Fullscreen HTML5Med Res (??MB)HTML4

AUDIO OUT AUDIO OUT L R By Jake Rothman Back to the buffers – Part 1 L ast month, we completed the Discrete Op Amp design. An obvious application for any op amp – precision or otherwise – is as a buffer. However, it’s generally not necessary to use the level of complexity we employed in the Discrete Op Amp design for a buffer, so this month so we’ll look at some high-quality discrete buffer circuits. Hitting the buffers While electronic engineers prefer the term ‘voltage follower’, audio engineers use the term ‘buffer’, since they are usually used to isolate one stage from another. The basic requirements for a good buffer are: • High input impedance • Low output impedance • Good current delivery. Since a buffer only has a voltage gain of one, or ‘unity’, the massive amount of open-loop gain of an op amp is not needed to obtain very low distortion. Most buffers rely on the 100% negative feedback that occurs inside the device. This works by holding the voltage constant between the input and output. The input voltage is reflected in the load resistor, as shown in the emitter follower in Fig.1. This means that there is an effective ‘open-loop gain’ for a single transistor. The transconductance (voltage in for current out) reduces the distortion and output impedance. The transconductance is less for JFETs V+ 15V BC546 VIN R1 10kΩ 0V 100% negative feedback between input and output –0.6V VO R2 10kΩ V– –15V Fig.1. Simple bipolar emitter voltage follower or buffer. 62 which means the buffer’s total harmonic distortion (THD) is higher when used as source followers (Fig.2). However, the JFET’s transfer function (voltage in versus current out) curve is less abrupt – essentially a square law initially (in theory) rather than the exponential curve generated by bipolar transistors. This curve difference is akin to the old triode vs pentode sound debate in Hi-Fi circles. The idea is that the greater profusion of lower-order harmonics generated by the slower rate of slope change in the JFET circuit (compared to the bipolar version) gives a less harsh sound. The clipping is also softer as well. I use this to good effect in the FET input buffer stage of the Colorsound Powerboost guitar preamp pedal. As far as Hi-Fi buffers are concerned, this is not good, Hi-Fi buffers need to be as linear as possible. Instability All buffers are prone to oscillation or high frequency instability at elevated frequencies (MHz). Often, the oscillation frequency is so high that it manifests itself as a thickening of the scope trace. This occurs when capacitance is present on the output. This capacitance can be from a screened cable or simply a scope probe. It can be isolated with a low-value series resistor of typically 22Ω to 100Ω. Sometimes this is bypassed with a small inductor to get the output impedance at audio frequencies back down to a few ohms. The root cause of instability is the 100% negative feedback normally present. Strangely, another factor is ‘negative output resistance’ at high frequencies, which can be generated in some transistors. This can cause the gain to rise above unity. Also, the input wiring inductance, along with the base-emitter capacitance of the transistor – JFET or bipolar, but especially the latter – can result in oscillation. This can be fixed with ‘grid-stopper’ R1 and a capacitor shown in Fig.2 One old buffer to another The ‘textbook’ two-component buffers shown in Fig.1 and Fig.2 have distortion in the region of 1 to 5% at line level. This is 0dBm or 2.2Vpk-pk/775mVRMS into 600Ω, and all distortion measurements are made under these conditions. Both of these approaches were used successfully in early solid-state Hi-Fi when the signal level was around 100mVRMS and loads were high, typically 47kΩ. Another common buffer problem is asymmetric current delivery, where the circuit sources or sinks current more easily in one direction rather than the other. When this happens one side of the output waveform clips first into low-impedance loads. A buffer may only have a voltage gain of one, but the internal gain required may be over 400, and this cannot be achieved by a single transistor if low distortion is also required. Bi-polar transistor also have a non-linear input impedance which causes distortion when fed from high source impedances, such as when following a volume control or as part of a filter. JFETs do not suffer from this defect, but come with extra circuit complexity. Fortunately, all these problems can be overcome. Constant current The first improvement to the basic buffer shown in Fig.1 and Fig.2 is to upgrade the load resistor to a constant-current load. The basic load resistor is effective+9V Input impedance 2.3MΩ VIN 4.7MΩ BF244A R1 4.7kΩ VO C1 470pF 4.7MΩ 22kΩ 0V Fig.2. Simple JFET buffer used on input to Colorsound Powerboost pedal. Distortion is very high, around 5%, but gives a bright sound on high-inductance guitar pick-ups due to its high input impedance. Practical Electronics | February | 2024 V+ BC546 VIN + 47µF 10kΩ VO 3.5mA Constantcurrent load 0V IDSS for JFET typically 3.5mA for BF244A Can vary ±50% V– VO BF244A V– Fig.3. The addition of a constant-current load gives multiple improvements. Distortion is 1% with the BC546B. A FET or 3.5mA current regulator diode can be used. ly in parallel with the load to be driven, resulting in over half the available output power being wasted. Changing to the constant-current load shown in Fig.3 means this loss is eliminated. Also, the circuit becomes immune to power supply variations, and has a stable predictable quiescent current. Another advantage is that the gain is closer to unity. With just a resistor the gain can be significantly less than one. Do remember though, to get the full benefit of the active load, its output needs to feed a high impedance. This is the main reason the distortion of a resistive load buffer is poor when driving the standard 600Ω test load. 3mA BF244A +15V VIN 1MΩ 100Ω 0.3V 0V +50mV BF244A VO 100Ω –15V Fig.4. Two JFETs can be combined to make an effective buffer. Distortion is 1.1%, but it has a lovely soft clip. I varies according to IDSS of JFET Suffix IDSS (mA) Y 1.2–3.0 GR 2.6–6.5 BL 6.0–14 2SK2145 dual JFET (Commoned sources) G 1MΩ 0V S G D Using a JFET results in a linear high input impedance of over Fig.7. Peak analyser shows why high-power Darlington 1MΩ, defined by transistors are not suitable for input devices. The gain is low at the input resistors. low currents (5mA). This quality is particularly useful for active filters using V+ low-value polystyrene capacitors along with high-value resistors. C2, Typically RC 0.6V 100pF to 1nF A popular JFET follower configuPNP ration is shown in Fig.4. This circuit uses a couple of BF244s. However, NPN CIN it exhibits disappointing distortion VIN into a 600Ω load, 1.1% at 0dB. If the VO RIN load impedance had been higher, say Constantaround 10kΩ instead of 600Ω then it current sink 0V would have been reduced to 0.1%. The V– problem seemed to be mismatching of the JFETs. When these were replaced with a dual matched pair device, as Fig.8. The CFP or Sziklai (pronounced shown in Fig.5, the distortion dropped ‘sick-lie’) follower gives lower distortion to 0.07% into 600Ω. than the Darlington version, but is more Darlington configuration A good compromise if JFETs are to be avoided for cost, availability or distortion reasons, is to use a small-signal Darlington transistor, which only costs around 30p. This increases the input impedance to 90kΩ with a 100kΩ input bias resistor, as shown in Fig.6. An MPSA13 Darlington works here, up to 30V total rail voltage. For higher rails, the MPSA29 can be used, which is rated at 100V. More common high-power Darlington devices, such as the BD680, won’t work here because of their low +15V +15V D VIN JFET followers VO 18Vpk-pk max into 600Ω 0.07% THD –15V Fig.5. Dual-JFET buffer has much lower distortion (0.07%) due to the better matching, causing some kind of curve cancellation. Practical Electronics | February | 2024 VIN 470nF 100kΩ MPSA13 Darlington –1.2V 0V VO 3.5mA –15V Fig.6. Darlington input can increase input impedance compared to a single transistor. prone to instability and often needs a capacitor across the first transistor’s collector load resistor. Hfe at low operating (emitter) currents of 1-5mA. This is because they have internal resistors shunting the base-emitter junctions. The Peak DCA75 gives an H fe reading of around 16 to 40 with these devices (as shown in Fig.7) and around 16,000 for the MPSA29. Complementary pair An improvement on the Darlington transistor circuit is to use two opposite polarity transistors, multiplying together each individual device’s current gain. This is called a compound follower pair (CFP) or ‘Sziklai pair’, as shown in Fig.8. In this case there is 100% negative feedback across two transistors. This improves linearity almost a hundred-fold compared to a single device, but makes oscillation more likely, so stabilising capacitor C2 is often necessary. The input device can also be a JFET, where its higher distortion is greatly reduced. 63 are almost exact solid-state versions of the valve circuit in Fig.9. Fig.12 shows a dual-JFET version of the modulated buffer. In this configuration the distortion was greatly improved and I would say this is our first Hi-Fi buffer. It’s also small and could be made into an 8-pin DIP module to directly replace IC op amps wired as buffers. +300V + 10µF 50V VIN 470kΩ 1.5kΩ 1W 510kΩ V1 ECC82 1kΩ 100nF 400V 1 V1a 100nF Sense resistor 4.7µF 250V 6 2 V1b 1kΩ 7 3 4 9 5 VO Next month 8 We will continue our in-depth investigation of high-performance discrete buffers in the next issue. 12.6V, 150mA heater 1MΩ 300Ω 1MΩ 0V Fig.9. The valve-based White follower has a modulated current load (V1b). Unfortunately, V1a can suffers from a high heater-cathode voltage possibly causing leakage current, resulting in noise. Note the 1kΩ grid-stopper resistors. Modulation The constant-current load’s operating current can be modulated to improve efficiency and distortion. One modulation technique is to take a signal from a sense resistor in the power feed to the follower, which gives the required 180° phase reversal. This configuration was called the ‘White follower’ in the days of valves, and is shown in Fig.9. I first saw this technique applied to transistors in Nelson Jones’ Wideband Oscilloscope Probe (Wireless World, August 1968). Theoretically, the same output current or load-driving capability can be achieved with half the operating current since it is a form of push-pull operation. Interestingly though, when testing this with my Audio Precision THD analyser, I found there was an optimum amount of modulation for lowest distortion. This point did not coincide with maximum efficiency and +15V *Used to minimise output offset VIN 470Ω Sense resistor 470Ω 1MΩ +50mV Modulated constantcurrent load BF244A VO VIN 10kΩ 3.66mA 82Ω BF244A 100nF 3V 3.66mA –15V Fig.10. Solid-state version of White follower using two JFETs (which are very similar to triodes). Modulation of current source gives a large distortion reduction down to 0.06%. 64 VO –0.7V 0V 390kΩ 82Ω VIN 2SK2145-BL 22nF 22Ω 1MΩ +0.33V 0V 390kΩ VO 4.15mA 82Ω 0.34V –15V Fig.12. Dual matched-pair JFET modulated buffer. It has an impressive THD figure of just 0.006%. NEW! 5-year collection 2017-2021 All 60 issues from Jan 2017 to Dec 2021 for just £44.95 BC546B 1µF 100Ω* 0V 390kΩ +15V BF244A 22nF 100nF was dependent on load resistance and rail voltage. I’m not sure what the relationship is, but there is a distortion null point as the sense resistor is adjusted. Most audio circuits are mains powered, so maximum efficiency is not an important criterion, thus these circuits are tweaked for minimum distortion at normal levels. The distortion was several times greater without current load modulation. This configuration has the advantage of easy optimisation and does not affect stability. The sense resistor also provides short-circuit protection when sourcing current. The disadvantages are the loss of 1V to 3V of headroom and injection of power supply noise (especially at clipping). However, with a regulated power supply, which any respectable audio system will use, this is not a problem. The modulated current load can be a FET or bipolar device. With a highimpedance device the coupling capacitor can be small and non-polarised. This is illustrated in Fig.10 and Fig.11 which +15V 19Vpk-pk max into 600Ω 330Ω 0.005% THD at 0dBm (1% not modulated) 100nF 3V –15V Fig.11. Bi-polar transistor version of White follower. Distortion is about 0.06%. PDF files ready for immediate download See page 6 for further details and other great back-issue offers. Purchase and download at: www.electronpublishing.com Practical Electronics | February | 2024