Fullscreen HTML5Med Res (??MB)HTML4

Ultra-LD Mk.4 200W RMS Power Amplifier Module, Pt.2 By NICHOLAS VINEN This month, we provide the construction details for our new ultra-lowdistortion amplifier module. Most of the parts on the PCB are surfacemount types, keeping it compact and allowing for unprecedented low levels of distortion. We have avoided difficult-to-solder parts. B Y NOW, you should be familiar with the features, specifications and circuit details of our new UltraLD Mk.4 amplifier. This month, we’re going to discuss some aspects of the PCB design, describe how we tweaked it to refine the performance and then 90  Silicon Chip go through the module assembly procedure. PCB design One advantage of the new PCB over the Mk.2 and Mk.3 designs is that we’ve totally eliminated all high-current vias, so there is no more concern about vias fusing under fault conditions and no wire feed-throughs are required. All high-current paths stay on the same side of the PCB. The front-end control section is routed entirely on the top layer, with just siliconchip.com.au MJE15030 BD139 MJE15031 Q6 FZT696B 622 Signal input D1 BAV99 100k 100k 104 104 68k 683 333 100k 511 CON1 102 104 1 1k 33k 104 47R 68R 68R 100k 100k Q16 ZD2 D5 Q14 ZD1 Q15 D7 D6 ZD1Q15 104 100k CON4 A A LED4 CLIPPING 47k CON3 –57V F2 M205 6.5A FAST BLOW +57V (2x27 Ω UNDER) 27Ω 27Ω 1W 1W K D4 A 0V 100nF 200V NP0 or PP POWER 331 121 AIR CORE (13.5T 1.25mm ECW) 123 47R 1M 1 101 10R 105 10Ω 68R 222 12k 1 Q1 511 12k 47µF L1 100Ω 510Ω 47 µF 1000 µF 16V 6.3V NP 1 µF 1nF1 1nF L2 2.2 µH SILICON CHIP 222 123 2.2k 2 x 0.1Ω 3W (UNDER) TP7 222 150k 154 150pF 15pF 1nF Q4 2.2k 150pF 222 K A 2.2k 6.2k 622 LED1 1 µF 12k VR2 47µF 35V 330Ω 331 2x47Ω2 47Ω 2x68Ω 68Ω Q2 Q3 2.2k 6.2k TP4 1µF 100V A LED3 K 47k – SPK + 39 0Ω 1W 391 + HP K – D3 CON2 A OUTPUTS D2 TP6 1µF TP4 100V 1000 µF 63V LOW ESR (OPTIONAL) F1 M205 6.5A FAST BLOW FZT796A Q5 TP2 100Ω VR1 120Ω 1k 330Ω 1µF 220Ω 100V (UNDER) 47 µF (UNDER) 6.3V 47 µF 63V Ultra-LD Mk.4 200W Amplifier 101 473 1000 µF 63V LOW ESR (OPTIONAL) NJL1302D 473 TP5 2 x 0.1Ω 3W (UNDER) NJL1302D Q13 473 100V 100Ω 101 TP5 1µF 1µF 100V A LED2 47k K GREEN= FUSE OK 473 47k Q9 TP1 TP3 Q12 Q8 27R Q7 27R NJL3281D 104 NJL3281D Q11 101 Q10 100Ω 1W K 01107151 RevB Fig.6: follow this layout diagram to install to parts on the top of the PCB. Fit the SMD parts first in the order listed in the text, then flip the PCB over and install the SMD parts on the bottom as shown in Fig.7. The remaining through-hole parts can then be fitted. Note that Q7Q13 are soldered to the PCB only after they have been attached to the heatsink. a few vias to connect components to the analog ground plane beneath. The remaining vias are arranged in pairs (or more) for redundancy and are mostly associated with either the clip detector circuit or the low-current power supply rails feeding the front end. The +57V and -57V pins of power input connector CON3 are soldered to top layer tracks which run to SMD fuse-holders F1 and F2 respectively, on the top side of the board. These then connect to two further top layer tracks which go to the output transistor collector pins. The output current at the emitter pins then runs along tracks on the bottom of the PCB to the 0.1Ω SMD emitter resistors, which are fitted directly underneath the fuse-holders. The current then feeds into another bottom layer track which combines the current from all four output transistors to through-hole air-cored inductor L2. A bottom layer track from the opposite end of L2 then connects to speaker output terminal CON2. Construction The double-sided PCB on which siliconchip.com.au the Ultra-LD Mk.4 is built is coded 01107151 and measures 135 x 93mm. The output transistors mount on a diecast aluminium heatsink using the same layout as the Ultra-LD Mk.3 amplifier. The easiest way to build the module is to first fit most of the SMDs on the top side, followed by the eight SMDs on the bottom, then the remaining SMDs and on-board through-hole components and finally the heatsinkmounted transistors. All the SMDs can be soldered using a regular soldering iron (ie, a pencil type) and solder wire as long as you have some solder wick and flux paste. Depending on your eyesight, you may also need a magnifying lamp or visor. A hot-air rework station or reflow oven can also be used although you will need to solder the fuse-holders and LEDs by hand as they can be damaged by excessive heat. If you’re assembling the unit from a kit which has the SMDs pre-soldered then you can skip this next section. Soldering the SMDs Fig.6 shows the parts layout on the top of the PCB. Begin by installing transistor Q2. This has a lead pitch of 0.95mm which is the finest of all the parts but it isn’t particularly difficult. First, remove the HN2C51F from its packaging (don’t drop it!) and inspect it under magnification to locate the pin 1 dot on the top. Place it near its mounting location with the correct orientation. Make sure it’s the right way up; the leads should be in contact with the PCB. Then flow a tiny amount of solder onto one of the corner pads on the board, without getting any solder on the other pads. Clean off the iron, grab the part gently with angled tweezers using your other hand, heat the solder on that pad and slide the part into place. Put the iron away and use a loupe WARNING! High DC voltages (ie, ±57V) are present on this amplifier module when power is applied. In particular, note that there is 114V DC between the two supply rails. Do not touch the supply wiring (including the fuseholders) when the amplifier is operating, otherwise you could get a lethal shock. September 2015  91 1µF 100V 220Ω 0.1Ω 0.1Ω 3W 3W 27R 27R 0.1Ω 0.1Ω 3W 3W 0R1 0R1 0R1 0R1 221 27Ω 27Ω 1W 1W Fig.7: once you’ve installed all the SMD parts on the top side, flip the PCB over and follow this layout diagram to install the eight SMD parts on the underside. Note that the four 0.1Ω resistors must be rated at 3W, while the two 27Ω resistors must be rated at 1W (don’t get these parts mixed up). Table 1 on the facing page shows the value code printed on the top of each SMD resistor. or similar to check that all six pins are correctly positioned over their pads, the pin 1 dot is in the right location and the part is sitting flat on the board. If not, reheat the solder joint and fix the problem by gently nudging the component. For example, if it isn’t sitting flat on the board, press down on it (not too hard) with the tweezers while heating the joint and it should drop into place. Alternatively, if it’s misaligned, carefully rotate or slide Parts List Errata In the parts list last month, the two VS-3EJH02 diodes were incorrectly listed as D2 and D4. They are D3 and D4. In addition, the bobbin specified for the 2.2μH air-cored inductor (L2) was incorrectly specified as having a 10mm ID. It should be 13mm ID. Finally, depending on how you choose to mount the transistors on the heatsink, you may need some additional hardware not listed last month, including three M3 x 10mm and four M3 x 15mm machine screws. 92  Silicon Chip it while heating the solder. Once it’s in place, solder the pins on the other side of the package. Don’t worry too much about bridging them; just make sure that the solder flows onto all three pins and their associated pads. Then do the same for the three pins on the other side, including the one you tacked down initially. Now it’s simply a matter of applying a small amount of flux paste along both sides of the IC, then using solder wick to remove the excess solder. Clean off using a flux remover (methylated spirits or rubbing alcohol will do in a pinch) and inspect under magnification to make sure all six fillets have formed correctly. Be aware that solder can adhere to the pin without flowing onto the PCB pad below. Once you’re happy with it, fit Q1 and Q3 which are in identical packages. Next, fit the 11 SOT-23 package parts: Q4, Q14-Q16, D1-D2, D5-D7 and ZD1-ZD2. These are similar to Q1-Q3 but with three widely-spaced leads. Use the same basic procedure; the correct orientation should be obvious as there is only one lead on one side of the package and two on the other. Do take care not to get the parts in the wrong place, though; if in doubt, refer to Fig.6. Transistors Q5 and Q6 can go in next. These are in larger packages with three leads plus a tab for heatsinking and are soldered to large copper planes so this will require a fairly hot iron. Smear a little flux paste on the large pad, then position the component on the PCB and solder one of the smaller leads at either end. You can then solder the tab and finish with the two remaining leads. Make sure that the FZT796A goes on the left and the FZT696B on the right. Now you can solder diodes D3 and D4 in place, with their cathode stripes towards the top of the board. These stripes are normally quite faint and you may need a magnifying glass to see them. The four LEDs can go in next. If you’re using the exact types we specified in the parts list last month, each will have a green cathode marking. However, some other SMD LEDs have similar markings at the anodes, so if using different types, check the data sheet or else use a DMM on diode test mode to figure out which end is the siliconchip.com.au Table 1: Resistor Codes 1 1MΩ 3-Digit Code (E24) 105 6 100kΩ 104 1003 2 68kΩ or 68.1kΩ 683 6812 4 47kΩ 473 4702 1 33kΩ 333 3302 3 12kΩ or 12.1kΩ 123 1212 2 6.2kΩ or 6.49kΩ 622 6491 4 2.2kΩ or 2.21kΩ 222 2211 2 1kΩ 102 1001 1 510Ω or 511Ω 511 5110 1 390Ω 1W 391 not applicable 2 330Ω or 332Ω 331 3320 1 220Ω or 221Ω 221 2210 1 120Ω or 121Ω 121 1210 1 100Ω 1W 101 not applicable 3 100Ω 101 1000 3 68Ω or 68.1Ω 680 68R1 2 47Ω or 47.5Ω 470 47R5 4 27Ω 1W 270 not applicable 1 10Ω 100 10R0 4 0.1Ω 3W 0R1 not applicable No. Above: this view shows the bottom of the PCB with all the SMD parts installed, before any of the through-hole components have been fitted. It’s easier to mount these components before the larger parts are installed on the top side so that the PCB will still sit flat on the bench. anode and which is the cathode. You can solder the SMD LEDs in place using a similar procedure as before, ie, tack down one side then solder the other. Don’t get the different types mixed up. Note that LED2 and LED3 each have four terminals so avoid bridging the two at each end. If you do, use flux and solder wick to clean them up. Also make the joints quickly to avoid heating them for too long; these LEDs are quite small and can be damaged by heat. In particular, the plastic lenses of the SMD LEDs can be damaged if the iron is held on them for too long or if the air temperature is too high. Be careful if using either hot air or infrared reflow. Trimpot VR2 can go in next. Try to avoid getting solder on its metal adjustment plate. After that you can solder the top-side SMD resistors in place. Each will be printed with its value as a 3-digit or 4-digit code (see Table 1). The only resistors which are not fitted at this stage are one 220Ω 0.5W, two 27Ω 1W and four 0.1Ω 3W types. Next, install all the SMD ceramics, except for one 1µF capacitor which goes on the bottom of the board. Ferrite bead L1 can also be fitted now. siliconchip.com.au Value There are a total of eight passive SMD components that go on the underside of the board; see Fig.7. Fit these now, using the same method as before. Returning to the top side, the SMD electrolytic capacitors can now be mounted. These consist of a metal can on a plastic base with two flat leads and all but one have a black stripe on the top of the can to indicate the negative lead and a chamfered base on the side of the positive lead. Orientate each capacitor as shown in Fig.6 and use a similar procedure as for the ceramic types to solder them in place. The last top-side SMD parts to fit are the two fuse-holders. These are quite large parts with high thermal inertia as they are soldered to large copper conductors. A fair bit of heat will be required but the procedure is otherwise similar to the other components. Note that the plastic portion can be damaged by too much heat. Through-hole components Now you can fit the through-hole components, other than the large transistors, in the usual manner. It’s best to start with trimpot VR1, then follow with CON4 (if fitted), CON2, CON3 4-Digit Code (E96) 1004 and CON1. For CON2 and CON3, we recommend that you orientate them so that when the terminal blocks are plugged in, the wire entry is from the right-hand side of the board. The easiest way to do this is to temporarily plug the terminals in just before soldering, to check the orientation. Note that depending on your amplifier chassis layout, it may be possible to mount these the other way around and have the wires come in over the PCB itself. However, we haven’t tried this. Now you can install the optional through-hole capacitors, if you are using these, with the exception of the 1000µF types which we’ll leave for later. You will definitely need to fit the 47µF 63V electrolytic type if you have not already mounted its SMD equivalent, in the lower right-hand corner of the board. Similarly, if you are using a polypropylene capacitor for the output filter, rather than SMD NP0 ceramic, install it now. You may fit PC stakes to the test points if you want to. This does make adjustments slightly easier as you can clip alligator leads onto them. However, if you do so, you will need to be September 2015  93 Making A Winding Jig For The 2.2μ 2.2μH Inductor ➊ START ➌ Wind wire on bobbin clockwise The winding jig consists of an M5 x 70mm bolt, two M5 nuts, an M5 flat washer, a piece of scrap PCB material (40 x 50mm approx.) and a scrap piece of timber (140 x 45 x 20mm approx.) for the handle. In use, the flat washer goes against the head of the bolt, after which a collar is fitted over the bolt to take the bobbin. This collar should have careful to avoid accidentally shorting to adjacent components. The inductor goes in next but first you will need to wind it. Winding the inductor This is easiest to wind if you make up a winding jig as shown in the accompanying panel. You only need a few cheap and easy-to-obtain items and it will come in handy any time you need to wind a small air-core choke. The inductor is wound using a ~1m 94  Silicon Chip ➋ These photos show how the winding jig is used to make the 2.2m mH inductor. First, the bobbin is slipped over the collar on the bolt (1), then an end cheek is attached and the wire threaded through the exit slot (2). The handle is then attached and the coil tightly wound onto the bobbin using 13.5 turns of 1.25mm-diameter enamelled copper wire (3). The finished coil (4) is secured using one or two bands of heatshrink tubing around the outside. a width that’s slightly less than the width (height) of the bobbin and can be wound on using insulation tape. Wind on sufficient tape so that the bobbin fits snugly over this collar without being too tight. Next, drill a 5mm hole through the centre of the scrap PCB material, followed by a 1.5mm exit hole about 8mm away that will align with one of length of 1.25mm diameter enamelled copper wire on a 10mm wide, 13mm inner diameter plastic bobbin. Fit the bobbin to the jig, or if you don’t have a jig, wind some electrical tape around a bolt or dowel so that it is a firm fit through the centre of the bobbin, to prevent the plastic breaking while winding on the copper wire. For a neat result, the wire can first be straightened by fastening one end in a vice and pulling hard on the other end with a large pair of pliers. This requires ➍ the slots in the bobbin. The bobbin can be slipped over the collar, after which the scrap PCB “end cheek” is slipped over the bolt (ie, the bobbin is sandwiched into position between the washer and the scrap PCB). Align the bobbin so that one of its slots lines up with the exit hole in the end cheek, then install the first nut and secure it tightly. The handle can then be fitted by drilling a 5mm hole through one end, then slipping it over the bolt and installing the second nut. a fair bit of strength so be careful in case the pliers or vice let go. Make a right-angle bend in the wire 25mm from one end, then insert this end through one of the slots in the bobbin and wind on seven closely-packed turns, which should fill the width of the bobbin. Since the winding direction affects performance, we recommend that you wind in the same direction as we did, as shown in the photos. Once that layer is complete, wind another 6.5 turns on top, again closelysiliconchip.com.au Drilling & Tapping The Aluminium Heatsink CL (SCALE 50%) 50.75 50.75 30.5 A 30.5 A A A A 75 A 42 Tapping A 30 25 10.25 10.25 200 100 HOLES A: DRILL 3mm DIAMETER OR DRILL 2.5mm DIAMETER & TAP FOR M3 SCREW. DEBURR ALL HOLES. Fig.8: this half-size diagram shows the heatsink drilling details. The holes can either be drilled and tapped (using an M3 tap) or can be drilled to 3mm and the transistors mounted using machine screws, nuts & washers. Fig.8 above shows the heatsink drilling details. If tapping the holes, they should be drilled to 2.5mm diameter right through the heatsink plate and then tapped to 3mm. Alternatively, the holes can be drilled through using a 3mm drill and the transistors mounted using screws, nuts and washers. It’s somewhat more work to tap the holes but it makes mounting the transistors quite a bit easier (no nuts required) and gives a neater appearance. Before drilling the heatsink, you will have to carefully mark out the hole locations using a very sharp pencil. That done, use a small hand-drill fitted with a 1mm bit to start the location of each hole. This is important as it will allow you to accurately position the packed and in the same direction, then bend the wire through the opposite slot and cut it off 25mm from the bobbin. To hold the windings in place, cut a 10mm length of 20mm diameter heatshrink tubing and slip it over the bobbin, then shrink it down gently using a hot-air gun on a low setting. Trim the two protruding wires to exactly 20mm from the base of the bobbin then strip 5mm of the enamel from each end using either emery paper or a hobby knife/scalpel and tin the leads. To get the specified performance, you must mount the inductor as shown in Fig.6, Fig.9 and the photos. Two slots are provided for a cable tie to hold it in place. Bend its leads down through 90° to fit through the PCB pads, then fit and tighten the cable tie before soldering and trimming the siliconchip.com.au Don’t try drilling the holes in one go. When drilling aluminium, it’s important to regularly remove the bit from the hole and clear away the metal swarf. If you don’t do this, the aluminium swarf has a nasty habit of jamming the drill bit and breaking it. Re-lubricate the hole and the bit with oil each time before you resume drilling. holes (the locations are critical) before stepping up to larger drills in a drill press. Be sure to use a drill press to drill the holes (there’s no way you’ll get the holes perfectly perpendicular to the mounting face without one). Use a small pilot drill to begin with (eg, 1.5mm), then carefully step up the drill size to either 2.5mm or 3mm. The holes have to go between the fins so it’s vital to accurately position them. In addition, you can drill (and tap) three holes in the base of the heatsink so that it can later be bolted to a chassis. Be sure to use a suitable lubricant when drilling the holes. Kerosene is the recommended lubricant for aluminium but we found that light machine oil (eg, Singer or 3-in-1) also works well for jobs like this. leads. Note the way we’ve orientated it; each wire from the PCB runs up to and then under the coil former. Drilling & tapping the heatsink If you are upgrading an earlier version of the module, or if you are building this from a kit, you may already have a drilled and/or tapped heatsink. Otherwise, refer to the accompanying panel and the drilling diagram (Fig.8). Ideally, the seven transistor mounting holes should be tapped with an M3 thread. Take your time doing this since it’s quite easy to strip out a hole in aluminium, in which case you may have to start again with a fresh heatsink (or drill the hole right through, as described below). If you don’t want to tap the holes, you can drill all the way through the To tap the holes, you will need an M3 intermediate (or starting) tap (not a finishing tap). The trick here is to take it nice and slowly. Keep the lubricant up and regularly wind the tap out to clear the metal swarf from the hole. Re-lubricate the tap each time before resuming. Do not at any stage apply undue force to the tap. It’s easy to break a tap in half if you are heavy-handed and if the break occurs at or below the heatsink’s face, you can scratch both the tap and the heatsink (and about $25). Similarly, if you encounter any resistance when undoing the tap from the heatsink, gently rotate it back and forth and let it cut its way back out. In short, don’t force it. Having completed the tapping, deburr all holes using an oversize drill to remove any metal swarf from the mounting surface. The mounting surface must be perfectly smooth to prevent punch-through of the transistor insulating washers. Finally, the heatsink should be thoroughly scrubbed cleaned using water and detergent and allowed to dry. Fig.9: bend inductor L2’s leads and fit it to the PCB as shown here to ensure that you get the best performance. L2 2.2 mH September 2015  95 MAIN PLATE OF HEATSINK MAIN PLATE OF HEATSINK MAIN PLATE OF HEATSINK SILICONE INSULATING WASHER SILICONE INSULATING WASHER M3 FLAT WASHER INSULATING BUSH M3 x 10mm SCREW M3 FLAT WASHER M3 x 1 5 mm SCREW M3 x 10mm SCREW M3 TAPPED HOLE M3 TAPPED HOLE A AMPLIFIER PCB M3 TAPPED HOLE NJL3281D OR NJL1302D TRANSISTOR (TO-264) BD139 TRANSISTOR (TO-225) MJE15030 OR MJE15031 TRANSISTOR (TO-220) AMPLIFIER PCB (HEATSINK FINS) B AMPLIFIER PCB C Fig.10: this diagram shows the mounting details for the TO-220 driver transistors (A), the BD139 VBE multiplier (B) and the four output transistors (C). After mounting these transistors, use your multimeter (switched to a low ohms range) to confirm that they are properly isolated from the heatsink – see text for details. heatsink and use longer machine screws (fed between the fins) and nuts to secure the transistors. However, you must drill the holes with a high degree of accuracy, otherwise the screws may not fit between the fins. After you have drilled and tapped the transistor mounting holes, you will also want to do something about mounting it in the chassis. Our pre- ferred method is to drill and tap three additional holes along the bottom of the heatsink, as shown in the photo on the following page. However, it’s also possible to fit right-angle brackets to the fins at either end of the heatsink. That can be done by drilling right through the fins and using screws and nuts to hold the brackets in place. Once all holes have been drilled, de- Three M3 or M4 holes can be drilled and tapped in the base of the heatsink so that it can later be attached to a chassis. Make them about 10mm deep. 96  Silicon Chip burr them using an over-sized drill bit and clean off any aluminium particles or swarf. Check that the areas around the holes are perfectly smooth to avoid the possibility of puncturing any of the insulating washers. Fitting the heatsink Now it’s time to mate the PCB with the heatsink but first re-check the face of the heatsink. All holes must be deburred and it must be perfectly clean and free of any grit or metal swarf. Start the heatsink assembly by mount­ing Q7, Q8 & Q9. A silicone rubber washer goes between each of these transistors and the heatsink. Q7 and Q8 also require an insulating bush under each screw head. Fig.10 (A & B) shows the mounting arrangements. We specified a TO-126/TO-225 insulating washer for Q9 as it is smaller than the TO-220 packages for Q7 & Q8 but if you can’t get one of these, you can always cut a TO-220 washer down to size. Just make sure it’s still large enough to cover Q9’s exposed metal pad completely, taking into account any slop in the screw hole. Be careful not to get Q7 & Q8 mixed up as their type numbers are similar. If the holes are tapped, these transistors can be secured using M3 x 10mm machine screws. Alternatively, if you have drilled non-tapped holes, you will need to use M3 x 15mm machine siliconchip.com.au Scope 1: amplifier output for a 1kHz square-wave into a 4-ohm load. As you can see, there is a small amount of overshoot (around 5%) but it recovers quickly with very little ringing. Scope 2: the same test as Scope1 but with a 2μF capacitor across the load. This results in more overshoot (~20%) and some ringing but it’s well under control. This is a standard test for amplifier stability. Scope 3: here the amplifier is delivering a 1kHz sinewave into an 8-ohm load at around 150W, ie, well into clipping. As shown, negative recovery is quite clean. Positive recovery has a small step due to the high open loop gain but it resumes its normal slope after about 25μs with only a small amount of ringing. Scope 4: distortion residual at 100W into 8Ω at 10kHz. The distortion level is so low that a significant fraction is noise even at this power level and frequency as shown by the display persistence. The distortion mainly occurs around the negative-most part of the waveform, hence it is even less significant at lower power levels. screws, with the screws coming through from the heatsink side (ie, the screw heads go between the heatsink fins). Make sure the three transistors and their insulators are properly vertical, then do the screws all the way up but don’t tighten them yet; ie, you should still just be able to rotate the transistors in each direction. The next step is to fit an M3 x 9mm (or 10mm) tapped spacer to each corner mounting hole on the PCB. Secure these using M3 x 6mm machine screws. Once they’re on, sit the board down on the spacers and lower the heatsink so that the transistor leads pass through their corresponding PCB pads. Note that you’ll probably have to bend Q9’s leads away from the heatsink as shown in Fig.10. screw with a flat washer (or M3 x 20mm for untapped holes). That done, hang the insulating washer off the end of the screw and then loosely screw the assembly to the heatsink. The remaining three devices are then installed in exactly the same way but take care to fit the correct transistor type at each location. Once they’re in, push the board down so that all four spacers (and the heatsink) are in contact with the benchtop. This automatically adjusts the transistor lead lengths and ensures that the bottom of the PCB sits 9-10mm above the bottom edge of the heatsink. Now adjust the PCB assembly horizontally so that each side is 32.5mm in from its adjacent heatsink end. Once you are sure it is properly positioned, tighten all the transistor screws just enough so that they are held in place while keeping the insulating washers correctly aligned. The next step is to lightly solder the outside leads of Q10 & Q13 to their pads on the top of the board. The assembly is then turned upside down so that the heatsink transistor leads can be soldered. Before soldering the leads though, it’s important to prop the front edge of the board up so that the PCB is at right-angles to the heatsink. If you don’t do this, it will sag Installing the output transistors The four output transistors (Q10-Q13) can now be fitted. Two different types are used so be careful not to mix them up (check the layout diagram). As shown in Fig.10(C), these devices must also be insulated from the heatsink using silicone insulating washers. Start by fitting Q10. The procedure here is to first push its leads into the PCB mounting holes, then lean the device back and partially feed through an M3 x 15mm mounting siliconchip.com.au September 2015  97 Improving The Distortion & Stability Our first prototype of the UltraLD Mk.4 incorporated a number of changes which we expected would lower distortion compared to the previous version. For example, the improved magnetic cancellation of the new PCB layout, the non-inductive surface-mount emitter resistors and the greater open loop gain provided by the new transistors should have each provided benefits. So we were disappointed to find that the distortion levels were initially very similar to the Mk.3 version. Convinced that it should perform significantly better, we investigated what might be holding the performance back. We made a number of interesting and important discoveries during this process. One was that using different load resistors affected the distortion measurements significantly, especially at higher frequencies. The output inductor’s impedance rises with frequency and it forms a voltage divider with the load. With a purely resistive load, this will only cause a roll-off in the frequency response. But if the load has any non-linearities, it will create distortion across the load even if the signal from the amplifier is perfectly clean. We use the Dummy Load Box described in our August 1992 issue for testing amplifiers and were assuming it was linear on the basis that it had given good results to date. But when we fed a 14V RMS signal from the Audio Precision System Two’s ultra-low distortion generator into one end of the load box and connected a polypro- pylene capacitor from the other end to signal ground, forming a low-pass filter, we found this wasn’t the case. Doing this test with a resistor we thought would be very linear (a 5W wirewound type) gave 0.00025% THD+N at 10kHz with an 80kHz measurement bandwidth. However, using our load box as the resistor gave a higher reading of around 0.0008%, ie, three times higher. Thus it’s likely the load box itself was contributing to the higher distortion reading from the amplifier. To determine the cause, we soldered a couple of wires directly across the resistor banks in the load box and repeated the test. The reading dropped to 0.00025%. We therefore believe the problem is in either the connectors or the relay switching arrangement in the load box. So we had to continue testing using the soldered connections as this was the only way we could find to get a true reading of the amplifier’s performance (we will need to further investigate the source of the distortion in the load box at a later date). under its own weight and will remain in this condition after the leads have been soldered. A couple of cardboard cylinders cut to 63mm can be used as supports (eg, one at each corner adjacent to CON1 & CON3). With these in place, check that the board is correctly centred on the heatsink, then solder all 29 leads. Make sure the joints are good since some can carry many amps at full power. Once the soldering is completed, trim the leads using a steel rule as a straight edge to ensure consistent lead lengths. That done, turn the board right way up again and tighten the transistor mounting screws to ensure good thermal coupling between the devices and the heatsink. Don’t over-tighten the mounting screws though. Remember that the heatsink is made from aluminium, so you could strip the threads if you are too ham-fisted. 98  Silicon Chip Tweaking the output filter We then measured the amplifier at around 0.0015% THD+N at 10kHz, a slight improvement on the Ultra-LD Mk.3 module under the same conditions (at around 0.002%). But we felt the new module should be more of an improvement than this and subsequently discovered that if we measured the distortion before the output filter, it was dramatically lower, Checking device isolation You must now check that the transistors are all electrically isolated from the heatsink. That’s done by switching your multimeter to a high ohms range and checking for shorts between the at around 0.0008% <at> 10kHz. Since the filter was still in-circuit and the load current was still flowing through inductor L2, this meant it wasn’t due to any interaction between the output filter and the front end. So it had to come down to the output filter itself; either the capacitor or SMD resistors were not linear enough or there was something odd happening to the signal in the inductor. We then separately tested a number of different resistors and capacitors, using a similar method as before, ie, hooking them up as RC filters and using the Audio Precision gear to test the performance. This gave the SMD resistors a clean bill of health as the four in parallel performed just as well as a 6.8Ω wirewound resistor. But the X2 polypropylene capacitor we were using on that prototype gave distortion of around 0.0006% in this test. We tested three other polypropylene capacitors, two other X2 types and an MKP. The MKP and one of the X2 capacitors got a clean bill of health (ie, reading around 0.00025%) while one of the other X2s also gave higher than expected distortion. We therefore put the better capacitor on the board but this only made a tiny improvement to its performance. Having essentially ruled out the capacitor and resistor as being the problem, suspicion fell on the inductor. But was it also possible that the connection routing on our PCB was not 100% correct, especially in the earth tracks? To rule this out, we removed the RLC filter from the PCB and heatsink mounting surface and the collectors of the heatsink transistors (note: the collector of each device is connected to its metal face or tab). For transistors Q7-Q8 and Q10Q13, it’s simply a matter of checking between each of the fuse-clips closest to the heatsink and the heatsink itself (ie, on each side of the amplifier). That’s because the device collectors in each half of the output stage are connected together and run to their respective fuses. Transistor Q9 (the VBE multiplier) is different. In this case, you have to check for shorts between its centre siliconchip.com.au mounted it entirely off-board, between the output connector and test load. This completely solved the problem, giving the excellent performance indicated in the Audio Precision plots last month. But why? We moved the inductor and resistors back onto the PCB but left the connections the same and the measured distortion doubled again. This pretty much ruled out a routing problem. So we mounted the inductor on short lengths of flexible wire and experimented with changing its position and orientation. Both the position and orientation of the inductor affected performance, however the mounting location mattered a lot less with the inductor rotated to rest on its side. Presumably this is due to its magnetic field affecting a plane orthogonal to the tracks on the PCB. And this is how we ended up with the final mounting arrangement. The only reason we can figure that this matters is that high-current pulses in the PCB power supply tracks were being picked up by the inductor and injecting a distortion signal into the load. This effect is greater at higher frequencies because the inductor’s higher impedance with these signals more effectively isolates the loudspeaker output from the low-impedance junction of the output transistor emitter resistors. By the way, we’re fairly sure that this amplifier has lower distortion than the 20W Class-A amplifier published in the May-September 2007 issues. The main advantage of a Class-A amplifier compared to a Class-B or Class-AB is that it doesn’t have any crossover distortion since all the output transistors are conducting all the time. Well, if this new Ultra-LD design has any crossover distortion, we certainly can’t detect it! In fact, if you make direct comparisons between the distortion curves in the July 2011 issue with those published last month (August 2015) you will see that the Ultra-LD Mk4 is a dramatic improvement on the previous design. Is there likely to be an audible difference? We think that is highly unlikely! We suspect what little distortion remains is mostly due to non-linearities in the front end – which a Class-A amplifier would suffer from equally. The bottom line is, there isn’t really any point in building a Class-A amplifier any more. You might as well build this one and get much more power, higher efficiency and less heat dissipation. Stability improvements While tweaking the amplifier’s performance, we changed some components which compromised stability and occasionally triggered oscillation, although no damage occurred as a result. This did, however, allow us to discover some ways to improve overall stability. This happened almost by accident. What we found was that when the amplifier was in an unstable condition and started to oscillate, touching certain components on the PCB would cause the oscillation to temporarily cease. We isolated this effect to two specific components: Q4’s collector resistor and the 2.2kΩ resistor from the junction of the two 150pF capacitors to the negative rail (part of the compensation network around Q4/Q6). We figured that connecting capacitors Thin Film SMD Resistor Values You may have noticed in the parts list published last month that we specified some odd value resistors. For example, 6.49kΩ, 332Ω, 47.5Ω and so on. As we explained then, many of the resistors in the circuit must be thin film types for good performance (many SMD resistors have thick film construction which is not suitable). The best SMD thin film resistors we found are made by a company called Stackpole Electronics. Besides being thin film, they also have a relatively siliconchip.com.au high wattage rating for their size (3.2 x 1.6mm) of 0.5W. However, this series of resistors (RNCP1206FTD) comes in the E96 series of values rather than the E24 series we are used to. The E24 series is as follows: 10, 11, 12, 13, 15, 16, 18, 20, 22, 24, 27, 30, 33, 36, 39, 43, 47, 51, 56, 62, 68, 75, 82, 91. It then repeats scaled up (or down) by a factor of 10. In other words, within every 10:1 ratio of resistances, there are 24 values to choose from and each value is roughly 10% higher across these resistors would have a similar effect on stability to touching them with a finger and proved this by modifying the prototype in this manner and curing the oscillation entirely. We explained the reason for the first of these two improvements in Pt.1: it eliminates the Early effect on Q4 which causes a form of local feedback. This change alone appears to make the amplifier much more tolerant and allows reduced compensation without prejudicing high-frequency stability. The advantage of the 15pF capacitor in the compensation network is less clear. Simulation suggests that it slightly reduces the phase shift around the VAS at very high frequencies while having a negligible effect on gain. But the combined effect of these two changes appears to be that if the amplifier does “misbehave”, it’s far less likely to go into damaging highfrequency oscillation. By the way, we tested all of the circuit changes in SPICE simulations to check that they were sensible but ultimately had to try them all on the prototype to verify their effect on performance and stability. Simulation is a good way of quickly finding out if a change is a bad idea without blowing the amplifier up, but when simulation shows that something should work, it’s far from certain that it actually will. One area in which simulation excels is the ability to see what’s going on in the circuit. For example, you can easily display the current passing through any component in the circuit whereas doing this on the real prototype would involve de-soldering the component and inserting a shunt which might upset the circuit’s operation. than the next one down. As you may have guessed, the E96 series has 96 different values for each decade. But while the E24 series contains all the values of the E12 series and simply adds new values in-between, the E96 series does not contain all the E24 series values. So the RNCP1206FTD series of resistors does not offer 6.2kΩ, 330Ω or 47Ω. In practice this does not matter as we simply picked close values; this circuit will tolerate values a few percent higher or lower, as long as all resistors of the same nominal value are closely matched. September 2015  99 This power supply board can run two Ultra-LD Mk.4 amplifier modules and will be described in Pt.3 next month. (collector) lead and the heatsink. In either case, you should get an open-circuit reading. If you do find a short, undo each transistor mounting screw in turn until the short disappears. It’s then simply a matter of locating the cause of the problem and remounting the offending transistor. Be sure to replace the insulating washer if it has been damaged in any way (eg, punched through). Completing the assembly The PCB assembly can now be com- This view shows the mounting details for the VBE multiplier transistor (Q9) and the two driver transistors (Q7 & Q8). Check that these transistors and the four output transistors (Q10-Q13) are all isolated from the heatsink 100  Silicon Chip pleted by installing the two 1000µF 63V capacitors – assuming you have decided to fit these. As stated last month, they can be left out as long as the power supply leads are kept short and made from thick wire. Otherwise the maximum output power will drop a little bit, due to losses in these cables, but performance should not be affected. One of the changes we’ve made in designing this PCB was to place these capacitors so they don’t interfere with access to the heatsink mounting screws to the same extent as they did on the Mk.2 and Mk.3 versions. However, working on the PCB is still easier if the large capacitors are not fitted and due to their proximity to the heatsink, they will probably dry out eventually (albeit probably after more than 10,000 hours of use, assuming they are goodquality capacitors). Now remove the two support spac­ers from the edge of the board adjacent to the heatsink. In fact, it’s quite important that the rear edge of the board be supported only by the heatsink transistor leads. This avoids the risk of eventually cracking the PCB tracks and pads around the heatsink transistors due to thermal expansion and contraction of their leads as they heat up and cool down. In short, the rear spacers are in- stalled only while you fit the heatsink transistors and must then be removed. They play no part in securing the module. Instead, this edge of the module is secured by bolting the heatsink itself to the chassis. As previously stated, this can be done by tapping M3 (or M4) holes into the main plate on the underside of the heatsink or by using right-angle brackets. The front of the board is secured using the two M3 x 9mm (or 10mm) spacers fitted earlier. Power supply & speaker protection modules That completes the assembly of the power amplifier module. The next step is to build the power supply module (shown in the above photo) and we’ll describe how that’s done next month. We’ll also explain how to power up and test the amplifier and give some basic details on housing it in a metal case. Finally, we’ll present the revised speaker protector module, which can also monitor heatsink temperature. You will need it (or our previous design) to prevent an amplifier fault from destroying the speakers and poSC tentially causing a fire. siliconchip.com.au