Fullscreen HTML5Med Res (??MB)HTML4

AUDIO OUT AUDIO OUT L R By Jake Rothman A switchable Class-A/Class-(A)B amplifier S ometimes it’s worth making a simple experimental discrete circuit for fun rather than a paid job. I have an old Hacker radio, like the one shown in Photo 1, with an excellent loudspeaker, case and FM section where the audio amplifier PCB was a write off. It was an unusual design running on two PP9 batteries feeding an 8×5 inch elliptical 25Ω Goodmans speaker (shown in Photo 2). A pair of batteries costs over £10, which was more than a power supply, but I still wanted to retain battery operation for my Radio 4 fix in the bath where they last about three years. For this application, keeping the current draw to a minimum is important, so Class-B operation is required (where there is basically no standing current through the output transistors at idle). When a mains supply is being used, Class-A will draw more power but will provide better sound quality as it eliminates crossover distortion, where one output transistor stops sinking or sourcing current and the other takes over, which happens at every zero crossing. That is inevitable in Class-B due to the lack of standing current but is totally absent in Class-A, where both output transistors are conducting all of the time, regardless of the instantaneous output voltage. Ideally, in a radio that can run from batteries or a mains supply, the amplifier can switch from Class-B (strictly speaking, Class-AB, where there is a small quiescent current) to Class-A automatically when connected to the mains. There’s no cheap chip I can find that does this, and Douglas Self’s Trimodal amplifier would be overkill. That made designing a simple circuit worthwhile. This little amp is good practice to prepare for more expensive designs later, the parts cost being only about that of a cup of coffee, so a blow up is of little consequence. It’s also useful for other things, such as musical instruments. Other old radios that this circuit could be used with, having 25Ω speakers and 18V supply rails, include the Robert’s 404, Ultra Transistor Six and the Perdio PR25/36/27. The latter was also available as a Lasky’s/Henry’s Radio kit for Radio Constructor magazine. The circuit The circuit is shown in Fig.1. It could easily have been made around the time of the original radio in 1973; it uses a four-transistor topology that was popular at the time. This is about the minimum possible to make an amplifier that sounds okay. It gives about 1.3W RMS, sufficient for the high-sensitivity loudspeaker used. Since there is a very limited openloop gain of around 200, the distortion in Class-B is high, but in Class-A it sounds much smoother. That’s because the effectiveness of negative feedback in cancelling distortion is related to the open-loop feedback divided by the closed-loop feedback, so a higher openloop feedback (well over 1000 times) is desirable. Our closed-loop gain (ie, the amplifier’s voltage gain) is 25 times, so with an open-loop gain of 200, that only gives Photo 2: a Hacker radio; the original audio amplifier PCB is at the base of the cabinet between the two large PP9 batteries. 66 Photo 1: many old radios can benefit from this design. A Hacker radio is shown here next to the green telephone at RetroTech 2024 (photo by Harvey Rothman). Practical Electronics | May | 2025 JFETs are depletion-mode devices that conduct a little current until they are biased off (similar to valves), so in this case, a positive voltage on the gate will switch it off for Class-A operation. The gate is normally pulled to ground by R4 unless switched into Class-A. Normally, 90mA of standing current would result in thermal runaway of the output stage. This occurs because the Vbe necessary to bias a transistor on drops with increasing temperature. As the temperature increases, if the bias voltage remains constant, the collector current increases, as does power dissipation, increasing the temperature. This is positive feedback! This is prevented by a Nelson Pass designed feedback circuit derived from his A40 amplifier design (in Audio Amateur magazine, 1978) consisting of R9 and D2. This senses the current across the emitter resistor (R11) and switches on TR2 harder as the output stage current increases. This, in turn, reduces the bias voltage across the output transistors, reducing the current and ensuring thermal stability. The effect of this circuit is shown in Fig.2. One problem with these types of output current feedback circuits is that the audio signal can modulate the current, causing second harmonic distortion. This is reduced by using a large bypass electrolytic capacitor (C5) across TR2. This is not very effective at low frequencies, but the distortion of the loudspeaker is going to be much worse a feedback ratio of 200 ÷ 25 = 8. That means that any distortion generated by the signal path (the non-linearity of the transistors etc) will be reduced by about 87.5% (1 – ⅛). By comparison, if the open loop gain were 2500, the distortion would be reduced by 99% (1 – ⅟100). The high-impedance load helps; it would be very difficult to make this circuit sound acceptable with a 4Ω speaker. TR1 is a common-emitter voltage amplifying stage with a collector current of 4.5mA feeding a push-pull emitterfollower output consisting of TR3 and TR4. The output stage is biased into Class-AB (ie, slight conduction) for battery operation by Vbe multiplier stage TR2, which can be thought of as an adjustable zener diode. The output stage standing current is set to 3mA by adjusting VR1, giving a total supply current of 7.5mA, which gives good battery life. TR2/VR1 set the quiescent current by setting the voltage between the bases of TR3 & TR4 to a fixed bias of about 1.2V, sufficient to get them just into conduction. Mode changing For Class-A operation, around 90mA of collector current for TR3 & TR4 is needed, and this is set by potential divider R5 and R6 feeding the base of TR2. To switch from Class-B to Class-A operation, VR1 has to be switched out. This could be done with an ordinary switch, but in this case I’ve used a P-channel JFET so that the class switching can be voltage controlled. D3 BAT86 R8 1.5kΩ Take for Class A VCONT Class B Iq set VR1 5kΩ C1 2.2µF 10V + Input R1 10kΩ 0V TR3 BD135 D2 1N4148 R5 6.8kΩ R9 2.2kΩ + TR5 2N5460 P-channel C6 100nF 2.8mA Class B 0V 4.3mA R3 10kΩ R4 470kΩ + C9 1000µF 25V V+ 18V TR2 BC549C R6 2.4kΩ TR1 BC549C C5 47µF 10V R11 3.3Ω 9.2mV Class B 0.3V Class A +9V R12 3.3Ω TR4 BD136 Isupply / mA + 0V C4 47µF 16V R7 470Ω + C2 + 100µF 25V 7.5mA or 90mA R13 22Ω C7 470µF 25V C8 100nF below its resonance of around 120Hz. D1 exists to ensure a symmetrical path for negative and positive currents through the input circuit (along with TR1’s base-emitter diode junction). Otherwise, DC could build up on C1 under overload conditions, such as if it were used for a guitar amp. Resistors R7 and R8 form a collector load for TR1. This is bootstrapped by C4 to boost the voltage drive and increase the open-loop gain. R8 could be replaced by a 4.5mA current regulator diode or another constant current source for lower crossover distortion, but the maximum power output would be reduced by about 30%. Diode D3, along with capacitor C2, provides decoupling for the input stage. Normally, a resistor would be used in place of D3, but by using a low forward voltage diode here, such as a schottky type, more output voltage is obtained. Decoupling is very important to minimise distortion when running on battery since the pulsating output current would otherwise modulate the supply voltage. This is further minimised by placing a large electrolytic capacitor (C9) across the main supply rails. The closed-loop voltage gain is set by R1 and R10 to around 25× in a similar manner to an inverting op amp circuit, with R10 being the negative feedback resistor. R2 is necessary to bleed off some bias current to set the average output level at half-rail. This may need some adjustment depending on the Hfe and Vbe of transistor TR1. Lower distortion can be obtained by increasing the value of R1 but this reduces the gain, and with many radios, a preamp gain stage will be required to get a decent output volume. There are a few extra capacitors needed for high frequency stability, such as for phase lead compensation C3, Zobel network C8/R13 and high-frequency decoupling capacitor C6 in parallel with the bulk bypass capacitor, C9. Output Bang! No feedback network 180 170 Thermal runaway 160 LS1 25Ω 2W 150 140 130 D2 R9 2.2kΩ 1N4148 With feedback network 120 110 D1 1N4148 100 0V 0V R2 27kΩ R10 270kΩ 90 80 70 60 C3 27pF Fig.1: the switchable amplifier circuit. The transistors are all 1970s European types. Practical Electronics | May | 2025 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Vsupply / V Fig.2: the current stabilising effect of the Nelson Pass quiescent current sensing circuit as the supply voltage varies. 67 Powering it Parts List – Switchable Class-A/B Amp Many old radios using germanium transistors have a positive Earth, meaning that all the polarised components in Fig.1 have to be reversed. The PNP transistor becomes NPN and all the others PNP. The JFET has to be an N-channel type. The +18V power rail becomes -18V. The circuit in Fig.3 could be used to connect the circuit to a battery and power supply with automatic switchover. I used diodes, but for lower voltage drop, Mosfets could be employed instead. However, there would then be the possibility of static charge damage requiring protection (eg, a zener diode between each gate and source). 1 4.7kΩ trimpot (VR1) Semiconductors 2 BC549C 30V 100mA NPN transistors (TR1, TR2) 1 BD135 45V 1.5A NPN transistor (TR3) 1 BD136 45V 1.5A PNP transistor (TR4) 1 2N5460 40V 10mA or similar P-channel JFET (TR5) 2 1N4148 75V 200mA signal diodes (D1, D2) 1 BAT86 50V 200mA or similar schottky diode (D3) Capacitors 1 1000µF 25V electrolytic (C9) 1 470µF 25V electrolytic (C7) 1 100µF 25V electrolytic (C2) 1 47µF 16V electrolytic (C4) 1 47µF 6.3V electrolytic (C5) 1 2.2µF 10V electrolytic (C1) 2 100nF 50V X7R ceramic (C6, C8) 1 27pF ±5% 50V C0G/NP0 ceramic (C3) Construction As this amplifier is a one-off, for the moment, I just built it on perfboard, as shown in Photo 3. I have yet to design a PCB or rescale the component values for different voltages and output impedances. In Class-A mode, the output transistors dissipate almost 1W each continuously, so small heatsinks are required (I used flag types). It would also be a good idea to sandwich the bias transistor, TR2, between the output devices for proper thermal feedback. The only unusual part is the high-impedance loudspeaker, which I can supply (jrothman1962<at> PE gmail.com). Resistors (all ±5% ¼W carbon film unless noted) 1 470kΩ (R4) 1 6.8kΩ (R5) 1 270kΩ ±1% metal film (R10) 1 2.4kΩ (R6) 1 27kΩ (R2) 1 2.2kΩ (R9) 2 10kΩ (R1, R3) 1 1.5kΩ (R8) Qty 1 1 1 2 1 1 1 1 1 1 2 Value 4-band code 5-band code 470kW 270kW 27kW 10kW 6.8kW 2.4kW 2.2kW 1.5kW 470W 22W 3.3W 1 470Ω (R7) 1 22Ω (R13) 2 3.3Ω (R11, R12) Photo 3: a perfboard prototype of the amplifier. Note the output transistor heatsinks. +18V (new batteries max voltage 19.4V) D4 1N4001 D5 1N4001 + +19V regulated power supply V+ + Mains supply 9V – – + +18V 9V – Input from pre-amp on radio VCONT Amp Output 0V LS1 25Ω 2W 0V 0V Fig.3: steering diodes can be used to connect the battery, mains supply and amplifier for auto-switching to Class-A when the mains power supply is on. Photo 4. the underside of the board. This method of hard-wiring was a popular way of working out a PCB layout in the days of designing by soldering iron. 68 Practical Electronics | May | 2025