Fullscreen HTML5Med Res (??MB)HTML4

Build the Ultra Amplifier Mod A 100W class-AB amplifier with very low distortion This new amplifier module is a refined version of our highly successful “Plastic Power” module described in April 1996. The new version is quieter and has much lower distortion, particularly at the higher frequencies. By LEO SIMPSON T HIS AMPLIFIER MODULE has been under development, on and off, since late 1998. It was in July and August 1998 that we featured the ultra-low distortion 15W class-A amplifier. Since then, that amplifier has become our benchmark. Its distortion is so low that we had to resort to new procedures to be able to measure it. Inevitably, soon after the 15W class-A amplifier had been published, we wondered about producing a high16  Silicon Chip er power version. As good as the 15W amplifier is, it is still only 15W and on many types of music, particularly opera and classical piano, it simply does not have enough power. So we thought a 100W version would be really good. However, we shrank back from the idea of producing a 100W per channel class-A amplifier. After all, the stereo version of the 15W class-A amplifier dissipates about 100 watts at all times. If we produced a 100W stereo version, it would dissipate around 600 to 700 watts at all times. In other words, it would make a good heater for small rooms. So we wondered whether, with the lessons we had learned in the development of the class-A amplifier, we could apply them to a class-AB amplifier and get similarly dramatic results. That was the hope anyway, as we set out in late 1998 to produce this new amplifier. That we are publishing the results only now is a reflec­tion on how difficult the process has been. Is this new amplifier as good as the 15W class-A amplifier? Alas, no. As far as we can tell, using currently available semiconductors and circuit tech­ niques, it will never be possible to produce a class-AB amplifier as good as our 15W class-A module. However, all the development has produced an a-LD dule The final version differs slightly from this prototype module. It delivers 100W into 8Ω with very low distortion. amplifier that is a major improvement on the 125W Plastic Power module published in April 1996. This new module has much lower distortion at the higher frequencies from 5kHz to 20kHz and it is quieter although not dramatically so, since the Plastic Power module was very quiet anyway. Specifications The major performance parameters are listed in an accompa­nying panel but the graphs of Fig.1, Fig.2 & Fig.3 give a better picture. Fig.1 shows the frequency response at 1W into 8Ω. As you can see, it is about -0.3dB down at 20Hz and at the other end of the spectrum, about -0.5dB down at 20kHz. It would have been a relatively simple matter to make the response much flatter at the high end, say to 50kHz and beyond, AUDIO PRECISION FREQRESP AMPL(dBr) vs FREQ(Hz) 5.0000 26 JAN 100 08:27:56 4.0000 3.0000 2.0000 1.0000 0.0 -1.000 -2.000 -3.000 -4.000 -5.000 10 100 1k 10k 100k Fig.1: the frequency response at 1W into 8Ω. The response is virtually flat from 20Hz to 20kHz and tapers off above that to avoid EMI. March 2000  17 AUDIO PRECISION DIST-PWR THD+N(%) vs FREQ(Hz) 5 26 JAN 100 12:34:48 1 0.1 0.010 0.001 .0005 20 100 1k 10k 20k Fig.2: THD versus signal frequency at 100W into 8Ω, taken with a measurement bandwidth of 10Hz to 80kHz. AUDIO PRECISION SCTHD-W THD+N(%) vs measured LEVEL(W) 10 26 JAN 100 12:58:57 1 0.1 0.010 0.001 .0005 0.5 1 10 100 200 Fig.3: THD versus power at 1kHz into an 8Ω load, taken with a measurement bandwidth of 10Hz to 22kHz. as some commercial amplifiers do, but we regard that practice as undesirable. Not only is it likely to render the amplifier more suscep­ tible to EMI (electromagnetic interference) but it also means that it will amplify extraneous residual high frequency signals such as 38kHz from FM tuners and over-sampling artefacts from CD players. Amplifying these extraneous 18  Silicon Chip signals might not be a problem to the amplifier itself but they might then cause audible beats with the harmonic distortion products of the higher fre­quency audio signals. For example, a 38kHz FM multiplex signal (usually about 60dB down) could beat with the 32kHz second harmonic of a legiti­mate audio signal. The 6kHz beat would certainly be audible although it might be at a very low level. Most of the time such residual signals would not cause any audible problems but our philosophy is “Why ask for trouble?” and so we roll off the frequency response above 20kHz, as shown in Fig.1. The graphs of Fig.2 & Fig.3 tell the real performance story of this new amplifier. Fig.2 shows the harmonic distortion versus signal frequency at virtually full power, 100W into 8Ω. As may be seen, for all frequencies below 2kHz, the THD (total harmonic distortion & noise) is .002% or below. But from 2kHz to 20kHz, the distortion rises very gently, to .006%. These figures are taken with a measurement bandwidth of 10Hz to 80kHz. These are really excellent figures for any class-AB ampli­ fier and especially when compared to the vast majority of domes­tic hifi amplifiers which may be comfortably below, say, .005% distortion for the mid-frequencies but then rocket up to around 0.1% or more at 20kHz and full power. Even our popular Plastic Power module referred to earlier had a THD approaching .03% at 20kHz, so this new amplifier is up to five times better at high frequencies! Fig.3 shows the distortion versus power at 1kHz into an 8Ω load. This time the measurement is made with a bandwidth of 10Hz to 22kHz, to limit the noise content, and this shows the amplifi­er comfortably under .002% from 20W to 100W and rising gradually at the lower powers, solely due to the increased residual noise content. Finally, this amplifier is extremely quiet, at -117dB un­weighted with respect to 100W and -123dB A-weighted under the same conditions. This is a great deal quieter than any CD player and much quieter than the vast majority of domestic hifi amplifi­ers, regardless of price. By the way, we have made no mention of power output into 4Ω loads and in fact, we do not recommend operation with 4Ω loads. This is not to say that the amplifier could not drive 4Ω loads but there are two specific reasons for not recommending it. First, the distortion will be approximately double that achieved for 8Ω loads and in this respect it won’t be much better than the Plastic Power module. Second, the output transistors are connected as current feedback pairs Fig.4: the circuit can be regarded as a conventional direct-coupled feedback amplifier with compound current feedback tran­sistor triples in the output stage. The input and class-A driver stages are fed with regulated supply rails. and there is no intrinsic method of ensuring even current sharing between each transistor. This is not a problem with the lower currents delivered to 8Ω (or 6Ω) loads but could be a problem with 4Ω loads. A similar recommendation applied to our 15W class-A ampli­fier design. While it would certainly drive 4Ω loads, it would not do it in class-A mode and therefore the distortion would be considerably higher. In any case, the vast majority of hifi loudspeakers are 8Ω or 6Ω nominal. The module As can be seen from the photos, the amplifier module is assembled onto a PC board measuring 176 x 105mm. The four plastic output power transistors and three smaller power transistors are aligned along one edge to make it easy to attach them to a rela­tively large single-sided heatsink. The PC board has two on-board supply fuses and provision for temporary mounting of two 5W wirewound resistors which are used for setting the quiescent current. Circuit details The circuit of the amplifier module itself is shown in Fig.4 but that is not all there is to it. Fig.5 is the circuit of the power supply and that is one of the major factors in obtain­ing the performance of the amplifier. Compared with the Plastic Power module of April 1996, the major circuit differences of this new module are as follows: (1) Uses Motorola MJL3281A and MJL1302A output transistors which have improved linearity compared to the MJL21193/94 transistors. (2) Uses Motorola MJE15030 and MJE15031 driver transistors which have improved linearity, gain-bandwidth product and higher gain than the previously used MJE340/350 transistors. (3) Improved constant current source for the input differential pair and driver stages. (4) Use of current feedback output stages for improved linearity compared to conventional complementary symmetry emitter follower output stages. (5) Use of regulated power supply rails for the input and driver stages of the amplifier to obtain increased power supply rejec­tion ratio (PSRR). March 2000  19 Parts List AMPLIFIER BOARD 1 PC board, code 01103001, 105mm x 176mm 4 2AG fuse clips 2 2AG 5A fuses 1 coil former, 24mm OD x 13.7mm ID x 12.8mm long, Philips 4322 021 30362 2 metres 0.8mm diameter enamelled copper wire 11 PC board pins 1 large single-sided fan heatsink (Altronics H-0526; Jaycar HH-8546 or equivalent) 2 TO-126 heatsinks, Altronics Cat. H-0504 or equivalent 4 TO-3P insulating washers (for output transistors – see text) 3 TO-126 insulating washers 4 3mm x 20mm screws 3 3mm x 15mm screws 7 3mm nuts 1 200Ω multi-turn trimpot Bourns 3296W series (VR1) Semiconductors 2 MJL1302A PNP power transistors (Q13, Q14) 2 MJL3281A NPN power transistors (Q15, Q16) 1 MJE15030 NPN driver transistor (Q11) 1 MJE15031 PNP driver transistor (Q12) 1 MJE340 NPN power transistor (Q10) 1 BF469 NPN transistor (Q8) 1 BF470 PNP transistor (Q9) 3 BC546 NPN transistors (Q5, Q6, Q7) 4 BC556 PNP transistors (Q1, Q2, Q3, Q4) 1 3.3V 0.5W zener diode (ZD1) Capacitors 2 1000µF 63VW electrolytic 2 100µF 63VW electrolytic 1 100µF 16VW electrolytic 1 2.2µF 25VW electrolytic 1 0.15µF 400VW MKC, Philips 2222 344 51154 or Wima MKC 4 In most other respects, the circuit of the new module is virtually identical in configuration (but not component 20  Silicon Chip 5 0.1µF 63V MKT polyester 1 .0012 63MKT polyester 1 100pF 100V ceramic Resistors (0.25W, 1%) 2 220Ω 5W (for current setting) 1 12kΩ 1W 1 1kΩ 1 8.2kΩ 1W 1 390Ω 1 6.8Ω 1W 1 330Ω 8 1.5Ω 1W 2 150Ω 2 18kΩ 3 120Ω 1 3.3kΩ 4 100Ω 1 1.2kΩ 2 47Ω POWER SUPPLY 1 160VA or 300VA toroidal transformer with 2 x 35V 2.25A secondaries and 2 x 50V 0.1A secondaries 1 DPDT 5A 250VAC switch (S1) 1 3AG fuseholder 1 3A 3AG fuse 1 PC board, code 01103002, 61 x 92mm 6 PC pins 2 2kΩ multi-turn trimpots Bourns 3296W series (VR2,VR3) Semiconductors 2 TIP33B NPN power transistors (Q17, Q18) 1 LM317 adjustable positive 3-terminal regulator (REG1) 1 LM337 adjustable negative 3-terminal regulator (REG2) 1 PA40 bridge rectifier (BR1) 1 BR610 bridge rectifier (BR2) 2 1N4004 silicon diodes (D1,D2) 2 33V 5W zener diodes (ZD2, ZD3) Capacitors 4 8000µF 63VW chassis mounting electrolytics 2 470µF 100VW electrolytics 2 100µF 63VW electrolytics Resistors 2 6.8kΩ 0.25W 2 180Ω 0.25W 2 47Ω 0.25W 6 15Ω 1W values) to the Plastic Power module. However, for the sake of complete­ ness, we will now give the full circuit description. In all, the circuit uses 16 transistors and one zener diode, plus those semiconductors used in the power supply. The input signal is coupled via a 2.2µF capacitor and 1kΩ resistor to the base of Q1 which together with Q2 makes up a differential pair. Q3 & Q4 act as a constant current tail to set the current through Q1 & Q2 and thereby makes the amplifier insensitive to variations in the power supply rails. Current mirror The collector loads of Q1 & Q2 are provided by current mirror transistors Q5 & Q6. Commonly used in operational amplifi­ er ICs, current mirrors provide increased gain and improved linearity in differential amplifier stages. In a conventional direct-coupled amplifier, the signal from the collector of Q1 would be connected directly to the base of the following class-A driver stage transistor. In our circuit though, the signal from the collector of Q1 connects to the base of Q7, part of a cascode stage comprising Q7 & Q8, with Q9 pro­ viding a constant current load to Q8. Q4 does double-duty, providing the base voltage reference for constant current sources Q3 & Q9. In fact, the operation of the Q3/Q4 current source is a lot more complicated than it appears to be at first sight but let’s just simplify matters by saying that it is an improvement on the constant current tail used in the Plastic Power module. A 3.3V zener diode, ZD1, provides the reference bias to the base of Q8. In effect, Q8 acts like an emitter follower and applies a constant voltage (+2.7V) to the collector of Q7 and this im­ proves its linearity. The output signal from the cascode appears at the collector of Q8. A 100pF capacitor from the collector of Q8 to the base of Q7 rolls off the open-loop gain of the amplifier to ensure a good margin of stability. The output signal from the cascode stage is coupled directly to the output stage, comprising driver transis­tors Q11 & Q12 and the four output transistors, Q13-Q16. Actually, it may look as though the collector of Q9 drives Q11 and that Q8 drives Q12, and indeed they do, but in reality, the signals to the bases of Q11 and Q12 are identical, apart from the DC voltage offset provided by Q10. Vbe multiplier Q10 is a “Vbe multiplier”. It can be thought of as a tem­ p eraturecompensated floating voltage source of about 1V. Q10 “multiplies” the voltage between its base and emitter, as set by trimpot VR1, by the ratio of the total resistance between its collector and emitter (330Ω + 390Ω + VR1) to the resistance between its base and emitter (390Ω + VR1). In a typical setting, if VR1 is 100Ω (note: VR1 is wired as a variable resistor), the voltage between collector and emitter will be: Vce = Vbe x 820/490 = (0.6 x 820)/490 = 1.004V In practice, VR1 is adjusted not to produce a particular voltage across Q10 but to set the quiescent current through the output stage transistors. By the way, because we’re using a different output stage in this new amplifier module, the Vbe multiplier is set up differently to that in the Plastic Module where it was set to produce about 2V instead of 1V. Because Q10 is mounted on the same heatsink as the driver and output transistors, its temperature is much the same as the output devices. This means that its base-emitter voltage drops as the temperature of the output devices rises and so it throttles back the quiescent current if the devices become very hot, and vice versa. Driver & output stages Q11 & Q12 are the driver stages and they, like the output transistors, operate in class-AB mode (ie, class B with a small quiescent current). Resistors of 100Ω are connected in series with the bases of these transistors as “stoppers” and they reduce any tendency of the output stages to oscillate supersonically. As already mentioned, the output stages are connected as compound current feedback transistors. These are a development from the current feedback pair (CFB) configuration used in our class-A amplifier. However, that circuit used just one output transistor coupled to each driver transistor, with the emitter of the driver transistor connected to the collector of the output transistor. This config- This view shows the prototype amplifier module with the two outboard wirewound resistors in place for setting the quiescent current. Note that the paralleled 1.5Ω resistors will be laid out side-by-side in the final version of the PC board. The RCA input socket was for testing purposes only. uration acts like a very linear power transistor with only one base-emitter junction rather than two, as in a Darlington-connected power transistor. In this circuit, we have two paralleled power transistors, Q13 & Q14, connected to NPN driver transistor Q11 and Q15 & Q16 are connected to PNP driver transistor Q12. The four paralleled 1.5Ω emitter resistors for each com­ pound CFB transistor are there to help to stabilise the quiescent current and they also slightly improve the frequency response of the output stage by adding local current feedback. As already noted though, there is no intrinsic means in the circuit for ensuring even current sharing between Q13 & Q14 and between Q15 & Q16. What current sharing there is will depend on the inherent matching (or lack of it) between the transistors. Note that we did try the effect of small emitter resistors for each of the power transistors but these had the effect of worsening the distortion performance. So we left them out. Note that the current and power ratings of the output transistors are such that even if the current sharing is quite poor, there should not be a problem. Negative feedback is applied from the output stage back to the base of Performance Output power ��������������������������������������� 100 watts into 8Ω Frequency response ��������������������������� -0.3dB down at 20Hz; -0.5dB at 20kHz (see Fig.1) Input sensitivity ������������������������������������ 1.8V RMS (for full power into 8Ω) Harmonic distortion ����������������������������� <.006% from 20Hz to 20kHz, typically <.002% Signal-to-noise ratio ���������������������������� 117dB unweighted (20Hz to 20kHz); 123dB A-weighted Damping factor ������������������������������������ >170 at 100Hz & 1kHz; >60 at 10kHz Stability ������������������������������������������������ Unconditional March 2000  21 Fig.5: the circuit of the power supply. There are two sets of supply rails. The unregulated ±52.5V rails feed the class-AB output stages and nothing else. The fully regulated ±55V rails feed the class-A driver and input stages of the amplifier. Q2 via an 18kΩ resistor. The amount of feedback and therefore the gain, is set by the ratio of the 18kΩ resistor to the 1.2kΩ resistor at the base of Q2. Thus the gain is 16. This means that an input signal of just over 1.8V RMS is required for full power and this is less than -1dB with respect to the 2V maximum signal from a CD player. Thus under music conditions, the full signal from a CD player should not overload this amplifier. This approach is deliberate because we intend presenting a pair of these modules as a stereo amplifier, driven directly by a CD player for optimum sound reproduction. The low frequency rolloff of the 22  Silicon Chip amplifier is partly set by the ratio of the 1.2kΩ resistor to the impedance of the associat­ed 100µF capacitor. This has a -3dB point of about 1.3Hz. The 2.2µF input capacitor and 18kΩ base bias resistor feeding Q1 have a more important effect and have a -3dB point at about 4Hz. The two time-constants combined give an overall rolloff of -3dB at about 5Hz. At the high frequency end, the .0012µF capacitor and the 1kΩ resistor feeding the base of Q1 form a low pass filter which rolls off frequencies above 130kHz (-3dB). An output RLC filter comprising a 6.8µH choke, a 6.8Ω resistor and a 0.15µF capacitor couples the output signal of the amplifier to the loud- speaker. It isolates the amplifier from any large capacitive reactances in the load and thus ensures stabili­ty. It also helps attenuate EMI (electromagnetic interference) signals picked up by the loudspeaker leads and stops them being fed back to the early stages of the amplifier where they could cause RF breakthrough. The low pass filter at the input is also there to prevent RF signal breakthrough. Finally, before leaving the circuit description, we should note that the PC board itself is an integral part of the circuit and is a major factor in the overall performance. The board features star earthing, for minimum interaction between signal, supply and output currents. Note that the small signal components are clustered at the front of the board while all the heavy current stuff is mostly at the back and sides. Note also that the class-B current pulses from the two halves of the output stage are added symmetrically (adjacent to Q9) before being fed to the output RLC stage. The configuration of the output stage copper tracks is also very important because the magnetic fields associated with their asymmetrical currents are partially cancelled by the lead dress of the cables from the power supply. In fact, the arrangement of the power supply cabling to the module is quite crucial in obtain the low distortion figures, particularly at high frequencies. Power supply Fig.5 shows the circuit of the power supply. There are two sets of supply rails. The unregulated ±52.5V rails feed the class-AB output stages and nothing else. The fully regulated ±55V rails feed the class-A driver and input stages of the amplifier. Why have we gone to this trouble when just about every commercial domestic stereo amplifier uses unregulated supply rails for the whole power amplifier circuit? The reasons are twofold. First, when we designed the 15W class-A amplifier we found that we had to resort to fully regu­lated supply rails in order to get the residual hum to a reason­ ably low value. This was critical in the class-A amplifier be­ cause the constant high power supply current means a high ripple voltage which the amplifier circuit cannot fully reject. With a class-AB amplifier such as this, the quiescent load currents are quite low and therefore hum is not a problem but the very high asymmetrical signal currents (equivalent to half-wave rectified signal) are an even bigger problem because they cause a distorted signal voltage to be superimposed on the amplifier supply rails. By using a fully regulated supply, we avoid the possibility of these signals being fed back into the input stag­es. Furthermore, in a stereo version, the fully regulated supply also improves the separation between channels. Looking now at the circuit for the power supply, it is effectively split This power supply module provides the fully regulated ±55V rails for the class-A driver and input stages. The power transistors provide over-voltage protection to the regulators at switch-on. into two parts. The two 35V windings are connected together to drive bridge rectifier BR1 and the four 8000µF 63VW electroly­ t ic capacitors and this gives an unregulated supply of around ±52.5V (at no signal) to power the output stages of the ampli­fier. The 50V windings on the transformer drive the second bridge rectifier BR2 and this gives unregulated supplies of about ±72V and these are fed to the regulator circuits to provide ±55V to the input and class-A driver stages of the amplifier, as noted above. It’s not what it seems However, the regulator circuit is not quite what it seems. At first sight it may appear like a conventional 3-terminal regulator plus booster transistor arrangement, with the power transistor being slaved to the regulator. But that’s not how this circuit works. In fact, you will notice that we have used an NPN power transistor in conjunction with both regulators while you would expect a PNP transistor to be used with the negative regu­lator. So what is going on? Looking at the positive regulator for the moment, REG1 carries all the current, around 20mA for a mono version of this amplifier or 40mA for a stereo version. So there is no need for a booster transistor or even a heatsink. But the 3-terminal regulator cannot do the whole job. Its input voltage is about 72V and when the power is first ap­plied to the circuit this would appear directly across the regu­lator, causing it to blow. Its maximum input-output differential is only 40V. This is where the power transistor comes into play. When the voltage across REG1 exceeds 33V, zener diode ZD2 will be biased on via the associated 47Ω resistor. This causes Q17 to turn on and it limits the voltage to around 35V or so. The cur­rent through Q17 is limited to around 6.5A peak by the three paralleled 15Ω resistors in the emitter circuit. This peak cur­rent is very brief and occurs only while the 100µF capacitor at the output of REG1 is charged up to around 40V. From there on, the LM317 takes over and Q17 switches off. The same process occurs for the negative regulator REG2 and the NPN transistor Q18 takes care of the charging current for its associated 220µF output capacitor. The power transformer for a mono version of this amplifier can have a rating of 160VA or more while a stereo version will require a 300VA unit. In the next article, we will discuss the power supply and the construction of a stereo version of the amplifier in detail. SC March 2000  23