VCO.

The general use of Q, however, is not toproduce oscillation. The presence of a resonantpeak at fc not only makes for a new timbre,but it makes any movement of fc more obviousto the ear. Without Q, the moving shoulder ofthe cut-off curve is not directly perceptibleonly the rolling in and out of harmonics isnoticeable, and when fc is high compared tothe input fundamental, the amplitude of theaffected harmonics is not very great. With ahigh Q, however, sweeping fc causes everyharmonic present to be sequentially boostedto a high level. You can hear this happen byapplying a sawtooth wave to the signal input

of a VCF, turning the Q up just short ofoscillation, and sweeping slowly from low tohigh fc with the pot. If you listen closely andkeep a light touch on the pot, you can isolateevery harmonic up to at least the 10th beforethey get too close together to separate. Whenthis sweeping is done quickly say, with atransient on a control input, the result is acharacteristic boing or wah sound. (Two inquick succession would give a wah-wah.)

An electronic description of Q may be of

interest at this point. A portion of the VCFsubmodule output is fed back, in an invertedsense, to the input circuitry through the Qpot. As the Q is turned up, the amount of thisfeedback increases. At most frequencies, theresult is merely a cut in amplitude, since thefeedback is negative. At frequencies close to fc,however, the phase of the output begins tochange relative to that of the input, as wasmentioned earlier. At fc

precisely, the phase

change is 180; as such, any signal present at fcgets inverted once by the filter and again by

the feedback loop, thereby feeding backpositively on itself and giving a resonance.Oscillation occurs when the gain around thefeedback loop reaches unity.

Voltage controlled Q can be obtained with the2100 VCF by creating this negative feedbackexternally through a VCA. For higher values ofQ, a Dual Preamp (q.v.) can be used to providemore gain in the feedback loop.

VCO VCF

Q control

VCA 2

VCA 1 Out


21/88


Voltage Controlled Highpass Filter

The function of the 2110 Voltage ControlledHighpass Filter is analogous to that of the2100 VCLPF, except the VCHPF passesfrequencies higher than its fc and cuts out (at

24 dB/octave), frequencies lower than fc. TheVCHPF has an inverting attenuable controlinput whose use will be discussed later. Inaddition, the VCHPF does not have the Qcontrol; details of the electronics dictate that anegative feedback loop which will produce Qin the VCLPF would give rise to distortion inthe VCHPF. The following is a gain curve forthe 2110 VCHPF when fc = 1000 Hz:

500

0

10

30

20

1000 1500 2000

frequency

fc

slope = 24dB/octavedB


22/88

22

Apply a control voltage simultaneously to acontrol input of the LPF and to a non-inverting control input of the HPF, and adjustthe attenuators to about the same controllevel. The band-accept filter will now respondto a change in this control voltage by chang-

Notice that the HPF can filter out the funda-mental frequency of an input while leavingthe harmonicssomething the LPF can neverdo. As a result, the timbres produced by anHPF tend to be unnaturally razzy, and thepitch of the output is often not clearlydefinable.

Though a useful source of new timbres in itsown right, the HPF is particularly interestingwhen used in conjunction with an LPF.Consider an HPF and an LPF in series, as in thefollowing patch.

500250

0

-10

-30

-20

1000 800040002000

frequency

dB

HPF fc LPF fc

slopes = 24dB/octave

VCO VCLPF VCHPF VCAOut

Any frequency which reaches the VCA must beable to pass through both the LPF and theHPFthus, we obtain a band-accept filter(a band pass filter is a special case of the band-accept filter where the fcs of the LPF and theHPF are the same). Clearly, if the fc of the HPFis above the fc of the LPF, no signal of anyfrequency can get through the series pair.

Here is an attenuation curve for a band-acceptLPF/HPF pair:

VCO

VCLPF

VCHPF

VCAOut

ing the center frequency of the band withoutaffecting the bandwidth. If the band is fairlynarrow (LPF fc is slightly higher than HPF fc),the effect is much like Q, in that narrow bandsare sequentially heard as the center frequencyis swept up and down. Experiment with atransient generator as the source for such a

control voltage. Notice the effects of HPF fc,LPF fc, transient attenuation, and Q from theLPF. With either filter used alone, changes in fcwill affect the overall volume considerably.However, in the case of a band-accept filter,certain settings of the two control attenuatorswill allow nearly constant volume to beobtained over a wide range of centerfrequencies.

In the same patch, switch the transient on theHPF to the inverting control input and set the

attenuators the same as before. The responseto a transient is now to vary the bandwidthwithout appreciably affecting the centerfrequency, since the HPF shoulder movesdown when the LPF shoulder moves up. Avariable bandwidth filter such as this doesaffect the overall volume. The signal will becompletely cut out if the control voltage goestoo low, since the shoulders will then cross.

An LPF and an HPF can also be patched inparallel, as follows:

Here, the only frequencies which cannot reach

the VCA are those which cant pass either theLPF or the HPFthe result is a band-rejectfilter. In this case, the LPF fc is lower than theHPF fc. If vice-versa, all frequencies are passedand some are boosted (this could be called aband-boost filter; not really a filter at all, andwe havent looked into it much (maybe itsjust the thing for your application).


23/88


The notch filter is a special case of the band-reject filter where the LPF fc is the same as theHPF fc. You might guess that the output at fcwould only be down 6 dB, since fc is attenu-ated 3 dB by each filter. However, the LPFintroduces negative phase shifts in frequenciesclose to fc, while the HPF introduces positivephase shifts; in the case of a notch filter, thephase shifts from the two filters are such thatfc itself is entirely cut out. (More about notchin the UAF chapter!). In any case, true notchcan be produced with an LPF and an HPF inparallel by tuning both to the same fc andinverting one of their outputs before mixing.

This inversion occurs automatically if the twoattenuable inputs on the VCA are used for themix, since the lower of the two inverts. The fcof the HPF is hard to set precisely, since the Qfeature is not available. However, in the case ofnotch, the final effect is so unmistakable thatthe HPF can simply be tuned until the effectoccurs.

Both notch and the more general band-rejectfunction give rise to very interesting timbralchanges as the center frequency is swept up

and down. The result is much like the phase-shift/flanger effects produced by tape delaysand commercially available phase-shifters.(Commercial electronic phase-shifters typicallyproduce three notches in the spectrum, whilea tape delay can produce a tremendousnumber of notches in the audio portion of thespectrum). Try using a transient to move thecenter frequency of a band-reject filter. A sub-audio sine wave used instead will give an effect

500250

0

-10

-30

-20

1000 800040002000

frequency

dB

HPF fcLPF fc

slopes =24dB/octave

similar to the phase-shifter vibrato. A variablebandwidth band-reject filter is obtained byusing the inverting control input on the HPF.The effect is somewhat obscure, but you maywant it someday!

Clearly, very complex spectra and changes inspectra can be obtained when several LPFs andHPFs are used in various series/parallelcombinations. If you familiarize yourself withthese filter functions now, the UAF andresonant filter will be more intelligible.

A sample band-reject attenuation curve is asfollows:


24/88

24

Universal Active Filter

The 2120 Universal Active Filter is a multi-mode resonating filter with simultaneoushighpass, bandpass, lowpass, and notchoutputs (all at the same fc). It has threeattenuable signal inputs, two attenuable fccontrol inputs and one at 1V/octave, oneattenuable Q control input and one at 1V/factor of two, and several features which willbe discussed in a moment.

First, though, something about filters whichhasnt been mentioned yetany filter,whether highpass, lowpass, or a combinationis characterized by a certain number of polesElectronically, a pole is produced by oneresistor/capacitor pair. A one pole filterproduces a cut-off slope of 6 dB/octave and anattenuation at fc of 3 dB. When several polesof the same type are strung in series, the chainacts in a multiplicative way; for example, the2100 VCLPF has four lowpass poles in serieswhose fcs are always at the same value

as such, when this value of fc

is, say, 1000 Hz,the fc of the whole filter is also 1000 Hz, butthe cut-off slope is 6 dB/octave per pole, or 24dB/octave total. Similarly, the attenuation ofthe VCLPF is 12 dB at fc, since each poleattenuates 3 dB at fc.

The UAF, on the other hand, is a two-polefilter. The lowpass and highpass outputseffectively use two poles in series, so their cut-off slopes are both 12 dB/octave. The bandpassoutput effectively uses one pole for each sideof the band, so its two cut-off slopes are both 6dB/octave. Notch, being a special case becauseof the phase considerations discussed in theprevious chapter, has a very steep 60 dB notchcentered on or near fc.

Notice the difference in sound resulting from aUAFs lowpass and a VCLPF as a result of thedifferent cut-off slopes. Compare the UAFshighpass with the VCHPF. Sharper cut-offslopes are popularly regarded as being more


25/88


desirable; we tend not to make such valuejudgments, since you may like the shallowercut-off better for certain effects.

Q in the UAF is quite different from Q in theVCLPF. Notice that the UAF will not oscillateeven at maximum Q, which can be over 500.Also, turning up the Q does not lower theoverall volume of the output as it does in theVCLPF. Notice, too, that lowpass, bandpass,and highpass can all have Q with the UAF. Theexplanation for these features is that, whereasa VCLPF or a VCHPF just performs a singlefilter function on its input, the UAF actuallysolves several simultaneous filter equations togive its several outputs. With the extremelyhigh Q available from the UAF, the effect ofsequentially sweeping through the harmonicsof the input by varying fc is really spectacular.

Using a sawtooth for the input and Q atmidrange or above, you should be able toisolate harmonics well above the 50th, (as theystart to blend, just turn up the Q some more).

An impressive patch with the UAF is simply toinput a signal from a synthesizer voice or anexternal instrument, turn the Q way up, andsweep fc around with a slow triangle wave, say,0.1 Hz. Listening to any of the outputs(lowpass may be most pleasant), you will hearan ascending and descending arpeggio which

can have very intricate intervals if the inputsignal is a chord. If a keyboard-controlledvoice is used for the input and the UAF isswitched to track the keyboard, the arpeggiowill maintain the same interval sequenceregardless of the key depressed.

You may find certain wolf notes in thispatch which result when the Q gives a lot ofgain to a frequency which is already present athigh amplitude. The voltage-controlled Q isideally suited for eliminating these wolf notes

by lowering and raising the Q as needed. Inthe above patch, you might hear wolf noteswhen the fc of the UAF coincides with thelower harmonics of the input, since they aremuch higher in amplitude than the highharmonics. You could eliminate the boom-ing of the low harmonics by using the sametriangle to control the Q (through theattenuated Q control input)thus, at lower fcsthe Q would also be lower.

Another application of voltage-controlled Q isin the synthesis of human speech, which isdiscussed in detail in a later chapter.

The standard howling wind patch is done verywell on the UAF by the lowpass filtering ofnoise from a noise source with high Q. Whenyou want more howling, turn up the Q controlvoltagethe UAF will boost any of fc presentin the noise, and since pink noise contains allfrequencies, howling at fc will always be heardif pink noise is used.

If white noise is used, the howling will soundlouder at higher frequencies. Pink noise willgive constant volume over all frequencies.)Similarly, a sharp tick (say, a keyboard trigger)applied to a signal input of the UAF will causea ringing at fc when the Q is high. As the Qis raised, the ringing has a longer duration.

Dont mistake these effects for oscillationtheUAF merely boosts any of fc present in theinput signal; when this signal is prolongednoise, a prolonged howling is heard at fc.When the noise is very quick (a tick), the pureringing at fc is all that can be heard. Theability of the UAF to ring is made available as apre-patch by means of the keyboard percus-sion switch. When switched to one of thekeyboards, this switch patches the trigger fromthat keyboard through a shaping circuit into

the signal input summing node, and the gatefrom the keyboard into a transistor switchingnetwork which selects which of the two Q potsis active at a particular time. When the gate islow (no keys depressed), the Q which is ineffect is set by the final Q pot (this is onlytrue in keyboard percussion mode. In normalmode, the final pot is inert, and the Q is set bythe pot). The Q pot, in keyboard percussionmode only, determines the which is in effectwhen the gate is high.

Thus, when a key is depressed, the trigger fromthe keyboard introduces a tick into the signalinput (this occurs even with all the attenuatorsfull left), causing the UAF to ring at fc if the Qis set high enough. Since higher Q gives alonger ring, the Q pot effectively determinesthe initial decay time of the ring. When thekey is released, the Q jumps to the value set bythe final Q pot (either higher or lower Q is


26/88

26

possible), and the UAF rings down at thatvalue of Q. Thus, the final Q pot effectivelydetermines the final decay time of the ring. Inits usual use, the fc is controlled by thekeyboard voice voltage through the KYBDswitch, making the keyboard percussion modea musically playable patch. It can be used with

all the signal inputs off, or the percussioneffect can be superimposed an input signal.(We use it most often to calibrate the1V/octave control!)

So what do you do with the four outputs? Youcan, of course, use them one at a time; youhave already heard the difference between thetwo-pole functions of the UAF and the four-pole functions of the other filters. In addition,the UAF bandpass and notch are moreconvenient to use than the LPF/HPF pairs

discussed earlier. In the case of notch, therelative mix of highpass and lowpass can becontrolled with a single pot, the notchfrequency pot. As the pot is turned left ofcenter, the fraction of highpass in the mixincreases, causing the center of the notch to beat a lower frequency than fc. As it is turnedright of center, the fraction of lowpassincreases, causing fnotch to exceed fc (fnotch candiffer from fc by at most a few semitones eitherway). With the notch precisely at fc, high Qcauses the notch to be defeated, while lower Q

values are themselves swallowed by the notch.When fnotch is different from fc, however, botheffects can occur, the notch at fnotch, and theQ peak at fc. An attenuation curve for such aknob setting might be helpful:

As this formation is swept up and down infrequency space, the effect is quite strange,being sort of a combination of wah-wah (Q)and phase-shifter (notch).

Given several different filter functions all withthe same fc, it is natural to think of mixingthem, and the possibilities are indeed musi-cally useful. Try patching highpass, bandpass,and lowpass into three channels of a QuadVCA. (Notch can be left out since it is the sumof H.P. and L.P.) With the initial gains of thethree channels set the same, the mix output ofthe QVCA should be equivalent in timbre(though maybe not in amplitude) to theoriginal signal input to the UAF. This expressesone of the basic equations solved by the UAF:H.P. + B.P. + L.P. = original signal. With the Qminimal, no change will be apparent in the

output as fc is varied, since all frequencies areattenuated the same amount. Q can still beadded, however, and the result is a resonantpeak at fc in an otherwise flat response, veryuseful for producing formants (q.v.) in anoutput spectrum. Experiment with thedifferent mixes possible from the QVCAsweep fc around and vary the Q. Rememberthat fc, Q, and the mix proportions can all bevoltage-controlled. The drawings on thefollowing page show some of the attenuationcurves available from the UAF/QVCA pair:

Notch frequency

pot turned left

of center, Q midrange

fc

fnotch


27/88


0

-5

-20

-15

-10

frequency

dB

fc

fc

LP = , BP = , HP = full

Q = +5 dB

12

12

14

0

-5

-20

-15

-10

frequency

dB

fc

LP = BP = HP = fullQ = +5 dB

-60

0

frequency

dB

fc

LP = HP = full

BP = off, Q = off

-60

0

frequency

dB

fc

LP =

HP = full

BP = off

Q = +10 dB

fnotch

12

0

-5

-20

-15

-10

frequency

dB

fc 2fc

LP = full, BP = , HP =121

4

Q = min

0

-5

-20

-15

-10

frequency

dB

fc

LP = full, BP = full, HP = off

Q = min

slope = 6 dB/octave


28/88

28

Resonant Filter 1140 Audio UAF submodule used in the RFwill not go as low as the UAFs fc, and the Qwill not go as high. Nevertheless, as a fixedfilter or formant filter, the resonant filter isjust as capable, much cheaper, and generallyeasier to use than the UAF/QVCA pair.

Several RFs can be used in series or parallel togive very intricate formant spectra. Weregularly use two RFs in parallel (stereo) onthe final output of our polyphonic system. Byusing moderate Q and adjusting the fcs by ear,we achieve an open vowel sound in theoutput timbre (in human speech, vowels arecharacterized by two major formants in thespectrum). An amazing range of timbres isavailable just by changing the two fcs. Wevebeen told that a violin can be duplicatedalmost precisely with twenty resonant filters

set to the right fc values! (There must be aneasier way).

The 2140 Resonant Filter is one module whichis capable of producing the same two-pole

mixed-function filter characteristics as theUAF/QVCA pair just discussed, but lacking thevoltage-controllability features.

Fc is determined on the front panel only by asingle pot (though there is an approximate 1V/octave Burndy input in the back), Q is notvoltage-controllable, and the mix is notvoltage controllable. In addition, the fc of the


29/88


Lag Processor

The 2340 Voltage Controlled Lag Processorperforms a rate-limiting function on its input;

it introduces a linear or exponential slide inthe output voltage if the input voltage changesfaster than a certain rate. Typically it is used toprocess control voltagesfor example, it cangive voltage controlled portamento when itsinput is a keyboard control voltage, it can turna gate into a voltage controlled attack/releasetransient generator and it can take the sharpjumps out of the output of a memory, a VSOU,or a sample & hold.

In linear mode (shape pot full left), a fastupward change in the input voltage producesan upward linear ramp in the output whoserate is determined by the sum of the upinitial rate pot and the attenuated up ratecontrol input. For larger values of this sum(initial rate pot clockwise, rate control input

more positive), the rate of the up ramp ismade faster. Similarly, a fast downward changein the input voltage gives a downward linearglide, whose rate is set by the down rate potand the down control inputs. In linearmode, a change in the input voltage whichhappens more slowly than the rate set by theappropriate controls is completely unaffectedby the LP, and appears at its own rate at theoutput.

In exponential mode (shape pot full right), the

LP acts somewhat like a VCLPF with a low fc;the use of the VCF for introducing exponentialglides in control signals was discussed in theVCF chapter. The big difference is that the LPcan have different fcs for upward and down-ward changes, and they can be separatelyvoltage controlled.

A few diagrams will explain the function ofthe LP completely. Consider the case wherethe linear slew rates are 50V/second up and10V/second down. (With no rate control

inputs, these are achieved at roughly midrangeand 3/4 left, respectively). The followingshows a sample input and the output whichresults in linear mode.

+5

0

+5

0

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 seconds

input

output


30/88

30

In exponential mode, with the up rate atmidrange and the down rate about 3/4 right,the LP responds approximately like this:

+5

0

+5

00 .5 1.51.0 seconds

input

output

The response to a gate, as shown in thisdiagram, is useful as a voltage controlledexponential attack/release transient generator.(With the shape pot in between the twoextremes, the response to a sudden change in

input voltage is a linear ramp initially, whichbecomes exponentially rounded, starting part-way through the slide).

With the LP in exponential mode, you mayhave noticed that the output voltage does notsettle out at precisely the same value as theinput voltage. This can best be observed byusing the LP to give exponential portamentoto the keyboard. Patch the keyboard controlvoltage into the LP input and use the outputto control a VCO at 1V. Youll find the VCO to

be out of tune with respect to the keyboard.This is an inherent limitation of the LP whichcan be minimized by selection of the compo-nents, but cant be eliminated. In linear mode,however, the portamento patch works verywellthe LP is, in fact, the only way to obtainaccurate voltage controlled portamento. Trythis patch with different up and down rates;try sinusoidal rate control inputs, or patch thekeyboard control voltage into one or both ofthe rate control inputs as well as into thesignal input, thus giving different portamentorates at the low and high ends of the key-board.

Other applications of the VCLP will beencountered in later chapters in connectionwith other control sources; rather thanmention them all here, they will be discussedas they come up.


31/88


Dual Transient Generator

The 2350 Dual Transient Generator containstwo identical, completely independent four-

phase transient generators in one module.Further discussion will concern only one ofthem; they are doubled up only because thepanel layout is most convenient that way. TheDTG may be confusing at first, particularly ifyou arent too keen on voltage. Its the firstmodule youve encountered that has no signalinput or signal outputits use is strictly as asource of transient control voltages:reproducibly varying, synchronously gatablefluctuations in voltage.

A TG (half of a DTG) performs a fairlystraightforward function: it has two inputs, agate and a trigger; and one output, a transient.In the external position of the triggeringswitch, the external gate jack is connected toboth the gate and the trigger inputs of theappropriate VCDTG submodule (for voltagecontrolled delayed transient generator).In the KYBD position of this switch, the gateand trigger from the selected keyboard are

pre-patched separately into the gate andtrigger inputs of the submodule. For the

moment, well consider the use of a keyboardto give the gates and triggers.

When the gate goes high and the triggeroccurs, the transient remains at zero voltsuntil the delay time set by the delay timepot elapses, and then begins an upwardinverse exponential increase at a rate deter-mined by the attack time pot. If the gatefalls before the delay time elapses, no transientis produced. Assuming the gate stays highindefinitely, the transient, now in the attack

phase, exponentially approaches a level setinternally which we will call the attackapproach voltage (about 12V) Being exponen-tial, the transient technically would neverreach the approach voltage, but it doesnt haveto at another internally set voltage, theattack cut-off voltage (about 10V), a-comparator is fired, causing the attack phaseto end and the initial decay phase to begin.


32/88

32

In the initial decay phase, the transient fallsexponentially at a rate set by the initial decaytime pot, leveling off at the sustain voltageset by the sustain voltage pot. This down-ward approach to the sustain voltage contin-ues indefinitely as long as the gate stays high.(For all practical purposes, the transient can be

said to reach the sustain voltage and stay thereas long as the gate is high). When the gatefalls, the transient enters the final decay phase,in which it exponentially decays to zero voltsat a rate determined by the final decay timepot.

Since the decays are exponential, they cant betimed from start to finish there is no finish.The attack can be timed precisely, since it iscut short at a definite voltage; nevertheless, allthese exponential approaches are conveniently

described by a time constant, which isdefined as the time required for the exponen-tial approach to get within 1/e (1/2.71828...)of its approach value. The following drawingsillustrate the concept of time constant:

Clearly, then, a decay is still significantly shortof its approach value well after the timeconstant elapses. For example, a final decaywith a time constant of 10 seconds will give atotal glide time of about 30 seconds. (The timeconstant is precisely definable, however, whilethe total glide time depends on the sensitiv-

ity of your ear). The attack phase, being cut offat 10/12 of its approach value, lasts about 1.7time constants.

The range of time constants available from thepots for attack, initial decay, and final decay is,in each case, about 1 msec. to 10 seconds. Youmay notice, particularly with a scope, that acertain rotation which gives a change in timeconstant from 1 msec to 5 msec at the far leftextreme will give a change from 2 seconds to10 seconds at the far right extreme. This is an

exponential response to the knobs; it allowsyou to achieve equal sensitivity at very shortand very long time constants. Electronically, itis accomplished by making the submoduleproduce an exponential change in timeconstant in response to a changing controlvoltage; this control voltage is then obtainedfrom a linear pot, giving an overall exponen-tial-response to the pot. The delay time is alsoexponentially controlled by its pot, and canvary from about 3 msec to 3 seconds. Thesustain voltage responds linearly to its pot,

and can be set between zero and +10V.

The following drawings show some of thetransients that can be obtained from TGs;first, the simplest cases, where a keyboard isused to initiate the transient, and only one keyis depressed at a time. Notice that if the gatefalls at any time during the transient, the TGimmediately enters final decay phase:

0

10V

10/e V

EXPONENTIAL APPROACH

timeconstant

0

10V

10-10/e V

INVERSE EXPONENTIAL APPROACH

timeconstant


33/88


When a monophonic keyboard (q.v.) is used,additional triggers are often produced withoutthe gate ever going low; specifically, thishappens either when a lower key is depressedwithout releasing an upper key, or when twokeys are depressed and the lower is thenreleased. It is necessary that a transient be able

to respond (with the appropriate delay) tosuch additional triggers.

The diagram at right shows most cases thatmight arise, and the response of the TG ineach case. The trigger applied by the keyboardis shown in solid red, while the trigger seen bythe circuitry after the delay time elapses (thedelayed trigger) is shown in dotted red. Thedelay time remains constant throughout, andis shown by the double-headed arrows:

00 1 sec.

+10

all controls midrange

volts

GATE

00 10 sec.

+10

no delay, fast attack

slow I.D., S.V. = 0, fast F.D.

(piano envelope)

volts

GATE

00

+10

no delay, slow attackfast I.D., S.V. = 10V, slow F.D.

volts

GATE

00

+10

no delay, fast attack,fast I.D.

S.V. = 10V, fast F.D.(organ envelope)

volts

GATE

GATE

Trigger =

Study carefully the conditions under whichnew attacks are initiated. Notice that a delayedtrigger has no effect if it occurs during anattack phase (or without a high gate). Also,many triggers in quick succession may have


34/88

34

no effect if the delay time is longer than theirspacingonly the last delayed trigger is seen.

The manual gate, like the external gate input,is patched to both the gate and the triggerinput of its TG. When you cant tell whatsgoing on with a TG, this push-button cannarrow down the range of your confusionconsiderably by allowing you to gate the TGby hand. Because of the extremes at which therate pots can be set, you may sometimes getno apparent transient even though you cansee the gate lamp lighting (indicating the gateis high). First make sure the transient ispatched where you think it is, and that it isnot attenuated too far. Next check to see if theattack is set extremely slow, or the delay verylongalternatively, the attack and initialdecay may be so short that the entire transient

becomes a blip (only a problem if thesustain voltage is low). You may also haveproblems with popping sounds if the attackor decay is set very fast; these can be elimi-nated by turning the appropriate pots slightlyclockwise.

The TG need not receive its gate and triggerfrom a keyboard. In the external position ofthe triggering switch, an input applied at theexternal gate jack is sensed as a high gatewithin the submodule whenever it exceeds

2.0V, and as a trigger whenever it passesupward through 2.5V. The difference inthreshold voltages assures that a trigger willnever occur without a high gate. It thusbecomes possible to initiate transients with apulse wave, a sine wave, or any other varyingvoltage with excursions above 2.5V. Slowlyrising voltages will, of course, give non-simultaneous gate and triggerattacks willbegin with the delayed trigger. A very impor-tant use for the external gate is the initiationof transients with digital signals (q.v.). By

patching a digital output of a memory into aTGs external gate, it is possible to have attacksonly on certain notes of a sequencethis ishow timing in a sequence is achieved. Inpolyphonic systems, the external gate is oftenthe most convenient way to apply keyboardgates to their respective TGs. This can be donewith the firm-wire external gate input, ifdesired.

Still another way to apply gate and triggersignals to a TG is through the keyboard busses.If a keyboard bus is free, the TG can beswitched to that bus and can then receive itsgate and trigger from the gate and trigger jackson the power supply front panel. It is thuspossible to use two completely independent

control voltages to provide independent gateand trigger to a TG (the threshold voltages arestill 2V and 2.5V for gate and trigger, respec-tively). The result is a type of and functionperformed on the two inputs, in that theconditions for a gate must be met at the sametime that a delayed trigger is generated for atransient to occur. The effect can be verycomplexsee if you can find a goodapplication!

The standard uses for a transient are pretty

obvious; in fact, you already know a numberof them. As an envelope, or loudness contour,a transient is indispensable. It was because ofits use as an envelope generator that it got thatnamewe have avoided the term because itrefers to only one of the many possibleapplications for sub-audio transients. The useof the TG for envelopes also explains thechoice of an A.D.S.R.-type function (attack,decay, sustain, release); fast attacks with longdecays give envelopes with a percussive effect(like a piano, for example), while slow attacks

with high sustain voltages can give brass-likeenvelopes. Generally speaking, the ADSRtransient is very versatile. In addition to givinggood likenesses of standard instrumentenvelopes, other envelopes not available fromany conventional instrument can be obtainedfrom the TG.

There are a few uses for transients that are lessobvioustransients on filters, VCO frequencycontrols (perhaps the most graphic way tohear the transient pattern), or using a transient

as the signal input to a Sample & Hold. It issometimes useful to invert a transient with aQuad Inverter (q.v.) before applying it to acontrol input. (One application of an invertedtransient was already discussed in connectionwith the VCHPF, but a quad inverter was notnecessary in that case). When you try this,remember to raise the initial setting of thecontrolled parameter, since the inverted


35/88


transient is always negative. Multiple tran-sients can be superimposed to control thesame module, with very complex fluctuatingfunctions resulting. Try patching two TGs intotwo of the control inputs of a VCAuse thedelay time to offset one transient relative tothe other. Several transients can also be used

to simultaneously control different parameterswithin a single voice. Weve already men-tioned the use of separate transients on a VCFand a VCA.

Consider the following patch:

VCO 1 VCA 1

TG 1

VCO 2 VCA 2

TG 2

VCO 3 VCA 3

TG 3

VCO 4 VCA 4

TG 4

Mix

Out

QVCA

If the four VCOs are tuned to track together,say, at intervals corresponding to the funda-mental and the next three harmonics of anote, the envelope of each harmonic can be

independently controlled by its own TG. If allthe TGs are triggered from the keyboard, thedelay time allows any of the envelopes to havethe same shape but to be displaced in time,the only limitation being that the final decayswill all begin simultaneously when the gatefalls.


36/88

36

VCTGIU

PLUG

UPPER

SKT

PLUG

LOWER

SKT

PLUG

BOTH

SKT

DTGDTG VCTGI1

VCTGI2

DTG

The 2355 Voltage Controlled TransientGenerator Input module makes it possible tocontrol all five of the TG parameters withinput voltages; the VCDTG submodule isalways voltage controlled, but it receives itscontrol voltages only from the pots on themodule when the VCTGI is not used. The

VCTGI simply adds attenuated controlvoltages to the voltages obtained from theappropriate pots.

The control inputs applied to the VCTGI arecarried by a length of ribbon cable (completelyindependent of the power and busses ribboncable) to a DIP socket on the TG moduleboard. The VCTGI inputs can control eitherthe upper TG in the DTG, or the lower, orboth, depending on how the DIP plug isplugged into the DIP socket on the VCTGI:

The DIP plug on the DTG itself is plugged intoits socket in the normal way. One VCTGI cancontrol any number of TGs; here is a wiringscheme which gives VCTGI 1 control over theupper TGs in three DTGS, and gives VCTGI 2control over the lower TGs;


37/88


Notice only one piece of ribbon cable isrequired. The responses of the time constantsof the TG to the VCTGI inputs are exponential(as you would expect, since the responses tothe pots are exponential). With the attenua-tors on the VCTGI full right, the delay, attack,and decay times vary roughly by a factor of

ten with a 2V change in the input. Thus, if thefinal decay pot an the TG is turned full left,giving a time constant of about 1 millisecondwith no input from the VCTGI, the timeconstant can be increased to 1 second byapplying about 6V to the unattenuated finaldecay jack on the VCTGI. As the attenuatorsare turned left, the response to voltage remainsexponential but becomes less sensitive.

With the sustain voltage attenuator full right,a voltage applied to the sustain voltage jack

simply adds its value (which may be negative)to the value set by the sustain pot on the TG.The response is thus linear, and may be madeless sensitive with the attenuator. The sustainvoltage of the transient will not exceed about12V regardless of the voltage applied to theVCTGI jack. It will go below zero, however,and this may be useful in some applications.(Though we havent thought of one!)

Notice that the time constants can be mademuch longer with the VCTGI than they can

with the TG pots alone. In fact, with the finaldecay pot turned full right and +10V on theunattenuated final decay input of the VCTGI,the time constant for the final decay istheoretically about 11 days! Less extreme timeconstants are perhaps morevaluable.

The patches that come tomind with voltage controlof TG parameters arenumerous. A foot pedal can

be used as a sustain pedalby having it control finaldecay time. A foot pedal onthe sustain voltage gives atype of sustain, also.

Try patching the inverted keyboard controlvoltage into the final decay input. (Turn thefinal decay pot on the TG high, since theoutput of the inverter is always negative). Thedecays are now shorter for high notes andlonger for low notesmuch like the behaviorof a piano keyboard. Use the four outputs of a

VSOU (q.v.) to control four of the parametersof a TG. (The TG pots can all be turned fullleft, so that only the VSOU pots determine thecontrol voltages). By setting up the eightcolumns of the VSOU appropriately, eightcompletely different transients are available,and the desired transient can be instantlyselected with the pushbuttons on the 8AG. Inthis same patch, any number of additionalTGs could be controlled by the same VCTGI,and all would have equal parameters if thepots on every TG were turned full left. This useof the VCTGI as a gang control for severalTGs is very powerful.

Try using sub-audio waveforms from VCOs tocontrol various TG parameters. For example, asine wave on sustain voltage will give atremolo (if the transient is used as an enve-lope) which begins only after the attack andinitial decay phases. (In this patch, initialdecay rate will have an effect on the depth ofthe tremolo). If the VCOs producing thecontrol waveforms are gated by the keyboard,

complex but reproducible transients can beobtained.

Heres a sample patch:

VCO 2

VCTGIHDI

VCO 1(audio)

TG

TG

VCO 3

KYBD

GATE

a.t.

s.v.

Out

slowgate

gate

slow


38/88

38

Noise Generator

Noise, in electronic terminology, has fairly

specific definitions and varieties; white noiseis a mixture of all frequencies in the audiospectrum at the same average level (analogousto white light in the visible spectrum).

This is obtained from the noise output jackwhen the spectrum pot is full left. Theattenuation of the theoretically flat spectrumat low frequencies is due to the necessity offiltering out any DC component in the output.Listen to white noise through a VCF with theQ set just short of oscillation. As you sweep fc,you will hear the VCF output whistlesmoothly through the entire audio spectrum,indicating the presence of every frequency inthe noise.

Using a VCA as an attenuator, listen to the NSnoise output and turn the spectrum pot fullright. The output is now called red or lowfiltered noise. In this position of the spectrum

pot, the noise output is white noise which hasbeen lowpass filtered at 6 dB/octave. (This isdone with a single pole filter whose fc is about34 Hz.) Its spectrum is as follows:

frequency

amplitude

1 Hz 16 Hz 20 kHz

frequency

amplitude

1 Hz 16 Hz 20 kHz

34 Hz

Notice that the low filtered noise is no lowerin volume than the white noise, even thoughit has been filtered. The circuitry boosts thelow filtered output to accomplish this, the

The spectrum of white noise available fromthe 2400 Noise Source looks like this:


39/88


result being that low frequencies are actuallylouder in the low filtered output than in thewhite noise, while high frequencies areattenuated

Documents

Eµ Modular OpManual