Fullscreen HTML5Med Res (??MB)HTML4

A high quality audio oscillator: Pt.1 Just where do you go to obtain an oscillator which will put out high quality sine and square waves at up to 10 volts RMS and with 3-digit frequency resolution? Well look no further. This unit will do stirling service in any workshop or laboratory. By LEO SIMPSON Finding a good quality commercial oscillator these days is pretty hard, particularly if you want low distortion and high output. Sure, there are plenty of function generators which will cover the range from DC to daylight (eg, from a fraction of a Hertz to 10MHz or more) and there is no doubt that for many laboratory and workshop tasks, function generators are perfectly adequate. Besides their wide frequency range they are quick to stabilise in amplitude and they often have a digital readout so that you know the exact frequency. However function generators can 42 SILICON CHIP rarely do better than about 0.5% harmonic distortion and for many applications, particularly those involving measurements of loudspeakers, amplifiers and other audio equipment, 0.5% harmonic distortion is just not good enough. In addition, the harmonic distortion of function generators is generally quite "spiky" in nature (ie, high order harmonics) and therefore quite audible, particularly when working with loudspeakers at the lower frequencies. A reasonably high output voltage is also a relative rarity as far as many commercial sine/square wave oscillators are concerned. Many oscillators will put out a maximum of 3 volts but we wanted a lot more than that - 10 volts RMS was our target. Why? Well, for example, when testing the input overload capability of preamps you want 5 or 6 volts or more. Also on our wish list was good envelope stability and quick settling. Many low distortion oscillators have poor envelope stability and long settling times. Partly this is due to the use of a thermistor in the feedback loop of the oscillator and partly it is due to the use of a less than perfectly matched dual ganged potentiometer for frequency setting. Other desirable points for a high quality oscillator are ease and repeatability of frequency setting as well as a wide range output attenuator. Having outlined some of our wishes, let us now discuss the features of our new Audio Sine/ Square Oscillator. As can be seen from the photographs, the oscillator is hous- ed in a standard Horwood instrument case made from black Marvipla te and measuring 307mm wide, 236mm deep [including handles) and 108mm high. On the front panel is a meter with three scales: O to lV, O to 3.16V and - 15dB to + ZdB. To the left of the meter are two BNC sockets which provide the main oscillator output and a sync output which can be connected to the sync input of an oscilloscope or to a digital frequency counter. Above the two BNC sockets is the main attenuator which varies the output in nine lOdB steps from 1 millivolt to 10 volts RMS. The attenuator also has a GND position [fully anticlockwise) which ensures that the oscillator output is fully off. This is handy when you want to remove the signal but you don't wish to disconnect the oscillator as, for example, when making signalto-noise ratio measurements. To the left of the BNC output sockets are two miniature toggle switches. One selects the sine or square wave modes while the other connects the case of the oscillator to mains earth or lets it float. Again, the ability to do this can be important in some measurement setups. Above the two toggle switches is a small knob which provides a variable output control. This allows the output to be reduced to zero on any position of the main attenuator. Frequency setting The four remaining knobs on the front panel are for frequency setting. One is the 4-position Range switch while the other three set the frequency with 3-digit resolution. Most equivalent audio oscillators use a dual gang potentiometer as a variable frequency control which has the advantage of being continuously variable but there are a number of disadvantages. First, the frequencies are inevitably cramped at one end of rotation. Second, the frequency scale must be carefully designed to suit a particular dual pot. If the pot varies at all in its overall resistance and its electrical rotation, the frequency scale is likely to be quite inaccurate. SPECIFICATIONS Frequency Range Frequency Resolution Frequency Accuracy Harmonic Distortion Squarewave Rise Time Squarewave Fall Time Output Level Output Impedance Load Impedance Protection Sync Output Sync Output Impedance Third, for lowest distortion, quick settling and best amplitude stability, the dual pot needs to be a high quality wirewound dual potentiometer with closely matched sections. Such a potentiometer is about as rare as a mint quality Ford Prefect - you just can't get 'em! Hence, we have gone for the more complicated but more readily available arrangement of three rotary switches for frequency setting. The switches give precise and accurate frequency setting, good envelope stability and contribute to the low distortion. We should note at this point though, that the multiwafer switches are expensive and a major part of the total oscillator cost. Also on the front panel is a red LED which functions as a power indicator. This is necessary to tell you that the oscillator is on because when the output is turned down, the meter will be indicating zero. The power switch is on the rear panel. It is there to keep all the internal mains wiring as far away as possible from the sensitive oscillator circuit and keep hum output to the absolute minimum. Specifications Most of the relevant perfor- 1 OHz to 109.9kHz in four ranges with 11 % overlap between ranges 3 digit Typically better than ±2% 20Hz (0.1% 1 OOHz (.02% 1kHz ( .0025% 10kHz ( .003% 20kHz ( .01 % 1 OOkHz ( .05% (all at 1 OV RMS) (30ns (20ns Continuously variable from 0-1 OV RMS in nine ranges 6000 (nominal) 6000 to infinity Short circuit protected 1 OV RMS sinewave 10k0 mance specifications are included in the panel accompanying this article. As you can see, the harmonic distortion figures are quite respectable although they are not extremely low. This is mainly as a result of the circuit not using a thermistor in the feedback loop for envelope stability. A thermistor could be used to give a better result but envelope stability would probably not be anywhere near as good. Apart from that reason, we have not used a thermistor because suitable units are now very expensive and at times they can be impossible to obtain. Some readers may wish that we had designed the oscillator to cover lower or higher frequencies. We have not done so for two reasons. First, we had to draw the line somewhere and increasing the frequency range would have required adding at least one more range. Second, operating an oscillator of this circuit configuration with a wider frequency range does not give particularly good distortion figures or good envelope stability a function generator is a better performer at very low and very high frequencies. Square wave rise and fall times JA NUA RY 1990 43 Cl Rl R2 THl R3 .,.. Fig.1: typical Wien bridge oscillator circuit. The thermistor (TH1) in the negative feedback path increases in resistance as it warms up and stabilises the output amplitude. Cl Rl R2 We '11 talk a bout these in some detail later. Before getting down to the nuts and bolts of the circuit operation, let us first explain how a Wien bridge oscillator works. Refer to Fig.1 which shows how a Wien bridge oscillator is usually connected. Fig.1 shows an operational amplifier with two feedback networks. The first network consists of resistor Rl and capacitor Cl in series from the output to the noninverting ( + ) input together with resistor R2 in parallel with capacitor C2 from the non-inverting input to the OV line. This network gives positive feedback from the output to the non-inverting input. The second feedback network consists of thermistor THl from the output to the inverting ( - ) input and then resistor R3 from the inverting input to the OV line (GND). This network provides negative feedback. ":' Pseudo resonance Fig.2: the thermistor can be eliminated by substituting an incandescent lamp and rearranging the feedback as shown here. are quoted at less than 30ns for the rise time and less than 20ns for the The network consisting of Rl, R2, Cl and C2 is known as a Wien network and it acts in a similar way to a tuned LC circuit. And just as a tuned LC circuit will give a resonance at a particular frequency then so does the Wien network give a "pseudo resonance". The frequency at which this pseudo resonance occurs is given by the formula: 1 Fo = 27!" jR1.R2.c1.c2 fall time. Verifying these figures can be difficult, depending on the CRO and probes used. To do it accurately, you need a CRO and probes with a bandwidth of at least 100MHz. The typical 20MHz CRO and its 10:1 probes are just not good enough. We used a 150MHz CRO and 250MHz probes. Circuit details Let's face it, the circuit is relatively complicated although it does have a number of elegant features. It is based on the conventional Wien bridge configuration but as noted above, it does not use a thermistor in the feedback loop for envelope stability. Instead, it uses one or two incandescent lamps. 44 SILICON CHIP -~;::::::::======= Now if Rl is made equal to R2 and Cl is made equal to C2, the formula is simplified to: F _ _1_ 0 - 27l"RC At this "resonance" frequency, the phase shift from output to input will be zero (or a multiple of 360°) and the transmission loss through the network is a minimum which is actually 3.0. Another way of saying this is that the gain is 0.33. Now for the system to oscillate with a steady amplitude, that loss of 0.33 via the positive feedback network must be exactly cancelled out. To achieve that, the negative feedback network must set the gain of the amplifier to precisely 3.0. When that happens, the circuit oscillates with a steady amplitude. In some books or magazine articles on Wien bridge oscillators you may see a reference to the "gain around the loop being equal to unity" if steady oscillation is to occur. This is correct but is not easy to understand. Think of it this way. The gain in the Wien RC network is 0.33. The gain from the noninverting input to the output is then equal to 3. If you multiply 0.33 by 3 the result is unity. Non-linear feedback element The problem in any Wien bridge circuit is how do you maintain the gain of the amplifier at exactly the right value? That is what the thermistor is there for. It has a negative temperature coefficient so that if its temperature rises, its resistance drops markedly. It works as follows . When the circuit first turns on, the thermistor will be cold and its resistance will be high. Therefore the negative feedback around the amplifier will be low and the oscillations in the circuit will build up rapidly. As the voltage at the output rises, current will pass through the thermistor and it will start to warm up. As it warms up, its resistance will drop and the negative feedback will increase. This means that the overall circuit gain will fall and so the speed at . which the oscillation is building up will be reduced. Eventually, the circuit will reach equilibrium and it will oscillate at a steady amplitude. The time which it takes to come to this steady state is the "settling time" . So we see that by using a nonlinear component such as a thermistor, we can stabilise the amplitude of oscillation in a Wien bridge circuit. Incandescent lamp But earlier we said that our circuit does not use a thermistor Fig.3 (right): the circuit uses a low► distortion amplifier (Q1-Q10) which oscillates due to the positive feedback components selected by switches S1-S4. 11.7k 1Bk// 33k 1.95k 3.9k// 3.9k 3.9k 1.17k 1.Bk// 3.3k 78011 1.1k// 2.7k 4151! 560!! 3250 560!!// 620!!// 1.6k 680!! 260!! 2.34k 300U// 2k 4.3k// SB A~ 5.1k D5 M1124 '-o-, ~2~4~V--~H-+------t ~A1C2V 240V ::) S1a 117k 180k// 330k 0 7.Bk 11k// 27k 26k 30k// 200k 4.15k 5.6k// 16k N ~~1-2v_-+----+---+---+---<t----+--ov 1000 35VW nfT7 0 .J: 1 25VW _ - 3.3k 100 15VW CASE S2a 1.17M 1.BM// 3.3M 390k 117k 18Dk// 330k 78k 110k// 270k 56k 41.5 k 56k// 160k 32.5k 62k// 68k 260k 300k// 2M 1N4004 S3a 0.68 15!! S4b 2.34k 4.3k// 5.1k S4c x1 ,10,0 3.25k 6.2 k// 6.8k * ADJUSTFOR 15mA QUIESCENT CURRENT ~1k q100 J. .006~ 0.68~06BI ,.~470pF // d_. 180pF , 10 ,100 4.15k 5.6k// 16k x1 5.6k 06 x1k 1xBC546 S4d 1.1k 2.2k 560!l -11v---...--➔--+-----------------➔--~--~ 780!! 1.1k// 1.7k 1.17k 1.Bk// 3.3k 1.95k J°.9k// 3.9k 10k 3.9k S1b SINE S3b SYNCcp 17k S5c 11.lk 18k// 33k SOUARE +21V s: OUTPUT 10k 014 BC639 470!! 012 51 O~! BC640 ..--1--+-~t--t 8 2.1 BP 160U 51\l 161 1 10k 7.511 -11v---<t----..,__ __ LM317 LM337 ~ 0 0 0 <at>J Ill BC546.548. 556,557 B EOc BC639 ,640 C aOE o<at>s 01 51U FLOAT~ VIEWED FROM BELOW ••~• ,ru~oo, OUT IN SG O m GND9 -d:,. Sl - 16!.? lflrl CASE AUDIO SINE/SQUARE OSCILLATOR JANUARY 1990 45 lkHz sine wave at maximum output: 0.2ms/div and 5V/div. Square wave rise time is less than 30ns; shown at 20ns/div. Square wave fall time is less than 20ns; shown at 20ns/div. 20V RMS square wave output at 100kHz. Note the lack of ringing. 20V RMS square wave output at lkHz. It has even less ringing. 20V RMS square wave output at 100Hz. Note lack of droop. because suitable units [such as the well known R53) are expensive and can be hard to get. So our circuit is a variation of Fig.1 and is shown in Fig.2. Here, the thermistor has been replaced by fixed resistor R3 and resistor R3 in Fig.1 has been replaced by incandescent lamp Ll. Incandescent lamps are also nonlinear but they have a positive temperature coefficient. When their temperature rises their resistance increases sharply. By arranging the negative feedback as shown in Fig.2, the incandescent lamp achieves the same result as in Fig.1. Oscillation is steady after an initial settling time. By designing an amplifier which has a very low distortion to begin with and then by carefully selecting a non-linear feedback element, in our case an incandescent lamp, the oscillator will operate with very low distortion. So the complete circuit of the oscillator is essentially an amplifier with very low distortion which is then provided with suitable feedback components around it to make it oscillate. resistors using Slb, S2b and S3b. These are equivalent to R2 in Fig.1 [or Fig.2). Circuit details Switch banks Now have a look at the complete circuit. Before we go into detail, let's just locate the main sections of the circuit. At the top righthand corner is the power supply which puts out ± 22 volts. Then below the power supply is the low distortion amplifier which uses BD139 and BD140 transistors (Q9 and QlO) in the output stage. To the left of the power section is a group of three switched resistance banks, using switch sections Sla, S2a and S3a. This is equivalent to Rl in Fig.1. Below the switched resistance bank are a number of ganged switch sections, S4a, S4b, S4c and S4d, which select capacitors. These are equivalent to Cl and C2 in Fig.1. Finally, to the left and slightly below the four section switch S4 is another group of three switched It won't be apparent just what is happening in the switch banks when you first look at them so we'll fill in some of the details. First look at the three resistor strings associated with Sla, Slb and Slc. Each switch section is wired as a variable resistor and the three sections are in parallel. Let's now assume that the frequency multiplier switch S4 is set to the "xl" range. Now any frequency combination selected by the frequency range switches [ie. Sl, S2 and S3) will be multiplied by one and we can select any frequency between lOHz and 109.9Hz. The lowest resistance string, at the top, is the most significant digit in the frequency, and in this case, sets the frequency in multiples of 10; ie. lOHz, 20Hz, 30Hz, 40Hz, 50Hz and so on up to lO0Hz. 46 SILICON CHIP PARTS LIST 1 Horwood instrument case , 305 x 1 02 x 203mm 1 24V centre-tapped transformer (Altronics Cat. M-7124) 1 240VAC 15A plastic bodied SPOT toggle switch (Altronics Cat. S-3220) 1 SPOT miniature toggle switch 1 3POT miniature toggle switch (Jaycar Cat. ST-0505) 1 single pole 12-position rotary switch 3 2-pole 12 position rotary switches with screen plate (from Farnell Electronics, see text) 1 4-pole 6-position rotary switch (from Farnell Electronics, see text) 4 21 mm collet knobs and caps for 6mm shafts 1 21 mm collet knob and cap for 6.4mm shaft 1 1 5mm collet knob and cap for 6.4mm shaft 2 lamps , 28V 40mA (Farnell Electronics Cat. CM 7 3 7 4; see text) 2 insulated BNC panel sockets 1 1 00µ,A MU-65 panel meter 31 1 .2mm PCB pins 4 6mm high spacers 4 10mm spacers tapped 3mm 5 3 x 1 2mm screws 8 3 x 6mm screws 5 3mm nuts 1 solder lug 1 3-core mains cord and plug 1 cordgrip grommet 1 metre 250VAC rated hookup wire 3 metres hook-up wire 4 stick-on rubber feet Printed circuit boards 1 oscillator board , code 04101901, 207 x 93mm The second resistance string, just below, is ten times higher in value and sets the frequency in units. For example, if S1 is set to 60Hz, S2 enables the frequency to be set anywhere from 60Hz to 69Hz. The third resistance string, is ten times higher again in value, and sets the frequency in multiples of 0.1Hz. So if S1 and S2 have been set for 65Hz, S3 enables the frequency 1 power supply board, code 04 101902, 108 x 64mm Semiconductors 2 1 N41 48 silicon diodes (01 ,02) 2 OA90 germanium diodes (03,04) 2 1 N4002 rectifier diodes (05,06) 3 BC557 PNP transistors (Q1 ,Q2 ,Q3) 2 BC556 PNP transistors (Q4,Q5) 2 BC546 NPN transistors (Q6,Q7) 1 BC548 NPN transistor (Q8) 1 BD139 NPN transistor (Q9) 1 BD140 PNP transistor (Q10) 1 VN 1OKM N-channel Mosfet (Q11) 2 BC640 PNP transistors (Q12,Q15) 2 BC639 NPN transistors (Q13,Q14) 1 7 4C14 hex Schmitt inverter (IC1) 1 LM78L 12 positive regulator 1 LM31 7T variable positive regulator 1 LM337T variable negative regulator 1 5mm red LED (LED 1 ) Capacitors 2 1 OOOµF 35VW PC electrolytic 1 330µF 25VW PC electrolytic 4 1OOµf 25VW PC electrolytic 1 1 OOµF 16VW PC electrolytic 2 1 OµF 25VW PC electrolytic 1 2.2µF 50VW BP electrolytic 6 0 .1 µF 63V polyester 1 22pF 50V ceramic Close tolerance capacitors 2 0 .68µF 63V 2% polycarbonate (Mayer Kreig NSR 680 VG 63) to be set anywhere from 65Hz to 65.9Hz. Note that many of the resistance values on S1a, S2a and S3a are parallel combinations of two resistors. This was necessary to give the precise values we needed. Note also that exactly the same resistor values are used with S1a, S1b and S1c. This is to be expected since Slb is ganged with Sla, S2b is 2 .068µF 1 OOV 1 % polypropylene (Mayer Kreig MKP 1837-368-013) 2 .0068µ,F 63V 2 .5% polypropylene (Mayer Kreig KP 1830-268-063) 2 470pF polystyrene 2% 2 180pF polystyrene 2% Potentiometers 1 1 kO linear Trimpots 1 1OkO horizontal mount 1 5k0 horizontal mount 1 1 kO horizontal mount 1 5000 horizontal mount 1 2000 horizontal mount Resistors (¼W, 1 %) 2 3 .3MO 6 3.9k0 2 2MO 4 3.3k0 2 1.8MO 1 3.3k0 ½W, 5% 6 39 0k0 2 2 .7k0 4 330k0 3 2 .2k0 2 2k0 2 300k0 2 270k0 2 1.8k0 2 1.6k0 2 200k0 4 180k0 2 1 .1 kO 2 160k0 2 1 kO 2 11 OkO 3 6800 2 6200 2 68k0 2 62k0 5 5600 2 5100 4 56k0 1 4700 6 39k0 2 3000 4 33k0 2 30k0 2 2000 1 1800 1W, 5% 3 27k0 4 1600 4 18k0 2 1500 2 16k0 1 1000 2 11 kO 1 820 7 10k0 2 510 1 8.2k0 2 160 3 6 .8k0 2 150 2 6.2k0 2 7 .50 5 5 .6k0 2 6 .80 2 5.1 kO 2 4 .3k0 ganged with S2a and S3b is ganged with S3a. Oscillator amplifier It is often said that any amplifier can be an oscillator and any oscillator can be an amplifier - it is just a matter of how the feedback works. In our case, we start with a low distortion amplifier and then make it oscillate by connecting the JANUARY 1990 47 The oscillator is built into a metal case. There are two PCB assemblies: an oscillator board and a power supply board. Wien network around it. The amplifier is very similar to some of the power amplifiers we have described in the past except that the output stage does not use high current power transistors. Transistors Q2 and Q3 form a differential input stage with their operating current set by constant current stage Ql which is referenced by diodes Dl and DZ. The outputs of the first differential stage are fed to another differential amplifier stage consisting of Q6 and Q7. These two transistors have their operating currents set by the "current mirror" consisting of Q4 (which is connected as a diode) and Q5. Q7 drives the complementary emitter follower output- stage consisting of Q9 and QlO. These two transistors operate in class AB with a collector current of 15 milliamps, The frequency determining components are all mounted directly on the rotary switch sections. The unit covers from 10Hz to 109.9kHz in four ranges. 48 SILICON CHIP as set by the "Vbe multiplier" QB. The whole amplifier is DC coupled throughout and has negative feedback set by the 5000 trim pot VRl in series with a 5600 resistor. The shunt part of the negative feedback network is provided by two 24V 40mA miniature incandescent lamps. High frequency compensation, to ensure that the amplifier is stable, is provided by the 22pF capacitor connected between base and collector of Q7. Some readers may wonder why we have used an output stage operating in class AB. Wouldn't class A give better distortion? As a matter of fact, it wouldn't. The reason is that the amplifier is an oscillator operating at a constant large output voltage swing of 28.28V peak to peak. At this large voltage, any small crossover distortion effects which may be present are vanishingly small. Well, that's about all we have space for this month. Next month, we'll complete the circuit description of our new oscillator and conclude with the full construction and setting up procedure. !b<