Fullscreen HTML5Med Res (??MB)HTML4

Part 1: by Nicholas Vinen Compact HiFi headphone Amplifier This Headphone Amplifier is easy to build, sounds great, doesn’t cost too much to make and fits into a compact instrument case. It’s ideal for beginners or just those who want to get the best out of a set of traditional wired headphones. It’s powered by a plugpack, so no mains wiring is required. I t has been a while since we’ve published a headphone amplifier. The reason I decided to design a new one is that my last design (in the September & October 2011 issues; siliconchip.au/ Series/32) had excellent audio quality, but was a bit overkill for many people. It was fairly large, somewhat expensive to build and consumed a fair bit of power, but you can’t really fault the resulting sound quality. Before that, we published the Studio Series Headphone Amplifier (November 2005; siliconchip.au/Series/320), which was not an integrated design (it required a separate power supply board), didn’t really fit into any particular case and was a fairly basic design with modest output power and had decent but not amazing audio quality. I thought there was room for something in between: an amplifier with excellent audio quality that fit neatly into a compact case and wasn’t too difficult or expensive to build. That’s precisely what this is. It’s also beginner-­friendly and has the handy feature of two stereo inputs that are mixed with independent volume controls. Fig.1: the Amp’s distortion versus frequency for four common headphone/earphone load impedances. Distortion is lower for higher load impedances due to the lower output current required; the 600W curve is higher mainly due to the lower test power due to voltage swing limitations. 44 Silicon Chip That means you can connect two sound sources such as a TV and a computer, a CD player and a TV or something like that. With the separate volume controls, it’s easy to account for different output levels from those devices, and you can also easily mute one if both are active. If you want to save time and money, you can build it with just one stereo input. You have the choice of 3.5mm or 6.35mm jack sockets for the output (or both, optionally connected in parallel). Power is from a 9-12V AC 1-2A plugpack, a type that’s readily Fig.2: this shows how distortion varies with the output power level, at a fixed frequency. The onset of clipping is around 0.9W for an 8W load, due to current delivery limitations; a little over 1W for 16W; around 0.75W for 32W; or 90mW for a 600W load due to voltage swing limitations. Australia's electronics magazine siliconchip.com.au Complete Kit (SC6885; $70) Features & Specifications 🎼 Drives stereo headphones with impedances from 8Ω and up 🎼 Two outputs to suit 3.5mm or 6.35mm jack plugs 🎼 Two stereo RCA inputs with independent volume controls 🎼 Powered by a 9-12V AC plugpack 🎼 Power on/off switch and power indicator LED 🎼 Signal-to-noise ratio: 103dB with respect to 250mW into 8Ω 🎼 Total harmonic distortion: <0.0025% <at> 1kHz, <0.01% <at> 10kHz (see Figs.1 & 2) 🎼 Frequency response: 10Hz to 100kHz, +0,-0.2dB (16Ω load; see Fig.3) 🎼 Channel separation: >70dB <at> 1kHz (see Fig.4) 🎼 Maximum output power (9V AC supply): 0.9W into 8Ω, 1W into 16Ω, 0.75W into 32Ω, 80-140mW (12V AC) into 600Ω 🎼 Class-AB operating mode (Class-A at lower power levels) 🎼 Inexpensive and easy to build 🎼 Fits into compact 155×86×30mm ABS instrument case available from most suppliers. There is an onboard power switch and power indicator LED. The headphone amplifier section is based on common low-noise, low-­ distortion op amps with transistor buffers to boost the output current. It will drive any headphones from 8W to 600W. It won’t deliver a ton of power, but should be more than enough for any headphones, up to a watt (or maybe more) per channel. If you really wanted to, you could use it to drive a pair of high-efficiency speakers to modest sound levels (eg, for use with a computer). While it isn’t really designed for that task, it will work as long as the speakers are efficient enough and you’re close to them. This design uses all through-hole parts and it fits into a really nice little snap-together compact case that’s just 155mm wide, 30mm tall and 86mm deep. So it takes up barely any room. The modest power consumption means it only gets a little warm during typical use, despite being unvented. There’s really nothing tricky to the construction. The only slightly fiddly Fig.3: the Amp’s frequency response is very flat for all load impedances within the audible range (20Hz–20kHz). The deviation above 20kHz is due to the output filter. The vertical shifts are due to the Amp’s output impedance (the level reduces slightly for lower load impedances). siliconchip.com.au Includes the case but not a power supply bits are winding the inductors for the output filter (which only takes a few minutes) and mounting the output transistors and heatsinks, which is only difficult because the thermal paste can get on your fingers. There is one adjustment per channel for quiescent current. It’s easy to make by monitoring the voltage between pairs of test points with a DMM while twiddling a trimpot. With a circuit that isn’t too difficult to understand and straightforward construction, this should be a good project for relative beginners. Performance At low signal levels, up to around 5mW (8W), 10mW (16W) or 20mW (32W/600W), the Headphone Amplifier operates in Class-A mode. Many headphones and earphones will produce reasonable volume levels at such powers. If your headphones require more power, or there are loud transients (like drum hits), the amplifier will automatically switch to Class-B (this is known as Class-AB operation). The resulting performance is pretty good – not as good as our very best amplifiers, but certainly well above average. It’s better than ‘CD quality’ under most conditions (which equates to about 0.0018% distortion at 1kHz with a 96dB signal-to-noise ratio). Fig.4: there’s a small amount of signal bleed between channels but it’s attenuated by more than 70dB at 1kHz and below, so it is unlikely to be noticeable. Most stereo content has less separation than this anyway. Australia's electronics magazine December 2024  45 The power supply section is on the left, signal input/ mixing in the middle and power output on the right. The performance was measured with a 9V AC plugpack; using a 12V plugpack will give the same or better performance. Fig.1 shows how the total harmonic distortion plus noise (THD+N) level varies with frequency at 250mW (a high level for headphones!) into four common headphone load impedances. The performance is excellent for 32W headphones, well below 0.001% even up to several kilohertz. It’s almost as good for 16W, reaching only around 0.0015% at 1kHz for 16W & 600W loads. Even for the relatively low impedance of 8W, more typical for loudspeakers, the THD+N is just 0.0025% at 1kHz for a fairly high output level (250mW) and remains below 0.01% up to 10kHz. Fig.2 shows how THD+N varies with power level. As the performance is essentially limited by noise, it is a steadily descending line until the point where it goes into clipping. That figure will give you a pretty good idea of how much power can be delivered with the 9V AC supply. Fig.3 shows the frequency response, which is basically flat across the audible spectrum. Fig.4 shows the channel 46 Silicon Chip separation, which we think is pretty reasonable. You’re unlikely to notice any signal bleeding between the channels. Note that the maximum power delivery into high-impedance loads will depend on the supply voltage. Testing with a 9V AC plugpack, we got around 90mW into a 600W load before clipping, but we’d expect closer to 150mW with a 12V AC plugpack. Most headphones and earphones are well below 600W, so they are unlikely to run into voltage swing limitations even with a 9V AC supply. more than annoyance. It didn’t always happen, but it’s still a good idea to take the headphones off before switching the amplifier off. We also tested it by plugging in the Exteek C28 Bluetooth adaptor (reviewed in the September 2024 issue; siliconchip.au/Article/16569). We connected it to one input using a 3.5mm jack to twin RCA plug lead. That worked fine, and the Amp’s gain was more than enough to drive the headphones to deafening levels from its relatively low-level output. Subjective testing The full circuit diagram is shown in Fig.5. We’ll start by describing the input section and volume control, then the power amplification section, then the power supply. This description is for the full version of the circuit; later, we’ll explain two ways it can be cut down. The stereo input signals are applied to either of dual RCA sockets CON2 & CON3. They pass through an RF rejecting filter comprising ferrite beads, 100W series resistors and 470pF ceramic capacitors to ground. This should help eliminate any RF (eg, AM radio or switch-mode hash) picked up by the signal leads that I tested the Amp with a pair of Philips SHP9000 32W headphones (which, in my opinion, are excellent). As expected based on the flat frequency response and low distortion, the sound quality was topnotch, with lots of punchy bass, plenty of treble and no audible noise or artefacts. There was no noticeable noise at switch-on with the headphones plugged in, although more sensitive headphones may make a noise. There was sometimes a modest crack or thump sound at switch-off, although it was not loud enough to cause anything Australia's electronics magazine Circuit details siliconchip.com.au could otherwise be demodulated by the following circuitry. The signals are then AC-coupled using back-to-back polarised electrolytic capacitors. This is a cheaper and generally more compact configuration than non-polarised electrolytic capacitors, and has no real disadvantages. We use high-value coupling capacitors to retain good bass response, it also keeps the source impedance low for the following stages, to avoid noise creeping in. The capacitor voltage ratings here are pretty high, so that if a faulty signal source delivering +18V or -18V DC (or more) is connected to one of the inputs, it won’t damage anything. It’s important to AC-couple signals to potentiometers to avoid crackle when they are rotated. The signal is applied to the top of the potentiometers, which act as variable voltage dividers, the attenuated signal appearing at the wiper. The potentiometers have a ‘logarithmic taper’, which is suitable for volume control since it better matches the way we hear loudness. Linear potentiometers tend to give poor control at the lower end of the volume range. From the potentiometer wipers, the signals are again AC-coupled to the following op amp buffer stages, so that the op amp bias currents don’t cause a DC voltage across the pots. Otherwise, that can also cause crackle when the pots are rotated. Here we only need a polarised capacitor because we know the op amp input will be slightly positive due to the bias current flowing out of it. That is true for either of the op amp alternatives specified (NE5532 or LM833, which should both perform well). 100kW resistors to ground both DC-bias their input signal to 0V and provide a path for that bias current to flow. The signals from the two pairs of buffers are then mixed using 10kW resistors and the mixed audio is fed to the power amplifier, on the right-hand side of the diagram. The 1MW resistors to ground provide a path for IC1’s input bias currents to flow without IC2 and IC3 having to sink it, although the circuit would still work if those resistors were left out. Parts List – Compact Headphone Amplifier This section is based on dual lownoise op amp IC1 and medium-power 1 double-sided blue PCB coded 01103241, 148 × 80mm 1 155×86×30mm ABS instrument case [Altronics H0377, DigiKey 377-1700-ND, Mouser 563-PC-11477] 1 9-12V 1-2A AC plugpack 1 PCB-mount right-angle miniature SPDT toggle switch (S1) [Altronics S1320] 1 PCB-mount barrel socket to suit plugpack (CON1) 2(1) dual horizontal white/red RCA sockets (CON2, CON3) [RCA-210; Silicon Chip SC4850] 1 PCB-mounting DPST 3.5mm stereo jack socket (CON4) [Altronics P0092, Jaycar PS0133] AND/OR 1 PCB-mounting DPST or DPDT 6.35mm stereo jack socket (CON5) [Altronics P0073 or P0076/P0076A] – not the taller version 4(2) small ferrite beads (FB1-FB4) 1 2-pin header with jumper shunt (JP1) 2(1) 10kW dual-gang logarithmic taper 9mm right-angle PCB-mount potentiometers (VR1, VR2) 2 2kW top-adjust mini trimpots (VR3, VR4) 3(1) 8-pin DIL IC sockets (optional, for IC1-IC3) Wire & hardware 1 2m length of 0.25-0.4mm diameter enamelled copper wire (for L1 & L2) 2 M3 × 16mm panhead machine screws 4 M3 × 10mm panhead machine screws 6 M3 flat washers 6 M3 hex nuts 4 No.4 × 5-6mm panhead self-tapping screws 2 TO-220 micro-U flag heatsinks (15 × 10 × 20mm) 2(1) small knobs to suit VR1 & VR2 4 small self-adhesive rubber feet Semiconductors 3 NE5532 or LM833 low-noise, low-distortion op amps (IC1-IC3) ♦ 5 TTC004B 160V 1.5A NPN transistors, TO-126 (Q1, Q3, Q5, Q7, Q8) 3 TTA004B 160V 1.5A PNP transistors, TO-126 (Q2, Q4, Q6) 1 3mm blue LED with diffused lens (LED1) 2 1N5819 40V 1A schottky diodes (D1, D2) ♦ only one is required for cut-down version (unbuffered or single-channel) Capacitors (maximum 20mm height) 4 1000μF 25V low-ESR electrolytic (5mm pitch, maximum diameter 13mm) 2 470μF 10V electrolytic (5mm pitch, maximum diameter 10mm) 8(4) 100μF 50V electrolytic (5mm pitch, maximum diameter 8mm) 4 100μF 25V low-ESR electrolytic (5mm pitch, maximum diameter 8mm) 4(2) 100μF 16V electrolytic (5mm pitch, maximum diameter 8mm) 2 10μF 50V electrolytic (2.5mm pitch, maximum diameter 6.3mm) 2 100nF 63V MKT 3(1) 100nF 50V MKT, ceramic or multi-layer ceramic 4(2) 470pF 50V NP0/C0G ceramic 2 100pF 50V NP0/C0G ceramic Resistors (all ¼W 1% unless noted) 2(0) 1MW 4(2) 100kW 7(3) 10kW 4 4.7kW 2 3kW 4 1kW 2 220W 4(2) 100W 2 10W 1W 5% 4 1W ½W (5% OK) n number in bracket refers to quantities for the single-channel version siliconchip.com.au Australia's electronics magazine Power amplifier December 2024  47 Fig.5: the full Headphone Amplifier circuit; the two stereo inputs are at upper left, the buffer and mixer left of centre, the output section at upper right and the power supply at lower right. It’s all pretty conventional, but note the use of capacitance multipliers rather than regulators to provide reasonably steady V+ and V− rails without requiring a specific AC supply voltage. transistors Q3-Q8. As the left and right channels are essentially identical, we’ll stick to describing the right channel, with the corresponding left-­ channel designators being given in brackets (parentheses). The incoming signal is fed into the non-inverting input, pin 3, of IC1a. IC1a is configured as a non-inverting amplifier with a default gain of four times (12dB), although that can be changed by varying the 3kW and 1kW resistor values between the output and the feedback point, the pin 2 inverting input of IC1a. The bottom end of the divider is connected to signal ground via a 470μF capacitor rather than directly, reducing the amplifier DC gain to unity. That way, the circuit doesn’t amplify the op 48 Silicon Chip amp’s inherent offset voltage (or any other offsets in the circuit). Most of the current to drive the headphones is supplied by NPN transistor Q3 (Q5) and PNP transistor Q4 (Q6), which are complementary emitter-­followers. As the base voltage of Q3 rises, it sources more current into the output via its 1W emitter resistor, reducing its base-emitter voltage until it stabilises. Similarly, when Q4’s base is pulled down, its emitter pulls the output down and it too stabilises at a more-or-less fixed base-emitter voltage differential. As Q3 and Q4 both have base-emitter voltage drops of around 0.7V when conducting a few milliamps, if we arrange for a difference of around 1.5V between the two bases (with Q3’s base Australia's electronics magazine voltage being higher than Q4’s), a small amount of current will constantly flow from the V+ rail, through Q3, the two 1W emitter resistors, then Q4 and back to the V- rail. This is called the quiescent current. By having a small quiescent current, we keep Q3 and Q4 in conduction all the time, and we only have to vary the amount of conduction to smoothly control the output signal, rather than switching Q3 or Q4 on when needed. This is called Class-AB (sometimes Class-B) and it has the benefit of minimising (and ideally, virtually eliminating) crossover distortion. Crossover distortion is an undesirable step in the output voltage as it passes through 0V, which an AC audio signal does frequently. siliconchip.com.au To achieve the required ~1.5V between the bases, we have NPN transistor Q7 (Q8), which acts as a ‘Vbe multiplier’. There are 4.7kW resistors from the V+ and V- rails connected to its collector and emitter, which provide a small bias current of about 3mA through it. Trimpot VR3 (VR4) is connected across the transistor such that we can vary the collector-base and emitter-­ base resistances. The ratio of those siliconchip.com.au resistances causes a multiple of its mostly fixed base-emitter voltage (again, about 0.7V) to appear between its collector and emitter. By adjusting the trimpot for a gain of a little over two times, we get the required 1.5V. You will note that its collector and emitter connect to the bases of Q3 & Q4, so that voltage appears across them. It is stabilised by a 10μF capacitor as the output swings up and down (and thus the bias in Q7 varies slightly). Australia's electronics magazine The 10kW resistor across the trimpot prevents Q7 from switching off fully if the trimpot is intermittent, which would cause a high current to be conducted by Q3 & Q4, possibly damaging them. Another thing you might notice is that Q7 is the same type of transistor as Q3, even though it only needs to handle a tiny current and power. That is because Q3’s base-emitter voltage will vary as it changes in temperature. By December 2024  49 mounting Q7 in contact with Q3, the bias voltage changes proportionally, so Q3 always receives the correct bias voltage. Q4 is the complementary type to Q3; while we are not tracking its temperature directly, its dissipation will very closely match that of Q3, so its temperature should as well, and thus its base-emitter voltage will be very similar to Q4’s. So the thermal tracking by Q7 will compensate for temperature changes in both output transistors and their required bias voltages. The 1W emitter resistors provide a little local negative feedback for Q3 & Q4 and also help to stabilise the quiescent current, by making the exact bias voltage across their bases less critical. The junction of these resistors is the amplifier output, which is fed to the headphone socket(s) via an RLC filter comprising a 10W resistor in parallel with a 4.7μH inductor and then a 100nF capacitor to ground. This filter is there to isolate the amplifier output from the headphones, so that any reactance at the headphone socket (eg, from cable capacitance or driver properties) cannot destabilise the amplifier and cause it to oscillator. The values have been chosen so the filter doesn’t change the overall frequency response when combined with typical headphone impedances. Finally, there is a 1kW resistor between the output of op amp IC1a and the junction of the 1W emitter resistors. That means the op amp’s output contributes a tiny bit of current to the amp output, helping to cancel out any small amounts of distortion caused by the output stage that the feedback loop is too slow to handle. CON4 gives you the option to use the smaller type of headphone jack, while CON5 is the larger and more robust type. If both are fitted, inserting a plug into CON5 will disconnect the ground path for CON4, unless there is a shorting block on jumper JP1. If there is, both headphones will be driven in parallel. JP1 must also be shorted if CON5 is omitted so that CON4 can be used. Output transistors We chose the TTA004B (PNP) and complementary TTC004B (NPN) because they are inexpensive, compact and designed for audio use. They have a high maximum collector voltage of 160V (not that useful in this application), a high transition frequency of 100MHz, low output capacitance and a reasonably high continuous current limit of 1.5A each. While they don’t have a super high current gain, it is pretty good at 140280 at 100mA (typically >200). All these properties combine to make them good as part of a feedback loop to deliver a reasonable amount of current while minimising distortion. The current gain (beta [β] or hfe) is still usefully high at 1A (around 100). They are also very linear, having a very flat hfe curve from 1mA to over 100mA. So overall, they are excellent medium-power audio transistors. Power supply Fig.6: we can omit IC1 & IC2 by coupling the signals from the wipers of VR1 & VR2 directly to the non-inverting inputs of IC1 & IC2 and removing the redundant pair of DC-biasing resistors. This will still work and save a bit of money, but the volume controls will have some interaction. Rather than an unregulated or a regulated supply, we have opted for a capacitance-multiplier type supply. This has the advantage of delivering much smoother rails to the op amps and output stage than an unregulated supply, without the power loss of a regulated supply or pinning us to a particular regulated supply voltage. The incoming low-voltage AC from the plugpack is converted to pulsating DC by the full-wave voltage doubler formed by schottky diodes D1 and D2. Schottky diodes are used here to minimise the voltage loss, so we can get decent output power from just 9V AC, and to improve efficiency. They achieve that by having a low forward voltage drop when in conduction. The result is about 12V DC across the two 1000μF capacitors (assuming a 9V plugpack), giving an unregulated ±12V supply. This will have Australia's electronics magazine siliconchip.com.au 50 Silicon Chip an increasing amount of AC ripple as the load on the supply goes up due to those capacitors discharging between peaks in the mains cycle. The ripple will be 50Hz, not 100Hz, due to the diode configuration. We measured over 300mV of ripple on our prototype with no signal, and obviously that increases as we load the output more. We could add two regulators to the output but they would need to be matched to the plugpack; for example, ±12V regulators might work well if the plugpack is 12V AC and thus develops sufficient input voltage for them to regulate, but they would be useless with a 9V AC plugpack. There’s also the problem that under load, the ripple could cause the regulators to enter dropout. Instead, we use capacitance multipliers formed by transistors Q1 & Q2, operating as complementary emitter-followers, with another set of 1000μF capacitors between their bases and ground. They are biased on by 220W resistors from each collector to the associated base. You can think of these as ‘variable regulators’ that produce a smoothed output but with the output voltage being related to the input voltage. That’s because the base capacitors charge to just below the average of the input voltage due to the RC low-pass filters formed by them and the 220W resistors. Keep in mind that, as they operate as emitter followers, the emitter voltage for a fixed load current is essentially a fixed amount below the base voltage (around 0.7V). So if the base voltage is steady, thanks to that lowpass filter action, as long as the collector voltages don’t drop too low due to excessive ripple, the voltage at the emitters will be essentially constant. As a result, with say ±12V DC at the collectors overlaid with several hundred millivolts of ripple (we measured around 350mV in our prototype), the outputs at their emitters will be close to ±10.5V DC with much lower ripple (10mV in our prototype). That’s a reduction of 35 times or 31dB. While the amplifier section has good ripple rejection, some may still be audible in the output with 350mV+ on the supply rails. We doubt any will be detectable with just 10mV of ripple on the supply rails, and the performance figures support that. siliconchip.com.au Fig.7: if you only need one stereo input, the circuit can be further simplified as shown here. Only one op amp, IC1, is required as there is no longer any signal mixing. There are four 100μF supply rail bypass/filter capacitors after Q1/Q2 although, two of which are physically located close to the output stages. Thus, they are shown on the circuit diagram at upper right. Putting them closer to the output transistors means less voltage drop during high-current transients. The power LED is connected between the two rails so it doesn’t ruin the symmetry of the device. Its current is limited to around 2-3mA by its 10kW series resistor. Variations There are two variations to this circuit that can be built on the same board. The first is the same as the full circuit shown in Fig.5 but without buffer op amps IC2 & IC3. The differences are only in that section, and they are shown in Fig.6. The signal path is the same as before up to the wipers of the volume control potentiometers. Subsequently, rather than being coupled to buffer op amps, the signals are coupled directly to the mixer resistors. This means that the signal sources are driving a lower impedance. Now the 1MW resistors to ground are required, as otherwise there would be no DC bias for the signals going to IC1a. The relatively high value of the 1MW DC bias resistors was chosen to avoid Australia's electronics magazine too much attenuation when combined with the higher source impedances due to the mixer resistors. This version has the advantage of retaining the two separate inputs but with fewer components and lower power consumption. However, due to the way the signals are mixed, there will be interactions between the two volume controls. That means that if you adjust the level of one source up or down, the level of the other source may also change a little. If that’s likely to bother you, or you bought a kit that came with all the op amps, you might as well just build the full version. But we thought we’d present this cut-down version as it doesn’t require any modifications to the PCB, just a few wire links need to be added to bypass the missing op amps. The other version is the simplest configuration, with just a single stereo input. It is shown in Fig.7. In this case, we don’t need the buffer op amps since there is no longer any mixing going on; the signal from the sole volume control can simply be coupled straight to IC1. Next month The second and final article on this Headphone Amplifier next month will have the PCB assembly instructions, case preparation, testing, adjustment details and some usage tips. SC December 2024  51