Wireless World, April 1977 Stereo coder 1—Choice of method /oscillator stability by Trevor Brook A practical design for a high quality coder suitable as a test instrument described. Apart from the audio filtering, inductors have been avoided and a compact board layout produced. Avis. unit for servicing checks on ‘eiver performance could also be lused by demonstration showrooms ‘0 food programmes of their own choice Part 1 examines the steroo multi plex system and establishes tolerance limits for signal components. Channel separation is considered as this would ‘assume increased importance if a matrix. system of surround sound brosdeasting were adopted. Part 2 gives construction and alignment details for the coder and Part 3 gives, modifications to the Portus and Haywood decoder to provide a low distortion reference decoder Work on this coder started originally out of curiosity’ as to whether an Inductorless design would be possible. Early experiments were promising and the design has been pursued to give performance of broadcast quality "The specifications of steren caders now in use at both national and independent local radio transmitters are given in Table 1 and most existing coders have similar figures. Particular objects of this design are to improve crosstalk at the higher audio frequencies and achieve mid-frequency distorion bet: ter than 0.05% Stereo signal specification The modulating signal in the Zenith-GE pilot tone system is defined as (ate, Ace 2a 09 sin2oft +0.1sin where A is pre-emphasized left channel, Bis the pre-emphasized right channel, and f, is 38H7, "(A +B) is called the Sunt or M signal, and S34 ~ B)sin2=f.¢ Is called the stereo difference or S Signal, and is a double sideband suppressed-carrier signal. O.sinaf,t is the pifot signal at I9kH Substituting the masinum values Az+1 and B= +1 or =I gives the maximum amplitude of 99% for the M. and S signals respectively. Monophonic receivers continue to produce only the IM signal as audible output thus giving the system ics compatibility. Decoding To retrieve the stereo information involves a decoder which can take the form in Fig. 1 The reduction in channel separation if a decoder adjusted to Fig. 1. How the stereo multiplex signal can be decoded. After de-emphasis the ‘A’ and B’ outputs become left and right ‘channels. 15 20 ero ” Gecode @ perfect multiplex signal is prevented with signals having the five following departures from ideal is shown in Fig 2: amplitude error between the Mand S| signals “phase error between the M and S signals phase error in the pilot relative to the 38kHz suppressed carrier. The require- ‘ment for pilot phase accuracy Is sub: stantially Tess than for M/S. phase amplitude error of one sideband only Of the $ signal, typical of the If. loss Fig. 2 Inherent crosstalk of the multiplex signal plotted against 4) Amplitude errar between the M and S signals 2) Phase error between the Mand § Signals 3) Error in pilot phase 4) Amplitude imbalance between the sidebands of the S signal 5) Phase shift in a sideband of the S signal Q ws orase error ccrosstack (v#) Stthuoe ence0 ‘Table 1. Some parameters of the broadcast coders in use today. BBC Ba, ‘Amplitude response +0548 ‘* 1aB ‘Channel separation Harmonic distortion (iki) 38kHz leakage encountered in receivers “phase error of one sideband only of the signal Forreasonably high channel separation, say better than 43dB, the above effects may be considered as algebraically additive. It is evident that extremely stringent amplitude and phase performance requirements are set for @ coder intended to give high channel separation. ‘Another problem in the multiplex system is distortion. Apart from the usual harmonic and intermodulation distortions. spurious beat tones can be produced in the decoded outputs. This is the result of intermodulation between the various components of the stereo signal and, though predominantly a receiver problem, could also be caused in the signal generation method or in a coder’s output amplifier. Beat tone distortion is worst at the higher audio frequencies and subjectively: produces an unpleasant ‘splashing’ sound on sibiliants, Most stereo receivers. will sive clearly audibie low- or mid-fre quency beats on the 10 and TkHz bbursts during the BBC stereo test zone transmissions even though these tones, are not at full level. On mono reception of @ stereo signal the effect is no: noticeable except on very poor receivers. Fig 3 shows the beat tone possibilities in both mono amd stereo reception, Generating multiplex signals ‘There are two principal Ways of pro- ducing the coded stereo signal. The first and almost universal are switching ‘methods, shile the second is the matrix method where the individual signal components. are generated separately’ ‘and then added together. Conceptually the simplest and also a common way of switch encoding is to ‘switch between the A and B signals with a diode ring or similar device driven by 38kHiz, Fig 4. For a square wave switching ‘signal the following output is produced ASB AB 4 Fe AE Sainzeh Hanser siningee of ‘The snag is that sidebands around odd harmonies of rhe switching frequeney are present in the output and, more difficult, the required difference signal 60H2 — LoKit2 40H2 — 15kH2 - oor — 10kH2 = 408 10kH2 — 15kH2=36dB 0.3% at 248 above peak level 40d 50H2 — 15kH2. 400H2 — SkHe 4848 30H2 ~ 15kH2=424B 0.5% at peak level =40dB @ Lysarmaunisn wy ap ane (0) \ on reaieney (we) Fig. 3. Principal possibilities for production of beat rone distortion in ‘any receiver (h} whenever @ composite stereo signal (a) undergoes intermodulation distortion. Additional tones are produced in a stereo receiver rs) hhas too high an amplitude. To remove the components above S3kHz as well as reduce the S amplitude by'=/4 requires a filter with the amplitude characteristic shown in Fig 5, and a linear phase response at frequencies below 53kHz! Broadcast quality coders typically use tenor so inductors in such a filter. Other switching arrangements can avoid the necessity for S amplitude correction and in other respects switching coder performance is largely determined by switching time and 1:1 accuracy of the 3akHz square wave. Whatever low-pass filter is required practical realisations need careful alignment involving sever al interdependent adjustments and Wireless World, ari 1977 aethe Fig. 4, Basic arrangement for the switch encoding method. Switel would typically be a ring of diades or fet switch driven by the S82 square Fig. 5, Desired charecteristic for the filter in the simple switch coder, Fig. 6 A linear phase response is required up fo 53k E12 Fig, 6 Block diagram for the matrix coding method. doficiences in the filter make such coders susceptible to all the forms of signal degradation listed earlier. A complete switching coder design has ‘been published by Mack! and the virtues of diode cross modulator circuits for applications including coding stereo extolled by the same author? ‘The matrix form of coding is shown in Fig 6. A point to nove is that the 38kHz fed to the multiplier in this case isa sine wave, Another altemative would be to use a conventional switching multiplier Fig, 7. Complete block diagram of the coder deseribed.Wireless World, Apel 1977 Fig. 8, Thermistor-contvolled oscillator with the R33 bead running at 205°C, and a resistance of 82 ohms. fed by a square wave, filter out the difference signal from the odd harmonic components, and then feed it to the adder. With the matrix arrangement in the diagram the last three degradations can be made neghgibie xo crosstalk performance mainly rests on achieving ‘200d gain stability and phase matching between the surn and difference signal paths However, some new problems arise ‘with this form of coder. The linearity of the multiplier to audio frequencies will hrave an effect om beat tone distortion performance and in a practical design there is the danger of impurities on the 38KH2, producing further beat tone ‘outputs, Common to all forms of coder is the need for a low distortion 19KH2 pilot of correct phase and good stability land audio pre-emphasis and filtering to limit the bandwidth of A and B signals to I5kHe. Choice of matrix method ‘With an instinctive loathing of induc- tors and poor prospects at the time of realising a sensible active filter meeting the stringent phase and amplitude requirements while introducing negli- gible noise and distortion, the matrix Approach looked more promising. Using the matrix prineiple only the first two signal degradations should be apparent and to meet a target channel separation ‘of 55dB implies an M/S amplitude error of less than 0.18% and an M/S phase terror better than 0.1 ‘The block diagram of the coder is, shown in Fig 7. Both 19 and 38kHz sine waves are required in this coder and starting with a I9KFL sine wave which fs then doubled wsing a linear multiplier to square its input (a sine-wave squared ‘equals a single wave of double frequency plus a de. term) and produce 38KHz of, Correct phase involves less filtering than starting with a square wave at a higher frequency (38, 76 or 152KHz) and dividing down, If the coder is to be fed from any practical signal source other than a Gistortionless audio signal generator then the 15kH2 filters are essential to Fig. 9. Oscillator with improved ‘amplitude/temperature drift. Tr, and ‘Tr; form a long:tailed pair comparator ‘ond Ry Ry equalize the signal voltages ‘acrois the fet. to linearize it and reduce distortion of the sine wave ‘output. prevent gross beat tone effects due to Ultrasonic components on the audio inputs. It is desirable that rejection of frequencies of 19kHz and above be at least 45dB so this means that when pre-emphasis is in use. giving around 20B boost at 19kMz, filter attenua: tion of 65dB is required by 19KHz. If a passband ripple of only 14B is allowed this implies the enormous attenuation rate of 200dB/octave between 15 and okt, Fortunately two cascaded Toko filter blocks can exceed the requirements in a ‘ery small space and at reasonable cost. ‘The drawback of a filker with such violent attenuation so close to the passband is that there is little hope of achieving a linear phase characteristic and this is a deficiency common to all stereo coders. Part 2 includes a spec: trum analyser photograph of the pre- sent filter response and a graph of ‘measured phase shift The audia difference is derived and fed via the balanced modulator to the ‘output adder while the sum signal is produced by feeding equal amounts of A land B directly to the output adder. The longer route of the difference signal ‘means that itis slightly delayed at the ‘adder compared to the A and B com- ponents and at the higher audio fre. quencies this would amount to a significant phase shift between the M and S components, hence the phase correctors inserted in the A and B lines tothe adder. Because the linear doubler used to produce 38KHz is not a perfect device Some leakthrough of 19kHlz may occur, particularly at extremes of the Kemperature range, so a I9kHz rejector {is placed before the multiplier’s carrier Input. The pilot must also arrive at the output adder at the correct amplitude and phase and a small phase shift is Fequired to equal the time delay through the doubler, amplifier, notch fier and balanced modulator. To provide balanced outputs a straightfor: Ward unity gain invertor is fed froma the adder output, Power supplies are entirely conven. tional, producing plus and minus 15, volts at around 100mA. Power take off points are provided for running the clipping amplifier and vhf. oscillator ‘described later, 19kHz oscillator The accepted frequency tolerance for the pilot tone is +2Hz so crystal control, if not essential. is certainly desirable, "AS a sine wave 15 Fequired anyway it seems sensible to start with a sine wave crystal oscillator. This is something which ‘often gives circuit designers a problem but a reliablé induetorless circuit. is easily formed at low frequencies by building a Wien bridge oscillator around the correct frequency and then putting the crystal across the series Clement of she bridge. Easy starting with reliable crystal lock is the result. The standard thermistor amplitude ccantrol method, Fig 8, proved adequate for an early prototse, but even running, the bead as hot as permissible, 200 deg. still means that its operating point is, determined roughly 9 parts in 10 by the oscillation voltage and 1 part in 10 by the ambient temperature. With this circuit I measured an amplitude drift of ~0.02dB/deg, C over the range +10 to +40 deg C and distortion was below 0.05%, Evidently some form of amplified control was needed to iraprave this rift figure tenfold and allow maintenante of good channel separation over @ wide temperature range. The circuit evolved isin Fig. 9 where an flex. replaces the thermistor as the82 igsin control element but with lineariz. ing components to maintain the distor. tion performance, Linearizing is achieved by equalizing the gate/drain land gate/source signal voltages and is done by Rand R,, The fet. is also only allowed to contribute a small amount of the total resistance between pin 2 of the ie. and common, and this fraction is determined by Ry in the source lead. Linearizing produces distortion better than 0.05% compared with around 0.4% without, Transistors Tr, and Te, form a long tailed pair which compares the oscilla tor amplitude with a direct reference voltage. Resistor 8 prevents loading of the oscillator output by changes in Tr, input impedance over each eycle. The direct error voltage feeds the fet. gate after filtering (R-, Cy) to remove oscillator components. The two transis. tors are identical types and mounted together so that thelr two base-emitter junctions provide temperature compensation; the use of a ‘matched pair in a single can does not seem to be justified. Stability of the de. reference is assured by using low ‘temperature coefficient resistors for R and R,, a8 well as a stabilized negative line Though a square wave oscillator followed by a filter could have produced similar amplitude stability simply. by defining the voltage excursion of the square wave generator there isa unique advantage in the method described. Namely, the fong-tailed pair compara tor need not look at. the oscillator ‘output directly: it could look at the level lof S8KHz which feeds into the multiplier and thus act as a servo, taking up gain Arif in the doubler, amplifier and notch filter Printed boards (a total of three) are available for this encoder for £7.50 inclusive from M. R. Sagin, 23 Keyes Road, London NW2, Appendix, Inherent crosstalk arising from def. cieneies in the coded signal. Crosstalk is expressed relative to the full evel on the decoded channels when A=B=1 as this is the most convenient reference when making measurements, Amplitude error between the M and S signals Ignoring the pilot signal and consider- ing an error 8 s0 that the composite signal becomes AB, 2 *as5 ALB sin2f.t se, Sis low in level if 8 is positive, After multiplication in the decoder, considering only the 38kHz component of the reinserted carrier waveform ASB. ACB. po=|AP sindefs [sinaofe Adding 4i(A +) to give the decoded A signal, and consideriag only baseband components gives ASB | ACB a tae Related to peak level, 's8, fractional ‘erosstalk is8/22 +8), Phase error between M and § signals, Suppose AsO. -B=l and Bit)=sin2efyt.tf a delay of At exists on the signal. the composite signal is, sin2sfat After decoding, adding (A+B), Le. ‘asindef hs and neglecting non-baseband components, the decoded A signal +) CCRT AY oe TT AT , ie ~ | = i ) Wireless World, April 1977 =sinafbt, so that fractional crosstalle is sinaf St. Error in pilot phase. ‘Suppose pilot is 0. 1sinf,¢ +1), sa in the decoder the regenerated d8kH2 Is sindaf.(t4+81), DO)= \AtB AB ' AOE ASB sinners inacf(r+ in. Add (A+B) and neglecting rnon-baseband terms the decoded A signal is 4 B 1+ cos2af.t)+ 1 -cos2af. x 13+ (0 Bt). _and fractional crosstalk is sin’=f.5t Amplitude imbalance sidebands of the S signal I A=0, B=1 and B(t)=sin2nfgt and a sideband imbalance exists then the ‘composite signal is between the sin2sj z ut _ 082 cosda(l + ft ars Al Considering only baseband terms B48 | was Adding 4(A +B) the decoded A signal is D(t)= ~sin2aft: 5 | a+ | and fractional crosstalk is thegefore 8 Tay Phase shift in the upper sideband of the S signal Taking equation AL but for a phase error inthe (, +f) component, signal is Fadl sin2ofyt sindefat cos ) p SOSPHE AG MCE 4 Do: |sinasfot cosanche+ fabe References 1. Z, Mack, Stereo service generator (in German). Circuit of a switehing type coder is given. Funk-Technile 1968 p.532, 2. Z, Mack, Comparison of transfor merless ring-modulators and cross modulators. Radio and Electronic Engineer vol. 44 1974 p07Wireless World, June 1977 Broadcast stereo coder 2 — Circuit description and construction by Trevor Brook Surrey Electronics The complete coder is shown in Fig. 10. IC, and IC, provide regulated and short-circuit protected plus and minus 15-volt lines. The output voltage of these ics has reasonable temperature stability. which is desirable for the negative line, since it provides the reference for oscillator amplitude. ‘Though short-circuit protected. the regulators cannot withstand reverse polarity at their outputs, so Dyg and Dy> prevent damage, should the two supplies be inadvertently shorted together. The 19 kHa sine-wave oscillator described in part 1. IC, has one addition, the chain of diodes Dy. 1. across the output. There is the chance that, when starting, the oscillator ‘output could hit the supply rails and thus go beyond the linear region of the multiplier, IC,, When the multiplier is overdriven its output, instead of rising further, distorts and begins 10 fall ‘which means that the comparator no longer receives an input in proportion to the oscillator amplitude and the oscilla tor stays locked into a condition where it oscillates atthe supply clipping point. Diodes 11 to 14 clip the oscillations below the multipliers serious non-lin- carity level without affecting the oscil lator distortion when running normally at the designed output of | volt rms. Multiplier IC, has its X+ and Y+ inputs tied together, so that it acts as a linear frequency doubler with Ray providing trimming of 19 kHz feed through rejection, The rejection figure obtainable worsens as the multiplie’s maximum permissible input swing is approached. hence the reason for driving at 1 vor Tne loss occurring in the multiplier is recovered by IC, and, since it must provide over 30dB gain. a wide band: width op-amp is used, a 531, A748 can just about manage the job but it introduces a significant temperature: dependent phase shitt, a very undesira ble characteristic in this part of the cireut. Noteh filter 1g has virtually: unity gain at 38 kHz’ and is within the capabilities of a 748. OF all the active notch arrangements T have tied, the ‘Wien bridge seems the most repeatable. No very high impedances are involved, the loss at double notch frequency is less than 0.2dB, the corresponding, phase shift is small and stable, and a ‘notch deeper than 30dB can be obtained at 19 kHz. Two adjustables set the time Constant of one bridge arm and the circuit Q and both are adjusted for the deepest notch. Perhaps IC, and its associated circuitry is alot of trouble to avoid a simple LC rejector; but cus- tom-wound inductors are also a lot of trouble, fave poor tolerance and the possibility of causing distortion if ferrite ‘cored, Capacitor 16 couples the 38 kHz into the balanced modulater and blocks the accumulated dc. offset. Though only a volt or so, itis unlikely to be tempera: ture stable so Ry establishes a stiff grounding for the multiplier. The value Of Cis chosen with Rj, to cause small phase shift, yet provide some welcome roll off at low frequencies, since the 531 isa disgustingly noisy little animal. The comparator sensing point is also taken from here, again with no worries about superimposed dc. Left-channel audio passes through Ry and Cy, where it receives pre-em: phasis of 50s. Capacitor 17 may be ‘omitted for a flat frequency response or alink could replace Ry oz éhe board and Ry be placed by @ switch bank with various capacitors to give a choice of pre-emphasis. A straightforward audio amplifier IC; drive the first filter section through its correct source impedance, Rys, The filter is terminated by’ Ryg and feeds into a compound emitter follower, ‘Try Try. single-transistor emitter fol lowers €ause too much distortion, even at signal levels below 1 volt as’ here Resistor 48 is the source impedance for Fig which Is terminated by Ry. At rangements on the right channel are identical apart from F's terminating, resistor which is split between a preset, Ry. anda fixed resistor. These filters are normally intended for use as a stereo pair, Bt on an experimental coder there appeared a surprisingly large phase shift berween the M and S signals as 15 kHz was approached. This turned out to be due co crosstalk (at ~GvdB) between the two halves of the filter which produced a spurious signal of different phase on the ‘silent’ channel. The cure adopted here is to feed each channel back through the second half of its ‘original filter block and keep the left ‘and right channel blocks well apart. ‘The A and B signals emerging from Fig and Fg are fed via their phase shifting networks, Ry. Cay Cop and Ros Cry, Cop 10 the output adder IC,,. The different values for Cys and Cy is explained by different paths through the differencing amplifier and differ- fence in circuit board capacity for the ‘wo channels. The differencing amplifier, IC, uses a 148 rather than a 74, since less phase shift is introduced at the higher audio frequencies and the change with tem- perature of the remaining phase shift is Tower. The second drawback of the multipliers used here is that they produce a small amount of second harmonic distortion and, though this is immaterial in the doubler configura- tion, it is relevant when using the balanced modulator configuration. Such distortion on the audio port will produce second harmonic distortion for difference signals below 75 kHz and beat tone distortion for frequencies between 7.5 and 15 kHz, On the 38 kHz port the effect will be to give an output, ‘with associated sidebands, at 76 kHz. Like feedthrough, these effects worsen fs the multiplier is driven harder and here the carrier level, and audio level for a full difference signal, are set 6dB below the multiplier’s’ non-linearity point. The audio takes precedence and foes to the X port, which has the better linearity specification. The ebjection to driving the balanced modulator at even lower levels is that noise would become obtrusive. The double-sideband, Ssuppressed-carrier difference signal from IC is fed to the adder at the correct evel via R, ‘The gain of 15dB required from IC. the output adder, for the S signal, is possible from a 748 and the noise level of these deviees is also good enough for this position. The signal components may be switched individually by the Gill. switch mounted on the board, S,.‘Wireless World, June 1977 oI and Ry is present to stop the 748 going = unstable should all the switches be tumed off. balanced output is pro- duced by IC, which is a unity gain inverter, ‘The sine-wave pilot signal is taken directly from the oscillator output, passed through a trimmable phase shift network Cp, Rig, Rie and then atten- uated suitably By Rio, Ryg before reach- ing the adder. Mono/Stereo switching is achieved by a reed relay mounted on the board immediately by the adder, which disconnects the pilot and S signal. The reverse diode and capacitor around the relay coil completely remove any click due to the switeh but some click remains as the reed contacts make or break. There is no de. offset being switehed and no capacitor charging as the contacts close and the click only ‘oceurs ifthe pilot is switched on at the iLL switch, The reason is that the 19 kHz sine wave is belng interrupted instantaneously: another way of think ing of it is 100% amplitude modalation, ‘and a continuum of sideband energy will extend from dc. to infinity. The peak level of the click at the coder output viewed on a 'scope with a 15 kHz filter and no de-emphasis is ~30dB, Some coders leave the S signal on when fm the mono mode but it is no trouble here to remove it and it seems good practice todo soif stereo performance is ‘not compromised, ‘The little arrangement around the red: and green led.sallows mono and stereo indicators to operate along with the reed relay, while only using a single: pole switch contact, which closes for stereo, This allows for easy remote switehing. The green Le. passes full relay current in stereo and the red Le, draws a small current in mono which is insufficient to hold the relay in, yet subjectively gives the same brighiness because of the greater efficiency of red leds. Construction ‘To achieve a compact layout, as well as to avoid links and Keep signal tracks short in the interests of reducing crosstalk within the coder, the pc. board has to be double sided, The whole coder. including its power supplies and mains transformer, is accommodated fon a board 165 by 165mm. To avoid hum pickup it is essential for the board- mounted mains transformer to be magnetically shielded, Though the ‘board track layout is designed 10 avoid ‘ground loops. many Les have built-in {oops which make them susceptible to hhum induction when in a magnetic field in the same plane as the ic. chip. This applies particularly vo the multipliers and regulators used here, and a cylin. Grical Mumetal can for the mains transformer provides over 30dB reduc- tion in its hum field, a more than adequate margin. The heatsinks pro: vided for the regulators run hardly ‘arm co the touch and only reach 30°C 10. Cirewit diagram of complete ‘should be taker to be the line marked ‘coder. (The junction of Cjy and Rye B) Fig2 above ambient under supply overload conditions. However, their sides provide convenient points for glueing down the ‘ange smoothing capacitors to prevent them from vibrating and thelr leads fracturing under severe mechanical shock. Clear Bostik | is suitable for the purpose. ‘All the trimmers are visible-setting, single-rotation types. None of them is doing more than providing a very fine trimming adjustment, so muhitura types are not justified. In -addition, being able to see the position of a preset ig extremely useful as an unusual setting frequently leads to discovery of ‘an incipient fault. Resistors which have a bearing on ain, phase or important time constants are 2%, with thick film types being preferred for the lower values where they are available, since they have a lower temperature coefficient (£100ppm) than the 2% metal oxide types (+= 250ppm). Similar comments apply to capacitors where 1% silver mica types are used for the notch filter and pilot phase corrector with a low temperature-coefficient. polycarbonate ‘ype for Cie Stripboard construction is ‘ot likely to be successful, but printed circuit boards are available from the address at the end, Ground tracks: radiate along the board from the output adder and there are in addition several, apparently redundant, ground tracks forming ground guards to reduce board leakage and intertrack capacity, The positioning of circuit sections on the board also contributes to minimal 19 or 38 kHz pickup along the audio paths ot by the output amplifier. The long-tall pair comparator transistors in the oscillator are mounted together and a drop of glue between them will do no harm, While the difference signal is ata fairly high impedance, the capacity of its ine has to be kept low to avoid loss for phase shift of the upper sideband and thisisdone by IC, being directly next to the adder. ‘The board pins connecting the plus and minus 15V lines through the board to their distribution tracks across the top can be omitted until correct func tioning of the power supplies has been checked, To simplify initial checking it {sa good idea to omit the pre-emphasis, capacitors as well, Cy, Ca, so the coder can be set up with a flat frequency response, Parts list 18k Re 38K So 330 33K fs Haw sae 470 Re 33k cn jor +5% Wein team Sopot wk Therm S0 Ye Teks2% Be 100K 2% Cy 500e, Baza" Re B70 3 see 22k Ba A70k2% Cy 3.38 330k22% es B.2k Ce Ip 26% arate Ben Roksan 470 47022% ae a Gu 200 me 33 33 Bs 2a ae Gai SRM aS 8 ee 33k RO Bt, 14001 3.3K ® i N4001 100k+2% RE ed Red lod, 150k. 2% Rt Dn Greene. 470 RS, on 62V' zener 15k2% Be = (ask ; Bt ae 47k 2% Ty C309 100K * 196 to hee Te 8C239C_ é Aine ie 531 Wireless World, June 1977 Printed eireuit boards, Auset of p.cbs comprising one double-sided board, which measires 6 X Blin, and two smaller single-sided boards is available at £7.50 inclusive from MR. Sagin at 22 Keyes Road, London N.W2. 2X 19kie exystal, RE 19U Surrey Eletrones, The Forge. tucks Green, Canlegh. Suray) Mains transformer Surey Electonics £5, Fa, GLR2O1N titer Harrogate Radio Lr, 23° Grove, Harrogate, W. York Heat sinks. Redpoint 1v3 Eleerovalye, 26 St tiudo's" "Road. Engieheld Green,” Egham Suey Relay, dil. switch, trimmers sd trimmer Capatvers ean” 'be° gbtsined. om ‘Dora Electonics PO Box TAS. Wellington Rong Industral Estate, Welington Bradge, Leeds 12 The next article will describe the alignment of the decoder.__)ireless World, October 1977 Broadcast stereo coder 3—Setting up By Trevor Brook, Surrey Electronics In this setting-up procedure 0dB level refers to 0.775, —With the coder in mono, set A and B gains, by means of Ryy, Ri, $0 that (4B input at IkHz gives —TdB at the output. Check that the amplitude response is +0.5, —1.0dB from 20Hz, to 15kHz relative to the IkHzlevel for each channel. For measurements near 15kHz, a fre. quency counter will prove most useful if the audio signal generator calibrations are not accurate. —With a grounded crocodile lead on the “oscillator defeat” pin. IC pin &, check the distortion for each channel with OdB output at IkHz. A reading of better than 0.039% will confirm that all is well —To align the 38kHz path, set presets Ra, Ry and Re; to mid-position and remove the “oscillator defeat” ink. —Looking at the output of IC;, pin 6 on ‘an oscilloscope, adjust Ry, for a rough full in 19kHz content on the 38kHz, waveform, Connect a nulling distortion meter, tunable to 38KHz, to pin 4 of IC (Many distortion meters only cover up to 20kHz, but generally they are easily modified by soldering an extra parallel resistor in each arm of the rhull bridge 80 that the upper fre quency becomes 40kiiz. For the job here accuracy is not very important; all that is required is good rejection of the 38kHz so that the remsining 19kHiz component can be nulled.) Looking at the distortion meter output on a scope adjust Ryy and Ry alternately to achieve the best rejec” tion of 19kHz, Final trimming of Ray as well should leave no 19KHz visible amongst the noise, and better than 60dB below the 35KHz level —The 38kHz amplitude at this same point may now also be checked as +848 20.548, —With the oscillator system now set up properly the distortion of the I9kH2. atpin 6 of IC, can be checked as below ols. —Switeh the coder to stereo and look at the 38kHz at the output, with only the ‘A practical design for a high quality coder suitable as a test instrument was described in the April & June inaues. Apart from the audio filtering, inductors have been avoided and a ‘compact board layout produced. A vhf. unit, for servicing checks on receiver performance, could also be used by demonstration showrooms to feed programmes of their own choice ‘to stereo tuners. Part 1 examined the stereo multi plex system and established tolerance limits for signal components. Channel separation was considered as this ‘would assume increased importance if a mateix system of surround sound broadcasting were adopted. Part 2 ‘gave construction details. and align- ‘ment details follow in this part. Part gives modifications to the Portus and Haywood decoder to provide a low distortion reference decoder. S switched on at the dil. switéh, ‘Adjust for minimum carrier with Ria. Using broadband metering the 38kFiz. pull will be masked by the residual TékHiz generated by ICie, which does not null out, Feed Iktiz at around 0 to +64 into the left channel and defeat the osellator. Still with only S switched on adjust Rey for @ null of audio leak through in IC. —Allow the oscillator to run and feed kHz at OdB into the let channel with A, B and § tured on at the dill ‘witch. Lock the "scope to the audio and adjust R, forthe roughly correct M/S amplitude relationship seen in Fig. 110). Repeat forthe right channel input but this time leave Ry, alone and adjust the B difference pot, Rey Switch th pilot on at the di. switch and, with no audio input, set its level to —21dB at the coder output, using Rs —Feeding IkHz at around —10dB into either left or right channels, turn on only the Sand pilotat thedill switch ‘Locking the ‘scope to the audio should display an “eye” pattern, as in Fig. 11(b). The correct pilot phase is when the eye appears symmetrical And this is more easily seen with some Yertical magnifietion arranged 5 shown in Fig. 11(@). Resistor Riy adjusts the pilot phase and the effect of & slighty incorrect setting is seen inFig. 1). Table 2: Measurements on prototype cod No pre-emphasis Frequency response +0.5 48, —1.0 68 Rejection of t9KHe Rejection of requercivs above 1 kHz Crosse at 20°C 20H2-15kte Crosstalk 10-40°C 20H. 1Bkite Residual 38H? Pilot phase accuracy Beat fone dstoruon, 1SeH fll M Tull Sor orRoverativen BoB 20H 1 15KHe 5808 5808 5508 2508 5008 + 0.1% Spurious esponses above S3KH», full M, ful § or L or R overdrven SAB: sicebands of 7kHe Carr and sideband at 76KI42 ‘aror and sidobands at 15242 6243 4688 esas Messurements using relerence decoder (part 4) and SOs de-emphas's harman distortion, Tei fll MS, Lor 0.08% 0, 20H to 15tHz, mean reading meter, unweighted 7968 7108Sa Fig. 1. Correctly setup coer with fll IkPt= A-signal and no pilot 1s seen at (2) which Indicates the lat'ero line The pilot phase ‘eye is at @), with only § end pilot, while (@) shows the zero crossing region of (2) magnified virtually by a factor 9f 100 ~(e)'s fhe same but with an incorrect pilot phase setting, Zero line ripple at (wih full TRB2 A signal, is obtained by X100 vertical gain and clipping amplifier shown in Fig. 12 Result on zero line npple of low S amplitude 'e at (g), while (h) shows too hit an S ‘amplitude in wrong phase. Photo (8 of composite multiplex signal sith pilot and Tull Ike signa. Zerotine pple at) stat ‘obtained for 15kHe with coder correctly set up. Spectrum analyser photo at () shows nolse spectrum when in stereo. mode bit With no audio, Pilot at [Oke and mat Spurious response ~ Taiz at ~48dB ean be Seen. Sterea spectrum with A overdriven by GAB with IH at () and with ISRFTS at). ‘Analyser measurements were performed by ‘Maroon instruments TF 2270, SOH= band width direct into S00 input via 33KO resistor, not sing high-impedance probe Pilot amplitude and phase adjust: ‘ments are very slightly interdepen dent, so repeat the last two adjust ments, Clipping amplifier Ifaltis well to this point, then channel separation will exceed 40éB at IkHz, Dut to see the M/S amplitude and phase error more easily for greater separa tions requires x 100 vertical magni cation compared with that in Fig. 11(a) Some ‘scopes may manage this without overloading, but most do not so a useful amplifier and clipper circuit is given in Fig. 12. This is simply a 20éB amplifier with diodes arranged to bring the gain below unity as soon as the output swing, exceeds 0.6 volts. The amplifier has quick recovery from the clipping so does not degrade the’ interesting zero voltage area of the stereo waveform. ‘The clipping amplifier can conven. Jently be built on a serap of Veroboara 40 Se. and placed inside @ metal 35mm film can. The output resistor stops rf. instability when driving capacitive loads in the clipping condition. Used in conjunction with a directly-coupled ‘scope giving 20dB gain (which should not cause overloads) and at least SMH bandwidth, the required x 100 magni- fication with low phase shift is achieved and a correct stereo waveform appears in Fig. 100. —Using the clipper arrangement repeat first two items in column 3, page 89, The only limitation to correct setting should be the noise along the zero line of the waveform. Figure 1(g) shows the in-phase zero-line ripple caused by low S signal amplitude corresponding to a lossin $ of 27%. ‘Now change the input frequency to I5kHz and adjust the M/S phase accuracy (Cy adjusts for the A and Cufor the B channel) Phase errors appear on the “scope Wireless World, October 182 geFig. 12 Amplifier with 2048 gain and clipping arrangement to allow large vertical magnifications of the zero-voltage region vathout overloading eslloscope ¥ amp display as a sine-wave zero-line ripple shifted in phase relative to the main pattern and Fig. 11(h) shows the ‘appearance of amplitude and phase errors combined —Check now that M/S accuracy is ‘maintained over the whole audio frequency range, Ian audio signal in antiphase and of precisely the same ievel is available the Initial adjustment can be improved upon, ‘Feed antiphase audio at 1kHz into left ‘and right channels at OdB and with the A and B signals only switched on make a very slight trimming adjust iment t0 either Ry oF Rye So that the output nulls. This gives a more accurate channel balance than set ting channel gains up on a millivole- meter, The audio leakthrough set on p.89 with the oscillator stopped can’ be adjusted under working conditions for the multiplier if a 15kHz low-pass filter is available, Connect to the coder output and with only the S signal turned on at the switch feed left or right input with 15kHz at +6dB (or feed both left and right with antiphase both at OdB). Potentiometer R,; is adjusted for a null in I5kH2 leakthrough, Correct setting of Ry is important, otherwise false settings in Ry, Cz and Cy, can bbe produced, Without a ‘spectrum analyser or decoder an estimate of the beat-tone distortion may be made with the aid of a distortion meter. Continue as above for the audio leakage check but switch on A, B and the pilot as well as S, Use the distortion meter to null the 15kHz and some of the beat tones will give a reading below 0.1%, (This reading is, only an indication that all may be well as no account has been taken of beat tones above 15kHz which will be heterodyned into the audio range in decoding. It a frequency counter is, available a check can be made that the pilot frequency is I9kHz + 2H.) Check temperature stability by feed- ing IkHz at 0dB into the left channel, with A, Band S turned on at the di. switch, Lock the "scope to the auc land with the x 100 vertical magnifi- cation arrangement view the change in relative M/S amplitude and phase. With a temperature rise from 20 to 40°C the S amplitude should fall by 08%, ie. 24mm on a pattern magni- fied to 6000mm. Allow at least half an hour for all components on the board to reach the new ambient temperature. What $ amplitude loss there is can be shown to be predominantly due to the balanced modulator i.c. by briefly holding a soldering iron on its case. With the ‘methods described here no phase error between the Mand S components should be visible over the whole tem- perature range +10to +45°C. Inciden- tally it is quite impossible to align a coder for channel separation by using a ‘decoder, as apparently good separation can be achieved on a particular decoder ‘with quite the wrong phase and ampli- tude settings. ‘Three checks on performance can be made using a suitable reference decoder, such as the modified Portus and Haywood design described in Part 4. Using 50ys de-emphasis the noise level referred to Ikt1z full level (—1aB it the coder output) should be =—70dB, unweighted, mean reading meter, 20Hz to. 15kHz. Again with ‘de-emphasis, readings of coder-decoder harmonic distortion for IkHz full A, B, Mor S should be 0.08% and the 15kHz ‘eat tone under the same conditions 0.35%. ‘Some of the distortion above is contributed by the decoder and the only satisfactory way of assessing the purity of the coder output is by spectrum analysis. Figure 11g) shows the coder noise spectrum when switched into stereo, The 19kHz pilot tore is at ~214B and the slight mark at 38kHz is the suppressed 38kHz carrier at —714B. The spurious 76kHz double frequency ‘output from the balanced modulator is at 48d, Figure 11(k) shows IkHz in left or right channels overdriven by 6B. The baseband signal is at —1dB, normally only reached for full M signal, i. full A and B in phase. After the pilot are the two S signal sidebands at_—7dB, normally only reached for full § signal i.e full A and Bin antiphase. Above this are two spurious responses, sidebands of S7ktiz and the 76kHz signal again. For IkHz, the S7kHz components are harmless, but for higher audio fre: ‘quencies the lower sideband of the pair falls into the $ signal band. On this photo itis also interesting to notice the slight noise modulation effect (about 444B) which only becomes visible when a the S signal is within 34B or so of full amplitude. Figure 11() shows the situation as above (6aB left or right overdrive) but with I5kHz input. Apart from the lower sideband of S7kHz, 42kHz at —64dB, other minor beat tones are visible at 4kHz, —674B, and 7kHz, —684B, The line at 27kHz seems to have been a noise peak, since it bears no obvious arith- metical relationship with the fre- quencies involved and does not appear Inother photographs taken at the time. ‘The 42kHiz component will demodulate to 4kH2 at —634B in the left and right channels and this would indicate a beat tone figure for the coder of 0.1%, and With 50hs de-emphasis 0.07%. Not covered on the photographs, the only component observed above 1OOkHz was 152kHz and associated sidebands at —84B, For decoder and receiver measurements the 76kHz out- puts are not troublesome — the pre- sence of odd harmonics would have been more worrying — but for some purposes the use of a precision multi- plier might be desirable. To be concluded, Correction. In the circuit diagram (on page 76, June issue) capacitors Cys and Gy should have been shown earthed, rather than returned to the —15V rail and Cj shown variable, The G lead should have Ryy inserted, and the junction of Cjg and Ryp should connect tolead B. Resistor Rys should be taken to theupper end of Rs, and not Try emitter, which itself should connect to Tr, collector through a 33uF capacitor. Emitter of Tr, should have Ry, connect- ing it to the —18V rail. Capacitor Ca should be short-circuited. In the com Ponents lst Ry is 47KO, Ry, is 3.3k0, and Gyis 47uF 1% and not 47pF. Resistor Rycan be 2%,WIRELESS WORLD, MARCH 1978 Broadcast stereo coder Three decoders assessed, a reference decoder circuit, filters, and a v.h.f. oscillator by Trevor ook, Surrey Electronics This article conclu the series on the high-quality ste7e0 coder design with a fow-sisorson decoder circuit. Perforrnance details of the coder. assessed using this decoder, were given in the October issue NEED FOR A REFERENCE DECODER for per- formance checks on the coder prompted fan investigation of some commonly available types of decoder. Some de- coders produce their best channel separation from a degraded multiplex signal, such as is likely to emerge from the demodulator of present receivers, and the crosstalk measured in Table 3 ‘using an ideal signal is given as a guide to what to expect when testing decoders {ed directly from a coder. The setting of the free-running frequency of the phase-locked loop ic. decoders can also have a considerable effect on channel separation and the best readings obtained are given in Table 3. The 1310, used was the best of seven selected for low mono distortion. All were very similar in stereo but two of the seven ave mono distortion readings of 045% n one of their outputs. The use of a low-pass filter pre ‘ceeding the decoder is bound to reduce channel separation if it does not have a linear phase characteristic and low amplitude ripple and this effect can be seen in the Skingley and Thompson circuit (WW May 1974 page 124) Though a sacrifice in channel separa tion results, such simple filtering does achieve its purpose of dramatically reducing "birdy" interference from ad: Jacent stations, which otherwise is sub- jectively far more irritating, Twa odd eftects appeared when tes ting CA3080 decoders using the RCA, data sheet circuit. The decoder would ‘sip out of stereo if full level LSke2 Mt signal was fed into it and limiting of the ‘audio outputs accompanied by large ‘beat tones occurred with ful S signal for ISkHz audio, These effects are presum: Fig. 13. Response af the audio filers in the coder and their measured phase response. The filters are two Toko BLR-2OLL.N units, each consisting of a modified = arrangement. Over 6548 rejection is provided at igkHz and the ‘Apple below 15hitz is less than 14 ‘Table 3. Stereo decoder comparison when fed with ideal multiplex signal. wpe Distortion (%) Crosstalk oN mone stereo a thie Skit 1kHe —_r6KHe METSTO casio 30000900907 a. a7 13108 iter goo - 40 20 casoao 180017 G18 7L0rw 43 30 33a “S008 Ports & Horwara 600 008 seta - Pat odie p00 oes a shes cor anne wearin Sle roca Foran andtaysedsncoce wreck Saha att ably due to the 15kHz, ar lower side band of the S signal, confusing the 19kHz phase locked loop. Finally. tested was the Portus and Haywood decoder (WW Sept 1970). Needing principally lower harmonic and beat tone distortion, ¥ devised the following modifications, included in the circuit of Fig. 14 © Change Trt and Te12, formerly BC108 types, for 2N2369, ZTX313 oF any high-speed switching transistor. © Change Tri4 and Trl5 for high: gain audio types, BCIO9C. ZTX108C, ete, © Convert the input amplifier to a ‘compound emitter fotibwer, now with a lower emitter resistor and a gain potentiometer at the input. This can De done neatly on the original inte rex pe. board using only one link, ‘This modification is only suitable if the inpat amplifier is not required to provide any gain. © Operate the decoder with only 1.4V at TP2, the pilot level test point, not 15v, These modifications brought the Vidz distortion im stereo 10 DEW and, with the further suggestion by Mr Portus of fitting pull-up resistors Ry, Ris ‘onto the bases of Tria ané 7:15, gives the excellent figures in Table 3 with the only penalties a couple of dB lower audio output and higher switching waveform on the outputs. Low, frequency channel sepsration is easily improved by paralleling 1000F 10V electrolyties across C, and C,,, Though icelevant for normal listening, good separation is desirable when measuring the coder's noise level, ‘All decoders proved sensitive to'eup ply hum ond noise and filtering along the lines shown, Fig. 18, is needed to regucethenoise varput from ie. voltage regulators to allow signal-to-noise ‘measurements beyond 64dB or so, VHF oscitator A simple v.nf oscillator with a varicap arrangement which has low enough capacity along the multiplex path to avoid hf. loss is shown in Fig. 16, The oscillator coit is printed on the pe. board alongside a coupling link which gives roughly 70 ohms output imped- ance through Re, Coupling is low er:Mid Su i Try Be209 Tee. TTT 74 Icy Icy Tras setae, 270096 ough to avoid frequency jumping with various loads. This device is only in- tended for use on a fixed frequency and there is no varieap sensitivity or linearity correction, Calculation for this circuit suggests distortion at full devia tion of less than 0.5%, ur a fully tune. able generator with calibrated attenua- tor the coder could be fed into the wideband modulation input of the Sound Technology FM1000 signal generator, ‘On stereo it is Important for the de- ‘viation to be set correctly. Without an analyser or deviation meter the best way is to measure the pilot tone level before deemphasis when tuned toa BBC stereo station iransmitting silence ‘They tune to the frequency selected for the oscillator and adjust its deviation to produce the same voltage, All the BBC Stereo stations | ean receive have pilot deviations within 1.34B of Wrotham Radio 3. The output from the oscillator at around 60mV is adequate to feed a passive distribution system or with Coaxial attenuators it can be used for receiver checking. Thirty decibels of attenuation (at 1.9mV) will still keep any reasonable fm. receiver in full A Fig. 14. Modifications to the Portus ‘and Haywood decoder to improve both distortion and channel separation, Faster switching times and high gain transistors in the matrix witha different input amplifier arrangement give IkHz distortion better than 0.04% Voltage levels of points A, B and C can be either +12, +6and OV or +6, 0.and ~6V respectively quieting on stereo while a further 68 attenuation (685mV) will quieten a ood tuner. Fig. 15, Stereo decoders proved susceptible to noise on the supply line ‘and filtering is needed to measure signal-to-noise ratios much above 0B. Regulator should be mounted ut of the transformer’s magnetic hum field. 2000.F capacitor should have low internal resistance. ¥neu yewono Mae #1 Rio + fev Dp eve The oscillator will run from either +12 or 415 volts so it can be run from the coder’s supply or tapped from the receiver under test. The capacitor types lused should be observed as they were chosen empirically to reduce the tem- perature drift. Wiring inside the box onto the p.c. board should use thin flexible wire with a slight slack left so that microphony is not transmitted from the input and output connectors ‘onto the board, ‘The phase and amplitude mangling of the § signal which occurs in most receivers is so large that degradation is, clearly visible on the demodulated ‘composite signal even without any ver tical magnification. Both low S amplitude and phase shift should be seen at I5kHz with $ amplitude loss being predominant for IkHz modula- tion, Oscilloscope synchronization will bbe helped by locking to the audio input to the coder or the deemphasized audio output from the receiver's active chan: nel 15kHz filter This is just @ convenient p.c. board which runs from 12 volts and will remove switching frequencies at de- coder outputs without introducing significant distortion, so allowing distortion and signal-to-noise measure ments, The resistor from pin6 to supply draws a small current to stop the crossover distortion which 741s other wise generate with only a 6-0-6V supply. To make distortion measurements be- low about 0.15% two such filters are needed to completely remove ultrasonic components 1 think the coder design presented here has reached a cost/performance plateau. Many of its identifiable de- ficiencies can be attributed to the balanced modulator i.c., and £80 or so spent on @ precision muitipiier will pro- vide some further improvements. The lack of induetors and single p.c. board make for a repeatable unit with stable performance. ‘The work described forms the basis of stereo coders for broadcast transmis sion, outside broadcast radio links and A Fig. 16. Circuit of a vhf. oscillator using a printed coil and providing a simply repeatable output level. Output voltage into 75 ohms is SSmV at 1OSMH2 and 65mV at 87.5MH2. Temperature stability over 20 to 57°C. at 964MHE is deHz / deg C. Deviation sensitivity at 104MHz relative to S8Mit2 ie + 5B, Fig. 17, The vhf. oscillator shown just fitting into the smallest diecast box ‘available (RS Components 509-923). Coaxial attenuators provide lower signal levels for receiver alignment. This series was written by ‘raver Brook, who te hoop aula forthe tie eng obo he iaxnt ten tng sath ot tedutng noun cette ops machines he hat cided to nproach manufacturers wth ft Iearng South London Gatoge ian Norwood ‘ech man va cominced tm tas ha cout ‘SurvyElecwonis fe youre ago with a apt of {raquoncyshiftre Mir interacts ae ot confined Sty cua nanan wt onl ne ‘dor, but never quit evereme the problem of ‘eeelver blanking with a good nie tigure, Y Fig. 18 Circuit of a convenient filter for removing ultrasonic signals when making decoder measurements, Distortion at +1148, 004%. Response =HdB at 19k, -45dB at 28hH2; ripple below 15kHz is less than 0.3dB. Crosstalle ~80dB at UkH¥=, ~S8dB at [5kli2, Noise ~96 co ~82dB over gain adjustmen¢ range of +4 to + 14a. av