Professional Documents
Culture Documents
Page 1 of 14
This article was originally published (in a slightly modified form) in the QST magazine, December 1998 and January 1999, and in the Radio Amateur's Handbook, 1999. Visit the American Radio Relay League for information on these publications, and a world of ham radio related things! Amateur radio has been somewhat slow to accept switching power supplies for powering communication equipment. This is a pity, because "switchers", as they are often called, offer very attractive features, like small size, low weight, high efficiency, and low heating. True, they are generally more complicated than linear power supplies, but this is easily compensated by the fact that they can be built for a lower cost. Some early switchers produced an objectionable amount of RF noise, bringing the whole switching technology into bad reputation. But by proper design techniques and careful EMI filtering it is possible to build very quiet switchers. In this article I will describe the construction of a switching power supply designed to power a complete ham station, with several radios and accessories. This power supply produces 13.8V regulated to better than 1%, at a continuous load current of up to 40A. It has current limiting, making it appropriate for direct connection to a 12V backup battery. If the current limit potentiometer is turned up, the power supply can deliver up to 60A on an intermittent basis, while maintaining regulation. No minimum load is required. The ripple on the output is about 20mV, and the efficiency is 88%. A cooling fan operates depending on the average current drawn, and a tricolor LED tells you if the voltage is normal, too high or too low. It produces no detectable RF noise at any frequency higher than the main switching frequency of 50kHz (I checked it with an antenna wire looped around the operating power supply, tuning my TS450 from 30kHz to 40MHz). And you get all this in a box that measures just 306 x 150 x 130mm, including all projections, and weighs only 2.8kg! Interested? Heat up your soldering iron and read on!
http://ludens.cl/Electron/PS40/PS40.html
3/15/2012
Page 2 of 14
Design decisions
There are several different topologies for switchers in common use, and the first decision a designer must take is which of them to consider. Among the factors affecting the decision are the power level, the number of outputs needed, the range of input voltage to be accepted, the desired tradeoff between complexity, quality and cost, and many more. For this power supply I decided to use the half bridge forward converter design. This topology connects the power transformer to a bridge formed by two power transistors and two capacitors. It is reasonably simple, puts relatively low stress on the power transistors, and makes efficient use of the transformer's magnetic capabilities. The second basic decision is which switching frequency to use. The present trend is to use ever higher frequencies. But by doing so it becomes more difficult to filter out the RF noise inevitably generated by the switching. So I decided to stay at a low switching frequency of only 25 kHz for the full cycle, which due to the frequency doubling effect of the rectifiers results in 50 kHz on the output filter. For the main switching elements, bipolar transistors or MOSFETs can be used. Bipolars have lower conduction losses, while MOSFETs switch faster. As in this design I wanted to keep the RF noise at an absolute minimum, very fast switching was not desired, so I used bipolar transistors. But these tend to become too slow if the driving is heavier than necessary. So, if the transistors have to switch at varying current levels, the drive to them must also be varied. This is called proportional driving, and is used in this project. The half bridge converter is best controlled by pulse width modulation. There are several ICs available for this exact purpose. I chose the 3524, which is very simple to use and easy to find. Any 3524 will do the job. It can be an LM3524, SG3524, etc. This basically ends the big decisions. From now on, designing the circuit is a matter of calculating proper values for everything.
Circuit description
For the following explanation, refer to the schematic diagram. Print it out, so you can follow the description on the drawing. Line voltage enters through a CEE-22 connector with included fuse and EMI filtering (P1). It is then passed through a 2-pole power switch, and an additional common mode noise filter (C1, L1, C2). Two NTC resistors limit the inrush current. A bridge rectifier delivers the power to a big electrolytic capacitor (C3), which works at the 300VDC level. The power oscillator is formed by Q1, Q2, the components near them, and the feedback and control transformer (T3). T2 and the associated components act as a primary current sensor. T1 is the power transformer, delivering about 20 V square wave to the Schottky rectifiers (D6..9). A toroidal inductor (L2) and a six-pack of low equivalent series resistance electrolytic capacitors form the main filter, while L3 and C23..24 are just there for additional ripple reduction. The 13.8V is delivered to the output through a string of ferrite beads with some small decoupling capacitors mounted directly on the output terminals. The control circuit is a 3524 IC (U1), powered from an auxiliary rectifier (D17). The IC contains a voltage reference, oscillator, pulse width modulator, error amplifier, current sense amplifier, flip-flop and driving circuitry. It senses the output voltage and the current level, and through transistors Q3 and Q4 controls the power oscillator. C37, C35 and R23 are used to implement a full PID (proportional-integral-derivative) response in the control loop. A quadruple operational amplifier (U2) is used for two auxiliary purposes: To control the cooling fan according to the average current level, and to drive the voltage indicating tricolor LED: It will glow green if the voltage is OK, orange if the voltage is too low and red if it is too high.
http://ludens.cl/Electron/PS40/PS40.html
3/15/2012
Page 3 of 14
Page 4 of 14
primary. This current, about 40mA at full load, is put into R12, producing a maximum voltage drop of about 7V. This is rectified, the half of it is taken at the center tap, divided down by R13 and VR1 and smoothed by C31. If VR1 is properly adjusted, on the top of it you will get 200mV when the power supply is running at full load. The 3524 pulse width modulation controller is an interesting beast. I suggest you take a manual of analog ICs from any of the manufacturers that make this chip, and read through the 3524's data sheet. But I know that you want me to explain how it works, so here I go: An internal voltage reference puts 5V onto pin 16. R22 and R21 divide this down to 2.5V, and apply this to the noninverting input of the error amplifier. The inverting input gets the sample of the output voltage. With 13.8V on the output, and VR2 properly set, you are putting 2.5V into pin 1. The error amplifier is a transconductance amplifier with 5M output impedance. Its frequency and phase response, and its gain, are set by R23 and C35. A second amplifier, specially included for current limiting purposes, has its inputs at pins 4 and 5. This amplifier can be ground-referenced, as done in this circuit, and has an internal offset of 200mV. That means that it will pull down the main error amplifier's output if the difference between pin 4 and pin 5 reaches 200 mV. The chip contains an internal oscillator, whose frequency is set by R24 and C36 to approximately 50kHz. The sawtooth output of this oscillator is connected to an internal comparator, which has its other input internally connected to the output of the error amplifier. The result of this is that the output of the comparator will carry a square wave whose duty cycle depends on the DC voltage at the output of the error amplifier. Finally, an internal flip-flop distributes the resulting variable length square pulses among the two outputs, which are uncommitted transistors. In this design, the emitters of those transistors (pins 11 and 14) are grounded, while the collectors drive Q3 and Q4. During operation at medium to high load, the duty cycle is about 70%. That means that at the cathodes of the Schottky rectifiers you will have a square wave that stays at about 20V for some 14s, and then slightly below ground level for 6us. L2, which has its other end at nearly constant 13.8V, will therefore see about 6V for 14us, followed by -14V for the rest of the time. Given its inductance of about 20uH, the current in L2 will increase by about 4A during each conduction cycle, and decrease by that same amount during rest time. As long as the current drawn from the power supply is more than 2A, the current in L2 will never stop completely. For example, if 20A are drawn, the current in L2 will vary between about 18 and 22A. As the ripple current stays basically constant while operating at up to the maximum current of the power supply, the filter capacitors C17 to C22 are never exposed to more than about 1.5A RMS total ripple current, assuring them a very long lifetime. This is an advantage over some other types of switching power supplies, in which the ripple current is much higher, forcing the designer to use more expensive capacitors or to accept reduced lifetime. If the load is less than about 2A, the current flow in L2 is no longer continuous. The duty cycle of the power transistors starts to drop, until at zero load the duty cycle almost becomes zero too. C37 serves several purposes. For higher frequencies it couples the first filter stage (L2 and C17..22) to the error amplifier, while for lower frequencies and DC the output of the supply is sampled. This is necessary because each filter stage introduces 180 degrees of phase shift at the higher frequencies, so after two stages the phase shift goes through a full 360 degrees, making it impossible to stabilize the control loop without additional circuitry. But for DC it is good to sample the output, in order to compensate for the voltage drop in L3. So I chose this arrangement. C37 was dimensioned to give the error amplifier a nice PID response, together with R23 and C35. This affords the best possible transient behavior and unconditional stability. In addition, C37 provides some measure of soft starting, so the voltage does not overshoot too much when switching on the power supply. R34 and C38 average out the current level over a period of about 2 minutes. U2C amplifies the resulting voltage by a settable amount, and U2D acts as a Schmitt trigger to switch the fan cleanly on and off when the current average crosses the trigger level set by VR3. R39 limits the speed of the fan to a rather low value which is more http://ludens.cl/Electron/PS40/PS40.html 3/15/2012
Page 5 of 14
than enough to keep the power supply cool. At this low speed the fan produces almost no noise, and can be expected to last longer than its owner.. U2A and U2B control the voltage indicator LED. They operate as follows: An independent voltage reference (U3) delivers 5V to the noninverting inputs of the opamps. If the power supply's output voltage is below 13.5V, both pin 2 and pin 6 of U2 will be below 5V. So the outputs of these opamps will both be high, lighting both the green and the red parts of the dual LED. It will glow orange, indicating low voltage. When the voltage is near 13.8V, pin 2 will still be below 5V, while pin 6 will be above 5V. This switches the red LED side off, leaving the LED glowing green, indicating correct voltage. When the voltage rises to more than 14.1V, pin 2 will rise above 5V. Pin 1 will go down, shutting off the green LED half, and pulling low pin 6 via D18. This will switch on the red LED side, so the tricolor LED now glows red, giving overvoltage alarm. If you use reasonably accurate resistors, this circuit needs no adjustment to operate well.
Page 6 of 14
at HF, so this coil like the others not only blocks, but also absorbs RF energy. Finally, the output leads are passed through a full dozen of 73-material ferrite beads. The filtering is completed by proper bypassing of the output leads to the cabinet. It should be noted here that the ground on the printed circuit board is floating, but the enclosure is connected to earth. This arrangement reduces HF current flow on the enclosure, which would lead the box to act as a noise radiator instead of a shield! The power transistors and diodes are mounted in a special way too, for the same reason. This will be described in the construction section.
Running on 110V
I live in a country where the mains supply is 220V, 50Hz, so I designed my power supply for this voltage. It will accept about 190 to 250V. But switching power supplies are very easy to convert from 220V to 110V, and it is even very easy to make them switchable between the two voltages. This drawing shows how to do it. Basically, you have to split C3 into two parts, connecting one side of the AC line to the junction of the capacitors. This makes D1 work as a voltage doubler. If you want only 110V operation, use the circuit in figure 2 A. As two diodes of the bridge are not needed for 110V, they were eliminated, leaving just two simple diodes. If you want your power supply to be switchable between 220V and 110V, you need to add a few more components. Part B of figure 2 shows how to connect a switch to run the supply from either voltage. In this case, voltage-equalizing resistors are needed across the capacitors, as shown in the diagram. For 110V operation the fuse must be rated for 8 or 10A. The rest of the components can stay the same, although you may replace the NTCs by lower resistance ones, in order to maintain best efficiency. A pair of 1.3R NTCs should be about optimum for 110V. If you make the switchable version, stay with the 2.5R NTCs, because they are needed for 220V operation. If you choose one of these options, you will have to make the necessary modifications to the parts list when you go shopping. But you will not need to modify the printed circuit board other than drilling an additional hole, because anyway C3 is not mounted on the board.
Switching power supply include Farnell, Newark, Spoerle (in Germany) and many more. Here is a parts list. For some of the more specific devices, RS stock numbers are given.
Page 7 of 14
Please do not ask me to get the parts for you. I'm glad to give you any advise and help needed to complete this project, but I do not provide parts for it.
Page 8 of 14
Now you will need some copper sheet 0.1mm thick, and some Mylar sheet of a similar thickness. Cut the copper in strips 22mm wide, and the Mylar in strips 28mm wide. If you can make long strips, say 2m, this is an advantage, otherwise you will have to solder individual copper strips together. In total, you will need about 7m of copper tape and slightly less Mylar tape. When you are ready with this, your epoxy has had ample time to harden, so rescue your bobbin from the oven and go on. Take off the rubber bands, the outer wood blocks, and the outer plastic wrapping (don't worry if it doesn't come off completely). Do not remove the plastic wrapping that separates the bobbin from the wood. You now have your wrapped wooden core and the epoxy-paper bobbin on it. Take a 60mm piece of #13 bare copper wire. Wrap the end of one of your copper strips around the wire, so that the wire protrudes only to one side from the copper sheet loop. Use a big soldering iron to flow some solder into the junction. Try to avoid getting solder on the outside, because this may later puncture through the Mylar insulation. Now the winding starts. Position the copper wire on one narrow side of the bobbin, so that the copper strip is centered on the width of the bobbin, leaving 3mm room on each side. Stick the strip start to the bobbin with some thin adhesive tape. Position the start of one piece of Mylar strip so that it covers all the copper and is centered on the bobbin, and tape it in place. Now wind 15 turns of this copper-Mylar sandwich, as tightly as you can, keeping the Mylar aligned with the bobbin sides, and the copper nicely centered. Don't loose your grip, or the whole thing will spring apart. If your copper strip is not long enough, fix everything with rubber bands or a clamp, and solder another copper strip to the end of the short one, allowing 2mm of overlap. Before doing this, cut the first copper sheet to a length such that the joint will be on one of the narrow sides of the bobbin, because here you have space, while the wide sides will have to fit inside the ferrite core's window. If the Mylar strip runs out, just use adhesive tape to add another strip. Make the overlap 5mm, to avoid risk of creepage between the sheets, and also try to locate the joint on a narrow side of the bobbin. When the 15 turns are complete, cut the copper strip to such a length that the second terminal will be on the same narrow side of the bobbin as the first terminal (the one you have already placed). Solder the second terminal (another 60mm piece of bare copper wire) to the strip, position it, and wind three or four layers of Mylar, to make a safe insulation between the primary and secondary. If you think this is a messy business, you are right. But it's fun too! The secondary is just a little bit messier: It is wound with a five-layer sandwich! Four layers of copper and the Mylar topping layer. But it's only four turns total, so take a deep breath and do it! First solder the four copper strips together around a piece of #13 copper wire. Don't be overly worried if the outcome is not very clean; mine was quite a mess too, and it worked well on the first try. Just be sure you don't create sharp edges or pointed solder mounds, because these may damage the insulation. Now position the start of your secondary conductor in such a way that the pin will come out to the same side as those of the primary, but on the other narrow side of the coil assembly. So you will get a transformer that has its primary leads on one extreme and the secondary on the other, and will fit the printed circuit board nicely. Wind two turns, solder the center tap wire between the four copper strips, wind the other two turns, solder the last pin, wind a finishing layer of Mylar and fix it in place with adhesive tape. Uff! This was the worst part. If you reached this spot, you will have no problem building the rest of this project! What you have now is a springy, messy coil assembly that will fall apart if you let it go. You have to seal it. This is easy to do: Wrap your two wooden blocks, the same you used to press the bobbin, in plastic film. Place them against the sides of the coil assembly, and apply hard pressure, using a clamp or a lot of rubber bands, so that the long sides of the coil straighten out completely, and any slack is displaced to the narrow sides. Now mix a fair quantity of epoxy glue, place the coil assembly so that the pins face up, and let the epoxy run into the coil. Continue supplying epoxy until it starts to set. If it drips out from the other side, no problem. Just don't do this work over your uncle's http://ludens.cl/Electron/PS40/PS40.html 3/15/2012
Page 9 of 14
persian rug. When the epoxy doesn't flow any longer, turn over the coil assembly, mix a new batch of epoxy, and fill the other side completely, forming a smooth surface. As the downside is now sealed, the epoxy will not flow out there. And when this epoxy has set, turn the assembly over again, mix epoxy, and apply it to form a smooth surface there. The idea is to replace all the air between copper and Mylar sheets by epoxy, and specially to fill the room left by the copper strip, which is narrower than the Mylar. This filling is necessary both for mechanical and for electric safety reasons. Now convince your kitchen's monarch that this devilish thing will add a nice scent to the next apple pie, and place it in the oven again. Let the epoxy harden completely, then remove the coil from the oven, remove the clamp, rubber bands, wooden blocks, wooden core and all remains of plastic film.. And now comes the big WOW!!! You will be surprised how your messy and springy assembly changed into a very robust, hard, strong and nice coil! Now test-fit the ferrite cores. See if they can be installed easily, so that each pair of facing E cores gets in intimate contact without pressing on the winding. If everything is right, the winding should have some playroom in the assembled core. But it is easy to get too much epoxy on the coil. If this happened to you, just take a file and work the epoxy down so that it doesn't disturb the ferrite. The ferrite core MUST close properly, otherwise you will later burn out the power transistors! When the sizes fit, prepare some epoxy (again...!), apply a very thin layer to all contact faces of the ferrite cores and mount them onto the coil assembly. You can hold them in place with adhesive tape until the epoxy sets. If you dare interrupting in the kitchen for a third time, use the oven to quick up the hardening! The last thing you have to do is bending the copper wires into the proper shape to fit the printed circuit board holes. Be sure that on the secondary winding the center tap is actually in the center position! The polarity of the other pins doesn't matter. This completes the manufacture of T1.
All the other transformers and coils are just child's play after making T1. The current sense transformer T2 has a lot of turns, but there is absolutely no need to wind them nicely side-by-side. You can use a winding machine with turns counter, or you can just wind it by hand. Get some #36 or other thin enameled wire, solder the end of it to one of the extreme pins of the EE24-25-B bobbin, and wind 100 turns. Don't worry if your winding is criss-cross and ugly, and don't feel guilty if you loose count and wind a few turns more or less. As long as you don't overdo it, it will just affect the position of VR1 when you align the completed power supply later. Solder the wire tothe center pin of the same side, then wind another 100 turns in the same sense. Solder to the other extreme pin on the same bobbin side, and apply one or two layers of Mylar, just to protect the thin wire. Now take a piece of #15 plastic insulated cable, wind one single turn over the Mylar and solder the two ends of the cable to the two extreme pins of the other side of the bobbin. It doesn't matter which end goes to which side. Install the EA77-250 core with a small amount of epoxy cement, and T2 is ready. http://ludens.cl/Electron/PS40/PS40.html 3/15/2012
Page 10 of 14
T3 is made using the same type of bobbin and core as T2. First you wind 26 turns of #27 enameled wire. The 26 turns fit nicely in a single layer. Wind a one layer of Mylar sheet, then put on the next 26 turns. Bring the wires out to one side of the bobbin in such a way that they will not be too close to the other windings. Wind 3 layers of Mylar tape, to give a safe insulation between primary and secondary. Now, wind 8 turns of #20 wire, and solder the ends to the bobbin pins. Look at the printed circuit board drawing to understand which wire to solder to which pin. Wind a single layer of Mylar, then make the other 8-turn winding over the first one. This will leave a space at one side of the bobbin which is big enough to take the single turn of #15 plastic insulated cable, which completes the assembly. Now glue the core in place with epoxy cement, and T3 is ready! L2 is wound on an Amidon T-157-26 iron powder toroid core. As it is too difficult to bend thick wire through a toroid, and tape winding it is not practical either, I chose to make this coil with 10 pieces of #16 enameled wire in parallel. Cut the wires to about 1.5m length, and twist them together. Then insert the bundle into the core, and starting from the middle of the wire bundle, wind 7 turns, using half of the core's circumference. Now wind the other 7 turns, starting from the middle towards the other end of the wire bundle. The 15th turn is the one you made when you inserted the wire bundle into the core! You will not be able to make a beautiful, nice winding, as the total of 150 wires passing through the toroid is too much to fit them in a single layer. But this doesn't matter at all, as long as you get 15 loops of the wire bundle through the toroid's hole, and there is plenty of space for this. To make L3 you must first get a suitable core. I used a fraction of an old ferrite antenna rod, which broke into several pieces when I let it fall down... So, this is the recipe: Take an old ferrite antenna rod of about 10mm diameter, throw it out of the window, go pick up the pieces and select one that is about 50mm long! If you live in a 40th floor, better don't use the window method. Instead, break the rod in some more controlled way. You can't wind L3 on a bag of ferrite dust! The winding itself is easy enough: Just wind 10 bifiliar turns of #11 enameled wire. This wire is quite stiff, but it is still no problem to handle. I suggest you wind the coil on a 10mm drill bit, then spring it open and place it on the ferrite core. Otherwise the ferrite could be cracked. Fix the core in the winding with some epoxy. Bend the wires in such a way that all four of them look down with the core straight up. That's the position L3 is mounted on the PCB.
Page 11 of 14
This is probably the easiest step. Using the parts placement guide, install and solder all of the parts except for Q1, Q2, and D6..9. Before installing D1, you must fashion a simple heatsink from a 30 x 80mm piece of 1mm thick aluminum sheet, bent into U shape. Drill a hole into it and screw it onto the rectifier bridge, using a locking washer. Then solder the D1 into the board. All of the parts are mounted flush against the board, as it is not good to have overly long leads. There are two cable jumpers on the board, drawn as straight lines on the components placement diagram. The short one can be made from a clipped component lead, while the longer one is made from a piece of thin hookup cable. Do not use IC sockets!!! These are an invention of the electronician's devil! They introduce additional contact resistance, inductance and capacitance, dramatically reduce the IC's cooling through the board, and cause trouble later when the IC pins oxidize, or the chips simply start falling out of the sockets. Don't tell me that you use good quality sockets. They do exist, they are very expensive, but they cause much the same problems. Keep in mind that modern ICs are made to be soldered! The tinned pins can be soldered easily and safely, but even in goldplated sockets those tin-plated pins oxidize, insulating themselves. I'm warning you so extensively because my job involves repairing electronic equipment, and I can tell you that more than half of all problems are caused by bad contacts, many of them in IC sockets!
Page 12 of 14
the front and rear panels. Q1 and Q2 are screwed to the rear panel, using M3 nylon screws and 3mm thick ceramic insulators. These thick insulators were used not only for safety reasons, but mainly because they reduce the capacitive coupling of the transistors to the enclosure. This is very important, because if the unwanted capacities at this place were significant, the switching spikes would be coupled to the enclosure, causing RF current to flow on it and noise to be radiated! It is of little use to filter the input and output leads, if you make your shield radiate the noise by using too thin insulators for the noise-generating transistors! Do not use metal screws with plastic washers, because this approach does not give enough safety margin to operate at line voltage. If you dislike nylon screws, a good alternative are steel clamps that press onto the plastic body of the transistors. The Schottky diodes are mounted using the same kind of insulators and screws, but there is a heat spreader, made from 6mm aluminum plate, between those insulators and the case. All surfaces requiring thermal contact are covered with heat transfer compound before assembly. When installing the diodes and transistors, first do all the mechanical assembly, then solder the pins. Otherwise you could stress them too much while fastening the screws. All cable connections are made next, and the output filter is assembled by sliding the ferrite beads over the output cables and soldering the bypass capacitors C25..30. Be sure to use a nice thick cable for the output! 40A continuous duty is no joke. I used a cable that has 4mm copper diameter, plus plastic insulation. It was a surplus from the installation of an electric winch in my 4WD vehicle. That cable is still a bit small for the winch, but fine for this power supply! Now you will need a big soldering iron (100 to 150 Watt) and a lot of solder. The tracks on the secondary side cannot be trusted to carry the 40 or 60A of this supply without some help! The necessary help comes in form of some lengths of #13 bare copper wire. Cut and bend it to fit the shape of all the high-current paths. Now take your big and very hot soldering iron, and flow generous amounts of solder onto all of those tracks, so that the reinforcing wire becomes buried in the solder. At the end, you will have tracks in which the copper wire does most of the conduction, while the solder couples the generated loss heat to the PCB, which dissipates it. When you are ready with this task, your board will look terrible with all that molten, burned and blackened flux residue! But don't despair: Place the power supply in such a position that any liquid on the solder side of the board can drip off, and now take a small paintbrush, a bottle of alcohol, and applying generous amounts of this magic liquid brush all of the ugly black rosin dirt away! You will be surprised at how easy this is, and at how nice the board looks after this treatment! This photo shows how mine looked after the alcohol bath. As the air flow from the fan will slightly shake the thin cables (those connecting the LED, and the jumpers), it is a good idea to place some drops of hot-melt glue at those places where these cables enter the board. This will prevent fatigue and breakage of these cables. Hot-melt glue is also an excellent material for fixing anything that would otherwise rattle, like ferrite beads.
http://ludens.cl/Electron/PS40/PS40.html
3/15/2012
Page 13 of 14
If you are building this project, you probably already have some experience in troubleshooting, so I don't need to teach you the basics. If you replaced critical components like the power transistors, or the cores for T1, T3 or L2, this may cause problems. The power transistors MUST maintain their beta up to at least 8A, otherwise they will cut short the conduction cycles when the load increases. If you happen to replace those transistors by some which have very low beta, the power supply may fail to start at all. If you replaced the magnetic cores and did a bad choice, the results can be quite dramatic. If T1 or L2 saturates, it can cause the power transistors to explode before the fuse has a chance to open. The light bulb in the AC line will http://ludens.cl/Electron/PS40/PS40.html 3/15/2012
Switching power supply avoid damage in this case, so by all means use that bulb for first testing!
Page 14 of 14
Another typical error is reversing the phase of some winding in T3. If you get one of the 8-turn windings reversed, the results will be explosive unless you have the light bulb in series. If you reversed the 1-turn winding, the power supply will simply not start. If you run into a problem which you can't solve, I will be glad to help. You can contact me from the homo ludens homepage.
http://ludens.cl/Electron/PS40/PS40.html
3/15/2012