No matter where you look, nowadays you're surrounded by switching power supplies (or "DC/DC converters," to use the latest jargon).  Sure, there're more efficient than the old linear supplies, but they really aren't any more reliable: the old saw that whenever equipment fails, "it's always the power supply," is at least as true today as ever it was.  The more cynical among us might say that we've traded higher efficiency and lower power supply temperatures for more circuit noise and higher complexity.  And it's pretty hard to argue with that statement.  But, you know, switchers have been coming on ever since the first television receiver was built (where did you think that high voltage for the picture tube came from?), and they're not going to be leaving us anytime soon.  So let's trade a few tips to make their analysis and repair a little less imposing…


There are several reasons why switching supplies can be a real bear to repair.  Firstly, there are often hazardous voltages involved.  And many of these new supplies play fast and loose with the notion of "ground," which adds to the danger as well as being a pretty important part of the functioning of our favourite test equipment, the oscilloscope.  Then of course, there's the fact that when switchers run into trouble, they generally react by stopping.  Once they've completely halted, the original cause of the stopping can be a real puzzler.  This is especially true if the equipment manufacturer uses the shutdown feature of the switcher to minimize component damage under fault conditions… the fault causing the shutdown may have nothing to do with the power supply itself.  And the fact is that many modern switchers are now operating at close to RF frequencies, which require us to analyze what's gone wrong a bit differently than the old 120 Hz linear supply.


Now, I'm going to try not to be too ridiculous here, and suggest that you should be repairing any power supply problem that comes along.  You have to keep an eye on the value of your bench time, but there are always exceptions.  Your PC power supply can be replaced for less than $40, and there's no way you can compete with that -- just change it out!  On the other hand, I just finished working on a small switcher, about 60 watts or so, in a microwave radio, which the original equipment manufacturer advised me to swap out -- at a cost of $2,500.00!  And often, in transmitting equipment for instance, the power supply is just too big to replace the whole thing conveniently anyway.  So you have to use your judgement.


First, let's talk about documentation.  Try and get yourself a schematic of the power supply.  This is pretty vital with a switcher, because of the circuit complexity; much more so than with a simple linear supply.   At the very least, get on the internet and try and get datasheets for the power transistors and switcher controller chips used in the supply.  You'll need them if you get to the point where you have to figure out how the darn thing was supposed to work!


Now switching power supplies come in several various flavours, and you need to be especially on your toes if there's no input transformer.  One of the available flavours takes the AC line, runs it through a bridge rectifier and a capacitor, and rams it into the supply.  I hate this type!  Be especially careful in this instance, because there is no ground reference on the input to this type of supply.  It may well be that the output side, however, is ground referenced!  Regardless, if you come into contact with either side of the input line, you and/or your oscilloscope are at risk!  Your best bet when dealing with this configuration of power supply is to get yourself an isolation transformer, so that you can ground one side of its output, and feed the supply under repair with this.  An isolation transformer fed by a variac is even better.  Of course if the supply happens to be fed by three-phase 208 VAC, this may not be practical. 


We're just getting started, and already I'm out of space!  More on the innards of recalcitrant

switchers next month.