up-to-date electronics for iab and leisure euEHTnr i equa amplifier quadro systems tap sensor steam whistle distortion meter mos clock Now you can change record-speeds without changing record-speeds. WeVe done away with the old turntable speed-control, on this very advanced Philips GA209 record deck. Simply by placing a record on the turntable the correct speed is electronically chosen and the pick-up lowered gently into the run-in groove. At the end of the record the turntable stops and the arm returns to the rest. This facility ensures that both the record and stylus are fully protected. In manual operation, the pick-up can be positioned over the grooves and lowered by means of a touch control. The mechanism permits very accurate positioning. Controlled by a servo motor via electronic touch controls, it can be operated whether the deck is used manually or as a fully automatic deck. Electronic control makes sure that the turntable speed is kept constant. Separate fine speed controls for 33'/» and 45 rpm. allow the record to be tuned to the pitch of any musical instrument. The photo-electric stop switch is completely soundless and frictionless. High stability and insulation against shocks and vibration are ensured by the floating suspension of the PHILIPS Simplyyearsahead turntable and pick-up arm. The tracking error of the practically frictionless pick-up arm is very small. Side thrust compensation is adjustable for all playing weights for both spherical and elliptical styli. The top cartridge from the Super M range, the GP412, is supplied as standard. Shown in manual position to illustrate control panel. Tekelec’s digital multimeters make everything "Liquid Crystal w A Tekelec A brand new portable from Telequipment The D32 Dual Trace 10 MHz Battery-Operated Oscilloscope Write now for full details and demonstration - you won’t be disappointed. TektronixU.K. Ltd., | ELEQUIPME nT Beaverton House, P.O. Box 69, Harpenden, Herts. Telephone: Harpenden 63141 Telex: 25559 Probably the smallest and least expensive ’scope of its kind in the world. Telequipmcnt’s D32 offers a generous performance specification yet remains in the realms of reality where price is concerned. Weighing 10 lb. and only 4x9x11 inches in size, the robustly built D32 can be carried comfortably on any assignment. Packed into its tiny frame is a specification with features normally associated with instruments twice its size. Priced at £250* (including rechargeable batteries) this dual-trace ’scope offers 10MHz bandwidth at iomV/div. sensitivity ; automatic selection of chopped or alternate modes ; automatic selection of TV line or frame displays ; the choice of battery or mains operation and a c.r.t. display covering a very large proportion of its total front panel art elektor decern be r 1974-5 BkeHTnr This is the first English edition of Elektor, a magazine that introduces a new way of presenting electronics. The Dutch edition of Elektor has been published for over 14 years and the German for over 4. Every month 120,000 copies find their way to readers ranging from enthusiastic amateurs to professional electronic engineers. Elektor's dynamic and practical application of new electronic techniques has stimulated the ever-present curiosity and imagination of designers. Modern components, active and passive and especially cheap digital and linear integrated circuits, are used in practical designs. Many of the circuits are developed in our own laboratories, and circuit building is greatly facilitated by using the ready-made printed circuit boards we produce for the more important designs. The availability of components is always considered, and when new components are needed every effort is made to ensure that they can be obtained through the normal retail outlets. On the continent, this practice has led to a modernisation of the retail trade so that now several retailers tend to base their stocks on the information in Elektor publications. This is very good for those firms of course, but it is even better for Elektor readers; it makes available for them a more comprehensive range of components at re- duced prices because of the greater demand. Elektor will not sell components, other than printed circuit boards, so that complete editorial independence is assured. Furthermore, the editorial staff cannot be influenced by advertisers, although it can sometimes influence them where it is important that certain components are made available to our readers. Elektor has always tried to be dynamic and informative; but it can occasionally irritate, as when it deflates technical imperiousness or indulges in a humorous self-criticism that has given it a 'British' image on the continent. In 1975, Elektor will appear every two months until August; from September on it will be published monthly. The July/August edition will be a large double issue. On the continent this has become known as the semiconductors guide, and its production is an established tradition. We shall be working on the first copies for 1975 even as you read this. Articles already accepted describe an electronically-compensated loudspeaker system, a high-quality pre-amplifier, an analogue-digital converter, gyrators, and further developments of the mos-clock, electronic drum and TAP. B. W. Van der Horst, editor. Editor Deputy editor J. Barendrecht T. Emmens T. Venema P. Appleyard C. Sinke L. Martin Mrs. A. van Meyel Advertising Art editor Drawing office Subscriptions Many elektor circuits are accompanied by printed circuit I designs. For those who are not inclined to etch their own printed circuit boards, a number of these designs are also available as ready-etched and predrilled boards. These boards can be ordered from our Canterbury office. Payment, including £ 0.1 5 p & p, must be in advance or by enclosed remittance. Bank account number: A/C No. 11014587, sorting code 40-16-11 Midland Bank Ltd., Canterbury. circuit number price distortion meter 1437 £ 1.50 tap sensor 1457 £ 0.55 equa amplifier 1499 £ 1.10 digital rev counter 1590 £ 0.90 mos clock 5314 clockcircuit 1607 A £ 1.10 mos clock 5314 display print 1607 B £ 0.80 aerial amplifier 1668 £ 0.85 ippsp ©pm pr'iiftl service context from din to equa-standards tup - tun - dug - dus larked TUP', TUN', 'DUG' or 'DUS', detriment to the performance of the sted. with tables of equivalent types. swinging inductor ands of circuits for transistor-amplifiers have been developed, all of w ter of hifi. The brands that meet the Equa-standards laid down in th. fingers of one - possibly two - hands. output power nomogram . divide by 1 to 10 mos clock 5314 The 'brain' in the digital clock described in this article is the clock-IC MMS314, which needs only a few external components. The time of day is indicated by seven-segment Ga-As displays. distortion meter The distortion in factory-produced or home-made amplifiers is frequently unknown; designers sometimes give specifications, but these are not always reliable. Since distortion meters are usually expensive. Elektor Laboratories have developed a simple, inexpensive, but effective instrument. quadro 1 - 2 - 3 - 4 .... or nothing? The phenomenon of 'quadrophony' has already been the subject of many publications, but the confusion only seems to increase with every new attempt to clarify the issue. This article may bring a little light into the darkness, by describing and comparing the most important systems that have been proposed so far. tunable aerial amplifier The aerial amplifier described in this article is characterized, among other things, by its low noise level (1-2 dB), a voltage gain of 10-20dB. and a wide tuning range (146-76 MHz). tap sensor An important alternative to the mechanical switch - rotating or push-button - is the touch switch. This has the advantages of greater reliability and a higher switching speed, as well as being noiseless and not subject to wear. flickering flame The simplest possible flasher device is a bimetal switch. This construction can be found in 'blinker bulbs' and in the starter-switch associated with a fluorescent lamp. The possibility immediately comes to mind of using a fluorescent-lamp starter as a flasher for Christmas-tree or he manufacturer has the resource his article explains that the do-it-yo le "electronic loudspeaker". loudspeaker diagnosis ctronic loudspeaker' in this issue, oi from din to equa- For more than ten years, manufacturers on the Continent have been measuring the quality of record-players, amplifiers, loudspeakers etc. against the West German industrial standard DIN 45500. Products that meet the published standards may be sold as "hifi according to DIN 45500". The editors of Elektor feel it is time to propose a more up-to-date norm. It seems reasonable to assume that ten years development of, for example, am- plifiers should lead not only to extensive miniaturisation but also to an improve- ment in the quality of components and circuits. This consideration gave the editors of Elektor the idea of checking these DIN standards against the present level of tech- nology. This in turn led to the formulation of the new quality standards that are now offered for discussion. The basis of the new Equa-standards is as follows: It must be possible in a sitting room to play back music with a recorded dynamic range of 40 dB which implies a required signal-to-noise ratio of at least SO dB and preferably 60 dB. The level of background noise in the sitting room is taken to be about 32 dB (sound pressure level); for headphone listening 20 dB SPL. The following standards can now be proposed: 1. The (minimum) output power of an amplifier, for use in a typical sitting room with typical loudspeakers, should be 10 watts; the equipment must be able to maintain this power level con- tinously for at least 10 minutes. This requirement is the same as the DIN standard. 2. The output power of an amplifier, driving the least sensitive headphones The requirements and designer's-aim values according to the Equa-standard, in comparison with the requirements laid down in DIN 45500. These standards apply to quality-amplifiers in- tended for use in domestic listening rooms. should be at least 0.2 watts; more sensitive units can however often manage with 1 milliwatt. 3. The signal-to-noise must be at least SO dB: one should aim at 60 dB. When a volume control is fitted, these re- quirements should also be met when this control is set at -20 dB. The DIN standard in this case specifies 50 dB or better. 4. The frequency response curve should be flat within 1.5 dB from 40 Hz to 16 kHz; in agreement with DIN. More- over, the curve must remain ‘smooth 1 outside these limits, although it may roll off gradually. 5. The peak amplitude (not RMS!) of harmonic distortion must be less than 0.3%; one should aim at 0.1%. DIN lays down a harmonic maximum dis- tortion of 1% RMS. This is a) too high and b) meaningless! (See the article “Equa-amplifier"). 6. The intermodulation distortion (measured as specified by DIN) should be less than 1% rather than the present 3%. 7. The stability must be unconditional, with any load. (The DIN standard says nothing about this.) 8. No damage may be caused (other than blown internal fuses!) by overdriving an input up to 20 dB (lOx) or by operating the output with a short or open circuit or with a reactive (in- cluding inductive) load. (This is not mentioned in the DIN standard.' 9. The crosstalk between different inputs must be at least 50 dB down from 100 Hz to 10 kHz; preferably 60 dB. The DIN requirement is 40 dB. 10. The suppression of crosstalk between a pair of stereo-channels must be at least 40 dB from 250 Hz to 10 kHz (DIN standard 30 dB). The table compares the requirements and designer's-aim values according to the Equa-standard with the DIN 45500 re- quirements. These standards were first presented by Elektor on the continent in 1972 as a starting-point for further discussion. It has since then become apparent that the usefulness of an IM distortion measure- ment (point 6) and the requirements for stereo crosstalk suppression (point 10) give rise to some queries. In addition, a need is felt for a relatively simple and precise measurement of tran- sient distortion and transient intermodu- lation distortion (slope overload, slew-rate limiting). H i-dugdus elektor december 1974 — tup- dug- dus B BBA Wherever possible in Elektor I HJBB I circuits, transistors and diodes are simply marked 'TUP', 'TUN', 'DUG' or 'DUS'. This indicates that a large group of similar devices Cc.n be used without detriment to ____ the performance of the circuit. In this article the minimum specifications for this group are listed, with tables of equivalent types. Also described are several simple measuring procedures that make it possible to find the connections and approximate performance of an unmarked device. As far as possible, the circuits in Elektor are designed so that they can be built with standard components that most retailers will have in stock. It is well-known that there are many general purpose diodes and low frequency transistors with different type numbers but very similar technical specifications. The difference between the various types is often little more than their shape. This family of semiconductors is referred to in the various articles by the following abbreviations: TUP = Transistor, Universal PNP, TUN = Transistor, Universal NPN, DUG = Diode, Universal Germanium, DUS = Diode, Universal Silicon. TUP, TUN, DUG and DUS have to meet certain minimum specifications - they are not just ‘any old transistor' or ‘any old germanium diode’ .... The minimum specifications are listed in tables la and lb. It is always possible, of course, to use a transistor with better specifications than those listed! Simple measurements It is advisable only to use semiconductors with a clearly legible type number, and with known specifications. However, transistors without a type number are often cheaper, and some simple tests can give an indication of their value. The first test serves to find out whether the transistor is a PNP or an NPN type. and to locate the base connection. A multimeter is used, switched to the lowest resistance scale. The plus lead of the meter is connected to one of the pins of the transistor (figure la). The minus lead is then touched to each of the other transistor pins in turn. If the meter shows a low resistance in both cases the transistor is probably a PNP type, and the plus lead from the meter is connected to its base. If the meter shows a low resistance at only one of the two remaining pins the transistor is probably an NPN type, and the minus lead from the meter is connected to its base. If the meter doesn’t show a low resistance in either case, the plus lead from the meter should be connected to one of the other two pins and the procedure repeated . Having located the base connection and the probable type (PNP or NPN), a double check can be made according to figure lb. For an NPN type, the minus lead from the meter is connected to the base and the plus lead is touched to each of the other connections in turn. The meter should show approximately the same (low) resistance value for both cases. After reversing the connections to the meter, the same test should show a very high resistance (little or no deflection) for both cases. For a PNP type, the first two measurements should show a high resistance and the second two should show a low resistance. The next step is to locate the emitter and collector connections. The multimeter is now switched to the highest resistance scale and the test leads are connected to the two remaining transistor pins (the base is not connected). If the transistor is an NPN type and the meter shows a very high resistance (figure lc), the minus lead is connected to the collector and the plus lead is connected to the emitter. On reversing the connections (figure Id) a relatively low resistance value should be indicated. If the transistor is a PNP type, the measurement results are reversed. If any of the tests show zero resistance between two pins of the transistor, there 125-260 240-500 450-900. NPN PNP BC 107 BC 177 BC 108 BC 178 BC 109 BC 179 45 V 45 V 25 V 20 V 20 V Veb 0 6 V 5 V 5 V 5 V 5 V 5 V l c 100 mA 100 mA 100 mA 100 mA 100 mA 50 mA P,„, 300 mW 300 mW 300 mW 300 mW 300 mW 300 mW 150 MHz 130 MHz 150 MHz 130 MHz 150 MHz 130 MHz F 10 dB 10 dB may 10 dB 10 dB is an internal short circuit in the transistor. It is then sometimes suitable as a diode, but usually can only be used as a very elegant kind of jumper wire .... It should be noted that in all the above tests the positive lead from the meter is the one connected to the terminal marked '+’. In practice the voltage on this terminal is negative with respect to the terminal marked when the multi- types (Vceo = 45 volts) and the BC109/ BC179 are low-noise. If these differences are not important in a particular circuit, the various types are interchangeable. The code letters A, B or C after the type number on these transistors denote various current amplification factors. For the A-types this is from 1 25 to 260, for the B-types it is 240 to 500 and for the C-types 450 to 900. A BC109C is therefore not a direct equivalent for a BC109B, for instance, although in many practical circuits it will make little or no difference. When using the equivalent types BC167, -168, -169, BC257, -258, -259 or BC467, -468, -469 it should be noted that the base, emitter and collector leads are in a different order (see table 6). meter is switched to resistance measure- ment, The measuring procedure is based on this polarity inversion. An indication of the current gain of the unknown transistor can be found in a similar way (figure 2). The multimeter is switched to the highest resistance scale, the plus lead is connected to the emitter and the minus lead lo the collector (if the transistor is an NPN type; otherwise the connections are reversed). If the previous tests were carried out correctly, the meter should show a fairly high resistance. The collector and base connections are now bridged with one finger, so that current flows via the skin resistance to the base of the transistor under test. The meter should now register a fairly low resistance. The higher the current gain (and the lower the skin resistance!) the lower the indicated resistance value will be. A comparative measurement with a transistor of known quality will give an indication of whether or not the ‘measured’ current gain was sufficient. Specifications and equivalents A number of transistor types that meet the TUN specifications are listed in table 2. This list is, of course, incomplete - there are far more possible types. Table 3 lists a number of possibilities for use as TUP, while table 4 gives equivalents for DUG and DUS. A further group of better quality tran- sistors are the BC107 - BC108 - BC109 (NPN) and BC177 - BC178 - BC179 (PNP) families. The minimum specifica- tions are listed in table 5, while table 6 gives a list of equivalents. As will be obvious from the specifications, the main differences between the types are that the BC107/BC177 are higher voltage iber 1974 swinging inductor using one op-amp The principle of simulating an inductor with a capacitor plus a gyrator is well known. With the usual gyrator circuits there is, however, the objection that one terminal of the resulting inductor is con- nected to circuit earth. A ‘swinging’ or free-ended inductor can only be obtained indirectly and with some complication. The accompanying diagram shows a swinging inductor that requires two capa- citors and one operational amplifier. The inductance appearing between points A and B is given by L = I’ 1 Xr, where r = R1 X Cl = (R2 + P2)C2. P2 will de- termine the ‘Q’ factors. The rules of the game are: the external impedance between point A and circuit earth must be less than 2 kfl, while the load on point B must be roughly equal to the value of PI (47 kfi in this case). With the values given in the circuit diagram, the inductance obtained is variable over a range of approximately I ... 1 00 Henries! M digital Until recently, the speed of a car engine (r.p.m.) was measured with an analogue system. It stands to reason that ^ a digital method would do equally well. In principle this can be W frequency meter. Since in this case the number of revolutions per minute (r.p.m.) is to be measured, the time base will have to be somewhat adapted. The contact breaker in every car (except diesels) and on every engine closes and opens a certain number of times per minute. This number is determined by the following factors: the number of cy- linders, the type of engine (two-stroke or four-stroke) and the number of revo- lutions per minute. If the first two data are known, it can be calculated how many pulses a certain contact breaker gives per second at a certain number of revolutions per minute. A one-cylinder two-stroke engine gives one pulse per revolution. A one-cylinder four-stroke engine produces one pulse per two revolutions. So a four-stroke engine gives half the number of pulses at the same number of revolutions. This leads to the formula for the number of pulses per second any type of engine produces at a certain number of revolutions (per minute): where p = pulses per second (p.p.s.) n= revs per minute (r.p.m.) c = number of cylinders a = 1 for two-stroke, 2 for four-stroke. By means of this formula we can now set up Table 1 which immediately shows the fixed r.p.m./p.p.s. ratio for each type of engine. For instance, a most common engine is the four-cylinder four-stroke. At 6000 r.p.m. this engine produces 200 p.p.s. To express the r.p.m. in four digits will therefore take some 30 seconds. This is, of course, out of the question because within the time span of 30 seconds the number of r.p.m. is subject to variation. Consequently, the number of digits shown is reduced to two. The measuring time is then only three tenths of a second. The engine speed can thus be measured with an accuracy of < 1%, which is amply sufficient. Nobody will care whether an engine makes 3418 or 3457 r.p.m. The circuit The pulses produced by the contact breaker are usually a bit frayed due to contact ‘chatter’, and the voltage pro- duced is variable because of the resulting inductance voltages. Since electronic circuits in general have a severe dislike of inductive voltage peaks, these voltages will have to be suppressed, or at least limited. A zener with a capa- citor in parallel for the sharp peaks provides sufficient protection. This pro- tective network is formed by Ri , Ci and Di (see figure 1 ). Thus the inductive peaks, and to some extent also contact chatter, are suppressed. The remaining chatter is suppressed by means of a monostable multivibrator, which uses half of a 7400 IC. This one-shot responds to pulses with a width of 50 /js or more. In addition, the one-shot passes pulses wider than the characteristic pulse time for their entire length, so that spurious pulses have no effect. The timebase is provided by a simple, yet relatively stable UJT-oscillator. Its pulse width can be adjusted over a wide range by means of potentiometers R s and R 6 ; the first is for coarse adjustment, the second for fine. In some cases the value of R7 must be changed (larger or smaller) to enable the required pulse width to be set. In contrast to the usual circuits, the out- put pulse is not used to drive a counter gate. The signal to be counted is fed con- tinuously to the counter input of the digital counter used. This is possible because the measuring time is so long that the measuring error due to the latch- and reset time is negligible. The signal for the buffer memory used in the counter is derived from the discharge pulse the UJT produces across R9. The transistors T3 and T4 provide a level suitable for TTL circuits. The latch signal thus obtained is a positive pulse. The negative edge of this pulse is used for triggering a one-shot, so that a reset pulse can be produced after the latch pulse. The decade counter, type 7490 (generally applied in digital counters) must be reset with a positive pulse. However, the one-shot produces a negative pulse. Moreover, the delay 1974 - 13 between latch and reset is too small to ensure optimum functioning. Therefore, the positive trailing edge of the negative pulse is used. After differentiation with C5 and Ris a useful signal appears on the reset output. Diode D 2 suppresses the differentiated pulse caused by the negative flank. So far the overall control circuit. Its layout is shown in figure 2. In principe any digital decade counter can be used, and one that is eminently suitable is the minitron counter. This decade counter consists of a display board with several counter boards mounted at right angles to it. For this application the display board is shortened to about 5 cm, so that it can accomodate only two minitrons. The complete minitron counter with two decades is then a block of no more than 5 x 6.5 cm. The dimensions of the control circuit board are reduced correspondingly. The diagram of the minitron counter is shown in figure 3. The 7490 is connected as a normal divide-by-ten circuit. The buffer memory, or latch, is a 7475. This IC contains four D-flipflops that store the information from the 7490 or pass it on continuously, as required. When mounting the IC on the board, pin 8 must be cut off; or, if IC sockets are used, pin 8 can be removed from the IC socket. Via the 7475, the BCD information is fed to the 7-segment decoder 7447 which drives the minitron directly. The board is shown in figure 4. By means of soldered connections the display and counter circuit boards are joined to form a kind of block. Figure 5 shows how and where the soldered connections must be made. The width of the control board matches that of the counter boards so that that, too, can be soldered to the display board. Supply The rev. counter operates on the usual voltage for TTL-ICs, that is 5 V. Adjustment Adjustment with the tone generator with the formula given above. So far so There are several ways of adjusting the For this method of adjustment, a tone good. rev. counter. The most accurate method generator with calibrated tuning scale for However, the circuit responds only to is by using the mains frequency or a reasonable accuracy is a first requirement. square wave voltages, so the tone crystal time base. Unfortunately, the Table 1 gives the frequencies corres- generator will have lo produce a square- latter will not always be available. ponding to a certain type of engine wave output, or the conventional sine- Another possibility is to use a tone gener- running at 6000 or 8000 r.p.m. Further- wave must be converted into a square ator. Both mains frequency - and tone more, each frequency corresponding to a wave. generator adjustment are discussed below. certain engine speed can be calculated This can be done with the simple circuit 1500 1000 2 cyl. 2-stroke 3 cyl. 2-stroke 1974 - 15 in figure 6. The output signal of this circuit is about 10 V, which is sufficient to operate the rev. counter. Adjustment with mains frequency Here again the auxiliary circuit of figure 6 is used, for the mains voltage is a sine wave. A simple bell transformer, or some- thing similar, will provide the required voltage of 6 V. The square wave output from the circuit is applied to the input of the control Table 2 shows what the rev. counter should indicate when used with a given type of engine, and operating on a 50 Hz input signal. While the input signal is applied, the counter can be accurately adjusted by means of Rs and R6 . Adjust- ment must be such that the reading fluctuates as little as possible between various values. As is usual for most digital counters, the last digit can jump plus or Engines with several ignition coils Some engines have more than one ignition coil and contact breaker. In this case the various channels from the contact points should be coupled with capacitors. Figure 7 shows how this is best done. A little of experimenting may sometimes be necessary to find the best values for the capacitors. equa amplifier Literally thousands of circuits for transistor- amplifiers have been developed, all of which marketed under the banner of hifi. The brands that meet the Equa- standards laid down in this issue can, however, be counted on the fingers of one - possibly two - _ hands. A feedback loudspeaker system ('electronic loudspeaker') places very strict requirements on the associated amplifier. This consideration, among others, led the editors to develop an equa-amplifier, with a circuit that could be easily adapted to give any output power up to 100 Watts. 1 A high quality amplifier must meet several requirements that are not laid down by I the DIN standard for so-called hifi- I amplifiers. With present techniques it is not very difficult to build an amplifier to satisfy these requirements. Quality requirements In the first place, the amplitude-frequency response curve of an amplifier should be flat over the entire audio-range, say from 30 to 20000 Hz. Outside this range the curve must remain ‘smooth’, which is actually the result of meeting a require- ment placed upon the phase-frequency response inside the range. (This latter point is the vital one; but the amplitude curve is easier to measure). A rolloff slope of, say, 12dB/octave below 30 Hz and above 20 kHz will not in itself influence the quality. (It will frequently prevent subsonic or ultrasonic overdriving, and produce an audible improvement.) Secondly, the distortion must be so low that it cannot be detected by ear. I The threshold for this is typically 0.5 to 1 1%. A problem here is that our hearing responds to the amplitude (i.e. peak level) i of a distortion component and not to its RMS level. Therefore, the amplitude of any distortion component must remain below 0.5%. The usual distortion measurement gives the RMS result of all unwanted components; this does not always give a meaningful, never mind accurate, impression. We will return to this point in a moment. ! Finally, we must also set up a requirement ' about reliability. This can be summed up in general terms as follows: the amplifier must be unconditionally stable, with any load; it must also be protected internally against overdriving, excessive loading and voltage surges by inductive loads. The output stage In principle, output stages can be built in many ways. With two or more transistors, a super-emitter-follower, the so-called Darlington pair, can be made. In figure la this is shown for two NPN transistors; figure lb shows the perfectly complementary arrangement using PNP transistors. Another possibility is to use complemen- tary transistors in each half of the output stage. This principle is shown in figure 2a with an NPN power transistor, and in figure 2b with a PNP power device. These circuits can be seen as amplifiers with fairly high open-loop gain, using 100% negative feedback to achieve a voltage gain of unity. This behaviour resembles that of an emitter-follower; the perfor- mance is however rather better, particularly with small signals. A very popular output stage configuration is the combination of figure la with figure 2a to form the ‘quasi-complemen- tary’ arrangement. This has the advantage that the power transistors are identical NPN types, which are usually easier and cheaper to get hold of than their PNP complements. It has the serious dis- advantage, however, that the two halves are not really complementary - which invariably causes increased distortion. The half stages of figures 1 a and b - two Darlington arrangements - can be com- bined to provide a perfectly complemen- tary circuit. The combination of figures 2a and 2b is, however, the preferred arrange- ment. The individual circuits themselves are better than Darlingtons, and the complete output stage is also complemen- tarity symmetrical. This arrangement therefore was chosen for the Equa- The Law of Cussedness requires that this circuit should also have objectionable aspects. Well, is has. One practical objec- tion is that the output is taken from the power-transistor collectors, which means that the device cooling surfaces carry audio voltage. To avoid stability problems the transistor must be insulated by mica washers, and the heatsink itself should be connected to circuit earth. Crossover distortion The distortion in a power amplifier is usually determined by the output stage. One well-known effect is (primary) cross- over distortion. This occurs with class B output stages in the neighbourhood of zero-crossing of the signal waveform Both halves of the stage are then operat- ing in the non-linear area close to cut-off. To avoid distortion it must be arranged that the stage-gain (actually its trans- conductance) does not vary with the position on the signal waveform. At greater excursions one half of the out- put stage is amplifying and the other is cut off. The active half will show its ultimate value of transconductance (or ‘slope’) over most of its working range. If the stage is sufficiently symmetrical, the ultimate slope will be essentially the same for both directions of swing. In the ‘crossover’ region near the zero-crossings both stage halve will conduct. This can lead to three situations (see figure 3): the sum of the two slopes can be greater, less than or equal to the ultimate slope of one half stage during greater excursions. Clearly, it is the third situation that is required for minimum distortion. This condition is most closely approached by arranging that both sections amplify with half their ultimate slope at the actual point of zero crossing. This is achieved by, among other things, setting the correct value of standing (‘quiescent’) Secondary crossover Less well-known is the so-called second- ary crossover distortion. This is caused by charge-storage in the bases of, mainly, the output transistors. The effect is that the output sections ‘cut off too late’ and ‘turn on too late’. It produces short distortion notches, shown for one half stage in figure 4 (exaggerated for clarity). This distortion is virtually ignored by the ‘normal’ distortion measurement ! The DIN standard specifies a measure- ment of the RMS value of the total of distortion products. Suppose now that the amplitude of these notches is 5% (!) of the signal amplitude. This is distinctly audible. During each cycle there will be only two notches, which are very short. Suppose now that the total notchtime is one fiftieth of a cycle. elektor decern ber 1974 — equa-amplifier fiber 1974 meter, a low-distortion oscillator and an oscilloscope. We hope to publish designs for such instruments shortly. Protection circuits Each half of the output stage is fitted with a protection circuit. Figure 6 shows | the arrangement for the upper half. The | circuit has three functions. Overdriving the input and/or excessively loading the I output will cause a large current to flow through the output transistors. The voltage drop across the emitter resistor Ri6 appears between the points B and C. If this voltage drop exceeds about 1 volt, T 6 will start to conduct. This short- circuits the drive to the output stage and limits the output current swing. The maximum output current is about l mav = - — , amperes for positive max R,® (or Ri 7 ) (or negative) swing. Talcing R| 6 = Rn - 1 1 ohm makes this current about I A; with the values Ri6 = Rn = 0.22 ohm it approaches S A. The third function is connected with the I experience that back e.m.f.s. produced by I inductances at the output can blow out I the driver transistors; the base-emitter I junction is exposed to an excessive reverse I bias and the resulting breakdown destroys I the transistor. In this amplifier, when the R base-emitter voltage of Ts goes negative, I the base-collector junction of Ts becomes I forward-biassed. This safely limits the 1 For high-power versions it is advisable I to add 1 k series resistors in the base I connections of Ts and T*. These are I shown dashed in figure 8. I An extra protection by means of a fuse I in the supply rail is not just luxury. Strictly speaking it is unnecessary, but it does provide a convenient measuring- point for the standing current. The milliammeter can be simply connected in place of the fuse. The complete amplifier Figure 8 shows the complete circuit of the amplifier. Several details meet the eye that have not been discussed as yet. The four capacitors C 4 , Cs , C» and C7 are included to control and improve the high-frequency performance of the circuit (stability and impulse response in par- ticular). The feedback resistors Rs and R 6 determine the amplification. This is set by the specified values at about x20. Reducing the value of Rs is allowed; it will increase the gain (and therefore the input sensitivity!) but will also increase the distortion. For this reason a minimum value of 100 ohm is specified for Rs . The distortion is then still accept- able while the gain is in the order of 100. Transistor T 2 controls the output stage standing current; the required value is set by adjusting P 2 . Before switching the amplifier on for the first time, P 2 should be set at minimum. The amplifier can then be switched on and the correct quiescent current set in accordance to table 2. The circuit around T 4 is unusual in this application. It is shown separately in figure 7a. Fundamentally it is a combi- nation of a current-source and a gyrator, providing a fairly high impedance for the collector load of T 3 . This enables T 3 to fully drive the output stage without 'running out of current’. The usual way Figure 4. The signal from one half of an output stage. The secondary crossover distortion is clearly visible as small notches superimposed Figure 5. The same circuit as figure 2, but now including the compensation-networks. The cor- rect component values depend on the charac- teristics of the power transistors. This arrange- ment is used in the equa-amplifier. Figure 6. The protection circuit. A network of this kind is added to each half of the output stage. It protects the amplifier against over- driving, excessive loading and inductive back- voltages at the output. Figure 7. To achieve a high collector feed- impedance for the pre-driver transistor T 3 the shown in figure 7a may be used. The classic solution is 'bootstrapping' as shown in figure 7b. We believe the first circuit it preferable, but the circuit board can be used with either. Figure 8. The complete amplifier. With the specified power transistors the maximum out- of providing this high impedance is the ‘bootstrap’ circuit shown in figure 7b. This latter circuit can be expected to have agreaterinstability-risk; but practical experience has yet to demonstrate any difference. The circuit board is suitable for either arrangement although, in our opinion, figure 7a is preferable. Finally, the loudspeaker connection is parallelled by a network consisting of Ris, R 19 and C». This guarantees the stability of the amplifier when it is operated without a load. The proof of the pudding . . . Several amplifiers were built according to this recipe, using randomly-chosen components. The worst-case measure- ment results were as follows: Amplitude-frequency response curve flat within 1 dB from 20 Hz to 60 kHz. Table 1. The required supply voltages and values of Ri6 and Rn, for various loudspeakers (nominal) impedances and output power ratings. Table 3. Test points (fig. 8) 1 60 40 20 ir 2 i ioo n 82 n 68 m 2 28 19 9.5 3 29 20 10.5 4 (+Vb - 0.7) 5 30 21 11.5 6 28 1 9 9.5 7 1.25 1.5 1.85 8 (+Vb - 0.65) 9 0.65 0.65 0.65 All voltages ±10% Peak distortion level below 0.07% (typ. 0.03%). Stability maintained for: resistive load (all values from dead short to open circuit), capacitative load from 10 pF to 1000 pF, inductive load from 10 pH to 200 mH, any combination of values. Output power The maximum output can be selected with the aid of table 1. As will be apparent, the absolute maximum is 100 watts (sine wave) into 4 ohms. For all normal listening in the sitting room how; ever, the 20 watt version is emphatically recommended. It has been extensively tested with electrostatic loudspeakers and as the ' driver for the ‘electronic’ (feed- back) loudspeaker, easily producing more than enough sound level. The various voltages, currents, loud- speaker impedances etc. can be found from the output power nomogram, else- Resistors: «1.»3 R 2 r 4 r 5 Re R 7 *8 R 9 Rio R 11 R 12- R 13- R 14- R 15 where in this issue. As will be obvious, the input sensitivity is equal to the output voltage Veff divided by the amplification. For the 20 watt/8 Q. version for instance, V e ff is found to be 12.5 volts. The input sensitivity is therefore approx. = 625 mV. | 1499] H?-- 22-i output pouter nomogram This nomogram has been prepared by the editors in response to regular requests from readers. When the required output power and the loudspeaker impedance are known, the nomogram can be used to find the associated voltage and current. It can actually be used as soon as any two of the variables are known-to find the remain- P is the continuous (sine wave) power Rt is the impedance of the loudspeaker V e ff is the effective (RMS) output voltage V is the peak value of the output voltage I e ff and T are the effective and peak values of the current swing The power supply must deliver at least 2 $ + 4 volts (measured to the lowest edge of any ripple waveform). For a stereo amplifier, it must be rated for at least I e ff. "Music power”-depending on the power supply and the output stage heat sink-can be anything from 1 to 20 x F . . . ! Example (see dashed line): For 20 watts into 8 ohms we find 9=18 volts and I e ff = 1.6 amps. So the power supply must be rated to deliver 2 x 18 + 4 = 40 volts at 1.6 amps. M WATTS S' *5 divide by 1 to 10 1974 - 23 divide by 1 to 10 electronic The starting point for design of this electronic candle was to reduce the fire risk associated with the Christmas season, at the same time pro- viding a candle which would not burn up so quick- ly. Naturally, the electronic candle can be lit with a match (but a pocket torch will do the job tool); it can be blown out or 'nipped out' with the fingers. Using the CD4017AE (COS/MOS inte- grated circuit RCA) it is possible to make a universal frequency-divider that will divide by any number from one to ten. If a square wave is presented to the ‘clock’ input while the ‘reset’ input is connected to circuit ‘ground’, a square wave output at one tenth of the clock frequency will appear at pin 12 (the ‘carry out’). Each positive-going edge of the clock signal will cause the outputs 0 to 9 in turn to assume the value M’ for a single clock period. Suppose for example that the first positive- going edge of the clock signal has caused output 0 (pin 3) to become ‘1’ - all the other outputs are then ‘0’ - the next positive-going edge will cause output I (pin 2) to become ‘1’ and output 0 to return to ‘O’. Since the outputs 0 to 9 act as a kind of shift register the circuit can easily be made to divide by any whole number from 2 to 9. All that is necessary is to interconnect the output having the desired number with the reset input (pin 1 5). If the reset is obtained from output 7 (pin 6) for example, the IC will always count up to 7. Any of the earlier intermediate outputs (in this example 1 to 6) can be used as the output of (in this case) the divide-by-seven. Note that the value of load resistance applied to any output must not be less than 47 kfi. If any output is required to drive TTL, the simple buffer stage shown connected to output 4 can be used. H dock 5314 The 'brain' in the digital clock described in this article is the clock-IC MM5314, which nee ds only a few external components. The time of day is indicated by seven-segment Ga-As displays, which are now offered at quite agreeable prices. Another attractive feature is that if no seconds reading is included in the design, a considerable saving can be made, whilst seconds indication can always be added at a later stage. The clock-IC The clock integrated circuit type MMS3I4 is designed to indicate the time in hours, minutes and seconds with the aid of seven-segment displays. In contrast to the MM53I3 it has no BCD output. Conse- quently, it is smaller (DIL 24 pins), has a simpler construction, and, what is perhaps even more important, is a lot cheaper. However, as appears from the circuit diagram of the MMS314 (figure 1), all the components needed for building a clock are available. The IC receives its clock pulse from the mains, and can be used for 50 Hz or 60 Hz drive. The supply voltage may vary from 8 V to 17 V and need not be stabilized. If not connected, all drive inputs are at ‘ 1 ’ level because resistors are incorporated which connect them to the plus pole of the supply voltage. As regards the clock design, the 1C offers the choice of various possibilities that depend only on a certain logic state of the drive input concerned. It is possible, for instance, to choose be- tween a 24-hour and a 12-hour cycle. With the 12-hour cycle the leading zero indication is automatically suppressed, which saves a lot of power. If in addition no seconds reading is required, two seven- segment displays and two transistors can be omitted, which gives a considerable saving. By means of the input ‘strobe’, read-out can be suppressed, and there are, of course, control inputs for re- tarding or advancing the clock. The clock can also be stopped for correct time set- ting. The table gives all possible settings of the control inputs. Figure 2a shows a top view of the pins of the MM5314 integrated circuit. Operation In the overall circuit of the IC two main sections can be distinguished: a. the counter with corresponding b. the circuits for decoding and driving the displays (surrounded by the dashed line in figure I ). Pulses to drive the counter are obtained from half cycles of the mains supply. The pulse shaper at the input of the counter changes the sine-waves into square waves by means of a Schmitt trigger. This trigger has a hysteresis of about 5 V. Depending on the logic state at pin 1 1 of the IC, the pulse signal is divided by 50 or 60, so that a signal of 1 Hz becomes available for the next divider. In the next three stages of the counter the pulse signal is divided into minutes and 12 or 24 hours, depending on the cycle chosen, and determined by the logic state of pin 10. Via the gates of the individual stages of the counter the clock can be set correctly. If pin 14 of the IC is at ‘O’, the clock will run at the rate of 1 minute per second. If pin 15 is at ‘O’, the hours will run at the rate of 1 hour per second. When pin 13 is at ‘O’, the clock is stopped. If a 12-hour cycle is chosen, the leading zero is suppressed by a special circuit in the IC. Counter read-out and display drive are achieved with a multiplex technique. The multiplexer senses the various counter positions successively in the rhythm of a multiplex frequency, and passes the value found to a decoder, and from there to an output memory (ROM-Read Only Mem- ory). The multiplex frequency can be varied by means of a simple RC network connected to pin 23. The multiplex oscillator is followed by a divider that, depending on the logic state of pin 24, produces four- or six-digit drive pulses (with or without seconds, respectively). Using the multiplex tech- nique implies that the displays are not driven in parallel, but in series. Parallel drive means that all counter positions can be read out simultaneously. To that end the counter reading of each decade is, at a certain moment, fed to a memory corresponding to each decade. The information thus stored drives the dis- plays of the counter readings via a decoder. This happens simultaneously for all decades; hence the term parallel drive. Multiplex technique, however, means that all counter readings are scanned quickly in successive order and are fed in the same order to an output memory (ROM), which for this IC is programmed for seven- segment displays. At the same time that the counters are read, each corresponding display receives the supply voltage via the drive logic of the block marked ‘Digit Enable’. This means that, with this clock, the counters can be read 1 out of 4 if a four-digit display is used, or 1 out of 6 for a six-digit display; the logic state of pin 24 determines the display mode. If, for instance, the one-second counter is read, the one-second display receives supply voltage via ‘Digit Enable’, and the reading of this decade becomes visible. Corresponding segments of each display are interconnected, but only the particular segments of a display that receive a voltage will light up. In spite of the fact that series drive is used, visual read-out remains constant, provided the multiplex frequency is higher than about 100 Hz. In the MM5314 the multiplex frequency can be chosen up to 60 kHz. If the read- out is suppressed via pin 1 (‘strobe’) of the IC, the clock will continue to run normally. Thanks to this feature it is quite easy to build an emergency supply. The circuit The complete circuit in figure 3 shows that apart from the MM5314 only few components are needed to build a com- plete clock. Perhaps somewhat unusually, the circuit description starts with the supply, because it is from there that the counter pulses are derived. Since the supply voltage for the IC need not be stabilized, the source has been kept as simple as possible. The d.c. supply voltage may be anything between 8 V and 17 V. The half cycles of the 50 Hz mains are fed to the pulse input via a decoupling net- work R22/C3. This input is protected against overloading by means of diode Di . The RC network (R23/C4), connected to pin 23 of the IC, determines the multi- plex frequency which, for the given values, is about 10 kHz. Because the integrated circuit cannot provide suffi- cient current to drive the seven-segment Figure 1. Block diagram of the MM5314 inte- grated circuit. From this it is clear that the entire dock, except the supply and drive for the displays, is incorporated in this 1C. Figure 2a. The pins of the 1C seen from the top. Figure 2b. Pin details of the Opcoa red GaP seven-segment display type SLA 1. With most other types of seven-segment displays separate hence, an extra connection is needed between these pins and pin 14. stop -O' 13 slow adjustment 'O' 14 quick adjustment ‘0’ 15 mains frequency 50 Hz 'V 11 mains frequency 60 Hz 'O' 1 1 12-hour cycle "O' 10 24-hour cycle 'V 10 with seconds 'O' 24 without seconds '1' 24 strobe 'O’ 1 •) An unconnected input is at state '1' be- cause within the 1C these inputs are connected to the plus of the supply voltage via resistors. display simple buffer stages are required. These use normal TUN’s and are connected between pins 3 to 9 and the display segments. The collector resistors provide current limiting for the segments, so their values determine the luminous intensity of the displays. The minimum permissible value for these resistors is 330 ft (+Vb = 17 V); in practice 470 ft gave satisfactory results for all supply voltages. A lower value produced no noticeable increase in luminous intensity, so that in fact only the life of the display is then unnecessarily shortened. Buffer transistors, acting as switches, are also connected between the ‘Digit- Enable' outputs and the anodes of the displays. These switches connect the second-, minute- and hour displays to the 26 - elektor iber 1974 5314 supply voltage at the correct moment. The switching transistors used here are TUPs. The circuit is mounted on two printed circuit boards: one for the displays, and one for the actual clock circuit with mains supply. Printed circuit boards Figure 4 shows the printed circuit board, and figure 5 the component layout for the mains-fed clock circuit. The boards are quite small, so that the whole unit can be housed in a small attractive cabinet. So much space has been reserved on the board for the supply transformer and electrolytic capacitor Cj that, if necessary, fairly large types can be used. All terminals and con- trols (50/60 Hz selection, strobe, etc.) are placed in a row on one side of the board, directly opposite the terminals they are connected to on the display board, which is shown in figure 6. This display board holds the displays and small push buttons for ‘stop’, ‘slow’ and ‘fast’. Displays The display board (figure 6) is mounted i clock 5314 -27 Figure 3. The total circuit complete with mains supply. If instead of TUNs, quality transistors are used for T ^ T7 (e.g. BC107), the resistors Rg R 1 4 can be omitted Figure 4. The printed circuit board of the clock circuit with mains supply. The pins are position- ed so that only very short connections are need- ed between clock and display circuit boards. Figure 5. Component lay-out for the clock cir- cuit. There is sufficient space for almost any type of transformer. Even a 40V electrolytic capacitor could be accommodated on the circuit behind the front plate of the cabinet. In- stead of the seven-segment LED displays used here (the Opcoa SLA 1 ), types MAN 1 , MAN7 and MAN 10 of Monsanto, T6302 of Texas, 5082 and 7730 of Hewlett Packard or Data Lit of Litronix can be used. Some of these even have two LEDs per segment, which gives a greater intensity at a slightly lower current consumption. Unfortunately, there are many displays where not all anodes are connected to pin 1 4, but have separate anodes connected to pins 3 and 9. The pins 3 and 9 (at the bot- tom of the displays concerned) must then be bent completely inward and connected to pin 14. With or without seconds If the ‘seconds’ indication is not used the expense of two displays, two sockets and two transistors can be saved. In this case there is no connection between pin 24 and earth. Since the board is designed for six displays, two more can always be added at a later time without much trouble. Connection between the boards In total (including the seconds) there are 1 3 control connections between the clock and the display circuit boards. The six pins of Digit Enable (h(, h e , mt, me, st, Se) are connected to the corresponding terminals on the display board. Furthermore, the terminals a to g of the clock circuit are connected to the same terminals on the display board. Three other connections run to the three small push-buttons for setting the clock. One side of each button is con- nected to the supply common. By means of time signals on the radio, TV, or telephone service, the clock can be start- ed properly and quite accurately. With the buttons ’’fast” and ’’slow” the clock is pre-set before the time signal comes, and the button ’’stop” is released the moment the signal sounds. The front of the cabinet must have openings for the four or six dis- plays which can be mounted behind per- spex, for instance. Further developments In Eleklor laboratories the following ad- ditional units have been developed for the - crystal-controlled time base with only one IC; current consumption complete with oscillator: about 90 fiA. distortion metetfl The distortion in factory- produced or home-made ampli- tiers is frequently unknown; designers sometimes give specifications, but these are not always reliable. Since distortion meters are usually expensive, Elektor Laboratories have developed a simple, inexpensive, but effective instrument. Low frequency pre- and power-amplifiers always produce some distortion. The various kinds are distinguished as follows: Linear distortion - the departure from a flat amplitude-frequency response curve. An amplifier which is flat within 1 dB from 20-20000 Hz has less linear distor- tion than another which only does this within the band from 100-8000 Hz. Intermodulation distortion - when two or more frequencies are fed simul- taneously into the amplifier and it pro- duces ‘sum and difference’ components. Harmonic distortion. This is real ‘visible’ distortion; if the input was a sinewave the output signal is definitely ’something else’. The output signal can then be shown to consist of the original sinewave (possible amplified), plus several overtones or harmonics. The ratio of the unwanted components to the total output signal gives the distortion percentage. This measurement can be made with the distortion meter described below. Design considerations A distortion percentage of 0.01% means that the fundamental in the output signal is virtually ten thousand times greater than the distortion. Therefore, if the dis- tortion is to be measured the fundamen- tal will have to be attenuated more than 10000 times. This is 80 dB! At the same time, the first overtone (second har- monic) must remain unaffected. This requires an exceedingly sharp filter. For normal low frequency work it must be possible to measure distortion in the frequency range 100 Hz to 10 kHz. The filter will therefore have to be tunable through this band. Transistorised power amplifiers frequent- ly produce spikes in the waveform at the zero-crossings as well as the normal dis- tortion components. These spikes can be as short as 10 fis or even less, implying the presence of frequencies in excess of 100 kHz. After the fundamental has been suppres- sed the distortion product then appears as in figure 1 . The spikes in this trace have an amplitude 1% of the total output! To enable these spikes to be measured the distortion meter will have to pass the high frequencies involved unattenuated. A passband to 500 kHz is therefore by no means an unnecessary refinement. For a distortion measurement according to DIN standards, the RMS value of the unwanted products - corresponding to their average power-contribution - is what must be determined. This requires an in- tegrating meter. However, since the human ear responds to the amplitude rather than to the power of a signal, a peak-level detector is what is really needed. This will often show a completely different (much ’worse’) result! An example of this is given in figure 2. Figure 2a is a trace of the distortion pro- duct from a reasonably good power amplifier. The RMS and the peak measurements give the same result - 0. 1 8% distortion. Figure 2b shows the distortion product from a similar amplifier. Along with ‘ordinary’ distortion however, this one also produces sharp spikes. The two measurement procedures now lead to totally different results: the RMS meter indicates a distortion increase to 0.21% (0.03% more than before). The peak meter on the other hand now indicates 0.95% distortion - an increase of about 0.75%! The latter value is a more accurate indication of the subjective increase of the distortion. Clearly, a universal instru- ment will have to be able to carry out both procedures. Finally, the measurement must be unaffected by hum and noise (which can be identified on the ‘scope’, but may cause a misleading reading on the pointer instrument). The design will therefore include hum and noise filters which can be switched out of circuit. The filter The design chosen for the rejection- filter is an unusual one. When two signals having the same frequency, amplitude and phase are presented to the inputs of a good differential amplifier, the output signal is zero. The signals are blocked. The block diagram of a rejection filter can therefore be as shown in figure 3. The input signal is first passed to a phase splitter (paraphase amplifier, with equal- and-opposite outputs). One of these output signals, the one which is 180° out of phase with the input signal, is applied directly to one input of the differential amplifier. The other output of the phase splitter is in phase with the input signal; it is passed to a phase shifter. This section imposes a phase rotation which, depen- ding on the frequency, lies somewhere between 0° and 360°. For one single frequency (f 0 ) this shift will be precisely 180°. The output of the phase shifter is now applied to the other input of the differential amplifier. For. an incoming signal of frequency precisely fo which will therefore be rotated exactly 180°, the output of the differential amplifier will disappear - the signal will be rejected. For every other frequency the output signal will be unequal to zero. The final step is to provide the required sharpness of the characteristic by means of overall negative feedback. The great advantage of this arrangement is that it does not require trimming, while at the same time it can be tuned over the entire working range using one stereo-potentiometer. The accuracy of tracking of the two halves of this poten- tiometer is completely unimportant. Circuit of the filter The filter circuit is given in figure 4. The transistors Ti and T2 form the phase splitter. The in-phase output signal is developed across R s , so that the circuit has heavy internal negative voltage feed- back like that of an emitter follower (but much heavier in this case). The anti-phase output signal appears over R4. This circuit is far better-behaved than any single- transistor arrangement and is used at all important points in this design. The phase shifter is built up around T3 to T 6 . It is actually a cascade of two simple phase shifters, each of which imposes a rotation between 0° and 180°. The frequency for which the total rotation >r december 1974 amounts to exactly 180°, is fo. This frequency is adjusted by means of P 2a and P 2 b- A fine adjustment is provided by P3. The capacitors C2 and C3 should have low thermal coefficients. The switches Sia and Sib enable the cir- cuit to be calibrated, in combination with Pi . When these switches are open the phase shift is 0° for all frequencies; the filter action is defeated and the input sen- sitivity can therefore be set correctly, T9 to T13 form the differential amplifier. The impedances in the circuit have been kept low so that it will also behave well at high frequencies. The inverted (180°) signal from the phase splitter reaches the plus-input via Ris and P s . The output of the phase shifter is taken from Pa and applied to the minus-input. These two signals must have precisely equal ampli- tudes at fo in order to cancel. This can be coarsely and finely adjusted using P4 and Ps- The potentiometer P 6 is a preset control for adjusting the DC balance of the dif- ferential amplifier, since this depends on the properties of the individual transis- tors. Set the DC levels at points A and B to be equal (about 4 volts). This is the only trimming point in the whole filter. Overall negative feedback is applied via R22 , R23 and R2 . Hum and noise filters The circuit of these filters is shown in figure S. They are active filters, contain- ing RC networks in their input, output and feedback paths. The turnover is fairly sharp and the rollof slope is more than 1 2 dB/octave. The hum filter is built around Ti$ and can be switched into circuit with S2 . The cut off starts near 250 Hz, the response being more than 20 dB down at 50 Hz. The noise filter (Ti6> is switched in by S 3 or S 4 and cuts off at 20 or 200 kHz respectively. Bear in mind that this filter will also suppress any spikes more or less completely. Figure 5 also includes a voltage amplifier (ICi ). This will boost the output signal by 10 or 100, so that a multi-meter can directly indicate distor- tion at 10% or even 1% fsd. A disadvan- tage here is that the response of the 1C - at a gain of 100 - already starts to roll off at about 20 kHz, so that the output contribution from the waveform spikes is lost. How to use the meter Measurements are taken with the equip- ment arranged as shown in figure 6. The sinewave generator must have very low distortion. We hope to publish a good cheap design shortly. The measurement procedure is as follows: Set Si to ‘calibrate’. Switch all filters and the xlO/xlOO amplifier out of circuit. Adjust Pi until the meter reading is 1 V; this is equivalent to a distortion of 100%. Set Si to ‘measure’. Adjust P2 and P4 alternately to obtain a minimum reading. S5 can be set to ‘xlO’ or ‘xl00’ as may be required for a useful deflection. When the adjustment of P2 and P4 becomes too imber 1974 - 33 quadra 1-2-3-4 ....or W W Wv 9 The phenomenon of 'quadro- phony'* has already been the subject of many publications, but the confusion only seems to increase with every new attempt to clarify the issue. This article may bring a little light into the darkness, by describing and comparing the most important systems that have been proposed so far. critical, continue fine adjustments with P 3 and P 5 . As soon as the minimum out- put has been found the distortion can be read directly. Just how this is done will depend on the indicating instrument used. If this instrument is a typical multi-meter, the normal harmonic distortion can be read with reasonable accuracy. The ‘xlOO’ position of S5 then corresponds to an fsd of 1% distortion. The contribution of waveform spikes will be lost, while there is no quarantee of the accuracy of the meter at higher frequencies. A more accurate result can be obtained if a good AC millivoltmeter is available. Set Ss in this case to *xP, otherwise the integrated amplifier with its early rolloff will be in circuit. Both of these methods have the objection that the indicating instrument integrates, so that its reading corresponds to the RMS value of the distortion. The amplitude of the distortion products can be measured using an oscilloscope. Connect this as shown in figure 6. The original signal from the sinewave genera- tor is applied to the X-input and the output from the distortion measuring circuit (at 'xl' gain!) is applied to the Y-input. The trace will now be of the kind shown in figure 2. Set the 100% level, during calibration, to indicate 3 volts peak-to-peak. 3 mV in the trace now corresponds to 0.1% distortion- amplitude. It may be possible to improve the reada- bility of the trace by using the hum or noise reduction filters. Remember, how- ever, that the noise filters will also suppress any spikes. Finally, a very good indicating instrument is an AC millivoltmeter that can be switched to operate as an RMS or as a peak detector. Beware of instruments that use a peak detector but have a scale calibration reading 0.707x the peak value - they only read the RMS level of a pure sinewave. The meters required here use some kind of square-law detector RMS value of the distortion, level). With such an instrument distortion can be read either according to the DIN standard or as a ‘genuine’ distortion- percentage. In order to simplify the comparison of the various systems, we shall proceed from a block diagram of the total sound signal path (figure 1). In this diagram, A represents the record- ing location (studio, concert hall, etc.) in which a number of microphones are placed. The type and number of micro- phones used and their position are, of course, significant for the maximum quality of transmission that is attainable. Many fundamental investigations dealing with these aspects are going on at the present time, but they will not be dealt with in this article. Block B represents the total chain of electronic devices that perform the coding, transmission (via gramophone record, tape or radio) and decoding. One of the possible quadrophonic systems is introduced into this chain. Block C forms the end of the chain as the living room in which the loudspeakers are usually placed in the four corners. The various systems in block B can now be compared to each other by relating the sound impression reproduced in space C to the original sound impression that was derived (by the recording technican) from the sound event. First of all, the basic methods of opera- tion of the various systems will be briefly discussed. Types of system In general, we can draw a distinction between three different types of system: quasi-quadrophony (or pseudo-quadro- phony, similar to pseudo-stereophony), ‘discrete’ quadrophony with four in- dependent transmission channels and, finally, quadrophony according to matrix systems. Quasi-quadrophony is based on the ex- perience that a ‘spatial impression’ enhances the reproduction — regardless of whether or not the reproduced sound actually corresponds to the original as far as the positioning of the various instru- ments or groups is concerned. Such systems can, for example, reproduce reverberation (or the difference signal from two stereo channels, which usually contains a lot of reverberation) via the two rear loudspeakers. This is sometimes referred to as a ‘2-2-4’ system, in other words a system that uses 2 original sound channels, 2 transmission channels and 4 reproduction channels. It goes under various banners, such as ‘Stereo- 4’, ‘Quadro-sound’ etc. However, it is not quadrophony in the true sense and will therefore not be discussed any further in this article. A discrete quadrophonic system contains four different channels that remain separated within section B of figure 1 - from the microphone to the loud- speaker (a ‘4-4-4’ system). An example of this is the CD-4 gramophone record-CD stands for Compatible Discreteness. An experimental radio transmission that used two stereo FM transmitters for one pro- gram could also be included in this group. Finally, matrix systems are based on the mixing of the original information channels. What were previously four channels of the total quadrophonic re- cording are now combined into two new, specially-coded channels. They can then be conducted over normal stereo systems, divided again into four channels at the destination and reproduced by the four loudspeakers in the listening room. These systems are classed as ‘4-2-4’. Since only two equations cannot be solved if they contain four unknows, the four resulting channels will in the last analysis never be identical with the original four: they must always contain crosstalk components. According to the choice of the mixing relationship, how- ever, the spatial sound impression during reproduction can correspond more or less satisfactorily to the original. CD-4 This system, advocated by Nivico and RCA, is a discrete system. On a gramophone record, the left ‘stereo’ channel now contains the sum signal of ‘left front plus left rear’, and, in addition, a frequency modulated 30 kHz carrier with the difference signal ‘left front minus left rear’. The right ‘stereo’ channel carries the two signals Tight front plus right rear’ and ‘right front minus right rear’ in the same way. For reproduction, the four original channels can (in principle) be regained by simple addition and subtraction of the respective sum and difference channels. The modulation of the left channel is shown schematically in figure 2. The sum signal with a bandwidth of 15 kHz is cut in the usual way. The difference signal is frequency modulated on a 30 kHz carrier. This modulation is asymmetrical (-10 kHz, +15 kHz), which easily gives rise to amplitude modulation and distor- The practical results with this system are discussed in the comparative section. SQ and QS SQ (by CBS and Sony) as well as QS (by Sansui) are matrix systems - the abbre- viations stand for ’Stereophonic Quadro- phonic’ and ‘Quadrophonic Stereo', respectively. Here the four original channels are mixed into two for trans- mission and are divided again into four before reproduction. In the case of SQ the mixing relationship (in amplitude and phase) is set up for optimal channel separation between left and right front, respectively, and between left rear and right rear. The front channels are cut in the same way as normal stereo channels. CBS chose this system because it was expected to produce optimal effects in the case of possible traditional stereo reproduction. From the compa- rative section, it can be seen to what degree this was achieved. The unavoid- able crosstalk takes place between ‘front’ and ‘rear’, audibly along both diagonals. In the case of QS, on the other hand, a mixing relationship that should make acceptable quadrophonic reproduction possible was chosen. A point-like sound source in the recording area is reproduced with an amplitude characteristic that is very close to cardioid. The sketch in figure 3 shows this characteristic for BMX, which will be discussed in the next section. For both QS and BMX this characteristic is always oriented towards the position of the original sound source. The Japanese ‘regular matrix’ standard (RM) is based on the QS system. UMX UMX is a 'universal matrix system’ derived by Professor Cooper (USA) in collabo- ration with Dr. T. Shiga (Japan). The practical development followed in cooperation with Nippon Columbia (trade name: Denon). This firm is a member of the Hitachi group. The point of departure was a thorough scientific investigation of the character- istics of matrix systems. From this the optimal two-channel matrix was derived: BMX. By the addition of a further, channel, the three-channel TMX was produced, while QMX works with an extra fourth channel. Of relevance here is the fact that the position of the phan- tom sound source during reproduction is not altered during transference from two, via three to four channel transmission. The localization does become more pre- cise: with BMX, a solo instrument sounds somewhat ‘mushy’ (spread over a distance of about 0.5 meters), but with the higher order systems TMX and QMX the sound seems to come from a precisely determin- able point. The characteristics of amplitude and phase, as they arise during the repro- duction of a single point source, are shown in figure 3. The amplitude characteristic of BMX is the same as for QS, and is always oriented towards the original position of the sound source. An essential difference from QS lies, however, in the fact that with BMX the phase characteristic also ‘rotates’: 0° corres- ponds to the direction of the sound source, while, for example, the sound coming at rightangles to the sound source is phased at ± 45°. This additional information gives a significantly better localization. With the QS-system,0° phase rotation always corresponds to the sound from the phantom centre front, so that sound sources in the front are drawn towards this point. In the case of gramophone records in UMX (called UD-4) the two basic channels of BMX are recorded in the same way as for stereo. One basic channel contains the mono signal (sum signal), while the other contains the difference information for the stereo or quadro effect. The third (TMX) and fourth (QMX) channels are frequency modulated on two 30 kHz carriers, similar to those used for CD-4. An essential difference from that system, however, lies in the fact that these two auxiliary channels can be contained in a fairly narrow band. An audio bandwidth of 3 kHz is com- pletely satisfactory, and this can be transmitted as symmetrical frequency modulation with a peak deviation of ± 6 kHz (figure 4). This limiting of the audio band is possible, because there is hardly any audible diffe- rence between BMX and QMX at fre- quencies above about 3 kHz! Since the orientation of the various sound sources is the same for all three systems, the transition from QMX to BMX at this cutoff frequency is almost imperceptible. jadro 1 -2-3-4 . 1974 - 35 . or nothing? TMX is mainly of interest for radio broad- casting: a third channel can be rather simply provided (for example, by quadra- ture modulation); however, four channels appear to be an impracticable process - at least in Europe. Greater bandwidths would be required for the transmission of four channels, and these would lead to unacceptable interference on neigh- bouring channels. Conclusions From the comparison of the four systems it is apparent that SQ seems to be based on a different conception of quadro- phony: to arrive with ‘logic’ at four 3 stressed ‘comers’ (and also ‘centre front’). This is succesful to the extent that presentations can be very impressive in spite of the noted shortcomings. The results of CD-4 and QS are adequate. Since several parameters are not optimal, the peripheral devices for noise reduction and image position stabilization are unnecessarily complicated. In spite of these additional devices, however, the results are not completely satisfactory. Finally, the UMX system combines the best features of both systems to give the best results. Therefore, from a technical viewpoint, this system is to be preferred. Unfortunately, the discussion of quadro- phony is at present clouded by confusion Figure 1. Block diagram of a complete quadro- phonic sound chain. A ■ recording area: B - Figure 2. Frequency spectrum on one record groove wall when recording according to the CD-4 system. The sum signal is recorded in the normal way in the base band (0 ... 15 kHz). A 30 kHz carrier is frequency modulated with the difference signal in the band from 20 to 45 kHz. of 8 the systems BMX. TMX and QMX. 0 dB of the amplitude characteristic and 0° of the phase of the sound source. If several sound sources are reproduced simultaneously, one can "piled on top of one another". Figure 4. Frequency spectrum when recording according to the QMX system (one groove wall). The two BMX channels are recorded in the base band (0 ... 18 kHz). The two auxiliary channels are each modulated on a 30 kHz carrier (FM). in the band from 24 to 36 kHz. of language and by commercial con- siderations. Partly because of this, the UMX system has often been practically ignored. It is often argued that UMX was developed too late, so that great invest- ments already lie in other systems. Professor Cooper argues strongly against this. In his opinion, the differences from the other systems (especially CD-4) are so slight that possible changeover offers no difficulty. The number of gramophone records already pressed according to a certain system should not (yet) be decisive either. It would be another matter if a company began to use a particular system for its entire record collection. Fortunately, this has not yet happened. In view of the rapidly increasing demand for quadrophony especially in the USA and Japan but also in Europe, there is still hope that a definitive choice will be made in the near future. In this event, it is to be hoped that technical arguments will be decisive, and from the technician’s standpoint this article could have been entitled: UMX ... or nothing! References: Y. Makita, 'On the Directional Locali- zation of Sound in the Stereophonic Sound Field', EBU rev., pt.A, no. 73, p. 102 I June 1962). R. Itoh, 'Proposed Universal Encoding Standards for Compatible Four-Channel Mixing'. Journal of the Audio Engineer- ing Society (JAES), April 1972, p. 167. D.H. Cooper and T. Shiga, 'Discrete- Matrix Multichannel Stereo ', JAES, June 1972, p. 346 and July 1972, p. 493. P.B. Fellgett, 'Directional Information in Reproduced Sound’, Wireless World, Sept. 1972, p. 413. P.B. Fellgett, 'The Japanese Regular Matrix', Hi-Fi-News, Dec. 1972, p. 2393. B.B. Bauer, G.A. Budelman and D.W. Gra- vereaux, 'SQ Matrix Quadraphonic Discs', JAES, Jan. 1973, p. 19. L 1 -2-3-4 alektor december 1974 - 37 38-i iplifier aerial amplifier The aerial amplifier described in this article is characterized, among other things, by its low noise level (1-2 dB), a voltage gain of 10-20 dB, and a wide tuning range (146-76 MHz). It is designed for use as an FM-aerial amplifier, although it is relatively simple to modify it for application as a TV aerial amplifier. Aerial amplifiers can be divided roughly into two categories: wideband and tuned. The main advantage of wideband types is, of course, to be found in the fact that a frequency spectrum of several decades can be amplified without anything having to be switched over or readjusted. On the other hand, there are some drawbacks that count all the more if the amplifier is ex- pected to provide maximum improvement in reception quality. Using wideband amplifiers entails the following drawbacks: 1 . Cross modulation soon occurs because the total amplitude offered can be fairly large. Furthermore, the entire amplified spectrum is fed to the receiver and this is another likely cause of cross modulation. 2. In most cases it is impossible to design a wideband amplifier for mini- mum noise contribution. This is be- cause the cable impedance (usually 60 fi) is not the optimum value for the amplifier. In addition, it is almost impossible to compensate fully for parasitic capacitances. Comparison of the noise contributions of TV tuners and of wideband amplifiers shows that both are usually of the same order of magnitude for the UHF band. In the VHF-TV and the FM bands, the tuner often has an even lower noise figure than the wideband amplifier. If the wideband amplifier gives better reception, this is due mainly to the fact that when the amplifier is placed between the aerial and the cable, the cable losses become far less important. Tunable amplifier A drawback of a tunable amplifier is that an extra cable is usually needed for the tuning voltage. By means of a simple cir- cuit, however, (figure I) it is possible to use a tunable amplifier without an extra cable. The stabilized power supply pro- vides the sum of the supply voltage and the tuning voltage, and within the ampli- fier the 12V supply is obtained by stabil- ization with a voltage regulator diode. By connecting a 1 2 V regulator diode in series with the supply voltage, the tuning voltage is 12 V lower than the supply voltage. If the variable stabilized supply is now adjusted from 14 to 26 V, the supply voltage for the amplifier remains 12 V, and a tuning voltage of 2 to 14 V becomes available. It goes without saying that the variable supply must have a very low hum and noise level to avoid amplitude and phase modulation via the varicaps. Therefore a large electrolytic capacitor is placed in parallel with D 2 . The circuit consumes about 100 mA, but offers the advantage that the amplifier always is at a higher temperature than ambient, so that water condensation and the resulting corrosion are avoided. Design possibilities for tunable am- plifiers A FET-amplifier can be based on two main circuits, to wit: the common-gate and the common-source amplifiers. Since the amplifier is tuned, the input and out- put capacitances of the semiconductors usually present no problems. Not so, however, the feedback capacitance, because this may give rise to instability. Another important quantity is the input impedance. If we tabulate the necessary design data, we get something like table 1 . Figure 1. With simple means the coax cable can be used for the signal-, the supply- and the tuning voltages. impedance manufacturer; viates no more can be any- than 20% from thing between 1/S 1 and 20 k at 100 MHz output specified by the manufacturer impedance and is usually of the same order feedback 1-10 p very low; capacitance usually 0.1-0.01 p The drawback of the common-gate amplifier is that its maximum gain is less than that of the common-source circuit. On the other hand, however, the common- gate amplifier has greater reliability and stability. A secondary advantage is that the difference in matching for minimum noise or maximum gain is much less than for the common-source circuit, and is in some cases even negligible. Radio recep- tion requires matching to minimum noise, TV reception requires matching to maxi- mum power gain to eliminate cable re- flection (picture “ghosts”). The circuit (figure 2) To obtain a wide matching range, the circuit is designed around discrete coils. This also offers greater freedom as regards using other types of FET, Often mistakes are made as regards the quality factor of such home-made coils; in this case a Q- factor of 100 or more can easily be achieved. Although the diagram shows the amplifier with asymmetrical input and output, it can easily be adapted for application with symmetrical aerials by providing L, and L 3 with coupling windings. To eliminate the problem of the (wide) tolerance in the pinch-off voltage, the gates are connected to a positive voltage so that each of the FETs draws about 10 mA. For a 12 V supply voltage, the gate-drain voltage is about 6 V, and for most types 40 — elektor december 1974 Mounting, construction and adjust- ment An important requirement is that all con- nections must be as short as possible. Photograph 1 gives a clear picture of the mounting. The FETs should have much shorter connecting leads than shown in the photograph (about 6 mm); long leads have distinctly unfavourable effects on stability and the signal-to-noise ratio; this was being verified when this photograph was taken. All capacitors, except for Cu, are of the low-loss ceramic disc type. Current types of Schottky diodes can be used for Di and D 2 . and types BB105A, BB105B and BB10SG are suitable for D3 to Dj. The .-oils are wound on Kaschke coil formers type KH 5/22, 7-560-8A, with a ferrite .ore, type K 3/12/100. Several other types of coil formers might be suitable as well, if the diameter is about 1/4 in. (6 mm) The ferrite core has to be a VHF-type! The winding data are given in table 3. T*le 3. ” b r6,PeC L, aerial 50/75 fi 2 240/300 £2 4 (coupling winding) source 2.5 L 3 output 50/75 Rl 240/300 fi 2 (coupling coil) The wire should preferably be silver-plated copper wire with a diameter of 1.2 mm. The spacing between the turns is 0.8 mm and is obtained simply by winding a so- called “blind wire” of a diameter equal to the spacing, i.e. 0.8 mm, together with the coil wire. Once the coil has been mounted, this blind wire is, of course, re- moved unless the 240/300 £2 connections are to be used. In that case the blind wire is 0.8 mm enamelled copper wire, and after mounting of the coil, this blind wire is wound off again until the above number of turns is left. As the coupling coils must be placed at the 'cold end', winding back takes place from the coil end that is connected to the varicap. This is illustrated in figure 4. Soldering the wires to the former pins is a time consuming job, particularly for the wire diameter quoted here. If more value is set upon efficient mounting than on appearances, the coils are mounted direct- ly in the circuit, as shown in figure 5. The coil formers will fit only after clipping, as can be seen in this figure. In this case the coils are wound on a drill with a slightly smaller diameter (about 0. 1 mm) than the outer diameter of the coil former. If the receiver used is not tuned by means of varicap diodes, the aerial amplifier should be adjusted as follows. Set the 42 - elektor decembar 1974 tunable aerial amplifier Figure 6. To obtain the tuning voltage for the amplifier from a tuner with a high-impedance tuning voltage, such as tap presets for instance, an emitter follower is required. If a low-impe- dance tuner voltage is used, the tuning voltage for the amplifier can be obtained directly via Figure 7. Layout of the printed circuit board. Figure 8. Component layout on the PC board in figure 7. ferrite cores half way in the formers. Tune the receiver to a weak station with a fre- quency of about 95 MHz and adjust the tuning voltage - the voltage applied to the varicap diodes — to obtain a maximum output. Tune Lj and Lj to increase the output still further or to obtain a maxi- mum; adjust Li to reduce the noise of the received signal to a minimum. If the vari- cap diodes are three matched diodes, the aerial amplifier will now track correctly over the range 76 to 146 MHz. If the receiver is tuned by means of varicap diodes, the voltage that controls them can also be used to control the diodes in the aerial amplifier. However, to prevent over- loading the receiver, the voltage should be applied to the diodes in the aerial ampli- fier through an emitter follower as shown in figure 6. The tuning procedure now is as described above, except that a weak station with a frequency of about 88 MHz should be used and P| is set to give a maximum tuning voltage. Next turn the receiver to a weak station at 100 MHz, and again adjust Pi to obtain a maximum output. Tune the receiver to 88 MHz and readjust the three cores to obtain a maxi- mum output (L-2. L 3 ) with the least noise (Li ). Tune the receiver back to 100 MHz and check that no further adjustment is required; the aerial amplifier should now track correctly over the band 76 to 146 MHz. If further adjustment is needed, then repeat the whole procedure until it is not. Results and application in the 2 m amateur band The sensitivity of F.M. tuners can be limited by: 1. the signal-to-noise ratio at the input, 2. insufficient amplification of the inter- mediate frequency. Most factory-made receivers are designed so that a combination of these two factors is operative. Although it is difficult to give an exact rule for the improvement obtained by using the amplifier, it may be expected that the sensitivity of the receiver will improve by about a factor of 3 for the same signal-to-noise ratio. If still great- er amplification is required, the amplifiers can be cascaded. An amplification factor of more than 10, however, will usually give rise to cross modulation in the re- ceiver; the same amplification can also be obtained by means of one amplifier equip- ped with FETs that have a steep slope. The coils described can be used in the two-meter band, but the varicaps must then be replaced by ceramic trimmers of 1- 9 pF. The bandwidth is more than sufficient to cover the entire band. Conclusions The aerial amplifier discussed in this article is suitable for many applications and has such a low noise figure that it will improve reception in all cases. Apart from the 76-146 MHz range, the amplifier, with modified coils, can also be used to great advantage in the following bands: 14,21 and 28 MHz amateur band, channel 2- 4 TV, channel 5-12 TV, and perhaps the U.H.F. band. These further applications may be discussed in one of the next issues of Elektor. H elektor december 1974 - ■ |gn An important alternative to the mechan- ical switch - rotating or push-button - is the touch switch. This has the advan- tages of greater reliability and a higher switching speed, as well as being noise- less and not subject to wear. Furthermore, front panels with touch contacts can be made available as printed circuits, so that it becomes much easier to build equipment with a neat appearance. Elektor laboratories have been asked to design a touch control switch with a single touching point and costing no more than its mechanical equivalent. Consequently, our laboratories have produced the Touch Activated Pro- grammer or TAP. Basic possibilities Operating a switch - touching, turning or pushing — is in effect feeding in a signal that must be stored somehow. The mech- anical switches do this by remaining locked in their new positions; a touch switch, however, cannot store a signal unless it is provided with a memory. If a switch is to be operated by touch, its input resistance must exceed the resistance of the finger if action is to be ensured. If it is a single-point touch switch, the signal fed in - the signal that activates the switch - must be the noise or hum picked up by the operator. Hence, the single- point touch switch consists essentially of an a.f. amplifier that has a high input im- pedance, a rectifier and a memory. This is shown in figure 3. In this system the input signal (hum voltage on the skin) is ampli- fied in the input stage, rectified and fed december 1974 to the clock input of a flipflop. Each time the input point is touched, the flipflop will change to another stable position. A practical circuit in accordance with the block diagram of figure 3 is fairly simple to design. A TAP (Touch Activated Programmer) that will replace a complete pushbutton unit needs a reset unit between the flip- flops of the respective switches. This will ensure that when there are several switches, all except the one operated are reset. This reset can be achieved with diodes as shown in figure 4 with a four-position switch. For simplicity the contacts are shown as push-buttons. S« is the total reset button. The three-position switch shown in figure 4 needs nine diodes. In general, the reset circuit requires a number of diodes equal to the square of the number of positions. Hence, an eight-position switch (plus, of course, a total reset) requires 64 diodes. So the system of figure 4 is rather expen- sive, and the circuit becomes complicated when there are more than four positions. A touch control switch operating without reset diodes is shown in figure 5, points A/A| and B/Bi being the touch contacts. Here reset is achieved by using a common supply resistor Ri . If one of the switches is ‘on’, it draws a current of about 1 mA. The voltage drop across Ri is then 3.3V. As soon as the second switch is operated, this one, too, will want to draw 1 mA. As a result, the voltage across R| drops al- most to zero, the non-operated switch is cut off and the last switch to be operated remains ‘on’. An advantage of such a switching system is that it can be easily expanded with more and more of the same units. There is the drawback, however, that extra components are needed to create ‘hard’ binary outputs. Consequent- ly, the cost of the switch becomes so high that the financial requirements can no longer be met. A better reset system uses a one-shot (monostable multivibrator). Each time a switch is touched, this one-shot circuit feeds a short reset pulse to each flipflop. This pulse must be so short that no audible interval occurs in low frequency applica- tions of the switches. Laboratory experi- ments have shown that touch-control switches with this reset system provide the most reliable circuit. It is for that reason that they are used in the TAP. Block diagram of the TAP Figure 6 shows the block diagram of the TAP, points A, B and C being the touch A separate overall reset is provided. Each touch point is followed by an input buffer circuit (IB-1, IB-2 . . . ). These amplify the hum voltage on the skin. The input circuits of the touch points A, B and C drive the set-(S)-input of the RS flipflops. Since driving the set input of such a flipflop several times in succession will only lead to one change in its binary state, the rectifier circuit shown in figure 3 is not necessary here. The input circuits also drive the one-shot. If, for instance, point A is touched, a 50 Hz square wave will appear on the S- input of the first flipflop (FF-1). At the alektor decern ber 1974 — 45 Figure 2. Photograph of the TAP. Figure 3. Block diagram of a simple touch Figure 4. A switching system with four digital (pulse) inputs and three binary outputs. The system is designed so that in all cases only one binary output assumes a set state whilst the other outputs are in the reset state or are being Figure 5. A touch control switching system set state. This system can be expanded with an unlimited number of touch control switches. Figure 6 . Block diagram of the TAP. The letters C F, OS and IB stand for FlipFlop, One-Shot = monostable multivibrator) and Input-Buffer. The 7400 The TAP is designed around the integrated circuit type 7400, a quadruple two-input NANO. Actually, the full type number will be SN 7400, S 7400, N 7400, SN 74H00. to name a few; the letters are not so important, how- ever. To gain a good insight into the functioning of the TAP circuit, it is necessary first to take a closer look at this integrated circuit. The part surrounded by the dashdot line in figure A represents the internal circuit of a NAND gate, and each 7400 comprises four such gates. The two emitters of T| are the inputs of the NAND gate. When both emitters of Ti receive a voltage +Vb, no current flows through its base of Ti rises and the P-N base-collector junction conducts. Hence, here transistor Ti can be regarded as an assembly of three diodes. potential drops sharply. Consequently, Tj no longer conducts and, at the same time, T 4 is driven into saturation. Point C, the NAND gate output, drops to aero potential (LOW). So when both inputs of Ti are at +Vb (HIGH), the output is LOW. It is also obvious that fact the same as applying +Vb. As soon as one of the emitters of T] becomes LOW (logic - 0'). the base voltage of Ti will also drop. As a result, the base-collector junction of Ti does not conduct, T 2 is no longer driven, and the output (C) will assume a HIGH level. When the output of the NAND gate is HIGH (logic * 11 , the output level is equal to the supply voltage +Vb minus the drop in the diode D, the collector-emitter saturation voltage Figure A. Circuit diagram of a NAND gate in a7400 1C. of T 3 and the drop in the 130(2 collector resistance. This output level therefore depends on the load current. If the output of the NAND gate is LOW (logic ' 01 , the load current is fed to the supply zero via T4. The maximum load current ('sink current') is then determined by the maximum permissable current through T 4 , which is 30 mA for a 7400 1C. touched, T 6 remains off and (he NAND gale sees (his as a M’ level. The circuit diagram of the TAP Figure 8 gives the circuit diagram of the TAP. It is designed around two ICs. The four NAND gates of ICi are used to form two RS-flipflops. The first one consists of the gates N1/N2, and the second one of N3/N4. A third is formed by the gates Ns /N* in IC2 . The two remaining gates (N 7 /N$) of IC 2 form the one-shot, which provides the reset pulse. Its pulse width is determined by resistor R 8 and capa- citor C 2 . Figure 9 shows an oscillogram of a reset pulse at the output of the one- shot (pin 8 of gate N7). The pulse width is approximately 400 ns! As appears from figure 9, the reset pulse is a ‘O’. The reset pulses are fed directly to the R-input of the three flipflops without diode coupling. This is possible because the emitters of the NAND gates are The set control for each flipflop takes place via the darlington circuit consisting of two transistors described earlier. For flipflop Ni /N2 these are the transistors Ti and T 2 . The collector of Ti is con- nected direct to the set input of the flip- flop. The negative-going pulse on this col- lector, when point A is touched, is used for driving the one-shot. To achieve a good switching edge, the collector of Ti is connected to ‘1’ level via resistor Ri (in the quiescent state). As soon as A is touched, the collector of Ti switches from ‘1’ to ‘0’ and back again 50 times per second. Via diode Di this signal arrives on resistor R9. Consequently, transistor Tg becomes conductive, and the drive input of the one-shot (pin 13 of gate Ng) is drawn to supply zero, so that the one-shot produces reset pulses 50 times per second. Resistor R4 in the base of T 2 prevents this transistor being damaged by static charges on the skin. To avoid instability of the TAP, a capaci- tor C3 is connected across the supply. Capacitor Ci is provided for automatic reset when the supply is turned on. This is achieved by feeding the positive voltage surge, occurring during switch on, to the base of T 7 via R 7 . Consequently transis- tor T 7 and Tg become momentarily con- ductive, and the one-shot produces a reset pulse. As well as having a Q and Q output, each flipflop also has extra S and S out- puts. These are intended as active attenuators. In the reset condition an S-output can be regarded as a relatively high-ohmic resistance relative to supply zero. Inversely, the ^-output is relatively low-ohmic. If. via a series resistor, a digital signal is fed to an S or an S output, this S or S’ output will function as a logic- controlled attenuator. The switching speed of the various out- puts is so high that nothing of the TTL character is lost. Figure 10 shows an oscillogram of a switching edge of one of the binary outputs of the TAP. As is seen from this figure, the rise time is less than 10 ns. The circuit shown in figure 8 can be considered a universal TAP. The points RB (Reset-Bar) and CB (Contact-Bar) provide an extra output for using several TAPs in conjunction with each other. Table 1 gives the truth table of the TAP, and table 2 gives various specifications. The printed circuit board Figure 12 shows the circuit board of the TAP. All the inputs are along the upper edge of the board, and the outputs along the lower edge. The supply terminals and the RB-CB rails are on one side. Screened cable should be used for the input connections. TAP applications A simple TAP application, an on/off switch for a 220 V lamp, is shown in figure 13. In figure 14 a similar circuit for operating three lamps is shown. If the diodes Di , D 2 and D 3 are omitted from the TAP in figure 14, the result is a triple lamp switch with one common reset. In cases where a triple touch control switch with a common reset is insuffi- cient, more TAPs can be used in con- junction. The RB- and CB-rails of all TAPs used must then be interconnected. Figure 15 gives a simple example. Of course, only one TAP need be provided with a one-shot reset circuit. flickering flame The simplest possible flasher device is a bimetal switch. This construction can be found in blinker bulbs' and in the starter-switch associated with a fluorescent lamp. The possibility immediately comes to mind of using a fluorescent-lamp starter as a flasher for Christmas- tree or other decorative lights. If one uses more than one starter in some combination of several lamps or lamp-groups, highly varied and interesting effects can be obtained. -Sure 1. Ph otograph of a partly dismantled -'Jure 3. Example of ^ —.p-st rings! of unequal - s+ily variable flickering- The basic idea is shown in figure 2. The starter is wired in series with the lamp or lamp-string (such as Tree-lights). When mains voltage is applied across the series combination the inert-gas mixture in the starter becomes conductive and a current-carrying glow-discharge occurs be- tween the electrodes. One of these elec- trodes is actually a 'bimetal’, two thin strips of different metals - having two different thermal expansion coefficients - welded together. Such a bimetal will curl (or uncurl) when it is heated. In the fluorescent-lamp starter the discharge cur- rent through the gas provides the heating, and the curling of the bimetal is arranged to cause a short-circuit between the glow- electrodes. This removes the supply of heat, so that the cooling bimetal reopens the circuit a second or two later. The lamp connected in our arrangement will therefore flash more or less regularly on and off. The current which may be switched by the starter depends on the rating of the lamp for which the manufac- turer intended it. The best place to find this rating is the label on the 'ballast’ device. Alternatively, assume that if the starter (e.g. Philips type SI 0, see photo) is intend- ed for fluorescent tubes up to 80 watt rating, that it will safely switch ordinary filament lamps to this amount. Note that the starter normally becomes ‘dormant’ when the arc-type gas discharge in the fluorescent tube ‘strikes’. This is because the voltage across the steadily burning arc is too low to allow the starter- glow to re-ignite. In our application there is no suoh effect, so that the ‘starter’ will flash its load continuously. It is however possible to dream up cir- cuits in which more than one starter is combined with a split-up load in a way which makes fuller use of the properties of a given type of device. As an example take figure 3. This circuit will do the wildest things, depending on the individual starters and on the load values. Suppose that Lj has the lowest wattage. When the mains is applied it will burn more or less brightly. As soon as one of the starters makes contact, either Li or L3 will come on full and Lj will go out. When the second starter makes contact all the lamps have the full voltage applied - but almost immediately the first starter will reopen . . . M electronic loudspeaker It is widely accepted that the loud- speaker is the weakest link in the high-quality audio chain. This is particularly the case at the lowest working frequencies due to the difficulty of providing a useful air- dimensions small compared to the sound wavelength. This compels the manufacturer to adopt clever but more or less expensive constructions for the loudspeaker unit and its enclosure. The manufacturer has the resources and facilities to tackle the problems at the mechanical-acoustical stage. This article explains that the do-it-yourself approach that provides the best results at the lowest price is invariably the "electronic loudspeaker". Methods of electronically compensating for the weaknesses of loudspeakers are by no means new. As Harwood recently pointed out, a patent granted in the early 20’s already describes a “motional feed- back” system. The basic idea is to somehow derive a signal that depends on the loudspeaker’s actual movement and to compare this with the original input signal. The resulting ‘error’ signal is used to modify the drive to the loudspeaker. One way of obtaining a feedback signal is to extract the voltage that is induced in the loudspeaker’s drive- coil when the cone moves. This extraction of the back-voltage has to be done with great care if the system is to remain stable. Also, not every loudspeaker is suitable for the technique. The design described in this article has, however, behaved itself properly during many demonstrations. Apart from the fact that the electronic loudspeaker does not need a specially- mounted pickup-device, which makes it simple to build up, it can be compared to normal applications of the same driver as follows: a) the lower limit of ‘flat’ amplitude re- sponse is independent of the fundamental resonance-frequency of the driver itself (or of the driver in its enclosure). b ) distortion due to certain mechanical non-linearities in the driver can be con- siderably reduced. c) although the frequency response re- mains ‘flat’ below the fundamental res- onance frequency of the driver in its en- closure, the maximum acoustical power output falls off below this frequency. It turns out however - as will be explained later - that a 20-watt amplifier produces more than enough sound level for domestic listening situations. d) a loudspeaker operating in this kind of feedback system can produce good sound at higher as well as lower frequencies, although optimum results can only be obtained when an extended circuit is care- fully matched to the individual loud- speaker. On the other hand, the greater cone excursions associated with extended bass response will aggravate the high-range (Doppler) distortion problem, so that it is desirable to use the electronic loudspeaker only for the woofer-range. The electronic woofer The behaviour of a moving-coil woofer in a closed box can be fairly accurately pre- dicted from simple theory (see ‘loud- speaker diagnosis'). This theory can be used to find a way to improve the bass response. If one ‘looks into’ the loudspeaker ter- minals one ‘sees’ a series-connection of two impedances - i.e. a voltage divider. One of these, called the static or ‘blocked impedance’, is the value measured when the voice-coil is prevented from moving (e.g. fixed with glue). The other impedance arises because of the movement of the coil in the permanent magnetic field and is called the dynamic or ‘motional im- pedance’. We will refer to them as Z s and Zd respectively. The radiated sound energy corresponds to the dissipation in a ‘radi- ation resistance’ which forms part of Zd- The objective in operating the loud- speaker is to arrange that this dissipation will be frequency-independently con- trolled by the input signal applied to the driving amplifier. The problem is that both Z s and Zd vary with frequency, that these variations are by no means the same, and that further- more the radiation resistance has neither a constant value nor is it a constant pro- portion of Zd- Pity the loudspeaker de- signer! Let us see what can be done about this state of affairs. The approach adopted for the electronic loudspeaker is to : a) note that the static impedance Z s con- sists essentially of the voice-coil resistance and self-inductance in series and that it is sufficiently well-behaved for elimination by means of an equivalent negative output impedance of the driving amplifier. b) use this technique to deal with Zd, and then apply a compensation to the driving signal, to take care of the frequency-de- pendence of the radiation resistance. This is not too difficult for a loudspeaker acting as a piston in one wall of a closed box : it turns out (see ‘loudspeaker diagnosis’ else- where in this issue) that a ‘flat’ frequency response is obtained when the voltage across the radiation resistance is made in- versely proportional to the frequency. This can easily be done using a 6dB/octave low- pass network inserted ahead of the ampli- fier in the bass channel. This network, to- gether with the negative output impedance of the amplifier, forms the basis of the ‘electronic loudspeaker’. Summing it all up it can be stated that the radiated sound energy corresponds to the dissipation in the radiation resistance; that for a constant voltage across this resistance the dissipation will increase in proportion to the square of the frequency; that for a flat frequency response this voltage must therefore be inversely proportional to the frequency - this calls for a 6dB/octave low-pass network; that this voltage can be forced to the required value once the series impedance Z s has been eliminated by means of a negative amplifier output impedance. The driving amplifier will then automatically deliver the required drive Negative output impedance A negative output impedance can be achieved by means of the arrangement shown as a block-diagram in figure I . ‘A’ in this diagram represents the gain of the driving power-amplifier. The loudspeaker is represented as Zl, consisting of the impedances Z s and Zj in series. Zf is a feedback current-sensing impedance, con- nected between the ‘cold’ loudspeaker terminal and amplifier earth return. The voltage drop across Zf is found from: (since the current through feedback net- work f is negligible) so that : The output impedance is worked out as follows: electronic loudspeeker lber 1974 - 51 V 0 = A*Vj-V z = A(v e +f‘v z )-v z = A-v e +(Af-l)v z = A-v e +(Af-I)-|t-v 0 After some tidying up: ,0 ' a ' , ''zl-(a(-i)Z( ■ A ' ,e 'zitz; in which the output impedance has been introduced as Z„ = -(Af-l)-Zf This is negative provided that Af > 1 . To compensate the static impedance of the loudspeaker we require: Z 0 = -Z s . Assuming that this is successfully done we find: v d Zd . Z s +Z d ° _?d_. A . v - Z L Zs+Zd e Z L +Z 0 Z d .a. v . Z s -i-Zd , Z s +Zd e Z s +Z d -Z s A Ve ' The voltage drop across the dynamic im- pedance (Vjj) is directly proportional to the incoming signal voltage (v e ). This achieves the first objective. Practical aspects For many moving coil loudspeakers the impedance Z s at low frequencies is pre- dominantly a resistance: the resistance of the driving coil (R s ). It is therefore suf- ficient to use a resistor (Rf in figure 2) as the sensing element for the current-feed- back (Zf). The compensation in this range s set up by adjusting the feedback attenua- tor (f) so that: R s = (Af— 1)-Rf This can conveniently be done using the circuit of figure 2. The amount of (posi- tive) current feedback is adjusted by P 2 . Starting with the slider of P 2 at the earth end, without any input signal, slowly turn up P 2 until a ‘howl’ from the loudspeaker heralds the onset of oscillation. A slightly lower setting, for which the system just remains stable, is optimal. One or two more practical aspects appear from the circuit diagram. The buffer stage (Ti ) has been included to prevent adjust- ment of the volume control Pi from up- setting the calibration by means of P 2 . Whether this stage is necessary or not will depend on where the volume control was placed in the original amplifier. The one place where the volume control may not be located is in the power ampli- fier itself! The gain factor A must remain constant. On the other hand, if the volume control is in one of the preamplifier cir- cuits the buffer stage will usually not be needed. Low- pass network We already indicated that a 6dB/octave low-pass network is required ahead of the power amplifier. The choice of rolloff point is a compromise. The rolloff point of the network deter- mines the lower limit of compensated re- sponse. If this rolloff point is placed at 40 Hz, for example, the response curve of the electronic loudspeaker will be essen- tially flat from 40 Hz to at least 300 Hz. On the other hand it is undesirable to place this lower limit unnecessarily far down the frequency range. This is because the exten- sion of bass response has to be ‘paid for’. If we assume that the maximum current which the power amplifier can pass through the loudspeaker is ‘matched’ to the amount of force which the drive-unit can handle without damage, then the ‘price’ for an extension of flat bass fre- quency response is reduced full-drive sound level throughout the whole working range of the woofer. As the lowest working frequency is re- duced past the ‘normal’ loudspeaker-in- box cutoff, compensation of the response requires rapidly increasing amounts of drive-power for a given sound level. Since the drive-power is limited, the power re- sponse must fall off. This is not so dra- matic as it may sound, however, since the maximum power level in any normal music spectrum (including organ pedal!) rolls off at approximately 6dB/octave below about 100 Hz, so that the maximum power that the loudspeaker can deliver matches the maximum power that is required over the whole frequency range. How many watts? What is the desirable loudness level - and therefore how much power is necessary - is probably the ‘cause celebre’ of hifi reproduction. The physical situation is suf- ficiently flexible to provide grounds for ‘objective’ justification of almost any subjective opinion, while opinions vary between the extremes of ‘shatteringly loud and the devil take the neighbours’ and ‘the loudest passages should not impede normal conversation.’ We will try to steer a middle-course — based on the requirement that the maxi- 52 - elektor december 1974 Figure 4. The frequency response of a 5 ’loud- speaker in a 5 X 5 " X 6 ' (!) closed cabinet. p o = 4irr 1 X Id = 4 jt • 9 • 1 0" 3 3! 1 OOmilli- watts, where we have inserted 3 metres for distance (r) and 1(T 3 watts per metre 2 for the direct intensity (Id), i.e. 90 dB. A loudspeaker with 1% efficiency will do this on 1 0 watts of electrical input and only ‘acoustic suspension’ woofers with heavy moving systems are less efficient than this! A 20-watt amplifier for each of two stereo woofer-channels is clearly sufficient. The driving amplifier The driving amplifier used in this system must reach a very high standard of per- formance. Not every ‘high fidelity ampli- fier’ automatically satisfies the require- The most important requirement is that the amplifier be unconditionally stable, with any load. In the compensated system, after all, the apparent amplifier load is the loud- speaker's motional impedance. This ap- pears as a parallel tuned circuit: in- ductance, capacitance and resistance all in parallel! Worse still, this apparent load is the result of applying positive current feedback around the whole system . . . We previously described the ‘Equa-ampli- fier’, which meets the requirements with an ample margin. It was indeed designed with the electronic loudspeaker in mind. This amplifier, like most ‘six-transistor’ circuits, has its input and output voltages in-phase. If an amplifier which reverses the signal phase is to be used, it will be necessary to insert a phase reversal in the feedback path. This can be simply achieved by replacing the figure 2 buffer stage by a so-called ‘virtual earth’ mixer. The loudspeaker In principle the loudspeaker and its en- closure do not have to meet any severe requirements. If the best results are to be obtained, attention must nonetheless be paid to one or two details. The volume of the enclosure will deter- mine the fundamental resonance fre- quency of the compensated system - and this is the point at which the power re- sponse starts to roll off. For normal dom- estic listening a volume of 1 5 litres is ade- quate. (15 litres = 15 cubic decimetres = 0.5297200050 . . . cubic feet ... if you must!) If only background music is to be reproduced, the enclosure will do as soon as the driver fits inside it! The enclosure should also be almost air- tight. One way of achieving this is to start with a completely-sealed box, then :o drill a small hole (about 2 mm 0) in the rear panel. This will enable variations of atmospheric pressure to equalise them- selves. The amount of leakage is correct when the cone of the mounted driver takes several seconds to recover position after it has been gently pushed a small amount inwards, momentarily held station- ary and then released. (N.B. Amplifier switched off!) Finally, the walls of the box must be suffi- ciently ‘solid’. They must not vibrate - and therefore contribute to the radiation - under the influence of the strong pressure changes in the driven box. Stiffening ribs may be applied if necessary. Damping material is not strictly necessary; but a angle pad of glass-wool or similar material, lath-mounted in the middle of the en- closed volume, will control standing waves m the box. The latter can give audible trouble, particularly if the enclosure is fairly large. The drive-unit itself should in principle j meet three requirements: it must be able ‘ o handle sufficient power input; the magnet must be large enough to guarantee an unvarying flux through the entire coil curing large excursions of the cone; the cone itself and the front-surround must be reasonably stiff. It must behave as a piston! special high-compliance woofers using a rubber front-surround are less suitable for rais application, particularly when in a small enclosure. When the cone is driven cotwards at high input levels there is a tendency for the surround to be sucked ■wards! The electronic multi-way system c is best to use the electronic loudspeaker is the woofer in a multi-way system. Figure 3 shows the block diagram of such an arrangement. The amplifier Ai is a small high-quality amplifier (6-10 watts) which drives only the treble loudspeakers). If desired the reproduction of mid-range and tweeter- range may be separated. This can be done by means of a dividing net work after Ai or by the use of a separate mid-range power- amplifier A3 (dotted). The bass drive-unit and amplifier A2 to- gether form the ‘electronic loudspeaker’. The low-pass step-network described earlier is installed ahead of this amplifier. The combination must meet the require- ments mentioned above. The block diagram finally includes a buffer stage with dividing networks for the bass and treble paths. These networks, like the low-pass step network, are built up from RC sections and buffer circuits. In a further article we will describe com- plete two- and three-way systems based on the use of ‘equa-amplifiers’. Details will be given of the dividing circuits and measurement results. ^ (to be continued) In the text, figures and unavoidable formulae the following symbols have been used: Zj = static {'blocked') impedance of the 2jj = dynamic ('motional') impedance of the drive unit 2(_ = total impedance of the loudspeaker —2, - negative (driving ) impedance 2 0 = output impedance of the amplifier 2f = feedback sensing impedance P 0 - radiated acoustical power v 0 = voltage across the speech coil v e = incoming signal voltage Vj = modified amplifier input voltage v 2 = current-dependent voltage across 2f vf = feedback voltage V0 = voltage across the motional impedance v s = voltage across the static impedance R s ■ copper resistance of the driving ('voice') coil Rf = feedback sensing resistor f = feedback factor A = gain of the driving amplifier proper Id = intensity of the 'direct' loudspeaker loudspeaker diagnosis Those who need to understand the underlying theory of the working of moving-coil loudspeakers usually try to read authoritative textbooks (which tend to be thick ones). Many others who really would like to understand are frightened off by these authoritative textbooks. The present short article, intended to accompany the 'electronic loud- speaker' in this issue, outlines the way in which a knowledge of the basics of electrical engineering can give access to the 'mysteries of the moving-coil'. For simplicity we will deal with the loud- speaker in a stiff airtight ‘acoustic box’ (sometimes called an ‘infinite baffle en- closure’). The mechanical quantities de- termining what goes on are: force (f), velocity (u), mass(M), compliance (C) and damping or radiation-resistance (D). The compliance is the reciprocal of ‘stiffness’ and describes, in this case, the springlike behaviour of the cone as it moves against the suspension to cause pressure-changes in the box. Electrical engineers describe their systems by drawing ‘circuit diagrams’ containing resistance, inductance and capacitance - in which applied voltages cause currents to flow (or injected currents cause voltage It would simplify matters a great deal if we could ‘translate’ mechanical quantities into equivalent electrical quantities, and draw a ‘circuit diagram’ of the mechanical To see whether this is possible, let us com- pare the formulae describing the mechanic- dii df f = M — ; u = C — ; and f = D-u; respectively: ; and v = R-i. dt' :s of formulae Force (f) ~ voltage (v) velocity (u) . ~ current (i) mass(M) ~ inductance (L) compliance (C) ~ capacitance (C) damping (D) ~ resistance (R) The textbooks call this the ‘electro- mechanical impedance-type analogy’. A mechanical circuit diagram can be drawn, in which the inductance symbol repre- sents the quantity that ‘behaves like’ inductance - the mass - and, similarly, damping is represented as resistance and compliance as capacitance. The units are newtons (force) and metres-per-second (velocity); so that circuit values are measured in kilograms (mass), kilograms- per-second (damping) and metres-per- newton (compliance). The mechanical circuit of the moving-coil loudspeaker (at low frequencies!) in a closed box is given in figure A. The force exerted by the voice-coil is shown as a force-generator (f) with an internal im- pedance (Zed) ar >d the ‘radiation load’ on the cone front as an air-mass (M 3 ) and a compliance (C 3 ) in series with a radiation resistance (D 3 ), which is what takes up the sound power). It is convenient to ‘lump’ impedance due to the enclosed volume of air in the box (Mi ,Ci ,D, ) together with the impedance due to the suspension of the drive-unit itself (M2 £2 ,D2 ). The mechanical circuit now simplifies to that of a series-tuned circuit with damping. The resonant fre- quency is the ‘fundamental resonance’ of the loudspeaker-in-box. (At frequencies above a few hundred Hertz, other reson- ances and anti-resonances start to appear - standing-wave modes in the box, the drive-unit’s ‘edge-dip’, flexural wave pat- terns on the cone surface or ‘break-up’ - but these complications are fortunately outside the scope of this article.) The next step is to couple the mechanical circuit of the loudspeaker to an amplifier. To do this we must succeed in replacing the mechanical force generator (f) by an electrical voltage or current generator. The coupling between the mechanical and the electrical system is described by the formulae: in which B is the magnetic flux and 1 is the wire length of the voice-coil. Using these formulae we can derive: f = M ^-Bli = M -£(5)-* _ _M_ dv " dt 'Bl' (Bl) 2 dt' Comparison with the electrical formula: i = cl v dt shows that in this case (BI) 2 ~ C ' Mass, which we originally translated as inductance, turns out to be equivalent to capacitance! In the same way it can be shown that compliance is equivalent to inductance, damping is equivalent to con- ductance ( -), force is equivalent to current and velocity is equivalent to voltage. Finally, a series circuit becomes a parallel circuit and vice versa. The ‘true’ electrical circuit diagram for the loudspeaker is shown in figure B. The final step is to substitute, for the current generator, a voltage generator with an additional internal impedance: the ampli- fier (figure C). For clarity, Ldi.Cdi.Ld 2 and Cd 2 are represented as one (‘dy- namic’) impedance Zd- The voltage across this impedance (vd) is proportional to the velocity of the cone (u) in figure A (vd = Blu!) provided B remains constant. This means that if the cone is held stationary (u = 0), this voltage vd = 0. Zd could be re- placed by a short circuit ! The impedance of the loudspeaker equals Z s in this case, the ‘static impedance’ or ‘blocked impedance’. The impedance ‘seen’ at the loudspeaker terminals therefore has two parts. The ‘static’ part - which is (theoretically!) independent of any movement of the coil - is simply the series connection of the coil’s copper (or aluminium) resistance and the inductance due to parts of the magnetic circuit behaving as an iron core. Since it can only be directly measured by Figure A. 'Mechanical circuit' of a loudspeaker in which the mechanical elements are represented by equivalent electrical circuit symbols. Figure B. Equivalent electrical circuit of a loud- speaker. This is derived from the 'mechanical circuit’ of figure A by a transition in two stages. Figure C. Equivalent electrical circuit of a com- plete system with the amplifier represented by a voltage source with an internal impedance Z a The frequency characteristic of this system is determined by the variation of vq and B[j2 with frequency. Figure D. This graph illustrates the total effect. The dashed line shows the influence of the radiation resistance ID3) on the radiated acous- tical power (P 0 ) : a rise of 6 dB/oct up to a certain ('critical') frequency, which is arbitrarily chosen in this graph as 500 Hz - The dotted line shows the influence of the low-pass filter: a drop of 6 dB/oct above the cut-off frequency (f 3 , arbitrarily chosen as 40 Hzl. Finally, the full line shows the result: a ‘flat' response between f 3 and * 2 - preventing coil-movements - for example with cement - this part is often called the ‘blocked impedance’ (Z s ). When the coil is permitted to move normal- ly the ‘electrodynamic’ coupling between the mechanical and electrical circuits give rise to the other part of the loudspeaker’s impedance: the parallel-resonant-circuit- with-damping described above. This part is called the ‘motional impedance’ (Zd)- The resistance in parallel to Zd (RD2) is derived from D 3 in figure A: the air radia- tion resistance. This is a true (mechanical) resistance, in other words the acoustical energy Po = u 2 • D 3 , while in the electrical equivalent P e i = i 2 • R. We have shown that u is proportional to vd (vd = Blu), so that: Po ~v D 2 -D 3 . Conclusions: The objective of operating the loudspeaker is to obtain a ‘flat’ frequency response. This means finding a way to ensure that the dissipation in the radiation resistance is independent of frequency. This dissipa- tion is affected in two ways: 1 ) The voltage V D zp Zp + Z s + Z 0 is frequency-dependent due to the im- pedances Zp, Z s and Z 0 . 2) Furthermore, the radiation resistance (P 3 ) is not constant : it rises proportionally to the square of the frequency up to a certain frequency (usually between 300 Hz and 1 kHz). Above that frequency it remains constant. The first problem can be countered by arranging for the power amplifier to have a negative output impedance, such that Z 0 = -Z s . In this case V P Zp Zp + Z s — Z s The variation in radiation resistance can also be compensated in a simple way: an increase in power proportional to the square of the frequency is equivalent to a rise of 6 dB/oct. This can be compensated by a simple 6 dB/oct low-pass filter in front of the amplifier. When both techniques are used, the result- ing frequency response rises at 6 dB/oct up to the cut-off frequency of the low-pass filter, and from there on remains ‘flat’ up to the frequency where P 3 becomes con- stant (somewhere above 300 Hz) (see figure P). This means an almost ideal bass response, independent of the volume of the cabinet! The volume only influences the efficiency of the system, not the frequency response. The demands placed on the loudspeaker are that the magnetic system must be ‘good’ (the flux must remain constant during all movements of the voice-coil); that the cone and its surround must be sufficiently stiff (to operate as a piston); and that it must be able to handle suffi- The cabinet is only of secondary import- ance, provided it is stiff and airtight - and provided the loudspeaker fits inside! M 56 - elektor de ibar 1974 Many owners of model railways want their H H 'world of trains' to be as realistic as possible. A w w w means of imitating the sound of a real steam train ^ is, therefore, more than welcome. This article describes a simple .v. ■ method of building an electronic ^ 9 ^0WwW w circuit of few components that will produce the required sound. To add even more authenticity, the rhythm of the steam train sound is regulated automatically and is practically proportional to the speed of the train. elektor december -57 cream whistle Many model railways still run on 'steam'. For greater realism the steam locomotives are nowadays often fitted with an artificial smoke device. They become even more realistic when an imitation steam whistle is also provided. In general, electronic imitation of sounds is not so easily done. Analysis of a specific sound by looking at an oscilloscope display, or, better still, with the aid of a spectrum analyser, will make clear just how complicated that sound can be. The spectrum analyser is the clearer, because it displays the various frequency com- cuit. A steam whistle produces a tone, so that the heart of the circuit must be an oscillator. Secondly, a steam whistle is blown - which means hiss. The circuit must therefore also contain a noise gener- ator. This noise generator must modulate the oscillator. Experiment will determine which method of modulation is to be used. The circuit Figure 1 shows the complete circuit dia- gram. The sound of a real engine is pro- duced by the regular escape of waste steam. This hissing sound is produced electronical- ly by a noise generator. The rapid increase and slow fading of the noise as well as its rhythm, is controlled by an astable multi- vibrator and a pulse shaper. The output of the noise generator T 6 is amplified by transistors T 7 and Tg. The amount of noise, or noise level, can be adjusted by means of potentiometer P 2 . The transis- tors Ti and T 2 form the astable multi- vibrator which produces a square wave. The rhythm of the steam sound can be varied by means of P,. By coupling the spindle of this potentiometer to the speed control on the supply transformer for the locomotive, the rhythm of the steam sound is automatically controlled by the speed of the train. Should this arrange- ment be too difficult, the potentiometer can be replaced by a light-dependant re- sistor (LDR); practically any type of LDR will do. A suitable lamp is then connected in parallel with the power supply for the train and placed with the LDR in an opaque envelope to ensure that other light sources, such as room lighting, have no effect. The light intensity now depends on the speed of the train; this controls the value of the LDR and this adjusts the rhythm of the sound to match the speed. To ensure satisfactory control, it may be necessary to try several lamps of different wattage. The capacitors C 2 , C 3 and C4 convert the square wave produced by the astable multivibrator into a certain pulse shape. This pulse drives transistor T s quickly into conduction, but cuts it off again at a much slower rate. For a short time, transistor T s then feeds the amplified noise signal to the output while amplifying it even more, after which the amplification is re- duced slowly. The output signal can be further amplified by means of an external amplifier or radio set. The supply Tne circuit can be fed from a 9 V battery. Figure 2 shows the circuit for a mains supply. M 2 2N1613 ponents with their relative amplitudes. But even given sufficient information about the composition of a sound, its electronic imitation is still no pushover. An accurate imitation usually requires a ‘truckload’ of circuitry. An acceptable imitation, however, can be achieved with less complication. The prob- lem in this case is nonetheless the same, how to dream up a suitable circuit. Any attempt to seriously calculate component values is futile, particularly when the sound produced is only an approximation to the original. Then there is always the consideration that a spectrum analyser is not normally readily available, never mind a genuine working steam whistle! One is forced to the conclusion that trial and error is the only available approach. The circuit We already know two aspects of the cir- Assuming that the brute-force excitation of the original steam whistle gives rise to strong overtones, the oscillator will have to be some kind of multivibrator produc- ing a fairly sharp-edged waveform. The selected square-wave oscillator is a 709 in a positive feedback arrangement (and in- cluding the usual compensation). The noise-generator is a reverse-biassed base-emitter junction of an NPN transistor. At the supply voltage of 1 5 V this junction operates in the breakdown region (Zener), producing plenty of noise. Re- sistor Ri limits the current to protect Ti. Since the noise is directly injected into the oscillator feedback path, it causes an irregular frequency-modulation of the square-wave. This irregular jittering of the waveform causes the output to sound piercingly shrill - very like a real steam whistle. The pitch of the note can be varied by 58-i changing the values of the capacitors. The influence of the noise generator is largely determined by R3. Varying R3 adjusts the shrillness of the note, but one must bear in mind that it will also affect the pitch to Keying possibilities Due to the fact that almost any disturbance of the circuit has an influence on the pitch, it is not possible to key the whistle by electronically switching the feedback. The best approach turned out to be short-circuiting the points A and B. This disturbs the biassing of the 709, causing the oscillation to stop immedi- This keying can be done, of course, with a push-button (break contact) - but it is much more interesting to let the loco- motive switch the whistle on and off. This can be achieved with a Light Dependant Resistor in two operating modes. The Figure 2. The optical keying switch for the steam whistle, which will respond to either illumination or shading of the LDR. whistle sounds either when light falls the sleepers, will greatly add to the realism | upon the LDR or when the LDR is of a model railway, shielded. Figure 2 gives the circuits for Sometimes a quite weak shadow is enough both modes. When the whistle is to be to start t he circuit. Some adjustment of started by illumination of the LDR, the the sen sitivity is possible with R n . circuit with T 2 is sufficient. If the When the ambi ent light level in the ‘play- 1 triggering is to be done by shadowing the room . js on the low side> it wi|1 be LDR. Tj and R 13 have to be added. The necessary to shine extra light on the LDR. board layout in figure 3 enables either The ^ applies to the circuit that arrangement to be used. In the first case, whistles upon illumination. To start the a jumper lead is required between the circuit it is necc essary to distinctly! base and collector connections for Tj. illuminate the LDR. H The positioning of the LDR is very important. When a shadow is to trigger the whistle, the illumination under ‘silent' conditions has to be very strong. A real train usually gives a warning signal just before entering and leaving a tunnel. An LDR positioned under the track will arrange for the model train to auto- matically do the same. The same applies to a level-crossing. Here once again an LDR mounted under the track, between Another example of how Hewlett-Packard technology brings down the price of professional instruments. /////» M HM. 13 ■ a uuuauL fe cog % SBBBB (. bLLLL ■ 6 f> to 8011A PULSE GENERATOR Output Amplitude: :16V Repetition rate: 0.1 Hz - 20 MHz Output formats: norm/symm/compl Square wave mode High z/ 50 n selectable source PRICE: £ 262* To select train of 1-9999 pulses: 8011 A opt 001 : internal preset pulse counter PRICE: £419* For more details please return coupon below, or contact Hewlett-Packard Ltd., 224 Bath Road, Slough, Bucks SL1 4DS. Coming soon in BLBHTDr Electronic loudspeaker systems MOSTAP Modulation systems Cybernetic beetle DIL LED probe TUP TUN tester A. Marshall (London) Limited Dept. E 42 Cricklewood Broadway London NW2 3HD Telephone 01-452 0161 Telex 21492 & 85 West Regent Street Glasgow G2 2QD Telephone 041-332 4133 /Marshall’s Everything you need is in our New Catalogue available now price 20 pence (100 pages of prices and data) You name the electronic part you're looking for... Van Dam Elektronica has got it for you... ex-stock! Like the contents of Elektor - always in the forefront of the very latest trends and develop- ments - Van Dam Elektronica is keeping abreast of the needs of the times by constantly updating its stocks. Be sure to follow the example of so many re- nowned companies and benefit from our huge stocks, our large call orders for prominent plants and of our international relations. Whether it’s a few thousand transistors, diodes, scr's, zener diodes, integrated units, resistors or capacitors you want, your supply problem is promptly solved by our special team of experts. So be sure to contact Van Dam Elektronica in Rotterdam, Holland and state your specific problem. You can do so by telex: 25336 darnel nl or by telephone. From Great Britain dial 01031 - 10670022. If required, write us a letter stating any electronics wishes you may have. BV Techn. Handelmaatschappij \/<»ri olomri ELEhTRDmCPi Spoorsingel 49, P.O. Box 450, Rotterdam, Holland. Telephone: 01 0-670022 -Giro: 295550. electronics IlHWill.l Sbitiii litronix i $ s & Dept 9. 56. Fortis Green Road. London. N103HN I telephone 01 883 3705 LIGHT EMITTIHG DIODES Our Special Our SpecW to RED £l " £&& £2.39 TO GREEN £)/<' £2 99 £*8* £3.99 10 YELLOW £ 9.99 £399 to ORANGE £3.10 £ 2.S9 £0.90 £939 OUK PRICE £2.59 £1.83 £2.95 10 RED 1 0.175 ‘ 9/a /// 1092 tin \ £**P £M9 Buck in stock again! F.M. STEREO T USERS DESPATCH NOW IN OPERATION AT 4 LAST [sj b 1 — -i ' 1 out met mytr.tt VAT l\ AT W. OVERSEAS CUSTOMERS DEOUCT VAT INVOICES ON REQUEST PA P On UK Orders mm 15p CALI.ERS WELCOME INDUCTORS TONING CONDENSERS DIODES TRANSISTORS REGULATORS DIGITAL SWITCH ALUMINIUM BOXES TNiRISTORS SCR's OVERSEAS CUSTOMERS DEDUCT ? VAT INVOICES ON REOUEST PAP On U.K. Orders mr 15p Overs Dept. 9 56. Fortis Green Road. London. N10 3HN. telephone 01-883 3706 Another example of how Hewlett-Packard technology brings down the price of professional instruments. 1220A OSCILLOSCOPE Bandwidth: DC to 15 MHz Sensitivity: 2mV/cm to lOV/cm Vertical accuracy: 3% Sweep: 0.1 ps/cm to 0.5 s/cm CRT 8 x 10 cm internal graticule TV-trigger, beam-finder PRICES: 1220Adual channel £ 409* 1221A single channel £ 322* For more details please return coupon below, or contact Hewlett-Packard Ltd., 224 Bath Road, Slough, Bucks SL1 4DS. The new Rank WOW& FLUTTER Meter Type 1742 Fully transistorised for high reliability Versatile Meets in every respect all current specifications for measurement of Wow, Flutter and Drift on Optical and Magnetic sound recording/reproduction equipment using film, tape or disc Full details, please, on the 1220A Oscilloscope and other cost effective professional instruments from Hewlett-Packard. Position Company/Institution Address High accuracy with crystal controlled oscillator Simple to use accepts wide range of input signals with no manual tuning or adjustment Two models available: Type 1742 ' A BS 4847: 1972 DIN 45507 CC1R 409-2 Specifications Type 1742 'B' BS 1988: 1953 Rank Kalee Specifications For further information please address your enquiry to Mrs B. Nodwell Rank Film Equipment, PO Box 70 Great West Road, Brentford Middlesex TW8 9HR Tel: 01 -568 9222- Telex 24408- Cables Rankaudio Brentford HEWLETT ihp] PACKARD T ^ I^BANK FILM EQUIPMENT EK-11 FOR FURTHER DETAILS SIEMENS Microwave tubes. Just a smaM part of our capability in electronics. Siemens high-power travelling wave tubes are specified in Aerials 1, 2 and 3 of the Post Office Earth Station for satellite communication at Goonhilly Down. The YH 1045 travelling wave tube can provide up to 6 kW CW RF power in the frequency range 5.925 to 6.425 GHz and is used as the final amplifier in the transmitter system. Siemens teams of professionals have been responsible for the development and installation of these and simular tubes in other earth stations around the world. With acknowledgments t c Post Office and Marconi Communication Systems This expertise enables Siemens to meet the growing demand for new generation microwave tubes - but this is just one part of the total capability in the design, development and manufacture of Siemens electronic components and equipment. Siemens now provide a major source of microwave tubes for customers throughout the world, as well as complete ranges of devices for all sections of the electronics industry. Siemens Limited, Great West House, Great West Road, Brentford TW8 9 DG, Middlesex, Telephone 01-568 8281. Telex 23176. Profit from the Siemens experience. The life and efficiency of any piece of electronic equipment can rest entirely on the solder used in its assembly. That is why for utmost reliability leading electronic manufacturers in the USA and in 106 other countries throughout the world insist on using Ersm Multicore Solder. It's the solder they have depended on for consistent high quality for more than 30 years. If you are not already using Ersin Multicore Solder it must be to your advantage to investigate the wide range of Specifications which are available. Besides achieving better joints - always - your labour costs will be reduced and subsequently savings in overall costs of solder may be possible. There are well over 1 .000 Specifications, made to all International Standards to choose from, and here are just a few of the special solders that we manufacture: Savbit Alloy - dramatically reduces erosion of copper wires and printed circuits and also reduces the wear of soldering iron bits. 96S Silver Solder - highest strength soft solder Melting point 221 °C. Bright and non-toxic. Replaces high temperature brazing alloys. 95A alloy - Melting range 236-243°C. For electrical connections subjected to peak temp, of approx. 240°C. H.M.P. alloy - Melting range 296-301 °C. Highest melting point soft solder for high service temperature applications. T.L.C. alloy - Melting point 145°C. Lowest melting point Ersin Multicore solder for making joints on top of other solders and for heat sensitive components. L.M.P. alloy - Melting Point 179°C. For soldering silver plated surfaces such as ceramic capacitors and soldering gold. Alu-Sol Multicore Solder -for soldering aluminium. Arax acid-cored solder - for non electrical applications or pre-tinning of parts of difficult solder- ability (flux residue must be removed) which can then be assembled with Ersin Multicore Solder. Write for Technical Bulletins, on your Company's letterhead, for products which interest you to: Multicore Solders Ltd. Maylands Avenue, Hemel Hempstead. Hertfordshire, HP2 7EP Tel: Hemel Hempstead 3636 'Telex: 82363 Why have leading USA manufacturers specified British made Ersin Midticoie solder tor over 30 years?