Fullscreen HTML5Med Res (??MB)HTML4

This amplifier is capable of delivering over 500 watts into 4Ω or around 280 watts into an 8Ω load. The large heatsink is mandatory and needs to be fancooled if it is to withstand the rigours of operating under maximum dissipation conditions. We envisage it as being used in high-end stereo systems and for musical instrument and PA work. 500W of audio power 24  Silicon Chip W N Pt.1: By LEO SIMPSON & BOB FLYNN O MATTER WHICH WAY you look at it, this is a big power amplifier. It’s physically big, it needs a big power supply and a big fan-cooled heatsink and it delivers lots of power. A pair of these amplifiers would be the basis of a magnificent stereo system for the home, especially if you have a large listening room. Perhaps you might think that a 500 watt per channel stereo system would be too much. The answer to that depends on what sort of music you like listening to and how efficient your loudspeak­ers are. If you like rock music with its fairly limited dynamic range (ie, loud all the time), then a 1000 watt system would be going over the top. But if you listen to a lot of classical piano music and your speakers are of only average efficiency, then 500 watts per channel might not be enough! One of the authors of this article has a large piano in his (large) loungeroom and often has the opportunity (every day) to compare the real piano with CDs played through the Studio 200 power amplifier published in the February 1988 issue of SILICON CHIP. That amplifier has a music power output of 120 watts per channel into 8Ω and 190 watts per channel into 4Ω. In a straight comparison for absolute loudness and dynamic range, the real live piano, played by an accomplished pianist, wins every time. We’re not talking about ridiculously loud music here – just a piano competently played. What is not commonly realised is that the piano is probably the most difficult musical instrument to accurately Do you want a big power amplifier for musical instrument or PA use? Something with real grunt? Well here it is, the big­gest power amplifier ever described in an Australian magazine and probably the biggest published anywhere in recent years. It delivers 500 watts RMS into a 4Ω load and 278 watts into an 8Ω load. August 1997  25 26  Silicon Chip Fig.1: the circuit uses 12 output transistors in a complementary symmetry arrangement, driven by an MJL21193/4 pair; ie, the same as the output transistors. Short circuit current limiting is provided by Q24 & Q25. The supply rails are ±80V so we have had to specify high voltage transistors for the input differen­tial pair, Q1 & Q2. record and reproduce because of its huge dynamic range – even when it’s not being played particularly loudly, most amplifiers and loudspeakers are not up to the task. But a pair of these new power amplifiers and large loudspeakers to match would certainly cope with any CD of classical piano! Without getting too much ahead of ourselves, this new amplifier design produces only about 5dB more power than the 1988 design so the difference in absolute loudness won’t be huge. On the other hand, it will be noticeably louder and will be far less likely to be over-driven than the older design. Background to the design It’s been a long time coming, this amplifier. It was first mooted more than 12 months ago in 1996 and we have made several false starts since, only to come to a stop as component availability or suitability stopped us from proceeding further. Also along the way we produced a full bridge design, effec­tively two power amplifiers on the one PC board which drive the single loudspeaker in anti-phase. The driving voltages from the two amplifiers add and so the power delivered is the sum of the power outputs from the two amplifiers. The advantage of the bridge design is that the amplifier supply voltages can be sub­stantially less than the equivalent large single-ended amplifier. The lower supply voltages mean that the electrolytic ca­pacitors in the power supply are less expensive and the transis­tors used throughout the amplifier can have a lower voltage rating. In practice, it was the rarity of suitable high voltage high current driver transistors that pushed us along this line of development. However, the resulting bridge amplifier proved to be not as efficient as a single-ended design and with the heatsink avail­able to us at the time, Fig.2: these are the load lines for 4Ω and 8Ω operation. The straight lines are for resistive loads while the arched lines are for reactive 4Ω (2.83Ω + j2.83Ω) and 8Ω (5.6Ω + j5.6Ω) loads. The concave lines show the 1200W power hyperbola (dotted) and the one-second SOAR curve for six MJL21193/4 power transistors. As you can see, the reactive 4Ω load comes quite close to the one-second SOAR curve. That is why a total of 12 output power tran­sistors is required. it proved impossible to cool it effec­ tively, even with two fans! After running up that blind alley, we went back to the drawing board. This time we were successful, with a bigger heat­sink, fan cooling and a thermal cutout. And instead of using conventional driver transistors, we used power output transistors in the driver stages. The power transistors specified have the advantage of being much more rugged and with a minimum gain-bandwidth product of 4MHz, their high frequency performance is just as good as many driver transistors such as the commonly used Motorola MJE340/350 pairs. The result of all the development Specifications Output power....................................278 watts into 8Ω; 500 watts into 4Ω Music power.....................................315 watts into 8Ω; 590 watts into 4Ω Frequency response ........................-0.3dB at 20Hz and 20kHz (see Fig.8) Input sensitivity.................................1.43V RMS (for full power into 8Ω) Harmonic distortion..........................typically less than .01% Signal-to-noise ratio............................... 117dB unweighted (20Hz - 20kHz); 122dB A-weighted Damping factor.................................>170 at 100Hz & 1kHz; >75 at 10kHz Stability.............................................unconditional August 1997  27 AUDIO PRECISION SCTHD-W THD+N(%) vs measured 10 LEVEL(W) 19 JUN 97 22:07:52 1 0.1 0.010 0.001 10 100 800 Fig.3: THD (total harmonic distortion plus residual noise) versus power at 1kHz into a 4Ω load. AUDIO PRECISION SCTHD-W THD+N(%) vs measured 10 LEVEL(W) 19 JUN 97 22:09:02 1 0.1 0.010 0.001 10 100 800 Fig.4: THD (total harmonic distortion plus residual noise) versus power at 1kHz into an 8Ω load. work is an amplifier capable of delivering 500 watts into a 4Ω load at .04% harmonic distortion and 278 watts into an 8Ω load at less than .009% harmonic distortion. Using the IF Music Power test conditions, the power output is 590 watts into 4Ω and 314 watts into 8Ω. Big power like this does not come in small packages. The amplifier uses 28  Silicon Chip 14 power transistors in all, from the Motorola MEL21193/94 series. These plastic power transistors are rated at 250 volts, 16 amps (30 amps peak) and 200 watts and have been featured in previous amplifier designs in the April 1996 and March 1997 issues of SILICON CHIP. As indicated above, two of the power transistors are used as drivers while the other twelve are used in the output stage. All are mounted on a large single sided heatsink. The PC board meas­ures 362 x 99mm. This month we are presenting just the PC board module itself but because of its sheer size and power output we strongly recommend that readers do not “do their own thing” and install the module with just any old power supply components and in just any old chassis. So next month we will present the full details of mounting the PC module in a chassis with a big power supply, fan cooling, the overload protection module presented in April 1997 and so on. By the way, we will be presenting it as a rack mounting mono amplifier only; if you want that magnificent stereo setup mentioned above, you would need two of these mono amplifi­ers. Performance The main performance parameters are summarised in the accompanying specifications panel and also demonstrated in a number of graphs. These indicate that just because a power amplifier deliv­ers a lot of power it does not mean that it cannot deliver high performance as well. This amplifier is very quiet (-122dB A-weighted with respect to full power into 8Ω) and has low distor­tion, typically around .01% or less. In fact, the amplifier is quieter than any CD player on the market. Note that there is not a lot of difference between the music power output and the continuous power output of this ampli­fier; ie, 500W continuous versus 590W music power. This amounts to a “dynamic headroom” figure of 0.7dB for 4Ω loads. This is a reflection of the fact that the power supply is very well regu­lated – a consequence of using an 800VA transformer and a filter capacitor bank of 80,000µF in total. While this may seem extravag­ant, cutting back on the power supply parameters does prejudice the performance. Note also that our power figures are quoted for a mains supply voltage of 240VAC. Typically, the mains supply is higher than this and so the maximum “unclipped” power output will be somewhat higher again. Bipolars vs. Mosfets In line with our philosophy of generally not using Mosfets in audio amplifiers, we have used bipolar tran- sistors in the output stages. Bipolar transistors have the advantage of requir­ing a lower quiescent current (to avoid crossover distortion) and for a given supply voltage they deliver more power than an equiv­alent design using Mosfets. Bipolars are also generally cheaper than equivalent complementary Mosfets (ie, N-channel and P-chan­nel pairs). Furthermore, as a result of our recent testing of this amplifier under conditions of maximum power dissipation, we are convinced that a Mosfet amplifier of this power rating would require considerably larger fan-cooled heatsinks if it was to be able to deliver its rated power on a continuous basis. Mosfet amplifiers are reputed to be almost “unburstable” because if they become overheated, they tend to shut down. While this is an advantage under overload conditions, this characteristic is a drawback when you want the amplifier to deliver lots of power on a continuous basis. As a Mosfet amplifier gets hotter, it deliv­ers less power. If it gets very hot, it throttles right back. By contrast, if a bipolar design becomes very hot, it still keeps on delivering the goods and the heatsink must prevent the output transistors from becoming overheated otherwise they will be destroyed. Overall though, a bipolar design is more efficient and requires less heatsinking. AUDIO PRECISION SCTHD-HZ THD+N(%) vs FREQ(Hz) 5 19 JUN 97 22:46:21 1 0.1 0.010 0.001 20 100 1k 10k 20k Fig.5: THD versus frequency at 250W RMS into a 4Ω load. AUDIO PRECISION SCTHD-HZ THD+N(%) vs FREQ(Hz) 5 19 JUN 97 22:44:35 1 0.1 Circuit details The full circuit diagram is shown in Fig.1. Aside from the large number of output transistors, the circuit is almost identi­cal in configuration to the lower power designs featured in April 1996 and March 1997. It also incorporates the same short-circuit overload protection circuit as in the March 1997 design. For the benefit of those readers who have not seen the previous articles and for the sake of completeness we shall go through the circuit description in detail. Note that the supply rails are ±80V or a nominal 160V in total, under no signal conditions. This very high voltage has required us to specify more rugged transistors than have been required in the past. This is particularly the case for the driver transistors, as already mentioned, and for the input transistor pair, Q1 & Q2. In 0.010 0.001 20 100 1k 10k 20k Fig.6: THD versus frequency at 150W RMS into an 8Ω load. the latter case, we have specified two 2N5401s rather than the BC556s we have used in the past. The 2N5401s have a collector voltage rating of 150 volts versus 80 volts for the BC556. The input signal is coupled via a 2.2µF capacitor and 1.2kΩ resistor to the differential pair of transistors Q1 & Q2. Q3 is a constant current source which sets the current though the differ­ential pair. The current through Q3 is set by diodes D1 & D2 and this sets the voltage across Q3’s 120Ω emitter resistor to 0.85V. This sets the current though Q3 to 7mA and so this is shared by Q1 & Q2 at 3.5mA each. Q3 is included instead of a common emitter “tail” for Q1 & Q2 because it renders the amplifier less sensitive to variations in the power supply rails. This is known as PSRR (power supply rejection ratio) and all good amplifier August 1997  29 AUDIO PRECISION SCFRQRES AMPL(dBr) vs FREQ(Hz) 5.0000 19 JUN 97 22:40:55 4.0000 3.0000 2.0000 1.0000 0.0 -1.000 -2.000 -3.000 -4.000 -5.000 20 100 1k 10k 20k Fig.7. frequency response at 20W into a 4Ω load. AUDIO PRECISION SCFRQRES AMPL(dBr) vs FREQ(Hz) 5.0000 19 JUN 97 22:42:24 4.0000 3.0000 2.0000 1.0000 0.0 -1.000 -2.000 -3.000 -4.000 -5.000 20 100 1k 10k 20k Fig.8: frequency response at 10W into an 8Ω load. designs, including op amps, feature a very high PSRR. Current mirror The collector loads of Q1 & Q2 are provided by Q4 & Q5 which operate as a “current mirror”. While it is a little hard to visualise just how a “current mirror” works, it is easier if you think of Q5 acting as a sharp 30  Silicon Chip cutoff diode, providing a voltage at the base of Q4 which is equal to the base-emitter voltage drop of Q5 (about 0.6V) plus the voltage drop across its 220Ω emitter resistor. What happens is that if Q2 tends to draw more than its share of emitter current from Q3, the voltage at the base of Q4 tends to increase and so Q4’s collector current tends to rise also. This forces Q1 to pull a bit more current and stop Q2 from taking more that its fair share. We say that Q4 “mirrors” Q5 and so Q1 “sees” a collector load which is a higher impedance than would otherwise be the case. The result is increased gain and improved linearity from the differential input stage. As a matter of interest, current mirror stages are very commonly used in op amp ICs, partly because they are easy to design in and partly because of their enhanced performance. The signal from the collector of Q1 drives a cascode stage comprising transistors Q7 & Q8, together with the constant cur­rent load transistor Q6 (top of the circuit). The cascode stage is another circuit which is a little hard to visualise but if you break it into sections, it is easier. Note that Q8 has a 3.3V zener diode ZD1 to hold its base voltage constant and so Q8 acts like an emitter follower to provide a constant collector voltage to Q7. This eliminates any gain variations (non-linearities) which would otherwise occur if Q7’s collector voltage was free to vary. The varying current drawn by Q7 becomes the input signal to the emitter of Q8 which is effectively operating as a “grounded base” stage. Q8 converts the varying signal current at its emit­ter into a varying signal voltage at its collector. The combined effect of operating such a cascode stage is improved linearity and bandwidth compared with a single common emitter stage. 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; ie, to eliminate the possibility of the ampli­fier oscillating supersonically. The output from the cascode stage is coupled to the driver transistors, Q10 & Q11. As mentioned previously, these are MJL21193/94 power transistors, the same as in the output stage. Note that the signals to the bases of Q10 & Q11 are identical, apart from the DC offset provided by Q9. Vbe multiplier In setting the DC offset between Q10 & Q11, Q9 is actually setting the quiescent current in the output stages. It provides a forward bias of about 2.3V or so between the bases of Q10 & Q11 so that they are always slight- ly turned on, regardless of whether signal is present or not; that is why it is referred to as “quiescent” current. Q9 acts as a “Vbe multiplier”, multiplying the voltage between its base and emitter by the ratio of total resistance between its collector and emitter to the resistance between its base and emitter. In practice, trimpot VR2 is adjusted not to give a particu­lar voltage between the collector and emitter of Q9 but to set the quiescent current through the output transistors. We’ll discuss how this is done in the setting up procedure. It is important that the bias voltage produced by Q9 tracks the temperature of the output stage transistors. As the output transistors become hotter, Q9’s collector-emitter voltage should drop, so that the quiescent current is reduced and the danger of thermal runaway is averted. Our prototype photo this month shows Q9 directly on top of Q12 but next month it will be shown above Q12. Output stage The output stage of the amplifier is effectively a comple­mentary symme- try emitter follower, comprising six NPN transistors and six PNP transistors. We need this many transistors to safely deliver the high peak currents involved (up to 17 amps peak) at high voltages. The load line curves of Fig.2 demonstrate that while 12 output transistors are adequate to cope with reactive 4Ω loads (typified by the 2.83Ω + j2.83Ω curve), there is not a lot of power capacity to spare when you look at the 1200W and SOAR hyperbola curves. In other words, while 12 big power transistors might look like a lot, every one of them is needed to safely deliver full power into typical 4Ω loudspeaker loads. Each output power transistor has a 0.47Ω emitter resistor and this more or less forces the output transistors to roughly share the load currents. If one of the power transistors tends to take more than its share of load current, the corresponding voltage drop across its emitter resistor will be proportionately higher and this tends to throttle the transistor back until its current comes back into line with the others. The emitter resistors also help to stabilise the quiescent current to a small degree and they slightly im- prove the frequency response of the output stage by providing current feedback. Gain setting Negative feedback is applied from the output stage back to the base of Q2 via an 18kΩ resistor. The amount of feedback is set by the 18kΩ resistor and the 560Ω resistor at the base Q2. These set the gain of the amplifier to 33. The low frequency rolloff is set mainly by the ratio of the 560Ω resistor to the impedance of the 100µF capacitor. This gives a -3dB point of about 2.8Hz. The 2.2µF input capacitor and 18kΩ bias resistor to Q1 have similar effect and give a -3dB point of 4Hz. The two time-constants combined give an overall rolloff of about 7Hz. At the high frequency end, the 820pF capacitor and 1.2kΩ resistor feeding the base of Q2 form a low pass filter which rolls off frequencies above 160kHz (-3dB). The overall amplifier frequency response is demonstrated in the curves of Fig.7 and Fig.8. An output RLC filter comprising a 5.7µH choke, a 6Ω resis­tor and a 0.15µF capacitor couples the signal to the loudspeaker. It isolates the am- SILICON CHIP This advertisment is now out of date. Please feel free to visit the advertiser’s website: www.emona.com.au August 1997  31 Parts List For 500W Amplifier Module 500 amplifier PC board 1 PC board, code 01208971, 362mm x 99mm 4 20mm fuse clips 2 5A or 7.5A 20mm fuses (see text) 1 coil former, 24mm OD x 13.7mm ID x 12.8mm long, (Philips 4322 021 30362) 1 2-metre length 1mm enamelled copper wire 1 200Ω trimpot (Bourns 3296W or similar) (VR2) 1 100Ω multi-turn horizontal mount trimpot (VR1) 7 PC stakes 2 TO126 heatsinks, Jaycar Cat. HH8504 or similar 1 single-sided heatsink, 400mm wide x 118mm high x 48mm deep, or two 200mm x 118mm x 48mm 14 TO-3P insulating washers 2 TO-126 insulating washers 17 3mm x 10mm screws 3 3mm nuts Semiconductors 2 2N5401 PNP transistors (Q1,Q2) 2 BC556 PNP transistors (Q3,Q25) 4 BC546 NPN transistors (Q4,Q5,Q7,Q24) 1 MJE350 PNP transistor (Q6) 2 MJE340 NPN transistors (Q8,Q9) 7 MJL21194 NPN power transistors (Q10,Q12-Q17) plifier from any large capacitive react­ ances in the load and thus ensures stability. Perhaps more importantly, the filter attenuates any RF signals picked up by the speaker leads and stops them being fed back to the amplifier’s input stage where they could cause audible breakthrough – no-one likes listening to radio stations when they are supposed to be hearing CDs. Overload protection & offset adjustment Two other circuit features need to be mentioned: DC offset adjustment and overload protection. Strictly speak­ing, the DC offset adjustment is not really necessary if the amplifier is not to be used with an output transformer, as 32  Silicon Chip 7 MJL21193 PNP power transistors (Q11,Q18-Q23) 4 1N914 small signal diodes (D1,D2,D3,D4) 2 1N4936 fast recovery diodes (D5, D6) 1 BZX55C3V3 3.3V 0.5W zener diode (ZD1) Capacitors 4 100µF 100VW electrolytic 1 100µF 16VW electrolytic 1 2.2µF 16VW electrolytic 1 0.15µF 275VAC (Philips MKP 2222 336 10154) 5 0.1µF 100VW MKT polyester 1 820pF MKT polyester or ceramic 1 100pF 500V ceramic (Philips 2222 655 03101) Resistors (0.25W, 1%) 4 22kΩ 1W 2 18kΩ 1 6.8kΩ 1W 1 1.2kΩ 1 560Ω 1 470Ω 2 390Ω 5W 10% 4 270Ω 2 220Ω 1 180Ω 1 120Ω 5 100Ω 2 30Ω 3 18Ω 1W 12 0.47Ω 5W 10% it would be if it was driving a 100V line transformer for PA work. However, because we envisage that some readers will want to use the amplifier for public address, we have included DC offset adjustment. This is provided by the 100Ω trim­ pot (VR1) between the emit­ters of the input pair, Q1 & Q2. VR1 is used to adjust the current bal­ance between the input pair and this, because it is a DC feedback circuit, causes the DC offset at the output to vary. The trimpot is adjusted to make the DC offset as close to 0V as possible; it should be possible to keep to less than ±5mV. Transistors Q24 & Q25 and diodes D3 & D4 provide the over­load protection feature. Q24 monitors the current flow through the emitter resistor of Q12, via a voltage divider consisting of a 300Ω resistor and a 270Ω resistor. Normally, Q24 & Q25 are off and play no part in the ampli­fier’s operation. However, if the current through the 0.47Ω resistor of Q12 exceeds about 3 amps, Q24 begins to turn on and shunts the base current from Q10, the associated driver transis­tor. This means that not only is Q12 throttled back, but so are the other five NPN output transistors, Q13-Q17, because they all must operate in an identical way. Hence the peak output current is prevented from exceeding about 18 amps. This means the amplifier can deliver full power into a 4Ω load but if a 2Ω load, for example, was connected, the power output would be heavily limited. The same process happens for Q25 which monitors the emitter current of Q18 (and thus Q19-Q23). The diodes D3 & D4 are included to prevent Q24 & Q25 from shunting the drive signal when they are reverse-biased; this happens for every half cycle of the signal to the driver transistors. Diodes D5 & D6 are included as part of the protection cir­cuitry and they absorb any large spikes which may be generated by the inductance of the loudspeaker when the current limiting circuit cuts the drive to the output transistors. D5 & D6 are fast recovery diodes, included to ensure their operation at high frequencies and high power. Thermal cutout Because the overload protection simply limits the current to the load, the output transistors and the fuses are protected from sudden death in the case of a momentary short circuit but if the overload (or short circuit) is maintained and the drive continues, the amplifier will very rapidly overheat and may still expire unless the fault condition is correctly quickly. To prevent failure of the output transistors, the circuit includes an 80°C thermal cutout. This is not shown on the circuit of Fig.1 but is a vital part nevertheless. It is part of the relay protection circuit to be presented next month. Next month, we’ll present the circuit of the power supply and for the relay protection circuit and give the construction details of the complete SC amplifier.