BUILDAHIGH PER-

PART ONEROBERT R. CORDELL

$ llX'*'*r q ' s o

:t *'::*' or$ls$raN

,.F

he Total Harmonic Distortion(THD) analyzer is probably themost widely used instrument for

evaluation of distortion in audio systems.Unforlunately, those analyzers which aregood enough to evaluate contemporaryequipment generally cost in the neigh-borfrood of 92,000. Because of a care-ful selection of leatures and a topologywhich lends itseli to realization with low-cost components, the analyzer de-scribed in this article can be constructedfor a fraction of that cost, while providinga measuremenl f loor on the order of0 001 percent across the full audio band.

A typical measurement setup of aTHD analyzer is shown in F ig. 1. l t con-sists of a low-distortion oscil lator feedingthe Unit Under Test (UUT) which in turnfeeds the THD analyzer; an oscil loscopeis optionally used to observe the wave-fornr of the distortion products. (ln moslhigh-performance THD analyzers, theoscil loscope and analyzer are in oneunit.) lt is well known that a nonlinearitv

The 'scope photo in Fig 2 shows thefundamental sine wave in the top traceand a typical distortion waveform in thebottom trace. Viewing the distortion inthe "time domain" is very useful in eval-uating the nature of the distortion mech-anism. Here, the disturbances near thezero crossings of the fundamental implythat crossover distortion exists. As far asaudibil i ty is concerned, lor a given mag-nitude of distortion, smooth or "soft"distortion wave{orms are preferable tojagged or "harsh" waveforms.

How THD Analyzers WorkWe can get a feel for some of the ana-

lyzer requirements by seeing what thedesired performance specification im-plies. Here, our goal is a measurementfloor of less than 0.001 percent. Thismeans that if we bypass the UUT, thatwhich remains after the fundamental isrejected is less than 0.001 percent. or100 dB down. The res idual wi l l consis tof distortion, unrejected fundamental,

n e , , ' ' " ' . . , " r o o \ .

*'iif,;{';,, -.,' ,* t'i.,a d . _ * . , l l o1\*

"l,, &.- *6p bo

o(oupur

SN?o" I 'NPT c

,ji:: '$,i

,., Ni i r

in the UUT, when adequately excited bya sine wave (the so-called "fundamen-tal"), wil l produce at the output harmon-ics of the sine wave in addition to thefundamental. The principle of the THDanalyzer is to remove the fundamentalentirely and measure what is left. Thisremoval of the fundamental is usuallyachieved with an extremely sharp anddeep notch filter centered on the funda-mental frequency.

noise and hum. Clearly, we need a verygood oscil lator, one with distortion atleasl 10 dB down from the floor, or bet-ter than 0.0003 percent . In addi t ion,low-noise, low-distortion electronicsthroughout the analyzer are mandatory.

Similarly, the notch fi l ter should prob-ably have at least 110 dB of rejection,while attenuating the closest harmonic(the second) by less than, say, 0.5 dB.Such a sharp notch is normallv diff icult

34

PARI OI\E

to achieve and even more diff icult totune manualiy. This problem is dealt wlthin most high-performance analyzers byautomatic tuning circuits, often termed"auto tune" [1]. Notch fi l ters are gener-ally realized by passing the signalthrough two paths with differing phaseand/ or amplitude characteristics so thatthe two resulting outputs are equal andof opposite phase only at the center fre-quency. When combined, the signalsthus cancel each other at the center fre-quency. lt can be seen that an automatictuning system must exert two types ofcontrol: One to assure the proper phaserelationship at the notch frequency, andone to assure that the amplitudes of thetwo signals are exactly equal at thenotch frequency so'that lull cancellattoncan occur. As we shall see later, thesetwo control circuits each function by ex-amining different characteristics of theoutput of the notch l i l ter and then adjust-ing notch l i l ter control elements accord-ingly. The auto{une circuits of a THD an-alyzer are generally responsible for asubstantial portion of the overall cost,but they are virtually indispensable. Withwell-designed auto-tune circuits, rejec-tion of the fundamental is so completethat analyzer internal distortion and noisegenerally dominate the residual and es-tablish the measurement tloor.

To make maximum use of the con-venience assocraled with auto tune, it isbest to have the oscil lator packaged inlhe analyzer so that its frequency can becontrolled in step with that o1 the ana-lyzer. This also tends to reduce the con-trol range required ol the auto{une cir-cuits, making them less prone to l imit an-alyzer performance.

An additional feature found in mostbetter analyzers is called "auto setlevel." In analyzers without this feature,a level-setting calibration adlustmentmust be made whenever the input levelis changed. An auto set level featurepermits the input level to the analyzer tochange over some range, say 11 0 dB,without afiecting the calibration of thedistortion meter. lt is achieved with atype of a.g.c. circuit which adjusts thegain in the measurement path (after thenotch fi l te0 downward in direct propor-tion to increases in input level.

Finally, many analyzers include high-pass and low-pass Jilters of varying com-plexity and flexibil i ty to optimize perform-ance. These distortion product filters are

placed in the measurement path afterthe notch ft lter. Generally, a high-passfi lter is used to minimize the eflect o1hum in the UUT on the measurement.Low-pass filters are usually employed tolimit the measurement bandwidth (tominimize noise) while passing all of theharmonics o1 interest (e.9., up to thetenth).

Practical Analyzer DesignMy goal here was to Provide a THD

analyzer desrgn capable of state-oJ-the-art oerformance which can be construct-ed at moderate cost and with readilyavailable parts. At the same time, I feltthat retaining the features of a built- in os-cil lator. auto tune, auto set level, andproduct f i l tering was essential. One com-oromise made to keep down costs wasto l imit the number o1 switch-selectableoperating frequencies to 1 0 evenlyspaced frequencies per decade from 20Hz to 200 kHz. Another cost saving re-sulted from the extensive use oi lCs,even in the most crit ical circuits. In par-ticular, the superb distortion perform-ance of the 5534 operatronal amplif iermade this possible

A major expense in many analyzers isdue or related to the control elements inthe auto{une circuits The use of so-called "state-variable l i l ters" in both theoscil lator and the notch fi l ter in thts ana-lyzer permitted the use of inexpensiveFETs as the control elements in thesecircuils. Ihe greater rmmunity to control 'element nonlinearit ies of these fi l ters hasbeen pointed out in [2] The topology otthese fi l ters also permits the use of 11 0percent capacitors where precision com-ponents would normally be required.

While several sets of four identicalrange-select ion res is tors must bematched to within 11 percent. the abso-lute value of these resistors is not crit ical.Thus, a patient constructor with a digitalVOM or simple resistor bridge can selectmatched sets from standard 5 percentcarbon-fi lm resistors. lt 's highly unlikelythat you won't f ind four or more resistorsmatched to within +1 percent among anrnr rn nf ten +5 ncrcent carbon-f i lmv , v v v v , r v ' '

resistors purchased at the same time.The exterior and interior views of the

completed analyzer are shown in Figs. 3and 4. The circuitry of the analyzer re-sides on one small and two medium-sized srngle-sided printed circuit boards;the regulated power supply is housed ina small metal box at the rear. As can beseen from the photo in Fig. 4, the malorpart of the construction effort is in therange and {requency switch wiring. As aguide to cost, the three circutt boardsand their associated components wil ltyp ical ly cost about $160. The cabinet ,power supply, meler, swilches, switch-mounted components, etc. wil l representadditional expenses.

Before proceeding, one word ol cau-tion is in order. This ambitious andmoderately expensive proiect should notbe attempted by anyone who has nolpreviously built and troubleshot electron-ic construction projects. Anyone whodoes not own or have access to an osctl-loscope should also think twice beforeembarking on lt. This is a challengingconstruction project requiring considerable effort and patience, but the payoffis great for anyone who wants a THDanalyzer with state-of-the-art perform-ance.

Fig. 1 - Typicalmeasurementsetupof aTHDanalyzer.

Fig 2 -'Scopephoto showing thefundamental sinewave (top trace)and a typicalcrossover distorlionwaveform (bottomtrace).

I

35A U D | O / J U L Y 1 9 8 1

l"/

PART OI\E

Fig. 3 - Exterior view of the completed THD analyzer.

ii r ,.ri

Fig. 4 - lnterior view of the THD analvzer.

l. ,s,

fed the input signal and a version of thissignal which has passed through a state-variable bandpass fi l ter whose lrequencyand garn are controlled electronically. Atthe center (fundamental) frequency, bothinputs to the differential amplif ier a;eidentical in phase and amplitude, leavingonly the distortion products at its output.The bandpass fi l ter is sufficienily narrowthat its output is very small at the secondharmonic and higher frequencies, resul!ing in less than 0.5 dB of attenuation atthe second harmonic. Operation of thenotch fi l ter is i l lustrated by the frequencyresponse curves in fig. 6.

The auto-set level circuit consists oftwo voltage-controlled amplif iers (VCAs)to which are fed rdentical control signalsso that their gains vary in the same fash-ion. One VCA is in the palh of the distor-tion product signal, while the other ("ref-erence" VCA) is fed the output (thefi ltered fundamenlal) from the state-vari-able bandpass fi l ter. The latter VCAfeeds a level detector whose d.c outputcontrols both VCAs. The a.g.c. loopformed by the level detector forces theoutput of the reference VCA to be at afixed reference level. Thus, if the inputsignal level doubles, the gain of bothVCAs wil l be halved, resulting in theproper gain correction being applied tothe distortion product path. Both VCAsare realized from inexpensive 1496-typebalanced modulator lOs and achieve acontrol range oI +20 dB with better thanl0 5-dB tracking. Notice that the dis-tortion performance of the VCAs neednot be exceptionally good because theone in the measurement path is passingonly distortion products. Placement ofan a.g.c. circuit ahead of the notch fi l terwould necessitate the use of an extreme-ly low-distortion VCA and might alsocompromise the noise floor of the instru-menL

The auto-tune control circuits, prod-ucl f i l ters, status indicator circuit, andmeter amplif ier are located on a thirdp,c. card, CP3. The auto{une contro lcircuits consist of amplitude and fre-quency detectors which produce d.c. er-ror signals to conlrol the gain and centerirequency ol the slate variable bandpassfi lter. These detectors function by look-ing for small amounts of fundamental inthe output ol the notch ft lter.

The principle of operation is as fol-lows: An amplitude error between thetwo signals at the inputs of the differen-

Overall DesignA block diagram of the complete ana-

lyzer is shown in Fig. 5. We'l l now pro-ceed with a brief overview of the ana-lyzer's operation at the block diagramlevel before moving on to more detaileddescriptions of the individual circuits.

The signal source resides on a singlep.c. card, CPI, and consists of a state-variable oscillator, the necessary ampli-tude stabil izing circuits, an output ampli-f ier, a level control, and an output atten-

uator. The amplitude control circuit is es-sentially a level detector which generatesa d.c. error signal to feed the amplitudecontrol multiplier. The latter controls theamount o{ positive feedback in the oscil-lator and consists of an FET and an oo-erational amplif ier.

The input circuits, controllable notchfi lter, auto-set level circuit, and distortionproduct amplifiers are located on a sec-ond p.c. card, CP2, The notch fi l ter con-sists of a differential amolii ier which is

36

T1 !

r N lqF

= l

L

I ra :o l rf u n r{ i l r ^ ,

j us t^ob \

a m carnar h ^

p h a .

n r p '

E+ h ^U t <

au l "f r rncf i l te .i n ' c -m a xverstneque '

ALJDIC

PART CNE

Fig 5 - THD analyzer block diagram

t ra l ampl i f ier wi l l resul t in a smal l amounlol In-phase (or 1 B0 degree out,of_phase)fundamental at the outpul of the notchfrlter The polarity of this error signal indi-cates which direction of amplitucle ad-lustment is necessary. Similarly, but lessotlvaously, a freqLrency error in' lhe band_pass l lter wil l result in a phase error bre-tween the lwo signals at the differentialampl i f ier This wi l l resul t in a smal lamount oJ fundamenla l component a lthe outpul of the notch f i l ter whosephase is lagging or leading 90 degrees(qrradrature) relative to the fundamentalWhether the component is lagging orleading indicales which direction of tre_quency adlustment is neoessary

Each detector functions by comparinglhe outpLrt ol the notch fi l ter (after thearilo-set level VCA) to a version of thefLrndamental suppl ied by the bandpasslrlter The amplitLrde detector looks forrn-phase f undamental components bymakinq i ts compar ison wl th an in-phaseversion of lhe fundamental suppliecl bythe bandpass f i l ter . S imi lar ly , the f re-quency detector looks for quadrature

lundamenlal components by making i lscomparison with a 9O-degree phase_shifted version of the fundamenlal, alsoconvenrenlly supplied by a different out_put of the bandpass frlter The readyavai labi l i ty of th is quadratLt re fundamen-ta l s ignal is an addi t ional advantage ofthe stale-variatrle fi l ter

The d.c. control siqnals from the leveland frequency detectors also feed a sta-tus indicator circuit. This circuit drivesfront panel LEDs which rndioate whetherthe input signal is too high or too low andalso whether the input frequency is toohigh or ton low (useful when a sourceother than the internal oscil lator is used)

The high-pass and low-pass produclfi l ters in this design are both seconci-or-der (12 dB/octave) desiqns whose cut-off frequencies track changes in the se-lected fundamental frequency Thesetracking fi l lers substantially improve themeasurement floor and simolifv use ofthe analyzer by automatically proviclingoptirnal f i l tering regardless of the funda-mental f requency setting

The high-pass filter falls otf aI j 2 dB /

AUDIO/JULY 198- I

I

octave below the second harmonic fre_quency. lts roll-off shape is chosen so asto have a minimal effect on distortionproduct frequency response. The low_pass fi l ter is a second-order Bessel cje_sign which is down 3 dB at the tenthharmonic f requency. The excel lentphase response characteristic of this de_sign minimizes waveform distortion ofthe distortion products, preservtng tneaccuracy of the visual displav. The ana_lyzer's overall distortion proclltct frequen_cy response is shown in Fig 7

While requiring l itf le in the wav ofelectronics. lhese trackrng ti l ters do addsubstanttally to the complexity of the fre_quency and range swrtches, and somebuilders may prefer a simpler and lessexpenslve alternative. Such an ap_proach, providing the fixed fi l terinq at400 H2 ,30 kHz , and B0 kHz found onmany commercial analyzers, is also de_scribed later in this article.

As a convenience, notice that with thefl ip of a switch (56), the input level canbe directly read with the ranqe selectedby lhe analyzer's input attenuitor.

J I

PqRT ONE

@

olt)o=

Fig 6-Notch filter operation

Fig B-Asimplestatevariable bandpass filter

o

- 5@

II

- ' ofF

J

-20

- 2 5

- 3 00

NORMALIZED FREQUENCY

T

t_I

(Rl : R2), a capacitor (C1 : C2) andan operational amplif ier. The gain of thelntegrator is equal Io 1 /(zrrRCIl, where fis the frequency of interest and R and Care the values o l R1 , R2, C1 and C2. Rand C are chosen here for an inteqratorgain of unity at our j -kHz operatin-g fre_quency. The inverter also has a qain ofunity. Notice that as the 1_V siqnal tra_vels around the loop it picks up gO O"_grees from each integralor and .l g0 de_grees from the inverter for a total of 360degrees, which ls the same as zero de_grees. Thus, at 1 kHz the srgnal from A3feeding the A1 integrator is exacfly thesrgnat requrred to provide and sustainthe assumed output We therelore havean oscil lator.

Looking at it from a slightly differentview, the feedback from A3 pioduces acurrent in R1 which is just equal to thecurrent required by C1 given the as_sumed signal at the output of A j . Thecurrents Into and out of an ideal oo_amp's virtual ground (the inverting inputwhen the noninvertrng input is grounded)must always equal each other

We can think of an oscil lator as abandpass fi l ter with infinite O. What if wewant finite O? We add resistor Ra, oftencalled the damping resistor, We nowneed an input, so R1 is also added. lf wemake the same assumptions as before,we see that the current in R1 sti l lbalances that in C1 at 1 kHz, The cur-rent in Re must therefore come from theinput via Rr, so the voltage gain at 1 kHz(the center trequency, fo) is Re/Rr, andthe phase shift is 1 80 degrees.

At higher frequencies the cunent de-manded by Cl increases and that pro-vided by R1 decreases. The input mustnow supply the extra currenl for C1 aswell as that demanded by Rs. A largerinput rs thus required for a given output,corresponding to decreased galn. By asimilar argument, the gain can be seento decrease at frequencies below fo aswell. We thus have a bandpass charac-teristic. The Q of this fi l ter is proportionalto the value of Re. The center frequency,f", wil l always be the frequency at whichthe gain around the main loop is unity,This means that the center frequencycan be conveniently changed by alteringR, C, the inverter gain, or any combina-tion of these.

With a bit more reasoning, it is easy todemonstrate that the output of the sec-ond integrator (A2) provides a j2 dB/

state-variable filter models a second_or_der differential equation in much thesame way as one would model such anequation on an analog computer. In thiscontext, the state variables are the oufput voltages of the two integrators.

A key to understanding the state_vari_able fi l ter is the recognition that an inte_grator has a frequency response whichoecreases with increasing frequency at6 dBloctave and introduces a 90-de-gree lagging phase shift at all freouen_cies. Let's temporarily ignore the inputresistor Rr and the bandwidth-settinoresistor Re and assume that we have i1-V, 1-kHz s ignal at the output of 41Each integrator is formed bv a resrstor

State-Variable FiltersBecause il is central to the operation

of the analyzer, a few words of exolana-tion regarding the slate-variable fi l ter areappropriate at this point.

A simple state-variable bandpass fi l teris shown in Fig. B. lt consists of two inte-grators and an inverter connected in aloop. An input resistor (R1) and a secondfeedback loop (Re) around the first inte-grator (Al )complete the fi l ter.

The term "state variable" comes fromlrnear system theory involving solutionsto l inear differential equations. The stateof such a system at a given point in timers completely determined by the valuesof the system's state variables. The

Ftg

octtha+ L ̂Lt r t r

a l lprosh t 'qu (

Th,

J t d

- t d

ASanc

d , u

eni

AUD

PART Of\E

^ --O4G 5K, lF

s r r a ' f f r s r x== lu5 / lr 'o i '

"c(xmlC ( L N )

+'' I !!3 .n-uiri' '^'l :+:l

As2a

c7lggrrFR5

6 .8X

R r 5750K

Fig 9 - Circuitry of the THDanalyzer's stgnal source

octave low-pass characteristic. Noticethat the phase of this output lags that ofthe bandpass output by 90 degrees atall lrequencres. This feature is used toprovide the required 9O-degree phase-shifted fundamental signal to the fre-quency detector in the auto{une circuit

The Signal SourceWith a solid understanding of the

state-variable filter, operation of thestate- variable oscil lator is quite simpleAs in the earlier example, we remove Rrand RB as a start.

As with any l inear oscil lator, we mustprovide a means to control the amplitudeof the oscil lations, r.e., provide a type ofa.g.c. circuit. Recognizing that the pres-ence o{ RB produced negative feedback

R23toK

R24I K

R25t o K

c5

R t 6loox

Cl3 - CzO - BYPASS IOT SHOwtl ,trlt

around the first integrator which tendedto damp out oscil lations, we can reasonthat a feedback circuit around A 1 whichcan provide variable amounts of nega-tive or positive feedback wil l allow us tocontrol the oscil lations. Negative feed-back wil l cause the oscil lations to decay,while positrve leedback wil l cause themto grow. This can be accomplished witha multiplier and a resistor in place of Re.To complete the oscil lator. we must adda control circuit to measure the ampli-tude of the oscil lations, then compare itto a fixed reference, and finally deliver ad.c. error signal to the control input olthe multiplier.

The complete schematic ol the signalsource is shown in tig. 9. First. a wordabout nolatron and conventions used on

C EIOO/aF

a 32N5904

s ? 0

MAII{OUT

R t 96.8K

the schematics in this article. For sim-plicity, power-supply bypass capacllorsand power-supply connections to stand-ard pin-out operational amplil iers (+15V to pin 7, -1 5 V to pin 4) are not shown.Oif-board interconnection terminals aredesignated with an "E" number. Termi-nals on dilferent boards which are con-nected together have the same E desig-nailon.

Each pole of a multi-pole rotary switchis designated with a letter, beginningwith "A" closest to the front panel.Wafers with two poles have the left-handpole (as viewed from the front panel)designated with the earlier letter in thealphabet; the letter for the right-handpole is next in the alphabet. Anows onswitch wipers indicate clockwise rotation

R35roox

- r r l

+ 1 5 + 1 5

J 2

t t +lOrrF

Rr5tox

c6

Il .{

cs2,, O

a

39A U D I O / J U L Y 1 9 8 1

PART ONEas seen from the front panel. Tuninoresistors are designated by lhe leiler o-fthe switch pole on which they are mounfed (each resistor is mounted betweenadjacent switch positions on a qivenpole). Tuning capacitors ure ,o*tedDetween two adjacent wafers on therange switch, and they are desiqnatedby the letters corresponding to the twoswitch poles to which they connecl.

Because of the large amount of wide_band garn packaged in this unit, soecificapproaches have been taken in lerms ofgrounotng and shielding strategy toavord oscil lations and interference pickup Much of this rs shown on theschematics to permit easy duplicationIne V symbol, one o1 lhree difierentgroundtng symbols on the schematics.denoles a single-point chassis ground located at the input jack. All leads with thissymbol are connected directly to this sin-gle point. The rfr symbol denotes asecondary single-point signal ground onCP2 (Figs 12 and 13, par t i l ) Thisground is connected to the single pointchassis ground via the shield of the inoutlead to CP2 lhe 4 symbol denoiesordinary mul t i -point 'c i rcut t grounding.Each circuit boar<l has its orcJtnarv ctrcuitground connected to the single_pointchassis ground, anci positive and neqa-tive supply voltaqe ls also drstributed ona single-point basis from the location ofthe single-point chassis ground

Returning to the circuit discussionthe oscil lator proper in Fig. 9 consists oflC1 (rnverter), lC2 and lC4 (integrators)Frequency-setting resistors R(F) ancj R(Etare mounted on frequency switch 53.ryhilg lhe corresponclinq capacrlorsC(KM) and C(t N) are mounreo on ranqeswitch S I Notice that the circuit hisheen slightly rearrangeci from thal o1 FigB, with the inverter "up front ' anO wiihthe amplitude control feedbaok ilC3) en_compassing both the inverter and theIntegrator. This anangement permits fre_quency changes to be made bv chano_ing the swilch-selected resistors withoitaltering other operating characteristics ofthe oscil lator. The analog multiplier foramplitude control is realized by op-amplC3 and FET Qi . Notice also that eachrange has its own frequency trimmer (R1to R4) to allow close matchinq of the os_cil lator frequencv to lhal of the analvzernotch fi l ter without requiring precisionIrequency-setting capacitors.

The output of the oscil lator is taken

from lC4, which would correspond to thelow-pass output of the fi l ter in Fig. B.This provides lower distortion and ptintsto a significant advantage of the state-variable approach [2.] Given low_distor_tron amplif iers, the domrnant distortion inan audio oscil lator results from nonlin_earity in the amplitude control element(here the multiplier) and ripple in the' d c control signal as a result of thedetection process in the amplitude con_trol circuit. In either case, this distortionis injected at the inverter (lC1) in this de_sign The two integrators between thrspornt and the output provide a j2 dB/oclave roli-off to these iniected harmon_ics, lhus af tord ing a 4- lo- l reducl ion ra_tio in second harmonic ancJ a 9_to_.1 re_duction in the more significant third harmonic. The inlegrators also tend to fi l terout norse, permitting the amplitude con_trol multiplier to operate at a lower siqnallevel, thus reducing its dislortion fromthe start. These same advantages arealso used in the analyzer notch fi l ter.

As mentioned above, the amplitudecontrol circuitry can be a maior source ofdistonion in an audio oscil lator. lt istherefore important that ripple in the con_trol signal be minimized by providing ad_equate fi l tering However, it must also bereallzed that what we are dealinq withhere is a feedback control svstem. andslabil ity must be considered Becauseheavy fi l tering of the control signal intro_duces delay in the feedback loop whichwill decrease stahrilrty we have conflict_ing requirements A further strain on sta-bil i ty is generated by the need to havevery high d.c. garn in the conlrol circuitto minimize output amplitude errors andthus provide a very flat frequencv re_sponse. The control circuit used in thisoesrgn deals with these problems.

The detector portion of the amolitudeconlrol circuit includes lC5 and e2 toQ4. Balanced oscil lator output signalsare provided hy lC4 and inverter lCS tothe full-wave rectif ier composed oftransistors Q3 and e4. Transistor e2provrdes a forward bias equal to onebase-emitter drop to e3 and e4 so as tominimize detection enors resultino fromtheir turn-on voltage requiremeni. Theuse of a full-wave rectifier greatly re-duces the magnilude of the ripple in thecontrol signal. The trimmer potentiome_ter in the feedback path of lCS permitsad1ustment for perfect balance of the twosignals feeding e3 and e4. This mini_

mizes distortion by minimizing ripple.The rectif ied signal is f i l tered bv the

capacitor connected to E7 whose valueis selected by the range switch (S1) tooptimize the conlrol dynamics for eachrrequency range. The fi l tered d.c. is thenapplied to differential amplif ier lC6where it is compared to a supply-derivedd.c. reference voltage to produce an er_ror voltage. The control circuit estab-lishes the desired operating level ofabout 1 .6 V rms at lC4's output

From lC6 the error signal proceeds toFET driver amplif ier ICZ through a non_trnear network. Under quiescent conrJi_tions the error voltage is small, and thegatn in this path is relatively low so thatripple transmifled to the tFT qate wil l beminimized Under large-error condil ions,such as just alter a range change, thegain in this path becomes large to speedup amplitude stabil ization. This is ac_complished by diodes D2 and D3, whichbypass R16 under these condi t ions.

For a c. error signal components, lC7operates as a simple inverting stagewhose gain is set by R15 This assuresstabil i ly of the feedback loop formed bvthe a g c. circuit. However, for d.c erroisignals capacitor C6 provides for an extremely large gain in lC7. In essence.lC/ looks l ike an integrator at low fre_quencies. The output from this rnteqratorwil l conlrnue to change and adjust thegate voltage of Q1 unti l the error voltaaelrom lC6 goes lo zero. Following

" r"ng"

change, it takes the inteqrator about f iveseconds to drive the oscll lator amplitudeto within 0.1 dB of its f inal value. DiocieD1 prevents a large forward bias lromever being applied to e1 .

The amplitude control multiolier is im_plenrented by tC3 and FET O j The FETacts as a variable resistance, with the re_sistance controlled by the d.c. voltaqeapplied 10 i ls gate by lC7 As the oatevoltage varies from zero to a neoitiu"value equal to its pinch-otj voltaqe(typically 5 to 1 0 V). its resislance froorall to source varies from about 25ohms to infinity. lC3 is connecred as adifferentral amplif ier whose positive andnegative inputs each receive a portion ofthe signal applied by lC2 Such a circuitcan provtde inverting or noninvertinogain depending upon the relative bailance between the circuits associatedwith the inverting and noninvertino in_puts of the op-amp. FET e j controlJthisbalance. When the FET resistance is

AUD|O/JULY 1981

PAR| ONE

i lI

I|l

low, the noninverting gain exceeds theinverting gain so that a noninvertingoverall characteristic results, When theFET resistance is high, the opposite ac-tion results. At some intermediate FETresistance, the circuit is balanced andthe gain is zero.

FETs are not perfectly linear resistors;

their drain{o-source resistance tends lobe a function of the drain{o-source volt-age. Such a nonlinearity can create dis-tortion, making this a major concern.One way of minimizing this distortion isto operate the FET at the lowest possiblesignal level across it that is consistentwith acceptable noise performance. The

excellent low-noise performance of theassociated 5534 op-amp makes it possi-ble here to operate at a level of onlvabout 40 mV rms across Q1 This is es-tablished by the voltage divider whichapplies a fixed level of about 40 mV rmsto the noninverting input o{ lC3. Becauseof the virtual short existing between theinput terminals of an op-amp operatingwith negatrve feedback, this voltage alsomust appear at the inverting terminal oflC3, and lhus across O1

A second way o{ reducing FET distor-tion involves a well-known techniquewherein half of the drain-to-source sionalis fed back to the gate of the FET. ihrsprovides a first-order correction {or thenonlinearity, practically eliminating sec-ond harmonic distortion and leaving onlya small amount of third and higher orderharmonic distortion. This feedback isprovided by the 750-kilohm resistor(R13) connected to the gate

The remainder of the signal sourceconsists of the level control (R30) theoutput amplif ier (lC8), and the output at-tenuator. The level control provides con-trnuous adjustment ol the output ampli-tude, while the output attenuator pro-vrdes maximum output ranges of 5 V,1 5 V, 500 mV, and 150 mV. An of fposition is also provided which effective-ly kil ls oscil lations, and it should be usedwhen a different signal source is beingemployed so that crosstalk from the os-cil lator into the analyzer cannol affect re-sults. The attenuator establishes a con-stant output impedance of 600 ohms. Afixed .1 .6-V rms "sync" output is alsoprovided by the signal source, which isuseful as the trigger source for an oscil-loscope used to visually monitor the dis-tortion signal output of the analyzer.

Having drscussed the overall analyzerrn general and the signal source portionin detail, we now conclude part l. Nextmonth, in Part l l , I wil l describe the re-maining circuits in detail. Construction.details and adjustment procedures wil lbe covered in Part l l l . A

Referencesl. Hewlett-Packard, Hp 3394 Distortion AnalyzerUser's N4anual.2. Holet, B. E., "A Comparison ol Low FrequencyFC Osci l lator Topologies," prepr int +1526, 64thConvention of the Audio Engineering Society, NewYork City, November 1979.3. Gef{e, P. R., "How to Buitd High-Ouatity FittersOut of Low-Ouality Parts," Electronics. November1 1 , 1 9 7 6 .

The new MXR Stereo Octave Equalizer lets you remix Vour music so thatyou can bring up that earth shatterinq bass line, screaminq quitar solo orfly part of the performance loud enough for everybody to-h6ar withoutboosting nor.se.Annoying problems like hiss, turntable rumble and other kinds of distortionaro easily overcome with the Stereo Octave Eq. lt can accommodate theextended dynamic range of the new high performance discs and be aninvaluable tool when recording car stereo cassettes.Featuring professional specs and the hiqhest qualitv componentshand-assembled in Rochester, NY, USA, the MXR Stereo'Octave Eoadds clean, noise-free power to specific parts of your music whilemaximizing your system's response. Check one out at Vour local MXRdealer and get to know your neighbors better.

gttss fE a r r u r t , 4 t | f i l t l '

you tho kind olan integral part of the contemporary horne musickind ol focussed music oower that ivill brino vourmusic power that ivill bring your

netghborb ovei for a late night get together.

tfXR lnnovstions, Inc.TrlO Driving Park Avenue, Rochester. New York 14613 (716) 254-ZglO

ConeumerPro&rcb Group

42 A U D I O / J U L Y 1 9 8 1

BUILDAHIGH PER

ast month we discussed the theo-ry of operation of total harmonicdrstortion (THD) analyzers, the

Input Circuit and Bandpass FilterThe input attenuator, input amplif ier.

and vollage-controlled state-variablebandpass f i l ter are shown in Fiq. 12.

The rnput ailenualor (S+) Js|..t, ,.appropriate input sensitivity for maxi_mum operating levels of j 00 V 30 V

features of the analyzer being describedhere, many of the under ly ing pr inc ip les,and finally the details of the signalsOUrce portion of the analyzer. Thismonth we wil l resume the circuit descrio_tion by starting wlth the input crrcuits andthe state-variable bandpass fi l ter. Beforewe go into the remainder of the analvzer.lel me present the printed wirinq boardlayout lor the s ignal source, Cpl , as Frq1 0 and lhe corresponding componentplacement as Fig. 11 | also repeat Fiq5, which is the block diagram of the ani-lyzer, for clarity.

PARTTY/OROBERT R CORDELL

Inpul impedance of the analyzer is 1 00kilohms on all input ranges.

Because of the high impedances inthe attenuator circuit, 54 and its associ_ated components must be placed in asmall shielded enclosure in order toavord stray pickup and oscil lations. De_penotng on stray switch capacitances.small capacltors, such as those showndotted in Fig 12, may have to be addedto obtain a flat frequency response at allattenuator settings.

The heart of the signal processing inthe analyzer is the voltage-controiledstate-variable bandpass fi l ter. lts orimarvfuncl ion ls to del iver a fundamental srq_nal to the d i f ferent ia l ampr i l ier (n6tshown in f g . 121 whose ampl i tude andphase are such that the fundamental wil lbe exactly cancelled, leaving only distor-tion products. In addition, very l it i le ofthe signal's distortion products shouldbe passed by the bandpass fi l ter so thatl itt le or no cancellation or attenuation ofdistortion products wil l occur at the dif_

n at t d

V E

ncccbaf rllf res retal

160.oo3

'*# cp

, ,.1ef. ,

{;

o"ln* lirltstltsN

$. lEvfr

ar

1 0 V , 3 V , o r 1 V . T h e " n o m i n a l , , o p e r -ating levels are a third of these values.Because of the auto-set level feature, ac-tual input levels which are above or be_low the nominal level by as much as 10dB can be accommodated; no inout ver-nier conlrol is required. Notice that onlfp 1-V range a gain-otthree amptif ier(lC9) is switched into the circuit to boosta nominal 0 33-V signal to the nominal1 -V internal analyzer operating level. The

ferential amplif ier. As mentioned earlier,in order to achieve exact fundamentalnull ing, the gain and center frequencv ofthe bandpass fi l ter are preciselv con_trolled by the auto-tune circuits (Frg 5)

The state-variabte fi l ter in Fig. i2 is abr1 more complicated than the simoleone il lustrated last issue in Fig. g, but iheoperating principle is exacfly the same. ltconsis ts of lCs 10 to 13 and 15. Theuse of two extra inverters in the loop

AUD

I

14

PART TV/O

i lC l 1 , 13) accounls for the addi t ionallOs The extra inverter at lC1 1 permitsre bandwidth-setting feedback pathiR57 and R62) to be connecled in sucha way that center frequency changes donot affect f i l ter Q or garn This lype ofconneclion was also used in the oscil la-' lor. but there the extra inverter was notneeded because of the sign l lexibil i ty af-forded by the mul t ip l ier .

The second inverter (lC1 3) is requiredto re-establish the correct polarity tor themarn leedback loop (R58 and R59) andalso provides a usefu l summing point forthe amplitude and lrequency control sig-

outpul. The center frequency and gain ofthe f i l ter are t r immed by R59 and R62respectrvely.

Control of the center frequency of thebandpass fi l ter is quite simple. Recallfrom our earlier discussion that the cen-ter frequency of a state-variable band-pass fi l ter can be changed by simply ad-just ing the loop gain. In F ig. 12 th is isaccompl ished by connect ing a mul t ip l ierand series resistor (R65) around the in-verter formed by lC13 to provide con-trolled amounts of negative or positivefeedback. When the multiplier provides anoninverted characteristic. additional

|C14. and i t is ident ica l to the one usedin the oscil lator (Fig. 9 Part l). As in theoscil lator, any distortion introduced bythe control circuit must pass through twointegrators before reaching the output(E18), and is thus substant ia l ly reduced.

Amplitude control of the bandpassfi lter is a bit more subtle. Based on lastmonth's discussion of the simole state-var iable f i l ter , one's l i rs t inc l inat ion is tochange the gain of the fi l ter by changingthe amount of the "damping" feedback,here provided by R57 and R62. This wi l lsurely accomplish the desired control,but notice that in this case any distortion

Fig. 5 -THD analyzer block diagram

STATEVARIABLEOSCILLATOR

DIFFERENTIALAMPLI F I€R

STATUSINOICATORCIRCUIT

REFERENCEV C A

LEVELOETECTOR

nals from the multipliers. In lurther con-trast to the simple slate-variable fi l ter, theinput s ignal rs here appl ied to the nonin-ver t ing input of lC1 0. This y ie lds a s l ightnoise advantage and also provides thecorrect signal polarity at the output of thebandpass fi l ter. Because the bandpassfi lter has a gain of about 5 at the centerfrequency, some attenuation of the inputsignal is necessary (R54 and R55) to es-tablish the proper operating level at its

negative feedback is provided aroundthe inverter, making its gain less thanunity and consequently lowering thecenter frequency of the fi l ter. The oppo-site action results when the multiplierprovides an inverted characteristic. Inthis manner the gain of the inverter canbe controlled over a !3.2 percentrange, resulting in a center frequencycontrol range of t1 .6 percent.

The multiolier consists of FET Q5 and

injected by the control circuit wil l be re-moved from the fi l ter output by only oneintegrator. To minimize distortion, wewould l ike to inject the amplitude controlsignal at the same advantageous pointas the f requency contro l s ignal . i .e . , a tthe input of inver ler |C13, For tunate ly , i tturns out that a small amount of feed-back (positive or negalive) around thesecond integrator (lCl 5) is equally effec-tive in providing control of the gain as

rt cAUDIO/AUGUST 1981

PART T\A/O-

ol.Er{

Iol

c P la

!

I!

ttI

Fig. 10 - Printed wiring board layoutfor CP1 , the signal source.

c P l

Fig. 1 1 - Parts placement for CP1 , thesignal source.

IIlr

E:t"lL+ $t +

I ,c5

| TJ5

*'* Il+r

) ;r*ilrrll t':I

Il r c r l lr r l

@icr tRssl

@

Av

- ts

oI

g l 3 ! f " 1 7

l r iG l t 0 |o

I

c'o{ ji,.l r +p+

O

"+Htr

@I f fos foz lort t a++lR rT lR rG | | |r t t t l

I

f*'R t 9

l"i*,"

@

long as the perturbation is small, as it ishere. The amplitude control multiplier,consis t ing of FET 06 and lC16 thusprovides amplitude control feedbackfrom the output of lC15 to the input of 'inver ter |C13. An ampl i tude contro lrange of about 13.5 percent results.

Dlfferential and ProductAmplifiers

The differential amplif ier which com-

pletes the notch fi l ter and the distortionproduct ampl i f iers are shown in F ig. I3 .Also shown here is the auto-set level cir-cuit which wil l be discussed shortly.

lC1 7 functions as the dilferential am-plif ier where the fundamental suppliedby the bandpass fi l ter is subtracted fromthe input s ignal . The combinat ion of th isdifferential amplif ier and the bandpassfi lter thus comprises the notch fi l ter. Thedifferential amplif ier also provides a gain

ol 1 0 to the distortion products. The in-put s ignal (E15) is appl ied d i rect ly to thehigh-impedance nonrnverting input oflC1 7 whi le the fundamental f rom thebandpass fi l ter (E1 8) is applied to the in-verting input through R84. Because ofthe attenuation which results from thevoltage divider formed by R84 and R85,null ing resulls when the fundamentaltrom the bandpass fi l ter is about 1 0 per-cent larger than the fundamental sup-

F t o

v o :

A . ^u y c

I U :

percwne'S w t : :

con :t u r Si s a s

0 0 3fo l l c '( r c1 :then

AUDi (

PART TV/O

R4690.9K

R446 8 . r K

- J L . Jr t 2

R 4 8t o o K

R 5 0toox

II

q - l - R47' - r - l l xI

R 5 22 . O 5 K

6.8K ;1- t anP|-

c ( p R )

cza 22

R 5 6 l o K I ? I R 6 l l o K R 6 2 c ( o Q ) E t B. + " ^ - 5 K I

nc l_ - c?6 .??E16 r - | E r7

S 4 A

c2l2.OpF

/FOURF E R R I T LE E A D S

c 2 ?47

Fig. 12-THDanalyzervoltage-controlled bandpass f ilter

R 8 l3 3 K

E24

F U N D .TOf r u . t J

T OS 6F r G . l 8

T OF I G . 1 5

R 8 0I O X

R 8 2I K

R70750K

C 2 9 T O C 4 O - B Y P A S S N O T S H O W N

ALL CAPACITOR VALUES ARE pF UNLESS OTHERWISE NOTEo.

R 7 3l a 62 N 4 0 9 |

R 7 lt o K

E22FCONT

E23ACONT

controlled by the 1496. As this ratio isvaried from a very small number,through unity and on to a very largenumber, the gain changes proportion-ately from small to large. In fact, in thiscircuit the gain is numerically equal tothe ratio. A d,c. control voltage appliedto the 1496 contro ls th is rat io and thusthe gain.

Beferring to Fig. 14, bias resistor R3sets up a bias current in Q7 and Q8 ofabout 1 mA. Transistors 05 and QG areconnecled as common-base stages andprovide a low-impedance point at theiremitters where a,c. srgnal currents can

17

r

plied directly from the input attenuator.The differential amplif ier is followed

by another amplif ier (lC1 8) whose gain is1 0 on all distortion ranges except the 1 0percent and 3 percent full-scale ranges,where it is unity. The gain of this stage isswitched by a FET whose gate voltage iscontrolled by the sensitivity switch (S5).lC18 is followed by an attenuator whichis also controlled by 35, establishing full-scale sensi t iv i t ies of 10, 3, 1 , 0 .3 0.1 ,0.03, 0.01 , and 0.003 percent . This isfollowed by a fixed gain-of-ten amplif ier(1C19). The distortion product signalthen oroceeds to the auto-set level volt

age-contro l led ampl i f ier ( lCs 20 & 21)

Voltage-Controlled Amplif iersBecause the voltage-controlled ampli-

f iers (VCA) are central to the operation ofthe auto-set level circuitry, we wil l dis-cuss their operation before proceedrngfurther.

A s imple VCA is shown rn Fig 14 l tconsists of a 1496{ype balanced modu-lator lC (whose internal circuit is shown)ano an op-amp.

The op-amp is connected as an in-verting feedback amplif ier, the ratio ofwhose input and feedback signals is

AUDIO/AUGUST 1981

PART TY/OI

+ t 5Rr00r 5 K

Rt02

F R O M E I 5LEVE L

sw rTcHF U N O . E l 8

R 8 4 3z . z K 2

R 8 6t o o x

2 N 4 0 9 |

R t 2 3t o o K

K

c44r.opF_l

R l 0 6l 5 K

R t04470

R 9 46 . 2 x R 9 8

roo

R 9 7t 5 K

FROMB P F

c46toyr

DIST.E 2 9

s 5 B

7 . 5 K

R 9 l2 . 1 5 K

c43tOltF

R 9 9r o o

R t 0 74.7 KR 9 5

t . t K l 0 0 K E 3 0

l--R 9 27 5 0

R tog4 7 K

R i l O4 . 7 K +

R t o 84 7 K E 3 l

c47topF

R i l ll 5 K

ALL CAPACITOR VALUES ARE oF UNLESS OTHERWISE NOTED

c o N T R O L F 2VoLTAGE lcio-

- 7 . 5 V

c llOtF

R 3t 5 x

2 _

Frg. 13 -THD analyzerproduct amplifierand level set.

F ig. 14 -A simplevoltage-controlledamplifier (VCA).

q

Rt?2

+ \3 t 8T a t q

! , )rc 23+ - /

be added to the d.c. bias currents. Theinput signai current established by R4 isapplied to the emitter of 05, while thefeedback current established by R5 isapplied to the emitter of Q6. The actionin the upper "quad" of t ransis tors (Q1-04) determines what portion of each ofthese a.c. signals reaches the output atp in 1 2 and thus the inver t ing input of theooerational amolif ier.

We see that at the collector of Q5 wehave d.c. b ias current p lus a.c. input s ig-nal, while at the collector oi Q6 we haved.c. b ias current p lus a.c. feedback s ig-nal. Each of these currents is aoolied tothe emitters of a differential pair, wheresome will l low in one emitter and the re-mainder wil l f low in the other emitter of a

1B

rruel l:

Qt'E I

_aS I

a 'ae:

: 't :a :

t .tf r

AU

PART T\I/O

r36roK Rt 58

22K

Rt42rooK

Rt29IOO 16

R t 5 l4 .7X

E234 ACONT

Rt30roo 4

Rt 39Z?K

R 1 3 7t . 5 K

R t 6 |22K

R r59r . 5K

R r 4 06 . 8 K

- t 5

R t 49I K

l t 4-15

c 5 3IOOpF 1Fr

C 5 7 = R 1 5 2l oo 4 .7 R

Fig. 15 -THD analyzeramplitude andf requency detectors

OREFE ? I 822c 5 6

loog.FR t 4 8roo

Rt626 .8 K

R t 5 3t o o

F C O N T

c60 To c7 r - BYPASS NOT SHOwN

ALL CAPACITOR VALUES ARE pF UNLESS OTHERIVISE NOTED

R22K

r-IIIt o

R 1 6 6 C 5 968 toopF

given pair. Each of these currenls thussplits in some proportron. The key oper-at ing pr inc ip le in th is c i rcu i t is that thea.c. currents into the emitters of a differ-ential pair wil l split in the same propor-t ion as the d.c. b ias currents f lowing in tothe pai r .

The attenuated control voltage ap-plied to pin 1 0 controls this split. Whenthe control voltage is zero, both transis-tors in each differential pair have thesame base voltage and conduct equally.In this case, half of the input signal andhalf of the feedback signal current reachthe oulout at oin 1 2 vra Q2 and Q4 re-spectively. Because of the extremelyhigh gain of A1 , the signal and feedbackcurrents at pin 1 2 must essentially can-

R 3roK-- vo

A U D I O / A U G U S T 1 9 8 1 19

f lJ

PART TV/O

F R O McPz

s5c

Rt67roo

Rt69 3

390 za a

(L-+.V E33

R t 7 92: lK

otsT.

Rt 80?KQ 8

2N4091

Rr7536 l ooc(BD) c(DF)Rt72

34 rOO

ALL cAPAc lToR VALUES ARE pF UNLESS OTHERWISE NoTED. F tg . 1 B -

Filter and metercircuits.

DIST.O U T

R(B) R(A) R t73roo Rt76 R r77

tK 9.O9K

R207rooE38 R 208

roK

E29D IST.

R t 8 3I K

R t 7 8t o K

R l 9 43 9 K

r c30L M 3 2 4

R204f .5K 847

R 2 0 02.7K E43

o?4

MF C O N T

R r95IOK

R t97toK

R r99t o K

R?O2T 2 . 7 K E 4 5

R t 9 83 3 K

f,I

+=so" J t

L""

+ = 4 5 "

cel each other for f inite output levelsf rom A 1 . They must therefore be virtuallyequal. This in turn indicates that the sig-nal currents in R4 and R5 must be equalfor a zero control vollage, implying unityga in .

A positive control voltage at pin 10will cause Q1 and Q4 to conduct moreheavily than Q2 and Q3, Thus, a greaterorooortion of feedback current reachespin 1 2 (v ia 04) than input current (v iaQ2) In order to maintain cancellation ofthe two currents at p in 12, the input s ig-nal must be larger to begin with than thefeedback current, implying less than uni-ty gain. As an example, it is well knownthat a potential of about 60 mV across

F ig . 17 -Phasecomparatoroperation.

R t 8 9?ox

R r 8 8t o K

\' ?:'? '1. {T

e.e {l'

20

PART T\\(]

4.3K SOKHz30KHr

E29 tOO

L P F

6 . 8 K

Fig . 19 -

Optional fixedproduct filters.

cs4ro0

K-'^\';+^t)> r c 3 9I C 4 0

Fig. 20 - Regulatedpower supply.

R ? r l

+ coo- 470- P f

the bases of a differential oair wil l causea 1 O-to-1 ratio of conduction in the twotransistors. In this application, a +60mV control voltage at pin 10 wil l thusresul t in a gain ot about 0.1. The s i tua-tion is the inverse lor a negative controlvoltage; a -60 mV level at pin 10 wil lresult in a gain of about '1 0

A particularly nice feature of this ar-rangement is that the gain in decibels isa l inear function of the control voltage.Here we get a variation of 20 dB/60 mVor about 0.33 dB per mV. The 1 00{o-1attenuator formed by R1 and R2 simplyallows for more manageable control vol-tages (3.3 dB per volt). The gain controlrange here is in excess oi 120 dB.

Auto-Set Level GircuitWith a knowledge of the operation of

the VCA, the auto-set level circuit in Fig..1 3 becomes easy lo understand. lt con-sists of two identical VCAs, each withthe same control voltage and thus eachwith the same gain. The first VCA (1C2021) controls the gain in the distortionproduct signal path and receives its in-put from |C19. lt produces the level-ad-lusted distortion product output at E29for use on CP3. The second ("refer-ence") VCA (1C22, 23) contro ls the gainin an automatic gain control (a S c )loop; its input is the fi l tered fundamentaloutput from the bandpass fi l ter (E1 8)

In addition to the reference VCA, thea.g.c. loop includes a level detector (D5)and an op-amp (1C24) connected as anintegrator. The d.c. output of the integra-tor is the control voltage for the VCAs,The a.g.c. circuit is ananged so that theintegrator wil l always adjust the gain ofthe reference VCA so that the level of thefundamental at i ts output is 1.1 V rms,Thus. if the inout to the notch fi l ter is thenominal 1-V level, the fundamental fromthe bandpass f i l ter wi l l be 1.1 V and thegain of both VCAs wil l be unity, as it

should be for th is s i tuat ion. 11 lhe inputsignal level were to increase to 2 V, thegain of both VCAs would drop to 0.5,implementing the proper gain correctionin the distortion product signal path.

Auto-Tune Control CircuitsThe amplitude and frequency detec-

tors which comprise the auto-tune con-t ro l c i rcu i t are shown in F ig. 15. l f thegain and center frequency of the band-pass fi l ter are not perfect, the notch fi l terwil l not produce a complete null of thefundamental signal. Some lundamentalwil l thus appear in the distortion-productsignal path. An amplitude error wil l pro-duce a "left-over" fundamental comoo-nent whose phase is zero (or 1 80) de-grees relalive to the input signal A fun-damental component with this phase re-lationship is said to be a ' 'normal' '

component. A frequency error wil l pro-duce a so-called "quadrature" funda-mental component whose phase is ei-ther leading or lagging 90 degrees rela-tive to the input signal.

The amplitude detector lunctions bylooking lor normal lundamental compo-nents in the distortion srgnal and adJUSfing the bandpass fi l ter gain up or downdepending on the phase relationship (0or 1 80 degrees). lt ignores quadraturefundamental components. Similarly, thefrequency detector lunctions by lookingfor quadrature fundamental componentsand adjusting the l i l ter center frequencyup or down depending on the phase re-lationship (lagging or leading). lt ignoresnormal fundamental components.

The special detector circuit whichpossesses the properties mentionedabove is called a "phase detector" be-cause it is sensitive to the phase of thesignal being detected. An ordinary enve-lope detector or rectif ier is not suitablebecause it wil l detect the signal regard-less of its phase. Here, the phase of the

"left-over" fundamental is a crucialpiece of inlormation.

A simplif ied schematic of a phase de-teclor is shown in Fig 16 Like the VCA,it is based on the 1 496{ype balancedmodulator lC. The phase detector is amore conventional use of this lC. Biasresistor R2 sets up an appropriate cur-rent l low in current sources Q7 and QB.The application ol a positive input signalat the Y input wil l cause Q5 to conductmore current than 06; the opposite wil loccur lor a negative input signal. A posi-t ive input at the X input wi l l cause 01and 04 to be "on" and Q2 and 03 tobe "ofl." The output (V,,) is taken dif-ferentially between pins 6 and 1 2

We can easily see what polarity ofoutput wil l be produced for any combi-nation of positive or negative X and Yinputs. For example, i l X and Y are bothposil ive, most of the current f low is in 05and 01 , pul l ing p in 6 lower than p in 12and producing a positive output. In gen-eral, a positive output is produced whenboth inputs have a l ike sign, and a nega-live output is produced when the X andY inputs have differing signs. In a sense,thrs circuit performs similarly to the "EX-CLUSIVE NOR" logic funct ion. We im-mediately see that two perfectly in-phasesignals at X and Y wil l always have thesame sign and thus produce a maximumpositive output.

The d iagram in F ig. 17 lends fur therinsight into the phase sensitivity of thistype of detector. Square-wave inputs areshown for simplicity ol i l lustration, butthe circuit functions similarly for otherwaveforms. In the top i l lustration we seewhat happens when the X and Y inputsare 90 degrees out of phase; the aver-age d.c. output at V" (which is what mat-ters) is zero, This i l lustrates how the am-plitude detector can ignore a quadraturecomponenl.

It is easy to see that if we slide the Y

AUDIO/AUGUST 1981 21

PART TV/O

!/t2

2<il*rd

Fig. 21 - Printed circuit board CP2 forthe input circuit, bandpass fllter, productamplifiers, and auto-set level circuits

input back and forth along the time axis,an asymmetry wil l be produced in the V"waveform, corresponding to an averaged.c. va lue,

The bottom il lustration depicts a 45-degree relationship between the X and Yinputs. By trigonometry we can arguethat the Y signal consists of equal por-tions oJ normal and quadrature compo-nents in this case. This would be reore-sentative of a situation where there wereboth amplitude and frequency errors si-multaneously. We see that in this case apositive average d,c. output is producedat V". lt is not, of course, quite as strong-ly positive as when X and Y are perfectlyrn Dnase.

We can see that if we apply an "un-known" signal to the Y input and a "ref-erence" signal to the X input, the detec-tor wil l respond to components in the un-known signal whose phase is similar to(or the inverse of) that of the reference.Components in quadrature with the reterence wi l l be ignored. In addi t ion, theeffects of noise and signals at other fre-quencies wil l average out to zero.

Return ing to F ig. 15, the ampt i tudedetector consists ot lCs 25-27 . The dis-tortion product signal ("unknown") istaken f rom the output of lC31 (Fig. 18)and passes through lC25 where it sees asmall-signal gain of f ive and is soffl imit-ed beginning at t0 .7 V swings by the

feedback diodes. The signal is then ap-plied to the Y input of lC26 where it isphase-compared with a "normal" funda-mental provided by the output of the ref-erence VCA (|C23)

The diflerential output of the phasedetector is f i l tered and applied to lC27Awhere it is converted to a single-endedsignal. This signal drives the integrator(lC27B\ to produce the appropriate gatevoltage for the amplitude control FET(Q6). The integrator wil l continue to ad-lust the gate voltage unti l the output oflC27A is driven to zero. Notice thatwhen the error from lC27A rs greaterthan 10 ,7 V , D14 and D15 conduc tand speed up the integrator to achieve

22 A I

PART T\A/O

Rr22 ,.-\q9I nez

| ( o z )r \__/

@

- t 5

I r \

l*l-'i I f'*l l l * "1 r l rn rL+ f iifel

R64 ._\

t?t R69 R72 R7?

l . r -K9 I i l - " , "? T ? ? ? T - ' -l c z t | | |

l '' ' r 'I

a .

; l kc47

-rc25

I

fR6l

Rt6 R60c 2 4 l i I

+ttit t l

R57

fT+

I-Ti

oG N D

tT'o

e+ t 5

e

1.utI

l n q t

\-'--i l l

, \ , lI J R2 r3 l

lbf**-t t?u

l =i c r z I I

1t \[ $nezFAIl-.--.ner

q9

| \4',

fneet lnsif

1lt l

ll'i"l'l'llR53 I

in lnlr-- tl r c s l f l r c r o l f l r c r r l f l r c r z lI l . t , l l rL . l l * l , l I

++R78 R74 (D

R65

ItI

Figfor

22 - Component placementCP2.

The fi l ter consists of a pair ol resistors(on S3), a pair of capacitors (on S1) andan op-amp (1C32) connected as a volt-age lollower [3]. This fi l ter is followed bythe second-order high-pass fi l ter (1C33)which is similarly realrzed. lts l i l ter Q ischosen to provide a slight gain bump atthe second harmonic frequency to partlyoffset the small loss in the notch fi l ter atthis lrequency, For reasons of high-fre-quency fi l ter stabil ity, the high-pass fi l terhas the same set of cutoff frequencieson the 200-kHz range as on the 20-k1zrange.

As mentioned in Part l, the trackingproduct f i l ters add a considerable por-tion of the switching complexity and ex-

faster tuning following a transient.Operation of the frequency detector

(lOs 28-30) is identical to that of the am-plitude detector except that it is suppliedwith a quadrature fundamental reference(lagging 90 degrees) instead of a normalfundamental relerence. The quadraturerelerence is supplied from lhe low-passoutput (|C15) o1 the state-variable fi l terand passed through amplif ier-i lmiterlC28 before application to the X input o{rc29.

Filter and Meter CircuitsThe fi l tering, metering, and status in-

dicator circuits are shown in Fig. 1 8. Thedistortion product signal from the auto-

set level VCA on CP2 (E29) is f irst ampli-fred by a factor of one or 10 in lC31 tokeep it at a reasonable level for use bythe auto{une circuits {or all sensitivitysettings of 55. Switch SSC sets the gainof th is s tage to 10 on the 0.03- through1 0-percent sensitivity ranges to compen-sate for attenuation introduced by theS5A attenuator on these ranges. Notethat S5C does not affect gain in the dis-tortion metering path.

The distortion product signal fromE29 is also applied to the low-pass prod-uct l i l ter. As mentioned earlter, the sec-ond-order Bessel low-pass product f i l teris set lor a 3-dB cutoff Jrequency equalto 10 times the fundamental frequency.

AUDIO/AUGUST 1 981

PART TV/O

frr! F

o{r- 7c3 1

I..1$

t , O

t,o

6

3

!

I

! !

t\;f,

Fig. 23 - Printed circuit board CP3 forthe auto tune control, product filter,meter and slalus lndicator circuits.

I

ator. Note that this provides a nar-rowband measurement of the input sig-nal level .

The distortion product signal is thenapplied to the meter circuit, consisting oflC36 and lC37 lC36 is connected as aunity-gain inverting feedback amplif ierwhich has two feedback paths, one forpositive output-signal excursions (D23)and one for negative output excursions(D22) Positive and negative haltwaverectif ied signals lhus appear at the twofeedback resistors. The negative half-wave rectif ied signal and half the unrecti-f ied input signal are added at the input oflC37 to produce a full-wave rectif ied re-sult. This amplif ier also provides appro-

pense (about $25.00) to the analyzer .Many professional analyzers provideonly a switchable 40O-Hz high-pass l i l terand a low-pass fi l ter which can be set to30 or 80 kHz or switched out, This sim-pler f i l tering approach can be imple-mented as shown in F ig, 19 wi th a SPSTswitch for the high-pass and a center-oflDPDT switch for the low-pass. Use ofthis simpler l i l tering wil l typically result inan increase of the analyzer's residual by0 to 3 dB and wil l make the readingsomewhat more susceptible to hum andnoise in the UUT.

Following the f i l ters, the distortionproduct signal is brought to a 100-mVfull-scale level by the switched-gain am-

oliJier comoosed of FET Q9 and 1C34.The total gain of this combination is ei-ther 0.33 or 3.3 dependrng on the set-t ing of S5D. Tr immer R180 provides cal -ibration for the distortion measurement.

Provision is also made to monitor theinput level with the meter crrcuits Thismakes it possible to make a completeTHD measurement on a piece o1 equip-ment without any other test instruments.Selection oi distortion or level as thequantity to be measured is accom-pl ished by 36, In the "Level" posi t ion,an attenuated output from the bandpassfi lter is applied to the meter circuits. Thefull-scale 'Level ' range then corre-sponds to the setting of the input atlenu-

AI

PART T\A/O

c P 3

Fig. 24 - Component placemenlfor CP3.

priate low-pass fi l tering of the result. Thesingle-ended output of lC37 rs then usedto drive the meter movement.

The status indicator circuits indicatewhether the incoming level is too high ortoo low for proper analyzer operation,and whether the incoming frequency istoo high or too low for the auto tune cir-cuits. A quad op-amp comparator (1C38)drives the four indicator LEDs,

The frequency indicator circult moni-tors the gate voltage on the frequencycontrol FET (Q5) to see if the frequencyis within the tuning range. lf the gate volt-age is zero or positive, lhe frequency istoo low, and the "Low" LED is l i t. l f thevoltage is more negative than -12 V, the

FET is pinched ofl and the frequency istoo h igh.

The input level indicator crrcuit moni-tors the a.g.c. control voltage in theauto-set level circuit. 11 this voltage goesmore negative than -3 5 V indicatingthat the input level is more than 11.5 dBabove the nominal 1-V internal operatinglevel , the "High" LED wi l l be l i t . s imi lar-ly, a level more than 1 1 5 dB below thenominal operating level wil l l ight the"Low" LED.

Power SupplyThe regulated tl 5 V power supply

for the analyzer is shown in Fig. 20. Thecircuit employs a lull-wave bridge recti-

f ier and standard three-lermrnal regulatorlCs

The input circuit, bandpass fi l ter,product amplif iers and auto-set level cir-cuits are realized on printed circuit boardCP2 lts layout is shown in Fig 21 andthe component placement diagram isshown in F ig. 22. The auto tune contro l ,product f i l ter, meter and status indicatorcircuits are realized on CP3. lts layout isshown in F ig 23 whi le componentplacement is i l lustrated in Fig. 24.

This completes the description of allof the circuitry in the THD analyzer. Nextmonth we'l l conclude with constructiondetails, the adjustment procedure, trou-bleshooting, and performance data. A

A U D I O / A U G U S T 1 9 8 1 25

BUILDAHIGH PER-

PART THREEROBERT R. CORDELL

In Parts I and ll the theory and opera-tion of all of the THD analyzer circuitswere covered. We will now oroceed withconstruction, adjustment, and troub-leshooting. lt goes without saying that athorough understanding of Parts I and llwil l help immensely in troubleshooting.

Circult Board AssemblyAssembly oJ the three printed wiring

boards should be quite straightforward ifthe component placement diagrams arefollowed carefully. Load all of the lCsockets first (socketing is slrong/y rec-ommended); this wil l make it easier toidentify locations of other components.Note that pin #1 of all lCs faces in thesame direction except lor lC3B. Nextload the off-board conneclion terminals("E" connections denoted by large do-nut lands on p.c. cards), including thosefor the power-supply leads. Any suitableterminals may be used here, but VectorType T-l8 terminals are a good choice.The boards come dril led with 0.042-inchholes at these locations for maximum

resistors and the resistance code on pre-cision resistors; many different codes ex-ist for the latter. lf you have any doubt,use an ohmmeter, Note that the twodiode st r ings on CP2 (D6-8 9- l1) areeach prewired and treated as a singlecomponent. Prior to install ing FETs Q1 ,05, and Q6, it is a good idea to measureand record their pinch-off voltage (Vo)with the simple circuit oI Fig. 24. Knowl-edge ot the pinch-otf voltage wil l permitthe best adjustments to be made. Use of27-gauge, insulated, solid wire is OK forthe long insulated jumpers. Note thateach of these lumpers connects twopoints on the board labelled with thesame letter of the alohabet

Complete circuit board assembly byinstall ing all lCs in their sockets, takingcare to avoid bent pins, which cansometimes be hard to detect. Again,note that pin #1 of all lCs faces in thesame direction except for 1C38.

Clrcult Board Bench TestBecause this is a fairlv comolex con-

flexibil i ty, but wil l have to be redril led to1 /16- inch for the T- '1 8 terminals.

Now load all resistors, diodes, barejumpers, capacitors, transistors, and in-sulated jumpers - in that order. Takespecial care to watch polarity on diodesand polarized capacitors. An elongatedpad on the p.c. card denotes the nega-tive terminal of most capacitors and thecathode of all diodes. Also take soecialcare in reading the color code on carbon

struction project which also involvesconsiderable chassrs wiring, it is stronglyrecommended that most of the electrrcaltesting, troubleshooting and adjustmentof the three circuit boards be done oneat a time on the bench. This wil l virtuallyguarantee success upon final assembly.

Before proceeding further, give eachof the boards a thorough visual inspec-tion, checking for incorrect polarit ies orvalues, cold solder joints, solder bridges

I

o r (pla,j unbYr

bedoroleno\

Sist

S O :

caaftc'rerc

E 50 'g r .1

(otE 5

S I T .

a lv r c ,

" Bmein," ,OU:^ + -

derorh ^

co'- i d

meJ t !

sl'rprc

ne(

( l i kmaprc

sht?l' lt

no'inginedi t 'gr€thrh ^Ug

AUt

_:-

ART THREE

or other shorts, improper lC insertion orplacement, missing components orlumpers, etc, This step should not bebypassed.

A t15V regulated power supply wil lbe required for bench testing. lf youdon't have one, you may wish to assem-ble the analyzer's power supply modulenow and use it.

Signal SourcePrepare CP1 for bench testing by

solder ing 0.022-pF, 1 10o/o polyestercapacitors from terminal E2 to E3 andfrom E4 to E5. Connect 3 9-k i lohmresistors from E1 to E2 and from E3 toE4. Connect a 2.7-k i lohm resis tor f romE5 to E6. Strap E5 to E9. Connect a0.1-gF polyester capaci tor f rom E7 toground Center the trimmer pot Connect1 1 5V to the power supply terminals

(observe polarity!). Connect a scope toE5

l f a l l is wel l , you should see 4.5V p-psrnusoid at about 2 kYz on the scope. Inany case, it is wise at this potnt to checkali important d c. voltages Table 1 pro-vides a l isting of all key d c. voltages asactually measured in the prototype.' 8Q3" means the base of Q3 Ge1means the gate of Q1 , -lcl means theinver t ing input of lc l , OlCl means theoutput of lC1 lc l -6 means p in 6 of lCl .etc Values indicated by an asterisk maydepend on the input s ignal , adjustmentor something s imi lar , and they may notbe close to those listed under certainconditions. All values assume a steady-state condition. A voltmeter with a 1megohm or greater input impedanceshould be used. Be very careful to avoidshortang adjacent pins on lCs whenprobing (especia l ly p in 6 on the 3t8s) .When possible, clip to a resistor con-nected to the desired point instead.When probing certain sensitive points( l ike the inver t ing input of an op-amp), i tmay be necessary to isolale the meterprobe with a 1O-kilohm resistor to pre-vent high-frequency oscil lations.

A check of the key voltages for Cp1should reveal any problems at this pointand aid in finding the cause. lt is worthnoting that the inverting and non-invert-ing inputs of an op-amp properly operafing with negative feedback should neverdiffer by more than a few mill ivolts. lf agreater difference is observed, there arethree possibil i t ies: First, the stage maybe over-driven by the input, which may

point to trouble further back in the chain.The second is a bad (or unpowered) oo-amp Thrs is general ly the case i f the po-larity of the input dif lerential is not con-srstent with the output polarity. lf this oc-curs, the output of the op-amp is usuallysaturated near one power supply rail. l fthe output is zero, a short from output toground may be present. The third possi-bil i ty is a fault in the associated input orfeedback circuilry ln thrs case, the po-larity of the input dif lerential wil l be con-sistent with the output polarity. In anycase, troubleshooting complex circuitsInvolvtng many possible interactions re_quires care, thought, arrd patience Thekey l ies in separating cause and effectOlten a particular portion of the circuitcan be made to look faulty by a problemersewnere.

lf the signal at E5 is large and clipped,and the gate voltage of Q1 is stronglynegative, the fault is probably associatedwi th the mul t ip l ier c i rcu i l ( lC3) . I f the gatevol lage is about +0.5V, the problem l ieswi th the a.g.c. contro l c i rcu i t (1C6,7) orlhe a.g c. detector (O2-O4) In this casea large positive level (greater than+2 5V) at E7 points lo the tC6,7 c i rcu i -t ry . A low level ( less than 1V) points tothe detector circuitry

lf the signal at E5 is zero and the gatevoltage of Q.1 is strongly negative, theproblem l ies wi th the |C6,7 c i rcu i t rv . l fthe gate vol tage of 01 rs aOout +0 5V,something is wrong with the oscil latorp rope r ( lC1 ,2 ,4 )

An unstable or varying level indicatesa problem with the lC6 7 circuitry involv-ing a.g.c loop stabi l i ly . Check R21 ,Rl 5 R16, Rl 7, CO and the capaci torconnected to E7.

Now check the main output at E11 l tsl 'ould be lhree times as greal as that atE5 i r the output ampl i f rer is operatrngproperly.

Replace the resistors from E1 to E2and E3 to E4 with 39-kilohm resistors.The frequency should drop to about 200Hz. Observe the full-wave rectif iersawtooth waveform at pin 6 of lC6.Make sure the signal is full-wave and nothalf-wave; the latter indicates trouble in-volving lcs, 03 or Q4, AdjusI RZ4 Iorperfect symmetry, i.e., so that adjacentpeaks are of equal level. Incorrect andcorrect adjustments oI R24 are i l lustrat-ed by the 'scope photos in Figs. 25 and26 .

Replace the capacitors at C(KM) and

Vp TOV T V M

Ftg 24 Test circuit for nteasuring J-FET pinch-off voltage

Fig 25 - lllustratton of tncorrectadJustment of R24. Bottorn trace ts f rompin 6 of lC6. (Scale 0 2 V/div.)

Fig. 26 - lllustratton of correctadluslment of R24 Botto;m trace ts frompin 6 of lC6. (Scale 0.2 V/dtv.)

+ l 5 v

l

AUDIO/SEPTEMBER 1981 6 ?

ART THRE

cPlo tc l +2 IO lC2 +0 B+ l c 3 + 0 . 1orc3 -4 3o tc4 +1 0orcs -4 .6BQ3 +602EQ3 +2 2VE 7 + 2 . 2 Vlc6-6 + 1 3E B _ 2 5 V 'G Q l - 2 . 5 V

CP2-A+ lcg -34o t c g - 1 0 1+ t c 1 0 - 1 2o t c 1 0 + 1 2o t c l 1 + 1 . 2+ t c 1 2 0 0o t c 1 2 + 4 3O l C 1 3 + 0 9o r c 1 4 - 5 . 6otc15 -27 5E 2 2 - 2 1 V 'E 2 3 - 1 . 5 V '

CP2-B

o tc17 -521+ l c l 8 - 1 5 3o l c 1 8 - 1 5 3o t c 1 9 + 5 0t c20 -1 -7 .52Vtc20-2 -B 1 9Vtc20-3 -8 1 9Vr c 2 0 - 5 - 1 3 . 9 Vl c 2 0 - 1 2 + 8 5 9 VO l C 2 1 + 9 2l c 2 2 - 1 2 + 9 . 3 VE32 +102ADs +450E 3 1 + 4 8 '

CP3-Ao lc25 +2 5l c26 -1 -7 6Vtc26-2 -B 2Vrc26-3 -8 2Vtc26-4 -7 6Vlc26 -5 -14 0Vlc26 -6 +5 6Vt c 2 6 - 1 0 - 2 2t c 2 6 - 1 2 + 5 6 Vlc27 -5 + 1 3Vt c 2 7 - 7 0 . 0 'o t c28 +7 2r c 2 9 - 1 0 - 1 0 4r c 2 9 - 1 2 + 5 6lc30-5 + I 4

CP3.BE 2 9 - 1 0o t c 3 l - 1 4E 3 5 - 1 4E36 -0 6E 3 B + 0 1E 4 0 + 1 3 6o t c 3 5 + 1 1 0orc36 0 0o l c 3 7 + 1 7 V 'rc3B-3 +3 3Vrc3B-6 -3 5Vlc38-9 -2 1V ' .l c 3 8 - 1 3 1 2 1 Vl c 3 8 - 1 4 + 1 3 3 VE 4 7 + 1 2 8 V

Table 1 - Key d.c. voltages as mea-sured in completed prototype ana-lyzer operating at 1 kHz with 1V rmsinput on the 3V input range, readingO.OOO5o/o distortion on the O.OO3o/osensitivity range. All voltages are inmill ivolts unless otherwise noted.The asterisk (') denotes voltages

C(LN) wi th 0.22 gF, t10o/o polyestercapacitors. Connect a 1.0-gF polyestercapacitor from E7 to ground. The fre-quency should now be about 20 Hz, thelevel should be the same as before, anda stable level should be realized within1 5 seconds after power is applied.

Input Amplif ierAnd Bandpass Filters (CP2)

Prepare CP2 for bench testing byconnecting 0.022-yF, l1 0olo polyestercapacitors lrom E.1 7 to E1 B and E20 toE21. Connect 3.9-k i lohm resis tors f romE16 to E17 and E19 to 820 . GroundE14 . E l5 . E22 and E23 Connec t 11 5V to the supply terminals, carefullyobserving polarity,

Check all d.c, voltages for this circui-try (Fig '1 2) in Table 1 . Pay less attentronto those with asterisks. lf problems arefound, they should be corrected now.Apply 1 V rms at 2 kHz to E1 2 and checkfor a 3V rms output at E13 Apply thesame signal to the bandpass fi l ter inputat El5. Observe the s ignal at E18. l tshould exhibit a bandpass characteristiccentered at about 2 kHz as the input fre-quency is varied. Set the frequency formaximum output. Adjust R62 for a levelo f 1 . 15V rms a t E18 . Ad ius t R59 fo r a

which depend strongly on operatingconditions, FET pinch-off voltages,etc. Keep in mind that many voltagesmay depend significantly on op-ampinput offset voltage or bias current,and may even have the opposite signin some cases.

center lrequency ol 2 kHz and retrimR62 .

Observe the s ignal at p in 6 of |C14. l tshould be approximately 1 0V p-p andin-phase with that at E19. Tempor.ari lyremove the short lrom E22 to groundand connect E22 Io -1 5V. The s ignal atpin 6 should now be about 1 0V p-p andinverted lrom that at E19. Repeat thisprocedure for pin 6 of lCl6 with phasechecked relative to that of E21 andchangrng the voltage on E23 fromground to -15V. The resul ts should beabout the same, lf either ol these proce-dures reveals a problem, one of the mul-t ip l iers ( |C14, Q5 or lC1 6, OO) should besuspected. Note that changing the volt-age on E22 should slightly affect thecenler frequency (about 11 6olo), whilechanging that on E23 should only s l ight-ly alfect the amplitude at E18 (about i3.50,6). Check for about 30mV rms atE24 .

Product Amplifiers AndAuto-Set Level Clrcuits (CP2)

Connect a 10-kilohm resislor fromE27 1o E28 and an 82-kilohm resistorf rom E26 to E28, and check the d.c.voltages for this circuitry (Fig. 13) inTable 1 . Voltage measurements at the

pins of lC20 and lC22 should be madethrough a 1 0-kilohm isolating resistor atlhe end of the meter probe lo preventosci l la t ions.

With the 1V rms, 2-kHz s ignal s t i l l ap-p l ied to E15, sweep the f requency andobserve the output at E26 A notchshould be observed at the cenler fre-quency of the bandpass fi l ter. Try to ad-just lhe generator frequency and the set-ting of R62 for a deep notch (less than1OmV rms). l f R62 doesn' t have enoughrange, connect E23 to -1 5V instead olground and try again. Now offset thegenerator frequency 1o obtain a 1 00mVrms output at E26 Ground E25 and ob-serve a level of I .0V rms at E26. Re-move the ground at E25. Check pin 6 oflC19 for a level of 1 00mV rms

Now check the auto-set level circuitry.A level of about 1 .1V rms should aooearat the output of the relerence VCA (E32),independent of input level over a 110-dB range. At the nominal 1V rms inputlevel , E31 should be at about 0V d.c. ,whi le p lus and minus 10-dB input levelsshould result in approximately minus andplus 3V respectively at E31 . Recovery tonominal output from a 2O-dB drop in in-put level should take about 1 0 seconds.lf the level is too large and E31 is strong-

54

l y

ICS

e(e(p(

ICcp 'I 'pSC

1 (S€p l'1 \

TNorc€C C

CIn€mIKRard l

a -

otro

AUt

ART THREry negative, the circuitry associated withlC22 and tC23 is probabty fautty tf E31is strongly positive, the circuitry associat_ed with D5 and lC24 should be suspec!ed. lf the level is too low, the oppositepolari l ies at E31 wil l point to the problemareas above.

Check the distortion product output atE29. A level of 100mV rms snoutO Oeobserved, independent of input level var_ra lons over a 11O_dB range about 1Vrms. This assumes that there is sti l l100mV rms a t p i n 6 o f lC l9 when theinput level is I V rms.

Auto.Tune Circuits (Cp3)

.^ Prepare this portion of Cp3 (includinglC31 Fig 18) for bench testing by con_necttng a 1O-kilohm resistor from 829 toground. Center tnmmers R135 andR157 Connect 1 j 5V to the supply ter-mrnals of CP3 (observing polarity) andcheck the relevent d.c voltages in Table1 . Voltage measurements to tne pins oflC26 and lC29 shoutd be made througha 1O-kilohm isolating resistor By offsJt_t ing R135 or R. l 57, i t should be possibteto get the respective jntegrator outputs atE23 and E2Z Io dnft slowly in a positiveor negative direction between aoproxi-ma te l y - 12V and +0 .3V

Apply 100mV rms al 2 kHz to E29and observe the same signal at pin 6 oflC31 Observe a somewhat sof|yclipped 1 5V p-p version of the signal aipin 6 of 1C25. Briefly short E33 toground and observe a 1V rms level atp in 6 of lC31 and a 2.5V p-p roundedsquare wave at pin 6 of 1C25. Apply100mV rms a t 2 kHz to E2 i and 'ob_serve a soffly clipped 0 gV p_p level atpin 6 of 1C28. Increase the input level to1V rms and observe a hard c l lpped 1 .OVp-p signal at pin 6 of 1C28.

.. Further b_ench testing of Cp3 requiresthe use of CPZ. l f any problems remainon CP2, correcl them now before oro_ceeding Prepare Cp3 by makinq rnter_connections E21 , E29 and E3i fromCP2 to CP3. Remove the existing con-nectlons at E22 and E23 on Cpl andmaKe lnterconnections EZZ and EZ3from CP3 to Cp2. Center R135 andR157 l t is assumed that E25 is openand that the attenuator between E26and E28 is s t i l l in p lace,

^ y lh the 1V rms, approximaretyz-kHz signal applied at E.l 5 as previ_ously, check pin 6 of lC25 for a 3V p_prounded square wave. lf the level is vervl

Fig 27 -- lr t tertor of the !1 5V powersupply rnodule A torrotclal powerlransforrner was utrlrzecl tn lhe prototyS:e,to rninimtze hun't

small, the analyzer may have tuned it_self. In this case, changing the frequen_cy oy about 10olo so that i1 is well out ofthe tuning range should yield the squarewave Also check for a 1V p-p squarewave at pin 6 of lC2B.

Adjust the input frequencv for a mini-mal oulput at E29 and measure the d.cvoltage af E22 Set the input frequencyto yield a voltage equal to one_half thepinch-off voltage measured for eS (3.5Vrf you didn't measure it). Now aOlusi nOZtor a d.c. voltage equal to one-half thepinch-off voltage of e6 at E23. A com_plete null of the fundamental should nowbe present at E2g, with only distortionand noise visible.

Now place a j 00-to-1 attenuator be_tween E29 on Cp2 and E29 on Cp3; a'l 00-kilohm series resistor and a 1_kil-ohm shunt resistor to ground wilt do. Al-lernate ly adjust R135 and R157 for thebest possible fundamental null as ob-served at E29 on Cp2. These adiust_ments should be made slowly, as thetime constants in the auro-rune controlcircuits are long.

Fllter. MeterAnd Status Circuits

Strap E35 to E36 and E4O to E41.Connect a '1 O-kilohm resistor from E34to ground, Connect a 1-k i lohm shuntresistor trom E4Z to ground. CenterR1 80 and Rf 92. Apply power and

Apply a 300mV rms, 2_kHz s ignal toE34. Check for 300mV rms levelJat p in6 of lCs 32 and 33. Adlust R1g0 forabout 100mV rms at p in 6 of lC34 and1V rms at p in 6 of 1C35. Drop the inputlevel at E34 to 3OmV rms and short E3gto ground. Observe aboul j 00mV rms atp in 6 of lC34 and 1V rms at p in 6 oflC35 Check for 1.4V p_p hal f_wave rec_rtrteo stgnals at the anode of D22 (nega_trve-goang) and the cathode of D23 (posi-tive-going). Check for a posrrtve d.c.level of about l 1V alE42

Remove the strap from E40 lo E41ano place a strap from E24 on Cp2 toE11 npflv a 3V rms, 2-kHz signar roE 1 5 on CP2 and check for t V rmj at pin6 o f 1 C 3 5 . A d j u s t R l 9 2 t o r a + I , t V d c .level at E42.

Check the status circuits by connecfing four LEDs from terminals E43through E46 to + 1 5V Remove thestrap from E24 Io E41 and replace thestrap from E40 to E4 1. Reconnect E29on CP2 direcily to E29 on Cp3. Connecta 1V rms 2 -kHz s igna t t o E15 on Cp2 .D26 and D2Z should be ext inguished,while D24 or D2S shoutd tight if the inputfrequency is tuned above or oetow thetuntng range respectively. They shouldboth be extinguished if a good notch isDetng observed at E29 Drop the inputlevel to 0.25V rms; D27 should l ight .Raise the input level to 4V rms: D26should l ight

Bench testing of the circuit boards ischeck the relevent voltages in Table 1 .

A U D T O / S E P T E M B E R 1 9 8 155

ART THREE

now complete. lf any problems surfaceafter f inal assembly, they are probabtvnot on the circuit boards

Chasels AssemblyThe first step in chassis assembly is to

bui ld the power supply module. l t is bui l trns ide a 2 'a by 3 by 5 n rnch a luminumutil i ty box. This affords some shieldinqagainst 60-Hz hum. Convent ional point_to-potnt wiring is used, employing a per_forated board for component mountinqThe voltage regulator lCs should be boit-ed to one wall of the enclosure and insu_lated with mica washers. Make sure thatmetal burrs don't cause shorts betweenthe enclosure and the lCs. The powersupply module is botted to lhe rear of theanalyzer enclosure, with the line cordpassing through both the power supplyand the notched-out analyzer coveiPower leads and switch wiring pass tothe interior of the analyzer through agrommet in the power_supply enclosure.A polarized l ine cord is recommended toguarantee that the power switch is al_ways on the neutral side of the a.c. l ineso as to minimize hum. The powertransformer should be mounted rn themodule so that it is at the extreme rioht_rear of the analyzer, far away from Cptand CP2. A photograph of the powersupply is shown in F ig. 27 .

Although the choice of the oowertransformer is not crit ical. use of a torror_dal design (as shown) wil l assure low in_duced hum. The Avel-Lin dberg 40 /3004 used here is available from Saoer

Electr ica l Supply Co. , 60 Research Rd. .Hingham, Mass. , 02043 at a cost ofaboul 924 00. In any case, the inputvoltage to the regulators under the fullanalyzer load should not be less than1 8V ( inc luding r ipple d ips) nor greaterthan 35V. l f necessary, adjust R21l andR212. fhe 40/3004 l ransformer doesnot have much extra current capacity, so'be parlicularly observant ol requlatorheadroom in thrs case

The three circuit boards should bemounted on 3/a-inch threaded 6_32 stan_doffs and placed as shown in Fig. 4 (partI , Ju ly issue) CPI is p laced so that tC1is closesl to the front panel. Cp2 isplaced so that lC9 is closest to the fronlpanel CP3 is placed so that lC3g isclosest to the front panel Make all of thepower-supply connections and then theconnections between Cp2 and Cp3. Cir_cuit board power and ground should bedistributed on a single,point basis from aterminal strip mounted near the inoutlack (J3)

Two types of shielded cable wereused in this project, low-capacitance mi_crophone cable (34pF/ft.) and high_ca_pacr tance min iature cable (1 Z4pF/ f t . ) .Unless specified, the miniature cablecan be used. Use a shielded cable forthe E29 interconnection (shield ground_ed only at E30).

Now mount and interconnect all of theremaining front panel items except therange and lrequency switches (S1 andS3) The resistors resrding on the oufputlevel, input level, and sensitivitv switches

(S2 54 and 55) should be wired ontothe switches prior to mounting of theSWIICNCS.

Level control R30 should be connecfed to E9 through a shielded cable. Theshie ld should connect to E10 and theuuvv eno or HJU I he output allenuator(S2) should receive rls ground direcilyf rom the srngle-point ground. R2O5 wi i lult imately be suspended between S2Ca n d S l l .

As mentioned earlier, the Input atten_uator should be mounted in a smal l .sh ie lded enclosure as shown in F iq 2g.Use shielded cable from the inpul iackand single-point ground to the attenua-tor. Four ferrite beads are placed on thelead from C21 to 54 for improved r.f. irmmunity. Connect the output of the at-tenuator lo E15 on CP2 wrth low-capaci_tance shielded cable Note that theshield supplies ground to the secondarvsrngle-point ground (E 1 a) on Cp2

Selection of the attenuator frequency-compensatron capacitors (shown dottedin Fig. 1 2) wi l l requi re exper imentat ion,as the required values and even topolo_gy (senes vs. shunt connection) wil l de_pend on particular parasitics. The bestapproach is to do the compensation be_fore install ing the module, using the ac-tual required length of low-capacitancecable at the output looking into a 1 00-kilohm low-capacitance load (or a 100_ki lohm 100-to-1 resrst rve at tenuator)Use an audio generalor and an audioa.c. voltmeter or oscil loscope to achievea flat frequency response on all ranoeswrth the module 's cover in p lace

High-capacitance shielded cable isrecommended for connection of the sen_srlrvity switch (S5A) to E28 Note that theatlenuator receives its ground trom EZ7vra the shie ld. The "hot" ends of R90and R94 on S5A can be connected to anearby unused switch position.

l l you have chosen to implement thesimple i ixed-product f i l ters of F ig. 19 in-stead of the tracking l i l ters, wire themnow. Use shielded cable for lhe connec_t ions to E29 (on CP2) E34, E3S E36and E38. The fi l ter ground connectionshould come via the shield of the cablegoing to E34, which shie ld can be con-nected at E37. The shield of the cabtegoing to E29 should be connected onlvat E30 on CP2 The 51O-pF capaci tor inFig. 19 should be made smal ler bv theamount of the cable capacitance in oar-allel with it.

f ig. 2B - lnterror vtew of the inputattenuator module.

I

ran 3mocanador^tancA ^ . -r t d \ e

SU n- r

are3 9

frorn l e e ;

the :, :2 '

S O t r ' :

( J3

? i u

Youhacsd u I

E 2 :c r l l a : :v r : ,

l lo ' - .wo ' /

ot t ,

K t i o . '

lwee -

SO L ' :d r d :

I U L -

tor ' l : '

) d L t > :

f es rC ,

corre:

Final

S W : : 'D r : .pur i l

a(e 'Wll u -s

wrtf 'swr lc 'Sawr'swr lc 'swrtc 'q r u 5

screw

l o n n t -

lrom :

tI

II

56AUDIO

ART THRE

It

tt

Intermediate Check-OutAt this point, prior to the wiring of the

range and l requency swi tches, amoderately thorough check-out of theanalyzer is recommended. This can bedone if the temporary tuning capacitorsand resrstors installed for bench testinghave been left on the boards. We as-sume here that the capacitors in placeare 0.022 pF and that the resistors are3.9 k i lohm. The 0 1-prF capaci tor f romE7 to ground and the 2.7-k i lohm resis torfrom E5 to E6 on CP.l should also be inplace. lf you have chosen to implementlhe tracktng product ft l lers. strap E29 toE34 and E35 to E36 Connect thesource output (J2) to the analyzer input(J3)

The analyzer can now be put througha full set of paces al 2 k1z. Note thatyou may have to adjust R59 and per-haps R62 to get a null and have reason-able FET control voltages at E22 andE23 (say, -1 to -4V) Also check the os-cil lator a.g c. FET gate voltage at EBCheck all functions of the analyzer undera variety of level and sensitivity condi-tions lo determine that everything isworking properly Look for evidence ofosci l la t ions.

"Distortion" from a separate signalgenerator can be iniected through a 62-kilohm resistor into the conneclion be-tween the source and the analyzer. lf thesource is producing 1V rms and the sep-arale signal generator is delivering'1 00mV rms to the resistor, a 0.1 o/o "dis-tortion" level wil l result Check the cali-brat ion of R180 and Rl92 Check torsatisfactory tuning time and instrumentresidual. lf any problems are uncovered,correct them now.

FinalAssemblyAssembly ol the range and frequency

swrtches may require some ingenuityS1 consists of nine five-posrtion, two-pole wafers if the tracking product f i l tersare implemented. The resulting lengthwill usually require that S1 be assembledwith parls from two or more rotaryswitches. The shaft can be extended bysawing off the shaft from a secondswitch and affixing it to the shaft of theswitch under construction with a stand-ard shaft coupler. The pair oI 4-40screws which hold the switch togethercan be extended by joining additionallengths oI 4-40 screw to them (obtainedfrom the second switch) with tapped 4-

Ftg 29 - C/oseup ol the spectailyconstructed range and frequencyswrlches.

40 standoffs. A close-up of the switchesassembled in this way is shown in Fig.29. Alternately, the Centralab swrtchcomponents specified in the parts l istcan be assembled inlo the requiredsw[cn.

The capacitors are mounted betweenadlacent wafers as shown in Table 2.For example, C(KM)-l mounts betweenpoles K and M at posi t ion l poles S1Aand S.l B are on the wafer closest to thefront panel. The capacitor and resistordesjgnations for switch-mounted compo-nents assume that the tracking productfi l ters are being implemented and thattheir components are mounted on theswrtch sections closest to the front pan-el. Note that the tuning capacitors for thehighest frequency range have a smallresistor in series with them. This resistorcompensates for a high-frequency phe-nomenon in active fi l ters known as "O_enhancement" which results from op-amp hrgh-frequency rolloffs. These resrs-tors can be mounted on the unused po-sit ion-5 switch terminals. Nole that thetuning capacitors are wired so that thereis always some capacitance connecteoeven when the switch is between posi_trons, preventing undesirable transients.This is accomplished by wiring the posi-l ion-4 terminals in parallel with their re-spective wipers. The use of shorting-typeswitches would also have accomplishedthe transient suporession.

The high-pass product f i l ter capaci-tors [C(BD) and C(DF)] for rhe ZO-kHzrange are also used for the 200-kHz

range by connecting positions 3 and 4of the associated switch poles together.This was done because stabil ity of high-er f requency active high-pass fi l terswould have been a problem, and the ad-ditional f i l tering is not really necessary onthe highest range.

As shown in F ig. 29, the four f requen-cy trimmers (R1 to R4) were mounted ona small prece of perforated board andmounled to poles J and H of Sl with 1 8-gauge solid bare wire R206 is mountedon the switch between the two wipers.

After 51 is assembled and loadedwith components, mount it and make allpossible interconnections to the circuitboards. Each of lhe four tuning capaci-tors [C(KM), C(LN), C(OO)and C(pR)] isconnected to the circuit boards by sevenInches of low-capacitance shieldedcable. The capacilance of the shieldedcable is directly in parallel with that of theassociated tuning capacitor. Instead ofberng grounded, the shield intercon-nects one side of the tuning capacitor tolhe output of the associated inteqrator,Thus. for example, the shie ld connectsone side ot C(KM) to E3 The capaci-tance of this length of cable has beentaken into account in the values of thetuning capacitors for the 200-kHz range,and appropriate alterations should bemade to lhese capacitors if other valuesol shielded cable capacitance are used.The shields are connected to switchpoles K, L, Q, and R.

Frequency switch S3 consists of eight1 1-position, single-pole wafers, and

AUDIO/SEPTEMBER 1981 57

ART THRET

to get all of the resistors which err on thehigh side placed on one wafer, and viceVETSA,

As with the tuning capacitors, 53 iswired so that resistance is present evenwhen the switch is between positions.This is done by wiring the position-1 ter-minal to the switch wiper. The free endof the lowest valued resistor connectedto position 1 1 can be tied to position 1 2if you have a 1 2-position switch or to asmall, insulated terminal on or near theswitch if you have an 1 1-position switch.

With 53 assembled and loaded withcomponents, mount it and make all oflhe remainrng interconnections. Wherepossible, make the interconnections wilhpositions on S1 that already have a wirerunning back to the appropriate circuitboard terminal.

Connection of the tracking-ftlter com-ponents on S1 and 53 to E29 E34.E35, E36, E37 and E38 can be donewith shielded cables as discussed earlierfor lhe fixed fi l tering option. In lhis case,however, the connections to E34 andE36 must be made with low-capacitanceshielded cable. The value shown lorC(EG) for the 200-kHz range assumesabout eight inches o1 this cable. Theproduct f i l ter interconnections between51 and 53 need not be shietded butshould be dressed as far away as possr-b le ( i .e . , against the f ront panel) f rom thesignal leads and components of lhe os-cil lator and analyzer sections. There is avery substantial amounl of gain betweenthe oscil lator/BPF circuits and the prod-uct f i l ters. Failure to get adequate isola-tion here could result in oscil lations, par-ticularly on the 200-kHz range Leadsassociated with the analyzer sectionshould also be dressed away from thoseof the oscil lator section. Assemblv of theanalyze( is now complete

Test And CalibrationThe signal source should be checked

first. Center trimmers R1 to R4. Applypower, and monitor the main outpul on a'scope and a.c. voltmeter. Also monitorthe d.c. voltage at E8, the gate bias forQ1 . Check operation at all frequencies.Check frequency response flatness andlevel stabil ization l ime. Check to see thatthe gate voltage of Q1 does not get with-in 0.5V ol either zero or the measuredpinchoff voltage at any frequency.Check the action of the output attenuatorand level control. lf you have a frequen-

should be constructed in the same wayas 51 above, lf possible, it is recom-mended that switch shields be installedbetween sections 4 and 5 and betweensections 6 and 7. The resistors on 53are mounted between adlacent positionson a given wafer. The wiring is docu-mented in Table 3. For example, R(A) 1-2 goes on S3A, the wafer closest to thefront panel, between positions 1 and 2.

Although precision 1 0,6 resistors areprefened for R(E) through R (H), the ontyrequirement is that the four resistors ofeach position be matched to each otherwithin 1olo. Thus, you may save somemoney or hassle by measuring andmatching 50,6 carbon-fi lm resistors. lfyou buy 1 0 o{ a given value at the sametime, there is a very good chance you'l lf ind Jour with the required match. Try not

Range20Q Hz

2 k H z20 kHz

200 kHz

Position C (AC)1 0 . 1 s F2 0 .01 pF3 1 0 0 0 p F4 1 0 0 p F

c (EG) C (BD, DF)0 .068 pF 0 22 pF6800 pF 0.022 pF680 pF 2200 pF

47 pF 2200 pF

c (Gr) c (KM, LN) C (OO, PR)1 0 pF 0.22 pF " 0.22 1tF'

0.47 1tF 0.022 1f ' 0.022 pF'

0 .1 pF 2200 pFt 2200 pFt0 .01 pF 200 pFt 200 pFt

+ 6 8 0 + 1 8 0 0

cy cc,th is cof t l^€creas(swttc"t ions :

Sel{ r e n t r

que . :freque

a s asourc i

the cand Eand E

l y ze r ' .hal f : 'U r E d >

l ne c lb r rn cL l t g - :

the ' l65Hzd l L l :

volta-:,R4 a :chec {{ .^^ ̂one -ctASSOC

R 4 t rn r n h a

con l r :

Now : 'i t woutor f.tShou r J

t f : -chang,a n a i . '

k1z e'catrc-sato. 'C / P R

q u i u ) .

bul r ' l :d ivrduiampsE23 gtnese ,goes :be toc

A n c

towarc

Table 2 - Qspssilel connections onrange swi tch S1. The aster isk ( ' ) in-d icates polypropylene, polycar-bonate or polyester of at least 1OO Vworking voltage. The dagger (f) indi-cates polystyrene or silvered mica,again of at least 1OO V working volt-age. The others are not crit ical.

Frequency2 02 53 04 05 06 5B O

1 0 0r 3 01 6 0200

Positionl234567B9

1 0t 1

Desig.R O 1 2R O 2 - 3R O 3 - 4R O 4 - sR O 5 - 6R O 6 7R O T BR O B - e

R O 9 r 0R O 1 0 , 1 l

R O l l - '

R (A, B)1 5 0 01 0001 3 0 07 5 06 8 04 3 03 6 03 3 02 2 01 8 0750

R (C)3 6 0 024003 0 0 01 8 0 01 6 0 01 0009 1 08205 1 04 7 0

1 800

R (D) R (8, F, G, H)9 1 0 0 7 5 0 06 8 0 0 s 6 2 07 5 0 0 6 1 9 04700 38304 3 0 0 3 4 8 02 7 0 0 2 1 s 02400 1 9602 2 0 0 r 7 8 01 3 0 0 1 1 0 01200 1 0004700 3830

Table 3 - f is5i5[e1 connections onthe frequency switch 53. The aster-isk indicates a tie-point, position 12if available; otherwise use an insulat-ed terminal. R (E, F, G, H) should be1o/o, th-watl carbon-film types whereeach group of tour l ike values shouldbe matched to within 'tolo. Others arestandard 5o/o, th-wall carbon film. Allvalues are shown in ohms.

AUDIO

-

ART THRE

I

tI

I

I

ttt

cy counter, check all the frequencies; althis point they should be within 120oloof the selected value, and should in-crease monotonically as the frequencyswitch is advanced. Check for oscil la-trons as well.

Set the analyzer to 650H2, and adiustfrequency trimmer R2 Ior a source ire-quency of exactly 650H2 if you have afrequency counter. Otherwise use a Lis-sajous pattern with a 65O-Hz oscil latoras a reference. Connect the signalsource to the analyzer input, and apply1V rms on the 3V input range. Moni torthe d.c. levels at E22 (frequency control)and E23 (ampl i tude contro l ) . Adiust R59and R62 as necessary to get ihe ana-lyzer lo tune and to set these voltages tohalf the measured pinchofi voltages o1the associated FETS.

The remaining frequency trimmers inthe oscil lator wil l now be adjusted tobring the oscil lator into alignment withthe center frequency of the analyzer forthe remaining frequency ranges. At65H2, adjust R1 so that the d.c vol taqeat E22 equals half the recorded prncf'offvo l tage for Q5. Adjust R3 at 6.5 kHz andR4 at 65 kHz in the same manner Nowcheck the voltages aI E22 and E23 at allrrequencres to make sure that neitherone gets wi th in '0.5V of e i ther zero or theassociated pinchoff voltage. Make com_promrse adlustments among Rj , R3,R4, R59, and R62 as appropr iate. Aproblem with the E22 vollage (frequencycontrol) may mean an incorrect frequen_cy-setting resislor somewhere on 53.Now that the oscil lalor lr immers are sel,it would be useful to recheck the oscil la_tor frequency calibration. Typically itshould be within 11 0olo.

lf the amplitude control vottage (E23)changes substantially toward the highend oi the frequency range (between 50kHz and 200 kHz), th js coutd be an indi_cation that the Q-enhancement comoen-salor resistors in series with C(OO) andC(PR) on the highest range need someadJustment. The value shown in theschematics is a good compromise value.but it does depend somewhat on the in-dividual rolloff characteristics of the op_a m p s ( l C s 1 0 1 1 , 1 2 , 1 3 , a n d 1 5 ) . t fE23 goes too positive (toward zero),these resistors may be too large lf E23goes too negalive, these resistors maybe too small.

Another phenomenon to watch fortoward the high end of the 200-kHz

Fig 30 - Analyzer residual at 20 Hz1V rms operattng level rn the 0 003,,,,set)sttivttv range (Scale 0 2 V/dtv )

Ftg. 32 - Anztlyzer resrduul at 2A kHz1V rms operattng level in i l te A 003,,, ,senstttvtty rartge (Scale A 2 V/r j tv )

Ftg 31 Analyzer restdual at 1 kHz1V rms operatrng level in the 0 003o,,sensttivtty range (Scale 0 2 V/diry )

Fig 33 Analyzer residual at 200 kHz.1V nln operattng level rn the 0 03,,, ,senstttvtty range (Scale 0 2 V/dtv )

range is "s lewing osci l la t ions. " Toomuch phase lag around an active-fi l terloop can cause Q-enhancement and. ul-t rmate ly , osc i l la t ion i f the s ignal beinghandled by an active fi l ter causes one ormore of the op-amps to slew-rate l imit,the effect is equivalent to a substantiallyincreased phase lag This can result inoscil lations which drive the amolif ierseven deeper into slew-rate l imiting. Theonly way to stop such an oscil lation is totune to a lower frequency or removepower. The normal signal levels in thisanatyzer cannot touch off such an oscil_lation, but a sufficiently large transientcan. That is why the range and frequen-cy swrtches are wired so as not to gen-erate transients when they are operated.However, occasionally a power-on tran-sient can trigger such an oscil lation inthis analyzer if i t is powered up whentuned to greater than about 1 00kHz.

The frequency response of the track-

ing product f i l ters should be checked atthas point by injecting a signal at theirinput in place of E29. lf the level at threetimes the fronlpanel setting is taken as0 dB, then the lower and upper 3-dB-down f requencies should be at 1.4 and1 0 times the set frequency, respectivelv(0 1 4 and about 8 t imes on the 200-kHzrange).

To calibrate the distortion reading,two signal sources should be used. Ap-ply 1V rms at 1 kHz to the analyzer fromthe internal signal source. put 56 in the"Level" position and adjust Rl 92 for afull-scale reading with the input attenua-tor on the 1V range. Apply 1V rms at 3kHz f rom a second audio generatorthrough a 62-kilohm resistor. The ana-lyzer wi l l now see 0.99V at 1 kHz and1 0.1 mV at 3 kHz, or a "d is tor t ion, ' leveloI 1 .02o/o Set the sensitivity switch (S5)to the 1olo fdrge and adjust R180 for ameter reading lust a hair over full-scale.

AUDIO/SEPTEMBEF1 9 8 1

ART THREE

*| .o l)o

u

I .oolF

)

kHz

\

- 2 0

l k

kHz

lOHz

H Z

;eI

J

6U

oIF

EI i l r

t r e :

l y L i '

o ' ' - ,

h '

o.5 t .o

RMS LEVEL _ VOLTS

8 5F .

R 6tr-F , :

F .F .

R : :R E :F 9 ;

R , :r :

: '

- ' -

S : '

R 2 a

F . : -R _ . - -F : 'o a ,

F 3 a -F 3

- -

R3 a-R 3 9o 1 '

l - 1 1 :

n 4 -

F i - -

B t a -R.1 : -t r :Q ; '

R 5. :R 5 an aF E :

F .

F .R B .R 8 9F 9 a -R 9 'R92 -R9 3 ,R 9 4 -

Fig. 34 - Restdual vs. level as afunction of frequency in the 3V tnputrange.

ftg 35 Besidua/ vs heeuettcv as af unction of input level in the 3V rnputrange.

To complete testing, measure the an-alyzer's residual at all frequencies and atvarious levels by connecting the signalsource directly to the analyzer input. lt isuseful lo observe the "Dist. Out" signalon a 'scope at this point.

lf the residual seems particularly noisyor jumpy on only one range, don't hesi-tate to suspect the trimmer pots and thetuning capacitors. This is a sensitive ap-plication and even minor deficiencies inthese components may cause trouble;I 've experienced trouble with both. Istrongly recommend the use of high-quality tuning capacitors.

Hum can be particularly insidious rnan instrument such as this where full-scale sensitivit ies on the order of 30 mi-crovolts are encountered, Although theconstruction details are intended to mini-mize hum, you may sti l l have to fight it. l twil l be most noticeable on the 200-Hzrange, Remember, hum can be pickedup capacitively by a high-impedance circuit (shielding helps here) or it can bemagnetically induced into any circuit, in-cluding grounds. Hum will probably bereduced when the instrument is fullvhoused in its enclosure.

PerformancePerformance of the prototype is i l lus-

trated in Figs. 30 to 35. Scope photos ofthe analyzer's residual at the nominal 1V

internal operating level tor 20 Hz, 1 kHz,20kHz, and 200 kP'zare shown in Figs.30 to 33 .

The total residual (noise and distortioncomponents) is plotted as a function oflevel for the lour frequencies above inFig 34. The residual as a lunction off requency for 0.3, .1.0 and 3.0V rms in-ternal operating levels is plotted in Fig35 Best performance is generallyachieved near the high end of its allow-able range of operating levels. At operatrng levels above about 1.5V rms, the re-s idual is below 0.001o/o scroSS the fu l laudio band.

Parte AvallabilityA serious attempt was made to de-

sign the THD analyzer with readily avail-able parts, and many constructors wil lhave no trouble finding most of the partsal normal outlets, As an aid. however.several dealers of varrous Darts are listedbelow, Because of the substantial num-ber of parts involved in this projecl, I rec-ommend obta in ing and searchingthrough the catalogs that many of theseand other comoanies make available, Al-though most ol these companies providebroad lines, a few are worth specialmention because they have some of theless commonly available parts TheNEs534AN op-amps and the Centralabrotarv switches are available from New-

ark Electronics ($25.00 min imum order) .The 2N4091 J-FETs and a suitable pow-er transformer are available from CFRAssociates. Precision one-percent resis-tors are available lrom International Elec-t ronics Unl imi ted.Digi-Key Corp.P.O. Box 677Thief River Fal ls , Minn. 56701800 /346-5144

Active Electronic Sales Corp.P O. Box 1 035Framingham, Mass. 01 701617 /879-0077

CFR Associates, Inc.Newton, N.H 03858

Jameco Electronics1 355 Shoreway RdBelmont , Cal i f . 9400241 5 / 592-8097

Newark Electronics146 Roule 1E d i s o n , N . J . 0 8 8 1 7201/572-2103(or contact nearest branch)

International Electronics Unlimited435 Fi rs t St , Sui te 19Solvang, Cal i l . 93463805 /688-27 47

AUD]C

ART THRE

R95-1 .1 k i lohmF 9 7 , R 1 0 0 , F 1 0 6 , F 1 1 1 , R l 1 4 , F 1 1 8 _ 1 5

ki lohm8 1 0 4 - 4 7 0 o h mR r 0 7 , R 1 0 8 , R 1 0 9 , F 1 1 O , R 1 2 5 , R 1 5 1 ,

R l52-4 7 k i lohmR 1 2 l - 9 1 0 k i l o h mR122-22 megohmR124-12 ki lohmR 1 2 7 , F 1 7 9 , R 2 0 0 , R 2 0 1 , R 2 0 2 , R 2 O 3 _

2 . 7 k i l o h mR 1 4 4 , R l 6 6 - 6 8 o h mF 1 69-390 ohmR1 70-3 3 k i lohmF1 7 / -9 09 k i tohm, I percenrR1 B0-2 kitohm tr irnpot (panasonic K4423)R1B4-9 1 k i lohmF 1 B 9 - 2 0 k i l o h mR 1 9 - 2 7 0 k i l o h rF1 93-8 2 k i lohmR 1 94-39 k i lohmF205-2 7 megohmR209, R210-3 .3 ohm, 2 wat t , remove i f us -

rng 40,23004 transformerR211 , 212-1 0 ohm, 2-watt; see textR(A) through R(Hl-See Tabte 3 and textC l , C25, C27-15 pF s i l ver mica (Arco

D M 1 5 - 1 5 0 )C2, C3, C26, C2B-22 pF s i tver mica (Arco

D M l 5 2 2 0 )c 4 , c 5 , C 2 3 , C 4 1 , C 4 3 , C 4 5 , C 4 6 , C 4 7 ,

C49-1 O pF, 25 V radial electrolyt ic (pana_son ic ECE,A lEVl00S)

C 6 , C / , C B , C 1 7 , C l B , C 3 5 , C 3 6 , C 4 2 ,c 5 2 , C 5 3 , C s 5 , C 5 6 , C 5 7 , C 5 9 , C 6 6 ,C 6 7 , C 7 2 , C 7 3 , C B 3 , C B 4 - 1 O 0 p F , 1 6 - Vr a d a l e l e c t r o l y t i c ( p a n a s o n i c E C EA r c v l 0 1 s )

C9-1 FF, 100,or 250-V metal l ized polyes,te r (Panason ic ECQ-E21 05KZS)

C10-0 47 sF, 100-V meta i l i zed po tyes ter(P lessey N,4 in ibox 0 47 l100 F box)

C11-0 1 ; rF , 100-V meta i l i zed po tyes ter(Plessey Minibox O t /1 00 C box)

C12, C76-0 .01 ; rF , >1OO-V meta i l i zed po__ Iyester (Plessey Mjnibox 0 01 /630 C box)c 1 3 , C 1 4 , C 1 5 , C l 6 , C 2 9 , C 3 O , C 3 1 , C 3 2 ,

c 3 3 , C 3 4 , C 6 0 , C 6 1 , C 6 2 , C 6 3 , C 6 4 ,C65, CBl , CB2-0 1 p rF , 25 ,V ceramicdisc (Panasonic ECK-DtE j 04ZFZ)

c 1 9 , C 2 0 , C 3 7 , C 3 B , C 3 9 , C 4 O , C 4 4 , C 4 8 ,C 5 0 , C 6 B , C 6 9 , C 7 0 , C 7 1 , C 7 4 - t g F , 5 O -

V radial electrolyt ic (Panasonic ECE,AIHVO_r 0s)

C21 -2 pF or 2 .2 UF, 250-V meta l i zed po-lyester (Panasonic ECe-E2225KZS)

C22 - 4 7 pF si lver nrica (Arco DM I 5-4 7 O)C24 - 1 O pF si lver mica (Arco DNfi S-1 O0)C51 - 0.22 pF, 1 00-V metal l ized polyester

(P lessey Min ibox 0 22 l100 D box)C54, C5B - 0 .47 sF, SO-V (Rad io Shack

2 7 2 - 1 0 7 1 \C 7 7 , C 7 8 , C 7 9 , C B O - 4 7 0 s F , 3 5 - V r a d i a l

electrolyt ic (Panasonic ECE-AlVV4 7 1 S)C (AC) through C (PR) - See Tabte 2 and

ICXI

Dl th rough D23 - Swi tch ing d iode (1N914,1 N41 48 or equ iv )

D24 through D2B - Red LED (Opto Etec_t ron ics XC209R)

D 2 9 t h r o u g h D 3 2 - 1 N 4 0 0 2 , 1 N 4 0 0 3 ,1 N4004 or equ iv .

Q 1 , Q 5 . Q 6 , 0 7 , Q B , Q 9 - 2 N 4 0 9 ] J . F E T(National)

02 , Q3, Q4 - 2N3904 or equ iv . gen pur_pose s i l i con ; NPN

r c r , t c 2 , t c 3 , l c 4 , t c 8 , t c g , t c l o , t c 1 1 ,r c 1 2 , t c 1 3 , t c l 4 , t c 1 5 , t C t 6 , t c 1 7 _NE5534AN (S ignet ics , T t )

r c 5 , r c 6 , t c 7 , t c l B , t c 1 9 , t c 2 3 ,t c 2 5 , t c 2 B , t C 3 1 , t C 3 2 , t C 3 3 , t C 3 4 , t c 3 5 ,lC36 - LM31 BN (Nat ionat , T l , e tc . )t c 2 0 , 1 c 2 2 , t c 2 6 , t c 2 9 _ L M 1 4 9 6 N ( N a _t rona l , l vo t . , e tc . )lC24 - LF356N (Nat iona l , Fa i rch i ld , e tc . )1C27, lC30 - t [ / ] 458N (Nat ionat , Mot . )1937

- UA741CN (Fa i rch i td , Nat ionat , e tc )lC3B - LM324N (Nationat, etc )l C 3 9 - L M 3 4 0 T - l 5 o r U A 7 8 1 5 C T ( N a t i o n _a l , Fa i rch i ld )l C 4 0 - L M 3 2 0 T - 1 5 o r U A 7 9 1 5 C U ( N a t i o n _a l , Fa i rch i ld )I1 - 32-48 V c.t . , 1 -amp power transformer(e.g., CFB Associates Tranny 1 or Avel Lind_berg 40 l3004)F1 - 1-A 3AG slow-blow fuseS1 - 5 posit ion, g,section, l B,pole rotaryswitch (Centralab PA302 6-inch index assy.,Newark #22F652 plus g pA-32 sections,Newark #22F842 or see text). ( l f using f ixedf i l te rs , 5 -pos i t ion , 6 -sec t ion , 1 2 -po le ,PA1 033, Newark #22F8337S2 - 5-posit ion, 2,section, 4-pole rotarys w r t c h ( C e n t r a l a b p A 1 0 1 3 , N e w a r k#22F813\53 - 1 1-posit ion, B-section, B-pole rotaryswitch (Centralab PA302 6,inch index assy.,Newark #22F652 plus B pA-30 sections,Newark #22F840 or see text) ( l f using f ixedl i l te rs , 1 1 -pos i t ion , 4 -sec t ion , 4_po le ,PA1 01 5 , Newark #22F81 5)S4 - 5 posit ion, 1-section, 2-pole rotarys w i t c h ( C e n t r a t a b p A l 0 0 3 , N e w a r k#22r803)55 - 1 1-posit ion, 4-section, 4-pole rotarys w i t c h ( C e n t r a l a b p A I 0 1 5 , N e w a r k# 2 2 F 8 1 5 )56 - Miniature SPDT switchS7 - Minature SPST switchM1 - 1-mA meter movement, preferablywith 0-1 and 0-3 scates (MURA pM-702)Misc - case; power supply and input atten_uator enclosures; l ine cord; BNC jacks;knobs; shielded cable, pWB terminals; Dlpsockets, mounting hardware, etc. A set o1three etched, dri l led and solder-plated circuit

boards (CP1 , CP2 and Cp3) is avai lable forS35.20 pos t -pa id ( in cont inenta l U .S.A. )f rom Ci rcu i t -Works , 11 tB 7 th Ave, Nep_t u n e , N . J . 0 7 7 5 3 . N e w J e r s e y r e s i d e n t smust add S-percent sales tax. please al low3 to 4 weeks for del ivery.

Al l resistors are 74-watt, s-percent carbon_fi lm unless otherwise specif ied. Those speci_f ied as "1 percent" are r. /4-watt metal-f i lmtypes. Substi tut ion of s-percent resistors fortne 1 -percent type wil l only degrade accuracyol the attenuators.R 1 , R 2 , R 3 , R 4 , R 5 9 , R 6 2 , R 1 9 2 - 5 k i t o h m

tnmpot (Panasonic K4A53)R 5 , R 1 9 , F s B , R 1 0 2 , F 1 1 6 , R l 4 0 , R l 4 1 ,

8 1 6 2 , R 1 6 3 , R 1 9 0 - 6 . 8 k i t o h mR 6 , R 1 1 , R 1 4 , R 1 5 , R ] 8 , R 2 2 , R 2 3 , R 2 5 ,

R 2 6 , R 2 7 , R 5 6 , R 6 0 , R 6 1 , R 6 3 , R 6 4 ,R 6 7 , R 7 1 , R 7 7 , R B O , R B 7 , R 9 6 , R 1 3 4 ,R 1 3 6 , R 1 4 3 , R 1 4 5 , R 1 4 7 , R 1 5 6 , R 1 5 8 ,R 1 6 5 , R 1 6 8 , R 1 7 8 , R 1 B B , R 1 9 5 , B 1 9 7 ,R 1 9 9 , R 2 0 B - 1 0 k i t o h m

R7, R66-1 50 k i tohmR B , R 7 2 , R 7 4 - 5 6 0 o h mR 9 , R 7 3 , R 7 5 - 3 3 0 o h mF 1 0 , R 6 B , R 7 6 - 2 7 0 o h mR 1 2 , R 6 9 , R 7 B , R 1 3 7 , R 2 0 4 - 1 5 k i t o h mR 1 3 , R 7 0 , R 7 9 - 7 5 0 k i t o h n rR 1 6 , R 3 3 , R 5 0 , R B 6 , R l 0 5 , R 1 1 9 , R 1 2 0 ,

R 1 2 3 , R 1 4 2 , R 1 4 6 , R 1 6 4 , R 1 7 1 , R 1 7 4 _1 0 0 k i l o h m

R 1 7 , R 3 1 , R B 2 , F 1 2 8 , R 1 4 9 , R 1 5 0 , F 1 8 3 ,R 1 8 5 , R 1 B 6 , R 1 B 7 - 1 k i t o h m

R20, R35-1 3 k i lohmR 2 1 , R 4 1 , R 4 2 , R 4 3 , R 5 3 , R 1 0 3 , R 1 1 7 _ 1

megohmR24, R135, R157-1 k i tohm t r impot (pana_

sonrc K4A 1 3)R 2 B , R 5 7 , R 1 2 6 , F 1 3 3 , R 1 3 8 , R 1 3 9 , R 1 5 5 ,

R 1 6 0 , R 1 6 1 , R 2 0 6 - 2 2 k i t o h mR29-5 .6 k i lohmR30-2 5 ki lohm potentiometerR32, R84-2 2 k i lohmR34-2 0 k i lohmR36-1 8 k i lohmR37-820 ohmR38-620 ohmR39, R40, R213, R21 4-220 ohmR44-68 1 ki lohm, i percentR45-42 2 ki lohm, 1 percent846-90 9 ki lohm, 1 percentR47-1 1 k i tohm, lpercentR4B-1 00 ki lohm, 1 percentR49-3. 1 6 ki lohm, 1 percentR51, R176-1 k i lohm, 1 percentR52-2 .05 k i lohm, 1 percentR54, R65-82 k i tohmR55, RB5-1 B k i lohmR 8 1 , 8 1 9 6 , R 1 9 B - 3 3 k i t o h mR B 3 , F g B , F 9 9 , R 1 0 1 , R 1 1 2 , R 1 1 3 , R ] 1 5 ,

R 1 2 9 , R 1 3 0 , F 1 3 1 , R l 3 2 , R l 4 8 , F ] 5 3 ,R l 5 4 , R 1 6 7 , R 1 7 2 , R 1 7 3 , R l 7 5 , F t B 1 ,R 1 8 2 , 8 2 0 7 - 1 0 0 o h m

RBB-1 .1 k i lohm, 1 percentRB9-1 0 k j lohm, 1 percentR90-7 5 ki lohm, 1 percentR91-2 15 k i lohm, I percentR92-750 ohm, 1 percentR93-316 ohm, 1 percentR94-6 .2 k i lohm

AUDIO/SEPTEMBER 1981 o l