Fullscreen HTML5Med Res (??MB)HTML4

E R U FEATJECT PRO Build the –– aa new, new, high high perform perform By LEO SIMPSON & PETER SMITH The transmogrification of an amplifier. Many readers will recognise the venerable (26-year-old!) ETI-480 top left. But the SC480 (bottom right) is as modern as tomorrow – with performance to match. Performance into 4Ω 50 watts into 8Ω; 70 watts Output Power tts into 4Ω 77 watts into 8Ω; 105 wa Music Power 1W – see Fig.1) B at 14Hz and 70kHz (at Frequency response -1d 0.875V for 50W into 8Ω Input sensitivity typically <.003% 5% from 20Hz to 20kHz; d, Harmonic distortion <.0 22kHz); -119dB A-weighte 4dB unweighted (22Hz to -11 io Rat ise -No -to nal Sig into 8Ω both with respect to 50W t PTC thermistor h respect to 8Ω and withou wit z, 1kH & Hz 100 at >140dB Damping factor thermistor es plus “Polyswitch” PTC Protection fus 26  S ilicon Chip onditional Stability unc www.siliconchip.com.au SC ETI-480 mance mance amplifier amplifier module module Have you built one – or more – of the popular ETI-480 power amplifier modules over the years? Here is the module to replace that old design. The SC480 produces a great deal less distortion, is much quieter and has inbuilt protection. It also sounds much better. B ack in the October 2002 issue we noted that we intended to produce a replacement module for the very popular ETI-480 amplifier module which was published back in December 1976 – 26 years ago. In the same note we stated, somewhat controversially, that the ETI-480 was a dog of an amplifier and that it was not a good performer, even by 1976 standards. Having made that outrageous statement (to some readers, at least), we had to come up with the goods. Fortunately, we were pretty confident that we could, and we are pleased to state that this new module is even better than we had hoped. It uses the same power output transistors as in the ETI-480 and just one more low-cost transistor has been added to the overall component count. Kit cost should be about the same as for the ETI-480. When this project was first mooted, we decided to base it on TIP3055 and TIP2955 plastic power transistors. These are 60V 15A 90W transistors in the TO-218 (SOT-93) encapsulation. We i n t e n d e d to produce a new version of a 100W module which was published in the December 1987 issue of SILICON CHIP. That design was based on a Hitachi amplifier circuit and used 2N3055/MJ2955 power www.siliconchip.com.au transistors in TO-3 metal encapsulation. Accordingly, we produced a PC board pattern for the new module and while we waited for it to be produced by RCS Radio Pty Ltd (thanks Bob), we realised that a substantial number of readers who had built countless ETI-480 modules would probably like to “graduate” to our new design but would wish to at least reuse the TO-3 power transistors from their ETI-480s on the new board. Hence, the idea of a TO-3 version of the new module also came to pass, as is featured here. The plastic version of the module has the power transistors lined up along the back edge, making it easy to mount them to the vertical surface of a finned heatsink. The TO-3 version is larger and has the four power transistors mounted on the horizontal shelf of a cast heatsink or on a rightangle bracket which can then be mounted on a vertical heatsink. Why publish both modules? The simple answer is that we had produced them both, so why not? However, each module has its own advantages. Version 2, using the TO-3 transistors, is rugged but takes up more space and is likely to be less convenient to mount in a typical chassis. There is also more work in assembling Version 2 with the TO-3 transistors. Version 1, with the plastic power transistors is quite a bit more compact and less trouble to mount in a typical chassis but the module assembled onto a heatsink is not quite as rugged to handle. By the way, if you decide to build Version 1, don’t be tempted to substitute the (usually) cheaper MJ“E”... versions of the transistors. These TO-220 transistors are rated lower (only 75W) and will inevitably cause you great pain and suffering. January 2003  27 Which version to build? Our preference is for Version 1 but we have a sneaking suspicion that Version 2, with the TO-3 power transistors, will be the more popular module (especially amongst those looking for a somewhat look-alike ETI-480 substitute). Depending on the particular brand of power and driver transistors used, both modules will give virtually identical performance. Regardless of which version you decide to build, the performance will be vastly better than the old ETI-480 design. And that is as it should be. After all, we should have learnt quite a bit about amplifier design in 26 years or so, shouldn’t we? Performance Power output is 50 watts RMS into a 8Ω load and 70 watts into 4Ω load, before the onset of clipping. Music power is around 77 watts into 8Ω and 105 watts into 4Ω. Hang on a minute! Wasn’t the ETI-480 claimed to be 100W into 4Ω? Well, it was but the distortion graph published by ETI back in December 1976 shows the amplifier heading well into clipping at around 70W RMS. This is to be expected since both the ETI-480 and the new SC480 use the same voltage rails and the same output transistors. A particular feature of the SC480 is low distortion. Distortion for all power conditions, up to clipping, into an 8-ohm load, is less than .05% for the full range of frequencies from 20Hz to 20kHz. Similarly, with a 4-ohm load, total harmonic distortion is less than .07% for the full audio frequency range. In reality, this is a very conservative rating as the distortion will typically be .003% or less for both load conditions. And for very lower power levels, less than 100mW, where noise becomes a significant part of the measurement, the distortion is really low, down to as low as .0005%. This is two orders of magnitude better than the ETI-480! Signal to noise ratio is better than -114dB (unweighted, 22Hz to 22kHz) with respect to full power into an 8Ω load. Frequency response is just 1dB down at 14Hz and 70kHz (see Fig.1). Fig.1: this is the frequency response of both versions of the new amplifier, taken at a power level of 1W into an 8Ω load. Transistor quality As in most things, you get what you pay for and it is no different with these modules. The plastic version (Version 1) of the amplifier was built with the output and driver transistors in what we would call the premium brands: Philips, Motorola (On Semi) and ST Micro. Version 2 was built with second rank power and driver transistors (Mospec). We did this to compare performance and we are pleased to report that although the premium branded transistors do give slightly better performance, there is a not a lot in it. Refer to the distortion graphs of Figs.2-9 to make the comparisons. Either way, the performance of these modules is very good, especially considering that we are not using expensive transistors such as Motorola MJL21193/4 or the even more expensive MJL1302A & MJL3281A. In fact, in some respects the measured performance challenges that of our popular and more powerful Plastic Power module published in the April 1996 issue. Interestingly, a These two oscilloscope screen grabs show just how clean this new amplifier is. The first screen (left) shows a 1kHz output waveform at a level of 40W into 8Ω at top. The lower trace is the distortion waveform which has been “averaged” by the scope to remove noise. Note that it is mostly second harmonic distortion. The same process has been applied to the screen shot at right except that it is a 10kHz signal. Again, the distortion is mainly second harmonic. 28  Silicon Chip www.siliconchip.com.au key part of that performance standard comes about because of improved PC board and wiring layout. We’ll discuss these vitally important aspects in more detail later in this article. Oh, and we should state that the SC480 Version 1 and Version 2 modules are a drop-in replacement for the ETI480 modules but will sound a great deal better. While nominally of the same rating, they will deliver more power, they’re quieter and as already detailed, much lower in distortion. By the way, these modules are not suitable for driving 2Ω loudspeakers as used in car sound systems. We do not have space to publish the load/line curves in this article but suffice to say that attempting to drive 2Ω loads will blow the fuses and may blow the output transistors as well. Protection The trouble with all high-power amplifiers is that, if a transistor fails, there is a big chance that the loudspeaker system could be damaged, despite having fuses in the power supply. The problem is that the fault condition may place a large DC voltage across the speaker’s voice coil and the resulting current may not blow the fuses. The speaker’s voice coil then gets red hot and may actually set the speaker cone on fire! Once that happens and if you’re not there to kill the power to the amplifier, you can have a raging fire in your home and enormous amounts of smoke being generated by the burning of the filling material in the cabinet. Our normal approach to this problem is to incorporate relay protection which will disconnect the loudspeaker in the event of a large DC fault condition occurring in the amplifier. Relay protection works as far as the speaker is concerned but it doesn’t protect the amplifier itself if the loudspeaker leads are shorted. Here again the fuses may not blow before the output transistors are damaged. Neither do fuses protect the speakers if you seriously over-drive the amplifier. This is a particular risk for tweeters but even woofers can have voice coil damage by serious over-drive. Complete protection The method we have used to provide protection to both the loudspeaker and amplifier is to connect a high current positive temperature coefficient (PTC) thermistor (known commercially as a “Polyswitch”) in series with the output circuit. This is the same method of protection as we used in the original module published in December 1987. The PTC thermistor normally has a very low resistance but when the current through it rises to high value, it immediately switches to a high resistance state and stays in that condition until the fault is fixed or power is removed. The resistance of the PTC thermistor is so low (typically 0.1Ω or less), it has a negligible effect on amplifier performance, apart from the fact that it does cause a reduction in damping factor. In practice, it works extremely well. It allows you to drive the amplifier to full power on program signals but the moment a short circuit is applied or the amplifier is seriously over-driven, the PTC thermistor goes high in resistance to cut off the fault current. After the protection thermistor has switched to its high state, it takes some time to revert to its low resistance condition, after the fault current has ceased. This depends on how much current is passing through it. If the drive level is maintained after a fault has occurred, the protection thermistor will stay high in resistance. Circuit description Now let’s have a look at the circuit of Fig.10. 13 transistors and three diodes make up the semiconductor complement. The input signal is coupled via a 1µF bipolar electrolytic capacitor and 2.2kΩ resistor to the base of Q2. Q2 & Q3 make up a differential pair. Q1 is a constant current source which sets the current through Q2 & Q3 and renders the amplifier largely insensitive to variations in its supply rails Just to confuse you, Version 2 of the SC480 amplifier (with TO-3 transistors) is on the left, while Version 1 (with TO-218 transistors) is on the right. There is only a small difference in performance between the two versions. www.siliconchip.com.au January 2003  29 Fig.2: THD versus power at 1kHz into an 8Ω load for Version 1 (TO-218). Fig.3: THD versus power at 1kHz into a 4Ω load for Version 1 (TO-218). (power supply rejection). Signals from the collectors of Q2 & Q3 drive another differential pair, Q4 & Q5 which have a “current mirror” as their collector loads. The current mirror, comprising D3 and Q6, ensure that this second differential stage has high linearity (ie, low distortion). The output of Q5 is then used to drive class-AB output stage consisting of drivers Q8 & Q9 and power transistors Q10, Q11, Q12 & Q13. Q7 is a Vbe multiplier, so-called because it multiplies the voltage between its base and emitter to provide a fixed voltage between its collector and emitter, regardless of the drive current delivered to the output stage by Q5. The voltage is adjusted by trimpot VR1. The function of Q7 is to set the DC voltage applied between the bases of Q8 & Q9. By doing this it sets the “quiescent current” in the output stage (ie, the current when no signal is present). This is to minimise crossover distortion. In fact, our tests did not reveal any signs of crossover distortion. The complementary output transistors are connected in parallel to give high output current capability. Each transistor has its own 0.22Ω emitter resistor. These are included to ensure that the output current is shared reasonably well between the output transistors. Negative feedback is applied from the output stage back to the base of Q3 via a 22kΩ resistor. The level of feedback, and therefore the voltage gain, is set by the ratio of the 22kΩ resistor to the 1kΩ at the base of Q2. The 47µF bipolar capacitor in series with the 1kΩ sets the DC gain to unity and sets the -3dB point of the frequency response to about 3Hz. The other determinant of the amplifier’s low frequency response is the 1µF input capacitor and the 22kΩ base bias resistor feeding Q1 and these set a -3dB point at about 7Hz. The 330pF capacitor together with the 2.2kΩ resistor feeding Q2 form a low pass filter to roll off frequencies above 200kHz. The 68pF capacitor between the base and collector of Q5 and the 10pF capacitor between base and collector of Q2 roll off the open-loop gain of the amplifier to ensure stability with feedback applied. Note that the 68pF capacitor can be a ceramic or polystyrene type and must a have a voltage rating of 100V or more. Other capacitor types are not recommended. Another important factor in the amplifier’s excellent Fig.4: THD versus power at 1kHz into an 8Ω load for Version 2 (TO-3). Fig.5: THD versus power at 1kHz into a 4Ω load for Version 2 (TO-3). 30  Silicon Chip www.siliconchip.com.au Fig.6: THD versus frequency at 40W into an 8Ω load (Version 1). Fig.7: THD versus frequency at 60W into an 4Ω load (Version 1). stability is the output RLC network consisting of the 6.8µH choke, a 6.8Ω resistor and a 150nF capacitor. Not only does this network ensure stability but the capacitor is an effective killer of any RF and mains-interference signals which can be picked up by long loudspeaker leads. As noted earlier, the design of the PC board is a very critical part of the overall circuit. The placement of the components and the way that heavy currents flow in the tracks is all arranged to minimise the radiation of harmonics into the input stage involving Q1 & Q2. This board is yet a further refinement of the topology we first introduced in the Ultra-LD amplifier featured in March, May & August 2000 and then again in November & December 2001. The PC board for version 2 and the component placement is shown in Fig.12. It incorporates “star earthing” whereby all earth currents come back to a central point on the board, thereby avoiding any flow of output, supply and bypass currents flowing in the signal earths. Furthermore, placement of the copper tracks to the output stages is arranged, as far as possible, to cancel the magnetic fields produced by the asymmetric currents drawn by each half of the output stage. By way of explanation, when the positive half of the output stage (Q10 & Q12) conducts, the DC current drawn is effectively a positive half wave (ie, rectification takes place) of the signal waveform. And when the negative half conducts (Q11 & Q13), the DC current is the negative half wave. A major cause of harmonic distortion in class-B amplifiers is the magnetic fields produced by these asymmetric Fig.8: THD versus frequency at 40W into an 8Ω load (Version 2). Fig.9: THD versus frequency at 60W into an 4Ω load (Version 2) Power supply The power supply circuit is shown in Fig.11. This uses a centre-tapped 56V transformer driving a bridge rectifier comprising four 1N5404 diodes and two 4700µF 50V filter capacitors. This produces unregulated supply rails of about ±40V. Depending on the mains AC voltage, the rails will drop to around ±32V or less when the amplifier module is delivering full power into a 4Ω load. We have also provided a ±15V DC supply for a preamplifier. This is derived with 2.2kΩ resistors and two 15V 1W zener diodes. PC board topology www.siliconchip.com.au January 2003  31 Fig.10: this direct-coupled amplifier module uses a differential input stage (Q2,Q3) with a constant current tail (Q1). This drives another differential amplifier (Q4,Q5) with current mirror load (D3,Q6). Quiescent current in the output stage is set by VR1 and Q7. The output stage is a complementary class-AB configuration using Q8 & Q9 as drivers and Q10 to Q13 as the output devices. Voltage readings are taken with no signal applied. currents inducing unwanted signals into the input stages, in this case involving Q1 & Q2. So we have tried to cancel these fields as much as possible (in a single sided PC board). For example, notice how the positive fuseholder (F1) is placed close and parallel to the emitter resistors for Q10 & Q12. So what happens is that the magnetic field produced by the asymmetric current in fuse F1 is more or less cancelled as the same current flows back in the emitter resistors. This is the main reason why the layouts for these two modules is much tighter than our designs of recent years. You will see the same method employed in the Version 1 of the board, with the heavy collector and emitter tracks 32  Silicon Chip placed close together but we think this has been more fortuitous on Version 2 than on Version 1. It is then most important to arrange the DC supply cables to the amplifier to further this cancellation process. We’ll detail this in the construction description. To make the input stage less vulnerable to spurious magnetic fields from the output stage, we have concentrated it into as small an area of the PC board as possible. Another trick is the location of the takeoff point for the 22kΩ resistor and its orientation at rightangles to the output stage emitter resistors. Finally, the signal earth for the input stage is separated from the main amplifier earth by a 10Ω resistor. www.siliconchip.com.au Fig.11: the power supply is very simple but adequate. The ±15V preamplifier supply is optional. This is not so important when a single module is in use but it is most important when two modules are used in a stereo system. In that situation, the joining of the two signal earths back via the input cables to a single program source such as a CD player will cause an earth loop and a resulting major degradation in the separation between channels and lesser degradation in the distortion performance. Well, that’s probably enough discussion of the PC board but suffice to say that the overall design has been carefully arranged to minimise distortion and leave as little to chance in the wiring layout so that constructors are certain to get excellent results. Next month, we’ll give the full details of assembly, wiring and setup of both versions, the parts list and the PC SC board patterns. Fig.12: version 2 of the SC480 amplifier with the TO-3 (steel) transistors. We’ll be presenting this again next month as part of the constructional details but it is reproduced here to demonstrate the attention we have paid to the PC board design to achieve the exceptional performance figures depicted on earlier pages. www.siliconchip.com.au January 2003  33