Fullscreen HTML5Med Res (??MB)HTML4

AUDIO OUT AUDIO OUT L R By Jake Rothman Low-noise Theremin Power Supply – Part 2 L ast month, we discussed the importance of a low-noise power supply for sensitive audio applications, such as a Theremin. This month, we will move on from theory and actually build one. In fact, we will offer not just one, but four circuits based around the same flexible PCB, supplying different voltage outputs. Basic design For linear power supplies we have to factor in all the losses, as shown in Fig.15:  Headroom required by the regulators, (the minimum difference between the input and output voltages is around 2V for the LM317 used here).  Regulation figure or the voltage droop of the transformer at the expected load current.  Total forward voltage drop (Vf) of the rectifier diodes (1.4V for silicon or 0.5V for Schottky devices).  Minimum input voltage at the bottom of the ripple waveform. This can be minimised by fitting the biggest smoothing capacitors (C10/13) physically possible in a small PSU, typically 1000 to 2200µF. Since it is a bridge-rectifier/capacitor circuit, the output voltage is theoretically equal to the peak AC (×1.4 the RMS input voltage), which gives us a useful boost. However, there is no such thing as free energy, this comes from a corresponding rise in the current drawn. All these factors limit the real-world current that can be drawn to around 120mA if the regulated voltages are set to around 9V and 15V. At least we won’t need any heatsinks on the regulators at this level. This circuit is thus ideal for powering the Theremin or other sensitive circuits and we also have an isolated 15V supply for the audio amplifier, which avoids interaction between the two sections. PCB layout C13 D3 Voltag e losses Peak vo ltag e at smoothing capacitor: 15.5V 1V mains f luctuation Transf ormer reg ulation (va ries w ith load current) 1.4V rectif ier drop(s) 1V ripple (va ries w ith load current) C10 Fig.16. PCB tracks from the rectifier to the smoothing capacitors. These have nasty pulses on them due to charging currents. Nothing else must be connected or 100Hz noise will result. Note the following on the above PCB: the track jumps from D3 to C10 on the other side of the board. (Red is the positive (+) line and the black dotted is the negative line (–) Getting low hum on the output is dependant on the PCB track layout. It is very important to ensure the charging pulses from the rectifiers to the smoothing capacitors do not get superimposed on the output. This can easily be achieved by making sure the two lines from the rectifier go directly to the smoothing capacitor pins with no other connections along these H eadroom vo ltag e 15.5V D8 tracks. The lines to the regulators then have to be taken directly from the capacitor pins. This practice is called ‘noding’ or ‘starring’. The two paths on the PCB are shown in Fig.16. There is no provision for the Zobel network on the PCB, since the components are rather large and the Theremin works fine without it. If you do want one, it will have to be hard-wired to the board, possibly using a small board as shown in Fig.31. Note that only one Zobel network will be needed for dual rails, since the two windings are well-coupled magnetically via the transformer. Regulators Fig.15. Taking account of the cumulative losses in linear power supply design. The low noise of the LM317 voltage regulator has been recognised for years by the audio community. In the 1980s it was expensive, now it is a cheap commodity component. It has lower noise because it uses a band-gap voltage reference rather than a Zener diode. The reference can also be decoupled with an external capacitor, which knocks the noise down by a further factor of 10. A classic example of a band-gap reference is the difference between silicon and germanium diode junctions being set to 0.43V by adjusting the currents flowing through them. The 0.43V is the point of maximum temperature stability. In the case of the LM317, the band-gap is made from 60 Practical Electronics | September | 2020 Total loss is approx 6.5V f or max load 0.5V Schottky drop 2V reg ulator headroom 9V Req uired reg ulator output vo ltag e: 9V Time Vout = 1.25 × (1 + R2/ R1) Input Vin O utput LM317 Adj Vout 1.25V R1 Adjust pin R2 (Radjust) n Vout –1.25V 0V LM317 f ront vi ew LM317 Input Adj O utput Fig.17. Basic LM317 voltage regulator circuit. A relatively low-resistance voltagesetting divider reduces variations due to variations in the regulators quiescent current that flows out of the adjust pin. two dissimilar silicon junctions giving a reference of 1.25V using one of the late Bob Widlar’s unique circuits. (He’s worth looking up on Wikipedia). By scaling this voltage using a resistor network, any desired voltage from 1.2 to around 32V can be obtained, as shown in the basic circuit in Fig.17. The equation is basically the gain of a non-inverting op-amp multiplied by the 1.25V voltage reference: Vout = 1.25[1+(R2/R1)] R1 is usually 120Ω to 240Ω; note that 120Ω uses more current, but gives a more predictable voltage. Secondary 1 Primary L 11V 29V 230-240V 240mA max continuous 18V N Secondary 2 11V 240V Link 29V 18V Top vi ew Fig.18. The transformer secondaries can be connected in series if needed to get a single-rail higher voltage supply up to 30V. Practical Electronics | September | 2020 Fig.19. A suitable internal fuse holder – this is insulated so you can’t get a shock off it. It must be placed in the live wire. If you want an on-line calculator then https://circuitdigest.com/calculators/ lm317-resistor-voltage-calculator is good. Note the overall tolerance of the set voltage of the LM317 is about 10%, so a bit of resistor tweaking may be needed. Make the calculated R2 value a bit bigger than required, and then put resistors in parallel with R2 to tweak its output down to create the required voltage. Here’s a list of some typical voltages and the associated value of R2 (R1 = 120Ω): 5V 6V 7.5V 9V 12V 15V 18V 22V 24V 25V 28V 30V 360Ω 453Ω E96 600Ω 750Ω 1kΩ 1.3kΩ 1.6kΩ 2kΩ 2.2kΩ 2.32kΩ E96 2.55kΩ E96 2.7kΩ Ω fuse in the mains plug. This can be wired as shown in Fig.19 or you can use a panel-mounted holder. Thermal A 127°C or 105°C non-resetting thermal cutout is placed in series with the primary to guard against the transformer overheating. It’s physical placement against the earthed laminations is shown in Fig.20. If it trips it has to be replaced. Over-voltage The absolute maximum input voltage to the LM317 is 37V. If the secondaries are connected in series (Fig.24), the off-load voltage can breach this so replace the standard protection diodes with Zeners. Components Ω Some values are from the E96 series. These are inexpensive now, about 3p from Mouser; eg, Part no. MFR-25FBF52-2K32. Above 15V, the transformer windings will have to be put in series (see Fig.18) unless the current drawn is very low. (With Schottky diodes and around 30mA load it is possible to get 18V.) The devices are not hard to source – see the transformer notes below and in Part 1. Explain versions. This list is for the +9V/+15V version. there are some changes for the +30V version marked with an asterisk, ‘*’. Resistors R1 1kΩ 5% 0.5W R3, R5 120Ω 1% 0.25W Protection Over-current The LM317 has a short-circuit current limit of 1.5A, assuming a massive heat sink. This is so high that it’s useless, something else usually blows up first. Here we will have to use a fuse in series with the primary. A low value (250mA) time-delay is about right. Don’t just rely on the 13A Fig.20. Location of the thermal cut-out; vital if you use a plastic or wooden case. 61 R2, R4 See above table for required values (use 0.25W, 1%) 33kΩ 2.2Ω, 0.25W, 5% (optional Zobel) R2a* Rz Fig.21. The output cable of DC power supplies is often the polarised figure-ofeight type, typically cheap speaker cable. Note the rubber grommet. Capacitors C1 10nF ceramic, 5mm X7R C2-C9 100nF ceramic, 5mm X7R C10 2200µF 25V (*1000µF, 50V) C11,12 100µF 25V (*35V) C13 2200µF 25V (*1000µF, 50V) C14,15 100µF 25V Cz 4.7µF, 63V any non-polarised type (optional Zobel) Mains input D9 D10 C2 L Thermal cutout D1 C3 D2 C4 T1 E C6 17V AC R 1 D3 C5 Secondary Primary N C7 C1 C8 C9 + IC1 + C10 D4 D5 A D6 C13 + Pad 5 +9V C13 Pad 4 15V C12 + Pad 6 R3 R2 B 25V R4 R5 + D7 C15 D8 IC2 0V Pad 7 C Pad 8 +15V Pad 9 C14 Pad 10 + 0V Pad 11 D12 D11 *For Theremin link earth pads B-B Fig.23. Standard low-noise Theremin +9/+15V power supply for the circuit in Fig.22. Note earth link B is linked. Note: the electrolytic capacitors determine the life expectancy of linear power supplies, so use long-life types (> 2000 hours at 105°C) such as those by Nichicon, Kemet and Panasonic. They don’t need to be low equivalent series resistance (ESR) types (see capacitor ‘tuning’ later). All electrolytics used are radial. Semiconductors IC1,2 LM317T positive adjustable voltage regulator D1-D8 SB40 1A 40V Schottky rectifiers (used where low voltage drop is required. If 1.4V can be accommodated then use 1N4002. For lower noise use UF4002) D9-D12 1N4001 (*22V BZY61, 1.3W zener) Miscellaneous  Fuse holder, Camden Boss single fused terminal block (Rapid 21-0740 with T250mA 20mm fuse)  IEC connector (Rapid Bulgin 51-4336 or Chinese 23-0100)  Rubber boot (Cliff Rapid 23-0350)  M3 × 40mm fixing screws (Rapid 513341)  Die-cast box, 115 × 90 × 55mm Eddystone. Hammond 120 × 100 × 60mm Rapid 30-7070.  Thermal cut-out SetFuse X1 (Rapid 260940 (127°C) or 26-0900 (105°C))  TO220 clip-on heatsink  Rubber grommet for enclosure cables exit hole (Rapid 04-0202) D9 1N4002 Vin C2 Thermal cut out 127°C T1 L – C3 D1 D2 D3 D4 Adj + +15V + 11V C4 C5 Mains input 230-240V IC1 LM317 C10 2200µF 25V Pad 4 Vout +9V to Theremin R3 120Ω 0.25W R2 0Ω D10 1N4002 + + C11 100µF 16V N D11 1N4002 D1-8: U F4002 C2-9: 100nF, X7R Vin E C6 RZ 2 2Ω R1 1kΩ 0.5W C1 10nF B CZ 4.7µF – C7 D5 D6 D7 D8 Adj + +25V + C8 Z ob el netw ork (optional) C9 IC2 LM317 C13 2200µF 25V (min) C12 100µF 16V Pad 6 0V B 18V Pad 5 Pad 7 Pad 8 Vout +15V to amplif ier R5 120Ω 0.25W R4 1 kΩ D12 1N4002 + + C14 100µF 25V Pad 9 C15 100µF 25V Pad 10 0V Pad 11 BB link f or Theremin Fig.22. Full circuit diagram of the power supply with the Zobel add-ons in red. +9V and + 15 dual power supply for Theremin. By changing the links and values of R2 and R4 to 1kΩ this can be made into a ±12V supply for op amp circuits (see Fig.27). 62 Practical Electronics | September | 2020 20V Tant Z 18V N 2 2Ω T1 Pad 6 –9V * C12 can also b e a 470µF polym er 0 Ω CZ 4.70µF 100V Film Pad 7 D9 BZ Y 61 22V D1,2,7,8: 1N4002 2 100 C2 Thermal cut out 127°C T1 L +39V C3 – D1 D2 D7 D8 + + 11V C8 C9 Mains input 230-240V B Link secondaries 18V in series N C1 10nF Earth link B E 1 1kΩ 0.5W Mains input C2 L Thermal cutout C3 Secondary Primary N E T1 C1 + IC1 LM317 Adj C10 1000µF 50V C8 +30V D10 BZ Y 61 22V + + 2 kΩ 2 kΩ C11 100µF 35V D7 D8 C Pad 8 Pad 9 Pad 10 Pad 11 Fig.25. Low-noise, single-rail +30V power supply. Very handy for single-rail preamplifiers using discrete circuitry. Note the three links to the left of C10/13. Transformer and PCB These can be obtained from the PE PCB Service. Other parts are available from the author. jacob.rothman<at>hotmail.com 01597 829102 Variations Power supply connectors The mains input socket is pretty much standard, thanks to the IEC (International Electrical Commission). However, for the output connectors, prepare to enter a standardisation minefield, it’s complete chaos. You can use screw terminals, 4-pin male XLR sockets, terminal blocks or one of the huge range of so-called DC connectors with an outer and centre pin. One so-called ‘standard’ in the audio industry for guitar pedals is a 2.1mm DC connector wired centre pin negative. Another ‘standard’ is polarised the cable. Normally, figure-of-eight thin speaker cable is used with the rib or printed line denoting positive. This can be seen coming out of the protective rubber grommet in Fig.21. 1) +9V and +15V dual-rail – this is the ‘standard dual-rail power supply shown in Fig.22 It is designed to power a Theremin and its associated amplifier driving the speaker. If the amplifier is not used, then the 15V power supply can be omitted and the circuit simply becomes a low-noise +9V supply. 2) Single-rail +30V – higher voltage design for discrete pre-amps and single rail op amp circuitry, as shown in Fig.24 3) ±12V general-purpose dual-rail –ideal for low-power op amp applications such as pre-amps and musical synthesisers. Shown in Fig.27. 4) Positive-earth –9V – designed for germanium circuits, such as Tonebender fuzz pedals and old 1960s radios, typically Practical Electronics | September | 2020 Pad 5 C12 47µF 35V Links on the PCB On the PCB overlay there are three link positions A, B and 0V C, as shown in Fig.23 These are 2 2 2 k Ω Pad 7 needed to connect the power supply DC ground or 0V to the ground lifted (by R1 and C1) mains earth connection. Link Fig.24. Single 30V rail designs for discrete A connects the 0V of the 17V pre-amplifier circuits. Note that the AC secondary winding power secondaries are in series and it’s worth supply rail to mains earth. Link putting a heatsink on the regulator. B connects the mains earth to the 11V-AC-fed power supply rail 0V line. Link C connects the positive of the 17V fed supply to the 0V of the 11V fed D9 supply. This is used to make centre-tapped D10 Pad 5 dual rail supplies. 30V + D1 IC1 Note that when using the Theremin C13 Small clip-on Pad 4 +9/+15V power supply the mains earth D2 heatsink C12 + is connected to the 0V of the 9V supply. Pad 6 + R3 0V C10 This is because the Theremin needs a stable R2 earth reference – more so than the amplifier. Pad 7 R2a A C13 + C7 Pad 4 Vout 120Ω 0.25W B 17V AC R 1 C13 1000µF 50V Vin those made by Roberts and Bush. This is a nice little extra I thought up after the Theremin supply was designed, built and finished. We haven’t got space to cover it this month, so it will be described in the next column. There are four variants of the power supply using the same board. This was originally designed for the PE Theremin, but with simple variations can be used to power a multitude of audio circuits. Pad 6 Capacitor ‘tuning’ Since voltage regulators are a negative-feedback control system with a falling open-loop gain at high frequencies, the reduction in output impedance becomes less effective at high frequencies. This means the output impedance of linear voltage regulators is slightly inductive. This can cause a resonant noise peak with the output capacitor (C12) if it is a low ESR (around 0.1Ω) type. National Semiconductor application engineer Robert Pease showed in his book Troubleshooting Analogue Circuits, that a bit of loss in the output capacitor was necessary. An ESR of 0.6Ω to 3Ω was needed to avoid noise peaks. (To see the noise Fig.26. Basic board before final voltage setting resistors R2 and R4 are added. Note the blue snubber capacitors across the rectifiers. 63 Fig.27. 12-0-12V low-noise supply uses the 18V winding on the transformer for the positive rail. This is because most synthesisers use +15V more current on the plus rail. C2 Thermal cut out 127°C T1 L – D9 1N4001 Vin C3 Adj + 11V C4 C5 Mains input 230-240V 18V N C6 – R1 1kΩ 0.5W C1 10nF A +25V Vin C7 D5 D6 D7 D8 C2 Thermal cutout C3 C4 Secondary Primary N T1 17V AC R 1 C11 100µF 16V R5 120Ω 0.25W C13 2200µF 25V (min) D12 1N4001 + + R4 1kΩ C14 100µF 16V A *SB40 E + Vout IC2 LM317 + C9 Mains input L + R2 1kΩ Adj + C8 Mains earth C10 2200µF 25V D10 1N4001 D11 1N4001 D1-4: SB40 D5-8: 1N4001 C2-9: 100nF, X7R E R3 120Ω 0.25W + D1 D2 D3 D4 Vout IC1 LM317 C1 C5 C6 C7 C8 C9 D9 D10 D1* Pad 5 0V C13 Pad 4 15V C12 + Pad 6 R3 Link R2 –12V D3* + C10 D4* A D5 D6 + IC1 D2* C13 + B 25V R4 Pad 7 C D12 IC2 9V To scope Load resistor 0Ω 2 2kΩ Pad 8 R5 + Pad 9 D7 C15 Small clip-on C14 Pad 10 heatsink + D8 dioxide (MnO2) solid-aluminium (SAL) 123 series capacitor (ESR typically 1.3Ω) but these stopped being made in 2015. Pad 4 This is shown in Fig.29. National Semiconductor used to Pad 5 recommend tantalum beads, but they don’t like surges due C12 100µF to shorting during testing and 16V powering up. In military work, special metal-cased surge-rePad 6 sistant tantalums are used, –12V but they cost serious money. 0V Pad 7 The new organic polymer SAL types have too low an ESR. The Pad 8 best noise results I’ve had so +12V far have been with a Nichicon Pad 9 470µF 16V RNE1C471MDN C15 (ESR = 0.01Ω) in series with 100µF 0.56Ω resistor, this giving a 16V more accurately defined total ESR which can be a variable Pad 10 capacitor parameter. 0V The adjust pin filtering caPad 11 pacitor (C11/14) can have a quite high ESR before it becomes ineffective, up to around 20Ω. I use a tantalum capacitor with a 10Ω series resistor to protect from surges. A value of 4.7 BC337 +12V Pad 11 D11 0V Sig nal g enerator 2V pk- pk, 100H z Sq uare w ave O utput 0V Connect across output of pow er supply Fig.28. Switched load for power supply transient-response testing. Fig.28. Low-noise, dual-rail 12-0-12V power supply. A useful general-purpose supply – handy for op amp circuits. Note the link from Pad 4 to Pad 11, and the A link. effect more clearly, disconnect the adjust pin decoupler capacitor C11). A link to the original 1989 article by E Dietz is here: https://bit.ly/pe-aug20-dietz Making C12 a large value, say 470µF, will swamp this noise problem, but then another problem emerges in that a lower frequency ringing of a few cycles (around 80mV pk-pk) can occur if the regulator load is switched on and off. This is because too big a capacitor across the output reduces the open-loop gain of the whole system, causing ringing as the negative-feedback loop seeks to correct its self in a sluggish manner, then overshooting. This effect is proportional to load current. Again, a damping resistance in series with the capacitor can reduce this problem. 64 Switched-load testing The way to test for poor dynamic load response is to put a ‘scope probe set to AC on the output and dab a load on and off. A better way, is to electronically switch the load on and off with a transistor, such as the circuit shown in Fig.28. By driving the transistor with a square wave from a signal generator at around 100Hz the sweep of the ‘scope can be synchronised to get a good view. By using these techniques I was able to get the noise down to 0.8mV peak-to peak. An output capacitor (C12/15) of around 47µF to 100µF with an ESR of around 1Ω works, and is cost-effective for both problems. A cheap wet electrolytic could be used, but this will have a short life. A good solution is a manganese Fig.29. The blue capacitor is a solidaluminium manganese dioxide (MnO2) Philips output capacitor (C12). This has a relatively high and consistent ESR, which gives a stable response in response to changing loads. The silver-coloured metalcased tantalum capacitor has a series resistor to increase reliability. (Note these photos are of the positive earth –9V PCB, which is covered next month). Practical Electronics | September | 2020 enough mechanically; it needs to be bolted down. I’ve had pieces of equipment where the transformer has sheared off the PCB when dropped. It’s not good having a lump of steel loose inside the box! Remember to employ the standard electronic construction technique of soldering after bolting, to the prevent the soldered joints being strained. A suggested mounting arrangement is shown in Fig.1, Part 1, where 3mm bolts are used. The core is also earthed by the bolts, essential to ensure safe operation of the thermal cut-out. Incidentally, it’s a good idea to use a smear of heatsink compound to aid thermal coupling between the thermal cut-out and the transformer core. This should be pressed firmly against the core prior to soldering. (see Fig.20). Remember to solder it quickly, otherwise it may overheat and go open-circuit! Testing Fig.30. Stages in cutting the hole for mains IEC socket in die-cast box (top to bottom): a) Mark out area to be removed and cut down the two vertical sides with hack saw. b) Grip ‘tab’ firmly with pliers c) Snap off ‘tab’ with pliers. d) Position the socket to mark out the screw holes. to 22µF works fine – see Fig.29. A polymer SAL capacitor works effectively as well. A diode (D10,D12) is used to protect the regulator adjust pin from discharge on turn off or output shorting. In the current-limited tantalum circuit, a 1N4148 is sufficient. If no resistor is used, a 1N4001 is needed, as previously specified. Construction Although it’s tempting just to solder the transformer to the PCB, that’s not strong Practical Electronics | September | 2020 The first thing to do is to make it safe to turn on. There is mains voltage on the board; if you are not scared of it, you should be. I always tape over exposed tracks and insulate terminals with silicon rubber sleeves. I also put a rubber boot over the mains input connector. Check there is earthing continuity from the earth pin on the plug to metal box. If there is no continuity you could be electrocuted. When turning on, listen for any buzzing noises from a stressed transformer or odd cracks and hisses from incorrectly inserted electrolytic capacitors swelling up. I know it sounds paranoid, but I’ve been on the receiving end of exploding electrolytic capacitors from inexperienced students, so I insist on them wearing glasses. If it passes this test, switch off and feel for any hot components. If okay, switch on again and measure the output voltage. Next, put on a suitable load resistor to give the maximum current draw required. This will normally be a big high-wattage type and will get hot. For example, if you are making a 12V power supply and need a 100mA output, then use a 120Ω 2W resistor. Finally, while loaded, check components for excessive heating. Note, it may take about 30 minutes for the transformer to reach its full running temperature. At this point it’s worth checking the electrical noise level. On the case Electronics belongs in metal boxes. They provide electrical safety when earthed, electromagnetic fields are screened, heat can get out without ventilation holes, they can’t catch fire and they don’t fall apart after two years. After a good 30 years life they can then be effectively recycled, like all metals. Steel is more effective than aluminium alloys at reducing magnetic fields, but it is harder to drill and it can buzz mechanically if the transformer is too close to any of the panels. Fig.31. Interior of power supply – the small board on the side is the Zobel – the ‘wacky circuit’ from Part 1. Fig.32. A view of the 30V board; note the links on the diode section and the regulator heatsink. Socket mounting The round DC connectors are relatively easy to mount with a 12.5mm (half-inch) hole. It should really have an anti-rotation flat on it but this needs a £400 metal punch set-up. It can be quite difficult cutting a rectangular hole in a metal box for the IEC socket. However, here’s a little technician’s trick I’ve used over the years to make it easier. It works very quickly with die-cast boxes, which although they are expensive, have always been the best choice for hand-built electronics. The stages involved are shown in Fig.30. The interior of the case is shown in Fig.31. The 30V board is shown in Fig.32. I have loads of these power supplies in my studio and they are not humming along nicely like the wall-warts used to. Next month In the next issue we will finish this power supply project with a handy positive-earth –9V variant. 65