Fullscreen HTML5Med Res (??MB)HTML4

A dirt cheap, high-current bench supply Got an old PC gathering dust somewhere? It mightn’t be much good these days but its power supply could be . . . especially if you want a high-current 13.5V bench supply! This article tells you how to modify one – at very little cost! By COL HODGSON, VK2ZCO This is NOT a projec t for the inexperien ced. DO NOT even think of opening the case of a PC switchmode po wer supply (SMPS) unless you have expe rience with the desig n or servicing of such devices or related high-voltage equipme nt. Note that much of the SMPS circuitry operates at full mains potentia l and contact with it could easily kill you. NEVER open up an SMPS case when it is connected to the mains, even if turne d off. Beware of any residu al charge on the ma ins capacitors, even if tur ned off for some tim e. DO NOT attempt to modify a SMPS unles s you are fully compete nt and confident to do so. T One of the nice things about using an old PC power supply is that it already comes in its own high-quality case, complete with fan. Some even include the mains switch – though none will have the large binding post terminals! This is a typical XT/AT-type supply, rated at about 230 watts (and therefore capable of 13.5V <at> 17A). You often find these PCs junked in council clean-ups, etc. www.siliconchip.com.au HE CONCEPT of converting a disused computer power supply to 13.5V operation was first mooted in the November & December 1998 issues of the no-longer-published “Radio and Communications” magazine. This article builds on that information. The process is relatively straightforward and involves removing all the components involved with the existing 5V and 12V outputs, rewinding the main transformer and then changing the feedback components to give an output of 13.5V instead of 5V. First, a few words on selecting the power supply to be modified. It must be an XT/AT type. It must NOT be an ATX type since they work quite differNovember 2003  25 Fig.1: inside a typical XT/AT-type switch-mode power supply (SMPS). Newer (ATX) supplies work a little differently. ently to the XT/AT types. Then, once you have a supply, check that it will maintain a constant output voltage under load; eg, one or two 12V 50W halogen lamps. Some SMPS may fail this test if the initial surge current drawn by the test load is too great (due to the overcurrent protection circuit being activated). In that case, switch off to allow the circuit to reset and retest it again, starting with lower wattage globes and increasing the load in steps. Reverse engineering and conversion to a new output is difficult at the best of times and nigh on impossible if the thing doesn’t work in the first place! Second, consider your power requirements. If you only need about 10A at 13.5V, you probably don’t need to change the main transformer as the original +12V outout can be modified to deliver +13.5V. This means that only the output voltage control sense circuits need changing. Third, choose a unit that contains the least amount of dust (possibly had the least use!) and check the fan for free movement and lack of “end play” in the bearings. Fourth, check if the unit uses two ICs in the control circuit: a TL494 and a LM339. Their IC pins and functions are easily identified, making analysis of the circuit much easier. If you can’t identify the ICs, you may still be able to modify the supply but you will be you will be very much on your own and the information in this article may not be of much help. What’s involved? (1) The main transformer will have to be removed, rewound and replaced. The only way you can tell from this angle that this is the modified supply is the absence of a 110/230V switch. This is removed because (a) it is quite superfluous here in Australia/NZ and (b) because over the years we have seen too many people flick switches like this with (briefly) spectacular results. 26  Silicon Chip Minor modifications are made to the mains circuitry. (2) The +5V, +12V, -12V and -5V output components are removed, with the exception of the +5V rectifying diodes and transient suppression network. A new output filtering circuit is installed. (3) The output voltage sensing resistors will need to be replaced. Jumper wires will need to be placed to supply the control circuit and the fan. Before we get too involved, some theory of operation is required. Basic principles The basic principles of typical PC power supplies can be described with reference to the block diagram of Fig.1. (1) The 240VAC mains input circuit contains the usual suppression components of chokes and capacitors before the four normal rectifier diodes in a bridge configuration. The rectified mains then passes to two storage capacitors connected in series. These capacitors will charge up to about 170V each and may be subject to ripple currents up to 5A or more. (2) Transistors Q1 & Q2 are alternately switched at 30kHz or more to provide a high-frequency alternating current to the main transformer primary. A small transformer with a single-turn primary winding senses the level of input current in the common line to the main transformer. www.siliconchip.com.au (3) The main transformer has three secondary windings providing the high current +5V and +12V outputs and a low current -12V output. (4) Large, fast-recovery double diodes (with a common cathode connection) in plastic TO-220 (or similar) packages rectify the high current +5V and +12V outputs, while smaller, fast recovery diodes rectify the -12V output. The -5V line is derived via a 7905 regulator from the -12V output. A large, multi-winding toroid provides initial filtering for the several outputs. Final filtering is provided by electrolytic capacitors and smaller inductors. (5) The main component in the control circuit is a TL494, Samsung KA7500B or equivalent IC. An RC network controls the operating frequency of the IC. The alternating drive to the switching transistors is pulse width modulated, depending on the load current demand, higher currents being supplied by longer duration pulses up to a maximum duty cycle of 45%. The output voltage feedback controls modulation width. The LM339 (and/or discrete transistors) senses over-current or over-voltage output conditions and shuts down the TL494. Features of the TL494 This is only a brief description of the operation of this IC. Further information is available from www. fairchildsemi.com The IC contains an oscillator capable of operating between 1kHz and 300kHz. The frequency is controlled by an RC network on pins 5 (C) and 6 (R) – see Fig.2. Two error amplifiers are included: pin 1 (non-inverting) and pin 2 (inverting) for amplifier 1 while pin 16 (non-inverting) and pin 15 (inverting) are connected to amplifier 2. The outputs from these amplifiers are commoned and internally control the pulse width modulation section of the IC. The common output is also connected to pin 3 to provide external control over the pulse width modulation. There are two output transistors with open collectors and emitters: Transistor Q1 has pin 8 (C1) and pin 9 (E1) while transistor Q2 has pin 11 (C2) and pin 10 (E2). These transistors can handle up to 200mA. The Dead Time control (pin 4) limits the duty cycle for each transistor to a www.siliconchip.com.au Fig.2: the two main chips you’ll find inside a typical SMPS are the TL494 and LM339. Here’s the pinout (and functionality) of both. maximum of 45% (0V to pin 4). This provides a 5% protection interval, preventing both output transistors being on at the same time. The Dead Time control is also used to disable the chip if an over-voltage or over-current condition occurs. Pin 13 (output control) may be used in some circuits to disable the TL494. The input supply (Vcc) is to pin 12 and has a maximum value of 42V. Pin 7 is ground. A reference voltage of 5V ±5% is available at pin 14. copied) to produce a grey scale image to fill an A4 page. The components can then be drawn on the page in a contrasting colour (eg, red) to assist tracing and identifying the various circuit features. By the way, if you haven’t already got the message, modifying one of these power supplies is not a quick or simple job but it does have the big advantage that you get a large output DC supply for very little cost. Make a drawing Some more recent PC power supplies derive their control circuit power from the +12V output. This feature allows the control circuit of these supplies to be powered and checked Before commencing testing and modification, I suggest that the underside of the PC board (track side) be scanned and printed (or photo- Pre-test before modification Here’s what you should find when you lift the lid on the switch-mode power supply. Usually it’s only four or so screws to get this far. All of the external cabling will be removed. Never run the supply with the lid removed unless testing – and then only with extreme care. These things can be lethal! November 2003  27 This waveform shows the ripple and noise output of the modified power supply. While it looks horrible it is only 67mV pk-pk. Note: measuring this waveform should be done on the external outputs, not inside the power supply (for safety’s sake!). without connection to the 240VAC mains. Connect a 33Ω 5W resistor between the +5V output (red) and ground (black) and a 47Ω 5W resistor between the +12V output (yellow) and the +5V output (red). This will maintain an approximate 5V to 12V ratio between the respective outputs. A variable DC power supply (8-14V range) is connected across the +12V output and ground. Check for power at pin 12 of TL494. It should be almost 0.6V less than the supplied voltage. In the absence of power, a jumper needs to be placed between pin 12 and the +12V line. An oscilloscope is used to view the waveforms and operation of the TL494 and LM339 as the applied voltage is slowly raised from 8V to 14V (no higher than 14V). A 30kHz (or higher) sawtooth waveform should be present at pin 5 and square waves should be visible on the ungrounded output pins 8 and 11 (or pins 9 & 10). These oscillations should stop as the voltage is raised to the level equivalent to the design output. The waveforms should reappear as the voltage is re- duced. If the over-voltage circuit has been activated, the waveforms will not reappear until the circuit is reset by removing the power. Careful adjustment of the power supply is necessary to demonstrate these two very similar voltage levels. If no oscillations are observed, pin 4 of the TL494 will need to be isolated from the circuit and connected directly to ground. Follow the track from pin 4, desolder and lift one leg of each component connected to this track. The track can then be grounded by a jumper wire. The over-voltage protection circuit will now be inoperative. Re-connect the variable DC power supply and a sawtooth waveform should now be visible at pin 5 and square waveforms at pins 8 & 11 (or pins 9 & 10). Do not exceed 14V in an attempt to demonstrate the over-voltage protection mode – you have just disabled this circuit! Use a multimeter to measure the reference voltage at pin 14; this should remain constant at about 5V, as the supply is varied. Make a note of this reference voltage. Next, measure the voltages at the input pins to the error amplifiers, pins 1 & 2 and 15 & 16, as the supply voltage is varied. Note: one of these amplifiers may not be used in the circuit. The pin with the constant voltage, pin 2 or 15 (inverting input), is connected to pin 14 via a resistor or a potential divider network and serves as the reference voltage for the error amplifier. Make a note of this voltage too. The non-inverting input, pin 1 or 16, is connected to the +12V and +5V outputs via another potential divider network to sense the output voltage. You will need to trace the connections to this pin to identify the voltage feedback network. The signal from the TL494 to the driver transformer can also be check­ ed. The primary of this transformer is a centre-tapped winding with the centre pin grounded. The signal to the other two pins should be identical in shape and amplitude (sketch these waveforms). A dual trace oscilloscope will show the phase relationship between these waveforms (no overlap at all). The waveforms at the five output pins of this transformer will vary, as the circuitry to the “chopper” transistors is not symmetrical. However, the waveforms should be roughly similar. Voltage measurements also need to be made at the input pins of the comparators in the LM339 IC. Usually only two comparators are used; the remaining inputs are tied to ground or Vcc. Two pins (the inverting inputs) should maintain a fixed voltage equal to the reference voltage on the input to the error amplifier in the TL494. The pin with the varying voltage (a non-inverting input) is connected to the supply output via a voltage divider network and senses an over-voltage It’s dunked in paint stripper overnight . . . The original transformer, as removed from the PC board. 28  Silicon Chip . . . allowing fairly easy disassembly. Make sure the ferrites and bobbin are very clean before going any further. Don’t worry about the wire – you won’t be using any of that. www.siliconchip.com.au Fig.3: rewinding both primary and secondary of the main transformer is arguably the most critical part of the whole exercise. The primary is rewound because its insulation will probably have been destroyed by the paint stripper. condition. This part of the circuit will also need to be identified and modified. The other non-inverting input pin is connected to the over-current protection circuit. This portion of the circuit does not require modification as the over-current condition is detected at the input to the main transformer. Take careful note of the results from the above testing procedure. The test will need to be repeated after the modifications and transformer rewind, as a final check before applying mains power. The only difference is that then there will be no output to the original +12V output, the new output appearing at the original +5V output. If your PC power supply cannot be tested with an external DC supply, you can still modify it but it will be far more difficult (and dangerous) to do any initial testing. However, you can still trace out the circuit and then follow the procedure within this article to make the necessary modifications. WARNING! The internal wiring of switch-mode computer power supplies is dangerous when powered up. Not only do you have bare 240VAC wiring to the IEC sockets but a good portion of the circuitry is at +340V DC and is also floating at half the mains voltage. It is POTENTIALLY LETHAL! Use extreme care if you do decide to take measurements on the supply when the case is open and DO NOT TOUCH ANY PART OF THE CIRCUIT when it is plugged into the mains (operating or not). Make sure that it has been disconnected from the mains for about 15 minutes before making any modifications and make sure that all high-voltage capacitors have been discharged before touching any parts. Transformer rewind The main transformer operates at a frequency of between 30kHz and IMPORTANT: although not shown here, fit PTFE sleeving over the primary wire ends (and to the inter-winding shield lead) before soldering them to the bobbin pins, so that no part of them will be exposed once the primaryto-secondary insulation tape is applied. PTFE SLEEVING Here’s what they should look like after disassembly. The next step is to wind on a new primary, as shown at right . . . www.siliconchip.com.au www.siliconchip.com.au Make sure it is a tight, neat winding – otherwise you might run into space problems. The original inter-winding shield is re-used. Note the layer of insulation between the windings. NO ovember ctober 2003  29 2003  29 be Farnell Cat. 753-002 (19mm) or 753-014 (25mm). Rewinding the primary There are a few modifications that you need to make to the PC board. These will vary according to manufacturer so be careful as you trace the circuit out. 85kHz and so is much smaller and has a surprisingly small number of turns compared to an equiv­ alent mains transformer operating at 50Hz. Begin by desoldering and removing the main transformer. Then submerge it in a container of ordinary paint stripper overnight, before any attempt is made at disassembly. Note: paint stripper is highly caustic and care should be exercised during this operation; use gloves and eye protection! The next day, carefully wash all traces of paint stripper from the transformer. The ferrite cores should now slip easily out of the bobbin. Keep careful WRITTEN notes of the windings (number of turns and pin connections on the bobbin) as the transformer is disassembled. In particular note the primary pin connected to the interwinding shield, if fitted. Note that ALL windings have to be removed as the primary has also been subjected to the effects of paint stripper. The ferrite core halves and bobbin should be thoroughly cleaned of all traces of adhesive, potting residue and paint stripper before rewinding. This may involve another overnight soak in paint stripper. Surprisingly, the paint stripper appears to have no effect on the bobbin. Care must be exercised during rewinding due to the space limitations imposed by the ferrites. All windings must be tightly and closely spaced. Do not overdo the application of insulation tape nor use larger gauge wire than suggested. Editor’s note: we recommend the use of a polyester tape when rewind­ ing the transformer, to ensure adequate high voltage and high temp­ erature ratings. A suitable tape would Rewind the primary with the same gauge wire and the same number of turns as initially used (usually 40 turns of 0.8mm enamelled copper). If the primary has been split into two windings (inside and outside the secondary windings) it should be replaced with a single winding. The primary is usually wound as two layers of 20 turns each. A single turn plus 10mm overlap of insulating tape is placed between the two layers during the rewind. The overlap must be located on a face of the bobbin not covered by the ferrite cores (see photo). After each primary layer is wound, install lengths of PTFE sleeving over the wire ends before terminating them at the bobbin pins.Suitable PTFE sleeving is available from Farnell, Cat 583-935 (0.86mm bore; other sizes are also available). Another single turn plus 10mm overlap of polyester tape is then applied over the final primary layer and the interwinding shield is then replaced. Note: this shield is approximately one turn and must be insulated so it does not form a single shorted turn. Terminate the primary winding and shield to the appropriate pins (in accordance with your written notes!) and cover them with two layers of insulating tape (trim to exactly two turns, no overlap). Again, fit PTFE sleeving over the lead to the inter-winding shield. Insulation at margins After terminating the primary wind­ ings and shield to the appropriate pins, use thin strips of insulation tape Then on go the secondaries. As with the primary winding, this should be nice and tight. The rubber bands are removed before adding the final layer of tape. As before, fit PTFE sleeving over the wire ends before terminating them to the bobbin pins. At right is one idea for the new output filter electros. 30  S 30  Silicon iliconCChip hip www.siliconchip.com.au www.siliconchip.com.au (trimmed to the appropriate width) to build up the gaps between the ends of the primary winding and the bobbin shoulders, to give a complete uniform layer the full length of the bobbin. Once you have a uniform cylinder, cover the entire winding (right up to the bobbin shoulders) with exactly two turns of insulation tape (no overlap). The idea here is to ensure that all possible points of contact between the primary and secondary windings are doubly insulated. WARNING: for safety reasons, it’s vital that the primary winding be correctly insulated, so that it cannot possibly come into contact with the secondary. If you get it wrong, the supply could be LETHAL if the earthing is incorrect. Do NOT attempt any of this work unless you know exactly what you are doing. Apart from the obvious output terminals, the changes made to the original supply are not all that obvious in this modified one. Winding the secondary A total of 10 turns, double-wound and centre-tapped, of 1.25mm enamelled wire forms the secondary. This winding is rather difficult to apply because the larger gauge wire has a tendency to spring open. Use a rubber band as a temporary hold after completing each winding. Start by selecting one of the outside four pins used to terminate the original 5V winding (largest gauge wire). Wind on five turns, tight and closely spaced, in the direction away from the other three pins, bringing the end of the wire up through the notch in the bobbin top. Leave about 20cm of free wire. Now select the adjacent pin and wind another five turns in the same direction and placed between the turns of the first winding. Allow the first coil to expand lengthwise along the bobbin as needed. Terminate this winding as above. Check and recheck that you have exactly five turns on each winding, otherwise you will effectively have a shorted turn. Firmly cover this layer with one turn plus 10mm overlap of insulating tape. The second layer begins from the outer pin of the remaining original 5V winding pins. Wind five closely spaced turns in the opposite direction to the first layer and terminate through the top of the bobbin. Again, leave 20cm free. Starting from the remaining 5V pin, wind another five turns placed between the turns of this second layer. Terminate as above. Again, check and recheck for exactly Here’s the stripped PC board with the rewound main transformer in place, ready for the new output filter components. Add a pair of polarised terminals on their own mounting plate and fasten it to the power supply case, as shown at right. www.siliconchip.com.au www.siliconchip.com.au five turns on each winding. Firmly cover this final double winding with two layers of tape. Refitting the ferrite core This is the real test of the rewind. Cautiously slide the ferrite core halves into the bobbin; remember, they are very brittle! If you are lucky and have been very careful, they will slip into the bobbin without any obstruction. If not, remove one turn of the outer tape layer and try again. If you are still unsuccessful, it may be possible to gently squeeze the windings in a vyce, padded with two pieces October 2003  31 2003  31 November Modified, checked, tested . . . ready for the lid to go back on. And at the risk of sounding boring, for your own safety don’t apply power while the supply is in this condition. of wood, to press the secondary into a slight oval shape. No vyce? Place the bobbin between two pieces of wood and GENTLY tap with a hammer. If the ferrites will still not fit, the secondary will have to be rewound Once the ferrite core halves have been fitted, with no spacing or foreign matter between the joining faces, two layers of tightly stretched tape will hold them together. Start across the base with the first length gently stretched, then tightly stretch the tape after the first corner. Finish with a gently stretched length across the base. Final assembly Gently twist the four 20cm centretap leads into a rope-like formation. Scrape the enamel off all wires and gently hook them around their corresponding termination pins and solder. Take care – the pins can be broken out very easily, particularly the pins for the secondary terminations. Replace the rewound transformer on the board and bend the flying centre-tap lead to its connection point 32  Silicon Chip on the board. This hole may need to enlarged slightly. Trim, clean and tin the end of this lead before soldering. PC board modifications After identifying the critical circuit features and rewinding the transformer, the PC board modifications are almost an anticlimax. First, re move, the input voltage selector from the board. Note: in the 230V position this switch is OPEN. Cover the vacant switch position with a suitable metal bracket. Next, connect three mains-rated 10nF capacitors (X2 class) across the back of the IEC socket to reduce rectifier noise imposed on the 240VAC mains. The capacitors are connected between Active & Neutral, Active & Earth and Neutral & Earth. Now we come to the output circuit. Do not remove the lower (earthed) output voltage sensing resistors. Starting from the output leads, work back to the transformer and remove all -5V and -12V components, including the spike suppression resistor-capacitor combination across the -12V winding. Repeat the procedure for the +12V components, including removing the double fast-recovery diode from the heatsink. Also, remove all +5V components back to the fast-recovery double diode. Leave the diode and the spike suppression components in place. The multiple-winding toroidal choke is also removed, stripped of its windings and then rewound with 14 turns of 1.25mm enamelled copper wire (ie, a single winding). Note that you will need two chokes of 14 turns each in the filter circuit – the second toroid can be scrounged from another power supply. This new +13.5V output filter is a low-pass “T” configuration, with the two rewound chokes in series and four 2200mF 25V electrolytic capacitors from their centre point to ground. Using the original +5V output copper tracks, insert and solder the rewound filter toroid (the original +5V output becomes the new +13.5V output). The placement of the remaining filter components depends on the physical layout of the original +5V output tracks. I used a small piece of PC board to hold the four 2200µF capacitors. This board was then mounted off the SMPS board using some spare 1.25mm wire. (Editor’s note: we strong­ly suggest that the four 2200µF 25V electrolytics should be low ESR types, such as those available from Altronics in Perth; Cat. R-6204). The second toroid was soldered to the +5V output pad and to the first toroid. A ceramic disc capacitor (100nF 63V) was also added to the SMPS circuit board in parallel with the four 2200µF electrolytics. The following jumper wires are needed to complete the circuit: (1) Between the common cathode of the fast recovery diodes and the supply circuit for the TL494 IC; and (2) Between the final output pad and the fan’s positive terminal (assuming, of course, that the fan is a 12V DC type). A resistor may be used for this jumper to reduce fan speed and noise. DO NOT make this connection if the fan is mains powered (rare). New values for the voltage and over-voltage sensing resistors now need to be calculated. These resistors are in divider networks and, in each case, you can leave one of the resistors in place and just change the value connecting to the output. For example, in the Seventeam ST230WHF unit shown in the accompawww.siliconchip.com.au And here’s the proof that it all works, with this test set-up following reassembly. The wooden contraption at right is a home-made dummy load (hey, don’t knock it: it works!). The DMM shows that we have achieved a perfect 13.5V output, while the ammeter (centre of pic) is reading almost 20A. Don’t even think about such a test before the lid is on the case! nying photos, pin 1 of the TL494 is the non-inverting input of the relevant error amplifier. It has a 3.9kΩ resistor from pin 1 to ground and its reference voltage (set by a voltage divider connected to pin 2) is +2.5V. We want an output of +13.5V, so we need to calculate a new value for the resistor from pin 1 to the new 13.5V output. From here it is a simple ratio calculation.     R = 3.9kΩ(13.5/2.5 - 1) = 3.9kΩ x 4.4 = 17.2kΩ So you merely have to replace the original resistor with 15kΩ and 2.2kΩ resistors in series. The over-voltage monitoring network to one of the LM339’s comparators may then need modifying to work with the new voltage output. The process of calculating the resistor is similar to that above; leave the resistor from the relevant comparator input to ground in place and calculate a new value for the resistor connected to the output. Note that the final output voltage may not be exactly 13.5V regulated due to resistance tolerances and the tolerance of the 5V reference from the TL494. Check that the potential dividers are connected between the new 13.5V line and ground. Jumpers may be needed to complete these connections. www.siliconchip.com.au If the supply proves to be sensitive to RF fields, 100nF monolithic capacitors fitted between ground and all used inputs and outputs of the ICs should fix the problem. (Editor’s note: the addition of these capacitors will severely reduce the transient response of the supply and so it should only be done if the unit is used in conjunction with a radio transmitter). The configuration of the final output connections is left to the constructor’s requirements. Remember that these connections will have to handle up to 18A or so. The board should now be ready for its first test. Note that you will still need a minimum load such as a 47Ω 5W resistor. Repeat the low-voltage pre-test procedure described earlier, using if necessary the 33Ω and 47Ω resistors connected in series across the output terminations. Hopefully, the earlier waveforms will be observed. If the connections to pin 4 of TL494 have been removed earlier, restore these connections and check if the oscillations cease as the voltage is increased to about 14V. If all is well and the modified board behaves as expected you are almost ready for the first big test but first, there’s one final safety check. Both the metal case and the ground (0V) output of the supply should be connected to mains earth. Use an ohmmeter to verify that these connections are in place. Check also that the centre-tap of the rewound transformer is connected to mains earth. Under no circumstances should the output be floated! Now reassemble the supply into its case. Make sure that all connections are correct and close the case. Place a test load, (eg, a 12V 50W halogen lamp) across the output, plug in to the 240VAC mains and switch on. If the globe lights, congratulations! Final testing can now proceed using a series of loads to measure the output current and voltage. If the globe does not light, switch off, unplug the unit from the mains and wait for at least 15 minutes to discharge the high-voltage capacitors, before opening the case. If the globe “blows” there is a good chance the output voltage sensing circuit is not correctly connected. Finally, note that PC power supply cases have ventilation slots. For safety’s sake, be sure to cover any slots or cutouts that give access to dangerous high-voltage circuitry (eg, by attaching aluminium panels) but make sure there is adequate ventilation overall. Further reading: “Making Use Of An Old PC Power Supply”, SILICON CHIP, Dec 1998. SC November 2003  33