On 12-10-97 Roy J. Tellason wrote to Day Brown... 
 
 RJ> Day Brown wrote in a message to Ian Woofenden: 
 RJ>  
 RJ>  DB> If I *had* to run a computer on an inverter, I'd use an  
 RJ>  DB> isolation transformer, and hope the hysteresis would damp the  
 RJ>  DB> ripple in the sin wave from the inverter, and pick up the slack  
 RJ>  DB> when the SWPS IGBT shuts on/off.  Of course, isolation xfmrs  
 RJ>  DB> are big, expensive,  and heavy. 
 RJ>  
 RJ> I have one here that handles a couple hundred "va" that cost me about 
 RJ> $60 or  
 RJ> so several years ago and occupies about a 5" cube in terms  
 RJ> of space,  weighs a couple of pounds (maybe 3 or so). 
 RJ>  
 RJ> Why would you assume that one of those would have hysteresis? 
*ANY* inductor would have hysteresis, which in this case, is of 
some considerable use.  The buildup of magnet flux in ferrite can 
not quite keep up with the steepness of a spike from a gate, or a 
spike on the input power line, and it is the *change* in that flux 
that induces current in the secondary- which is fed to a computer 
power supply. 
 
Looked at another way, if the input were an audio signal, we would 
notice a loss in fidelity, which in this case, we want.  I have a 
10 amp autotransformer (10 lb?) which is of course, resonant to 60 
hertz.  throw a spike at it, or a series along simulated inverter 
sin wave output, and the spikes, which are the result of a 150KZ 
switcher, just get soopped up with all those soft iron leaves- as 
that much mass cannot change it's magnetic flux so easily. 
 
The PWM chip of the new inverters and computer power supplies seem 
like a free lunch.  One will monitor the shape of the AC sin wave, 
and turn on/off the gate to different windings in a much lighter, 
and *much cheaper* transformer.  This works fine until you have a 
computer power supply with a PWM chip, that has *another* idea of 
the shape of the sin wave it wants, and turns on/off the gates in 
the SMPS to accomodate to what it sees as spikes on the line, or 
in the current draw of the motherboard. 
 
Of course, the two PWM chips might get along fine forever, but I 
am here to tell you, I have seen, and read about some that didn't. 
___ 
 * OFFLINE 1.58 * 1000 Half funny taglines w/ 1000 That Ain't, you pick, 
10$/BO 
--- Maximus 3.01
---------------