Fullscreen HTML5Med Res (??MB)HTML4

A low cost, high-quality audio amplifier – ideal for flat panel TVs, MP3s and more! “Tiny Tim” Part 1 – By Leo Simpson & Nicholas Vinen Stereo Amplifier Most flat panel TVs have mediocre sound quality from their tiny inbuilt downward firing loudspeakers. So how do you get get better sound? The short answer is that you need a good quality stereo amplifier with either a Toslink or S/PDIF digital input and some decent speakers. Our solution is to adapt the quality headphone amplifier from the September 2011 issue, increasing its power to around 10 watts per channel and adding digital inputs. E lsewhere in this issue we have featured the Tiny Tim loudspeaker system which is based on a 4-inch wide-range driver in an unusual horn-loaded cabinet. It only requires modest power to drive it to more than adequate sound levels. Combined with the amplifier described here, it is ideal for that purpose: for TV viewing or for a high quality music system in a small living room, study or bedroom. When we published the high quality headphone amplifier in the September 56  Silicon Chip 2011 issue we did indicate that it could comfortably drive 8-ohm loads to quite respectable power levels, more than 4W, at very low distortion. However, it was only equipped with a front panel headphone socket so you would have to use some sort of cable adaptor to connect the speakers to the socket. As a result, very few readers have probably bothered to do so but simply used it with headphones alone. That is unfortunate because it really is a very good performer, rivalling the sound quality of our now-famous Ultra-LD series amplifiers. But few people would bother to build a stereo amplifier capable of many hundreds of watts merely to listen to their TV; it would be over-kill. So that is part of the reasoning behind this project: to give the headphone amplifier a boost in power output to around 10 watts per channel while still retaining its very low distortion. At the same time, we are teaming it with a compact commercial DAC (digital-to-analog converter) to provide the required Toslink or S/PDIF input. siliconchip.com.au The perfect partner for our “Tiny Tim” speakers elsewhere in this issue. The Tiny Tim amplifier uses the same PCB as our high-quality headphone amplifier (September 2011) but has several component changes to allow it to produce around 10W per channel. Full construction details, including PCB component layout, will be published next month. While this isn’t as good as our own CLASSiC DAC project (SILICON CHIP, February, March & April 2013), it still has respectable performance while being significantly cheaper and much more compact. Elsewhere in this article are the performance specifications of the completed amplifier and a number of graphs illustrating its frequency response, harmonic distortion versus frequency and so on. Compact case One of the problems we have with presenting small projects such as this is sourcing suitable small cases which look good and are not frightfully expensive. For this project, we are taking the recycling approach and it involves using the case from a digital set top box which recently failed. The compact case is a good size and readily accommodates the headphone amplifier PCB, a small DAC and a 30VA (or 20VA) toroidal power transformer. No doubt other cases from compact DVD or CD players could also be pressed into service. In fact, some readers might take the approach of buying a set top box and removing the innards, just to get a cheap metal case. Either way, you should be able to use some of the existing hardware such as the power cord and power switch. That is what we were able to do. siliconchip.com.au We removed the existing PCBs from the STB case, a job which only took a few minutes. Then we unclipped the plastic front panel section so that we could do some surgery to it. This involved cutting away a section which was evidently provided for a model with some sort of card reader. We needed to do this as it would otherwise have interfered with the amplifier PCB. We also wanted to remove all of the existing screenprinted labelling. This was a matter of judicious cleaning with mineral turps. This slightly dulled off the shiny finish of the panel but it was easily restored with a light car polish. We then installed a dual gang volume control and a 6.5mm stereo headphone socket. This socket would allow headphones to be used instead of loudspeakers with automatic switching to turn the speakers off if a headphone jack plug was inserted. We also added a LED as a power indicator. • Easy to build • Uses common, low-cost parts • Suits 4-8speakers, 8-600hea dphones and ear buds • Ver y low distortion and noise • Short-circuit protected (bandwidth 20Hz-22kHz unless othe rwise stated; see Figs.1-4) Output power, 8 (THD+N < 0.01% ): 2 x 8W Output power, 4 (THD+N < 0.01% ): 2 x 6.5W Music power, 4/ 8: 10W THD+N: <0.0006% <at> 1kHz/1W Signal-to-noise ratio: -120dB unweigh ted with respect to 10W Frequency response: ±0.15dB, 20H z-20kHz Channel separation: 100dB <at> 100Hz, 83dB <at> 1kHz, 63dB <at> 10kHz NB: Pow er measurements made with a 20VA toroi dal power transformer; the alternative 30VA transformer would be expected to produce slightly higher power figures. October 2013  57 THD+N vs Frequency, 1W, 8W 04/09/13 13:15:30 0.01 Left channel, 20Hz-80kHz bandwidth Right channel, 20Hz-80kHz bandwidth Left channel, 20Hz-22kHz bandwidth Right channel, 20Hz-22kHz bandwidth Total Harmonic Distortion + Noise (%) 0.005 0.002 0.001 0.0005 0.0002 0.0001 20 50 100 200 500 1k 2k 5k 10k 20k Frequency (Hz) Fig.1: distortion when driving 8 loads is very low across the audible frequency range. The two lower curves include a realistic noise level however they do not show the rising distortion with frequency. The upper two curves do show this but the inaudible noise between 20kHz and 80kHz increases the overall readings. THD+N vs Power, 1kHz, 20Hz-22kHz Bandwidth 04/09/13 13:19:27 1 8W (both channels driven) 4W (both channels driven) 8W (one channel driven) 4W (one channel driven) Music power (8W, both driven) 0.5 Total Harmonic Distortion + Noise (%) 0.2 0.1 0.05 0.02 0.01 0.005 0.002 0.001 0.0005 0.0002 0.0001 .005 .01 .02 .05 .1 .2 .5 1 2 5 10 20 Power (Watts) Fig.2: distortion is slightly better driving 8loads than 4although the latter still gives a very respectable result. Distortion drops with level as the signal increases above the noise until the onset of clipping. Slightly more power is available with one channel driven than both due to power supply limitations (20VA transformer used). Inside the case we have mounted the PCB for the above-mentioned amplifier, the compact DAC and a 20VA toroidal power transformer plus a rectifier and filter capacitors on a small secondary board. But before describing the internal details, we need 58  Silicon Chip to describe the modified headphone amplifier circuit. Modified headphone amplifier circuit The main changes to the circuit involve the just-mentioned transformer which is part of a beefed up power supply in place of the original 12VAC 1A or 2A plugpack. Briefly, the other changes include increasing the capacitance of the power supply filter capacitors; increasing the voltage rating of other electrolytic capacitors from 25V to 50V, increasing the drive to the output transistors and increasing the gain of the power amplifiers. Rather than just describe the changes, we will give details of the complete circuit, for the benefit of readers who may not have seen the article in the September 2011 issue. Fig.5, the complete circuit, shows both channels. It is split into two sections, with the preamplifiers and power supply on the lefthand side and the power amplifiers on the righthand side. The preamplifier for each channel is based on three op amps in a classic Baxandall design so three LM833 dual op amps are used. The preamplifier is inverting and has a gain range from zero to -7. The Baxandall preamplifier circuit has the advantage that it varies its gain according to the setting of potentiometer VR1. As a result, the residual noise level is kept low at the low gain settings most commonly required. Like a traditional preamplifier, its gain can go all the way down to zero and up to some fixed number, in this case, -7, with the minus sign indicating that it inverts the signal. The two power amplifiers on the righthand side of the circuit are very similar to the 20W Class-A Amplifier (SILICON CHIP, May & June 2007) but with smaller output transistors and tiny heatsinks. The power amplifiers invert the signal again, so the unit’s outputs and inputs are in-phase. Since there is so much gain available in the preamps, the power amplifiers operate with low gain, (ie, -1.83). This improves the noise performance and maximises the feedback factor, keeping distortion exceedingly low even with run-of-the-mill output transistors. Since the headphone connector is a jack socket, the outputs can be briefly short-circuited by the plug if it is inserted or removed during operation. Because of this possibility, the design incorporates short-circuit protection to prevent any damage. Common mode distortion By lowering the gain, we get a higher siliconchip.com.au siliconchip.com.au Frequency Response, 1W 04/09/13 14:43:03 +3 Left channel, 8W +2.5 Right channel, 8W Left channel, 4W Right channel, 4W +2 Amplitude Variation (dBr) +1.5 +1 +0.5 -0 -0.5 -1 -1.5 -2 -2.5 -3 10 20 50 100 200 500 1k 2k 5k 10k 20k 50k 100k Frequency (Hz) Fig.3: the frequency response is ruler-flat between 20Hz and 20kHz. A slight rise is evident above 20kHz due to the RLC output filter however this drops off at frequencies above 100kHz (not shown). The difference in left and right channel level is due to the tracking error in the pot, which is less than 1dB across much of the range of the pot. Channel Separation vs Frequency, 3W, 8W 04/09/13 15:04:44 -50 -55 Right-to-left (8W) Left-to-right (8W) -60 -65 -70 Crosstalk (dBr) feedback factor (which is good) but we also increase the possibility of common-mode distortion. This can reduce the effectiveness of a high feedback factor so that the distortion reduction (due to the feedback) is not as much as would otherwise be the case. While the differential input voltage (ie, the voltage between the two inputs) of an amplifier operating in closed loop mode is very small, both input voltages can still have large swings, especially when the amplifier is being driven hard. This is the “common mode” signal, ie, signal common to both inputs. For a non-inverting amplifier, the common mode voltage is the output voltage swing divided by the closed loop gain. So at low gain, the common mode signal amplitude is similar in magnitude to the output signal amplitude, which for our amplifier can be around 28V peak-to-peak. Typically, if the common mode signal exceeds 1-2V RMS, common mode distortion can become the dominant distortion mechanism, marring its performance. This is due to “Early effect” in the input transistors (named after James M. Early of Fairchild Semiconductor). This is caused by the effective width of the transistor base junction varying with its collector-base voltage (see www.wikipedia.org/wiki/ Early_effect). If the common mode voltage is large enough, the result is modulation of the input transistors’ beta (or gain) and this reduces the overall linearity of the amplifier. These non-linearities cannot be corrected by negative feedback since they occur in the input stage. The solution is to use an inverting amplifier, as we have in this case. Its non-inverting input is connected to ground and so the inverting input is held at “virtual ground” too, regardless of the output voltage. This configuration has so little common mode voltage that it can’t suffer from common mode distortion. To make a power amplifier inverting, we rearrange the feedback network in the same manner as we would with an op amp. In fact, common mode distortion in op amps can be reduced using the same method. The main disadvantage of the inverting configuration is that the input impedance is low, as determined by the resistor from the signal source to the inverting input. For good noise -75 -80 -85 -90 -95 -100 -105 -110 20 50 100 200 500 1k 2k 5k 10k 20k Frequency (Hz) Fig.4: channel separation vs frequency, with a higher value being better. This is better driving speakers (shown here) than headphones because speakers do not have a shared ground return path. The coupling between channels is mostly capacitive, hence separation is better at lower frequencies. performance, its value must be low (minimising its Johnson-Nyquist thermal noise – again, see www.wikipedia. org/wiki/Johnson_nyquist_noise). In this case, the preamplifiers provide the amplifiers with a low source impedance, so it isn’t a problem. No driver transistors If you compare the amplifier circuits to our previously published amplifier designs such as the Ultra-LD Mk.3 or 20W Class-A Amplifier, you will find many similarities. As with the Ultra-LD Mk.3 ampliOctober 2013  59 10W +12V K D9 1N4004 100nF K D15 BAT42 A LEFT INPUT A CON1 L1 470nF 680W 8 3 2 4.7nF MKT 100k +11.8V -11.8V IC1a 1 OFF-BOARD 220mF VR1b 10k LIN 100pF NP0 100k 4.7k 22mF K 8 3 D16 BAT42 A 680W 1 IC2a 2 22k 6 5 -11.8V 220mF 7 IC2b 4 IC1, IC2, IC3: LM833 +11.8V K VOLUME RIGHT INPUT D17 BAT42 CON2 L2 A 470nF 680W 5 6 4.7nF MKT 100k 7 IC1b 22mF D18 BAT42 A 4.7k 8 3 -11.8V 1 IC3a 2 680W 6 5 22k 1k A -11.8V VR1a 10k LIN 100pF NP0 K N 100nF 4 100k *NOTE: MAINS EARTH IS NOT CONNECTED q THIS IS A DOUBLE INSULATED DESIGN 220mF 220mF 100nF OFF-BOARD 220mF 7 IC3b 4 D10 1N4004 N/C* K POWER -11.8V A MAINS PLUG K +20V 10W A D3 1N4004 F1 1A SLOW BLOW 15V K A IN K BR1 A W04M K GND 4700mF 4700mF 100nF +12V OUT REG1 7812 D4 1N4004 220mF A K K 4700mF T1 30VA TOROIDAL A 4700mF A IN A TO DAC POWER SUPPLY K K D6 1N4004 A 220mF K + 12V 30k -12V OUT D5 1N4004 -20V SC 100nF GND POWER SUPPLY PCB Ó2011 REG2 7912 A l LED1 230V 15V 22k NOTE: VALUES SHOWN IN RED HAVE BEEN CHANGED COMPARED TO ORIGINAL HEADPHONE AMPLIFIER DESIGN TINY TIM 10W STEREO AMPLIFIER Fig.5: The full circuit for the Tiny Tim Amplifier, including the mains power supply (lower left) which is built on a separate PCB. The onboard preamp is shown at upper left and this provides gain control and buffering to drive the power amplifiers, at right. These are based around a TIP31/TIP32 complementary transistor pair without driver transistors, driven by a more-or-less conventional front end. The supply voltage has been increased compared to the original headphone amplifier design and some of the component values have been changed to increase gain and current delivery, hence available power. 60  Silicon Chip siliconchip.com.au 10W K D11 1N4004 220W A Q5 BC559 E 47mF 2.2k B 100mF 50V E B E C Q7 BD140 -20V 22W 47mF 10k VR2 500W C B E 2.2k Q11 TIP31 TP1 + C Q10 B E C Q2 Q1 BC559 BC559 220mF 50V C 100W E 1.2kW 22W 2.2k B 10k 100W 100nF E C C 1.8k B Q6 BC559 +20V BD139 1.2W 0.5W 30mV 1.2W 0.5W TP2 3.3k B + 680pF NP0 220pF NP0 A 1.2W 0.5W 30mV D7 1N4004 1.2W 0.5W L3 4.7mH K B 1.8k C B 10k Q8 BC549 22W E B E Q3 BC549 C E D12 1N4004 K C B B E 68W Q4 BC549 C Q12 TIP32 2.2k C B E + Q9 BD139 HEADPHONE SOCKET 47W -20V 10W D13 1N4004 220W A Q17 BC559 E 47mF 100mF 50V B E C C Q14 Q13 BC559 BC559 Q19 BD140 47mF 22W VR3 500W + -20V C B E 2.2k 10k OFF-BOARD TO RIGHT SPEAKER 220mF 50V C 100W E 1.2kW E CON4 +20V 22W 2.2k B 10k 100W 100nF E C C 1.8k B Q18 BC559 2.2k B TO LEFT SPEAKER 2.2k 68W 10W Q25 BC328 150nF C A K E Q23 TIP31 TP3 + C Q22 B BD139 1.2W E 0.5W 150nF L4 4.7mH 30mV 1.2W 0.5W TP4 10W 3.3k B + 680pF NP0 220pF NP0 30mV A 1.2W 0.5W D8 1N4004 1.2W 0.5W K B 1.8k C B 10k Q20 BC549 22W B E Q15 BC549 C E D14 1N4004 K B B C E 68W Q16 BC549 E C Q24 TIP32 Q26 BC328 7812 C 2.2k B C E GND IN OUT GND Q21 BD139 7912 -20V 2.2k 68W E GND 47W A OUT IN D1qD14: 1N4004 A siliconchip.com.au K LED1 D15qD18: BAT42 A K K A B B C TIP31, TIP32 BD139, BD140 BC328, BC549, BC559 E IN C B E C C E October 2013  61 Here’s the integrated DAC we used, outside and inside. It comes from Jaycar Electronics. While you could use our CLASSiC DAC, it is much more expensive and would be overkill in this project. fier, this design uses 2-pole frequency compensation. As a result, the Tiny Tim amplifier has particularly low distortion at high frequencies. For a detailed explanation of the advantages of 2-pole compensation, refer to the article published in the July 2011 issue on “Amplifier Compensation and Stability”. The main difference is that the two output transistors are driven directly from the voltage amplification stage (VAS), with no driver transistors in between. This design decision is due to the original application of the amplifier being for headphones, where the current requirements are quite small and thus the Class-A VAS is easily able to supply it. This is still a feasible configuration for a 10W-per-channel amplifier but we have had to increase the VAS standing current to around 30mA, by using a 22resistors at the bases of transistors Q7 and Q19. Happily, the TIP31 and TIP32 output transistors have quite a good beta figure which drops as the collector current increases. For 10W output we need a peak output current of 1.65A and their beta at this sort of current is around 55. 1.65A ÷ 55 = 30mA, hence our choice of the 22 resistors. It’s only just enough current but we don’t want to use too much of the available power up in the driver stage. The TIP31 (NPN) and TIP32 (PNP) transistors are readily available and rated at 3A and 40W each; sufficient for our needs in this circuit. 62  Silicon Chip How it works Let’s start with the preamp stages and since both channels are identical, we will just describe the left channel. Any RF signals or ultrasonic noise picked up by the input leads are attenuated by a low-pass filter consisting of a ferrite bead, a 680resistor and a 4.7nF capacitor. The ferrite bead acts like an inductor to block RF. The signal is then coupled via a 470nF capacitor to pin 3 of op amp IC1a which is configured as a voltage follower. This provides a low source impedance to the preamp gain stages comprising IC2a & IC2b. IC1a’s output is fed to the following stage via a 220F electrolytic capacitor. This large value ensures good bass response and avoids any distortion that may arise from the typical nonlinearity of an electrolytic capacitor with a significant AC voltage across it. The signal passes to the non-inverting input of IC2a (pin 3) via volume control potentiometer VR1 and a 22F electrolytic capacitor. This capacitor ensures there is no DC flowing through VR1, which would otherwise cause a crackling noise when it is rotated. IC2a buffers the voltage at the wiper of VR1 to provide a low impedance for inverting amplifier IC2b. IC2b has a fixed gain of 7, set by the 4.7k and 680 resistors. The 100pF feedback capacitor is there to improve circuit stability and reduce high-frequency noise. Volume potentiometer VR1 is part of the feedback network from the output from IC2b to the input at the 220µF capacitor (from pin 1 of IC1a). Hence IC2a & IC2b form a feedback pair with the overall gain adjustable by VR1. When VR1 is rotated fully anticlockwise, IC2b’s output is connected directly to VR1b’s wiper. Thus IC2b is able to fully cancel the input signal (as there is zero impedance from its output to the wiper) and the result is silence (no output signal) from the preamplifier. Conversely, when VR1 is fully clockwise, VR1b’s wiper is connected directly to the input signal, which is then amplified by the maximum amount (7 times) by IC2b. At intermediate settings, the signal at the wiper is partially cancelled by the mixing of the non-inverted (input) and inverted (output) signals and the resulting gain is intermediate. The way in which this cancellation progresses as VR1 is varied provides a quasi-logarithmic gain curve. IC1 needs input protection Because the amplifier may be turned off when input signals are present, IC1’s input transistors can be subjected to relatively high voltages; up to 2.5V RMS or maybe 7V peak-to-peak. This will not damage IC1 immediately but over many years, it could degrade the performance. This is because normally very little current flows through the op amp inputs and so the metal traces within the IC are thin. If enough current passes through the inputs (5mA or more), “metal migration” can cause degradasiliconchip.com.au tion and ultimately failure. For that reason we have included small-signal Schottky diodes D15 & D16 to protect pin 3 of IC1a (and D17 & D18 for pin 5 of IC1b) when the unit is switched off but a large signal is applied. They clamp the voltage at that input to within ±0.3V of the supply rails under normal conditions, preventing current flow through the op amp input transistors should their junctions be reverse-biased. So if the unit is off and the supply rails are zero, the input voltages will be similarly limited to ±0.3V. The BAT42 diodes have been carefully selected to clamp the op amp input voltages appropriately without having so much leakage current that they will introduce distortion into the signal (Schottky diodes normally have a much higher reverse leakage current than standard silicon diodes). For more information on protecting op amp inputs, see Analog Devices tutorial MT-036, “Op Amp Output Phase-Reversal and Input Over-Voltage Protection”. We also tested BAT85 diodes (Al- tronics Z0044). These have slightly higher capacitance when reversebiased (10pF compared to 7pF) and a significantly higher reverse leakage current (400nA at -15V/25°C compared to 75nA). However, testing shows no measurable increase in distortion with these in place of the BAT42s so they are an acceptable substitute. Amplifier circuit Low-noise PNP transistors Q1 & Q2 are the differential input pair, with the base of Q1 being the non-inverting in- Parts List – Tiny Tim 10W Stereo Amplifier 1 integrated DAC (Jaycar AC-1631) 1 Mini-Reg kit or PCB & parts (SILICON CHIP, Dec 2011) 1 PCB, code 01309111, 198 x 98mm 1 vented metal case, 250 x 220 x 45mm or larger# 1 PCB-mount 6.35mm switched stereo jack socket (3PDT) (CON4) 6 PCB-mount 6021-type flag heatsinks (Element14 Order Code 1624531; Jaycar HH8504, Altronics H0637) 1 2.5mm DC power plug 6 M3 x 10mm screws and nuts 8 TO-220 insulating washers 6 TO-220 insulating bushes 6 PCB pins 8 M3 x 9mm tapped Nylon spacers 16 M3 x 6mm machine screws 1 35 x 15mm section of tin plated steel (eg, cut from a tin can) 3 8-pin DIL sockets (optional) 2 small ferrite beads 4 insulated binding posts: 2 red, 2 black 2 RCA plugs 2 plastic former bobbins (Jaycar LF1062, Altronics L5305) 1 2m length 0.8mm diameter enamelled copper wire 1 25mm length 25mm diameter heatshrink tubing 1 1m length light duty figure-8 cable 1 500mm length 2-core shielded cable 1 250mm length 4-core shielded cable 1 1m length red medium-duty hook-up wire 1 1m length black medium-duty hook-up wire 1 250mm length blue medium-duty hook-up wire Semiconductors 3 LM833 dual low noise op amps (IC1-IC3) 1 7812 positive 12V linear regulator (REG1) 1 7912 negative 12V linear regulator (REG2) 2 TIP31 3A NPN transistors (Q11, Q23) 2 TIP32 3A PNP transistors (Q12, Q24) 4 BD139 1.5A NPN transistors (Q9, Q10, Q21, Q22) 2 BD140 1.5A PNP transistors (Q7, Q19) 2 BC328 PNP transistors (Q25, Q26) 6 BC549 NPN transistors (Q3-Q4, Q8, Q15-Q16, Q20) 8 BC559 PNP transistors (Q1-Q2, Q5-Q7, Q13-Q14, Q17-Q19) 1 5mm LED (LED1) 12 1N4004 1A diodes (D3-14) 4 BAT42 Schottky diodes (D15-D18) (or use BAT85, Altronics Cat. Z0044)# siliconchip.com.au Capacitors 2 4700µF 25V electrolytic 2 220µF 50V electrolytic* 7 220µF 25V electrolytic* 2 100µF 50V electrolytic* 4 47µF 16V electrolytic* 2 22µF 16V electrolytic* 2 470nF MKT 2 150nF MKT 7 100nF MKT 2 4.7nF MKT 2 680pF C0G/NP0 ceramic 2 220pF C0G/NP0 ceramic 2 100pF C0G/NP0 ceramic * Low ESR 105° types preferred if their diameter is no more than 6.3mm for 22F/47F and 8mm for 100F/220F. # See text Resistors (0.25W, 1%) 4 100kΩ 1 30kΩ 3 22kΩ 6 10kΩ 2 4.7kΩ 2 3.3kΩ 10 2.2kΩ 4 1.8kΩ 2 1.2kΩ 1 1kΩ 4 680Ω 2 220Ω 4 100Ω 4 68Ω 2 47Ω 6 22Ω 6 10Ω 8 1.2Ω (0.5Ω, 5%) 1 10kΩ dual gang linear 16mm potentiometer. with knob (VR1) 2 500Ω sealed horizontal trimpots (VR2, VR3) Power supply board 1 PCB, coded 18110131, 75 x 100mm 1 30VA 15+15VAC toroidal transformer (Altronics M-4915A) or 1 20VA 15+15VAC toroidal transformer (Jaycar MT-2086) 1 M205 fuse holder with clip-on cover 1 1A slow-blow M205 fuse 2 3-pin headers, 3.96mm pitch, with centre pin removed # 1 250VAC switch with double-sheathed lead and sheathed   terminals, terminated with 3-pin, 3.96mm pitch header plug # 1 twin core mains lead, double-sheathed and terminated with   3-pin, 3.96mm pitch header plug # 1 3-way terminal block 4 M3 x 9mm tapped Nylon spacers 8 M3 x 6mm machine screws 1 W04M 1.5A bridge rectifier (BR1) 2 4700µF 25V electrolytic capacitors 2 10kΩ 0.25W 5% resistors October 2013  63 put to the amplifier and the base of Q2 being the inverting input. Q1’s base is tied to ground by a 1.2k resistor (to match the 1.16k source impedance at the base of Q2) and is bypassed by a 100nF capacitor to reduce highfrequency noise. The signal from the preamplifier is fed to the base of Q2 via a 3.3kfeedback resistor, so that the amplifier works in the inverting mode. This gives the amplifier stages a gain of -3.3k÷ 1.8k = -1.83. PNP transistor Q5 operates as a 3mA constant current source (0.65V ÷ 220) to feed the Q1/Q2 input pair. Negative feedback for current regulation is provided by another PNP transistor, ie, Q6. It has a bootstrapped collector current sink (two 10kresistors and a 47µF capacitor), so that it operates consistently. NPN transistors Q3 and Q4 form a current mirror for the input pair, with 68emitter resistors to improve its accuracy. Any difference in the current through Q1 and Q2 must then flow to the base of NPN transistor Q8. So Q1Q5 form the transconductance stage of the amplifier. Together, Q8 and Q9 form a Darlington transistor, configured as a commonemitter amplifier. PNP transistor Q7 acts as a constant current source for its collector load, sourcing about 30mA (0.65V ÷ 22). Q6 provides current regulation feedback for Q7 as well as Q5. The 680pF and 220pF capacitors between Q9’s collector and Q8’s base, together with the 2.2kresistor from their junction to the negative rail, form the 2-pole frequency compensation scheme mentioned earlier. Together, transistors Q7-Q9 are the voltage amplification stage. Because Q7 and Q9 have to handle significantly more voltage and current in this beefed-up version of the amplifier (compared to the original headphone amplifier circuit), their dissipation has increased beyond the capabilities of the small TO-92 signal transistor package. We calculate their dissipation as around 20V x 30mA = 600mW while the limit of a TO-92 package at 55°C is about 500mW. As a result, we have had to change them to BD139 & BD140 which are 80W transistors rated at 80V and 1.5A. These are in TO-126 packages which can dissipate just under 1W at 55°C with no heatsink. But they have 64  Silicon Chip a different pin-out to those originally specified (ie, BC337/338 and BC549) so it will be necessary to bend their leads when they are installed on the PCB. You can see how we did this in the photo of the PCB. VBE multiplier Between Q7 and Q9 is Q10 (another BD139) which functions as a VBE multiplier to set the quiescent current for the output transistors Q11 & Q12. Q10 is mounted on the back of Q11’s heatsink so that its junction temperature tracks the output stage. Thus, its VBE tracks that of the output transistors (Q11 and Q12), so the bias voltage varies to compensate for changing output transistor temperature, keeping the standing current through them more or less constant. VR2 is used to adjust this current, while the 2.2kresistor prevents the bias from becoming excessive if VR2’s wiper goes open-circuit, as it may do while it is being trimmed. A 47µF capacitor filters the bias voltage, improving distortion performance when the output voltage swing is large. The resulting bias voltage is applied between the bases of output transistors Q11 (NPN) and Q12 (PNP) via 22stopper resistors, which prevent parasitic oscillation. Each output transistor has a 0.6emitter resistor (two 1.2resistors in parallel) which helps to linearise the output stage and stabilise the quiescent current. Current limiting While it’s always a good idea to plug and unplug the headphones while the power switch is off, we can’t rely on that and we don’t want the output transistors to blow when it happens. Therefore, both Q11 and Q12 are protected against over-current conditions. Q11 is current-limited because the 30mA current source (Q7) sets a maximum limit for its base current. According to the TIP31 data sheet, at 25-125°C, the maximum collector current will be about 1.65A, well within its safe operating area (SOA) so as long as the short-circuit is brief. Q12 is more of a concern because Q9 can sink significantly more than 30mA. The 10kresistor at Q8’s collector ultimately limits how much current Q9 can sink as follows. Q8’s maximum collector current is around (12V - 0.7V) ÷ (10k+ 2.2k) = ~1mA. According to the BC338 data sheet Q9’s maximum current gain figure is around 160, so the maximum it can sink is about 160mA. However, if this much current were pulled from Q12’s base then it would fully saturate (turn on hard), exceeding its SOA and possibly causing it to fail. Q25 and D7 prevents this. Should the current flow through Q12’s collectoremitter junction exceed 2A (within its SOA), the drop across the 0.6emitter resistor exceeds 2A x 0.6 = 1.2V. At this point, Q25’s base-emitter voltage increases beyond 1.2V - 0.6V = 0.6V and so Q25 starts to turn on, shunting current around Q12’s base-emitter junction and preventing Q12 from turning on harder. Any current sunk by Q9 beyond that necessary for Q12 to pass 2A goes through D7 and Q25 rather than Q12’s base-emitter junction. Output RLC filter The output filter isolates the amplifier from its load at high frequencies, improving stability. Because this amplifier circuit is already fairly stable (thanks to its simple output stage), we can get away with slightly less inductance than usual (4.7H rather than 6.8H or 10H). We can thus use a thinner gauge wire which is slightly easier to wind, for roughly the same DC resistance. Ideally, the output filter should be optimised for the expected load impedance but because headphones have such a wide range of impedances, all we can do is compromise and specify an intermediate value. As a result, for higher impedance headphones, the amplifier has a slightly elevated response at above 20kHz. For 4-ohm and 8-ohm loudspeaker operation, the high frequency response is virtually flat and then for higher load impedances, up to infinity, the gain increases to as much as +0.13dB at 20kHz. The increase is slightly lower (+0.09dB) for the most common headphone impedances of 16 and 32. This deviation is so small as to be imperceptible. In fact, all our amplifier designs using this type of output RLC filter (devised by the late audio genius Neville Thiele) have such a response with higher than usual output impedances or no load. Power supply We have had to increase the voltage and current of the power supply in order to allow the modified amplifiers to siliconchip.com.au deliver the target of 10W per channel. Instead of a 12V AC 2A plugpack (ie, 24VA) we are using a 20VA or 30VA 150-15 toroidal transformer (T1). Significantly, we have also modified the PCB so that the amplifier sections run from the unregulated ±20V supply rather than the regulated ±12V supply, which was sufficient for driving headphones but a bit limiting for loudspeakers. Another benefit of using the toroidal transformer is that it has a centre tap which means we can use a bridge rectifier (BR1) to get full-wave rectification, recharging the filter capacitors at 100Hz rather than 50Hz. This reduces supply ripple and thus reduces resicual hum while increasing available power and dynamic headroom. T1 and BR1 are mounted on a small secondary PCB which forms a self-contained mains power supply. We have done this for a number of reasons; one is that it allows us to build the unit as a double-insulated piece of equipment. Most commercial devices that constructors are likely to “rat” for their amplifier housing will already be double-insulated (and thus have no earth connection). We pulled the pin headers off the power supply PCB of the recycled settop box and re-used these on our board, allowing the pre-existing mains cable and main power switch to simply plug in, as they did before. While we were at it, we stuck another pair of 4700F filter capacitors on the power supply board. This improves the power supply filtering and also means that very little 100Hz current passes through the wiring between the two boards, minimising hum coupling into the amplifiers. Switch-on/off behaviour You may notice that there is no speaker protector or de-thump circuit. Neither is really necessary in this case. The amplifier’s power supply can only deliver about 40W and this is unlikely to do much damage to a speaker in the case of a circuit failure, especially since some of this would be dissipated in the amplifier itself. As for switch-on and switch-off thumps, the headphone amplifier circuit was already designed to minimise these and since speakers are significantly less sensitive, these should be kept well under control. This was partly achieved by removing the capacitor which would typically be between Q5’s base and the positive rail (as present in the 20W Class-A Amplifier and the Ultra-LD Mk.3). Despite changing the circuit to run from an unregulated supply, virtually no ripple seems to make its way to the amplifier outputs, as demonstrated by the very good signal-to-noise ratio of -120dB (including the preamplifier!). Diodes D11 & D12 (D13 & D14 in the right channel) are important for proper switch-on behaviour. While the ±12V regulated rails are already protected to prevent the positive rail from going negative and vice versa, the RC filtered supply rails for the early amplifier stages can still suffer from this problem unless extra steps are taken. That’s because the filter resistors isolate the capacitors from the clamp diodes D4 & D6. Without D11 and D12, the positive filtered rail could be briefly pulled negative and this would cause an amplifier output excursion. This could cause unwanted noises in the speakers at start-up. The different positive and negative rail filter resistors (10 and 47 respectively) allow the positive rail to come up more quickly which also helps achieve a clean switch-on. Together, these details allow the amplifiers to operate normally just milliseconds after both filter capacitors are partially charged. Similarly, diodes D9 & D10 clamp the RC-filtered supply for the op amps in the preamplifier. Without these, the op amp input transistors may become briefly reverse-biased at switch on, causing supply current to flow into the AC-coupling capacitors and again causing a thump to be generated. Finally, the 1k resistor in parallel with D10 discharges the op amp negative supply rail faster than the positive rail when power is removed. The op amps are prone to oscillation when their supply capacitor is mostly discharged and this can cause a “chirp” at switch-off. With the 1k discharge resistor, this chirp is made very short and often eliminated entirely. SC Next month In November SILICON CHIP we shall present the construction details and describe the setting-up procedure. That includes details of the new power supply board and mounting both PCBs, plus the small off-theshelf DAC, inside the case. IN STOCK NOW Check out our SUPER SPECIAL BUNDLE PRICES For more information & to shop online, visit www.wiltronics.com.au Ph: (03) 5334 2513 | Email: sales<at>wiltronics.com.au Raspberry Pi is a trademark of the Raspberry Pi Foundation siliconchip.com.au October 2013  65