Path: utzoo!attcan!utgpu!news-server.csri.toronto.edu!mailrus!cornell!uw-beaver!zephyr.ens.tek.com!tekig5!brianr From: brianr@tekig5.PEN.TEK.COM (Brian E Rhodefer) Newsgroups: comp.sys.amiga.hardware Subject: Re: UnInterruptable Power Supply Keywords: power supply Message-ID: <6921@tekig5.PEN.TEK.COM> Date: 8 Sep 90 18:26:39 GMT References: <02284.AA02284@lemsys.UUCP> Reply-To: brianr@tekig5.PEN.TEK.COM (Brian E Rhodefer) Organization: Tektronix, Inc., Beaverton, OR. Lines: 124 In article <02284.AA02284@lemsys.UUCP> clemon@lemsys.UUCP (Craig Lemon) writes: > As I said earlier, I use a fair bit of power so the DC/AC convertor >would have to be built to rediculous specifications. I circuit would have >to alternate about 20 to 30 amps (depending on what is on) at 60 Hz and >then step it up with a transformer. I then though of tapping into the DC >side of the power supply ONLY WHEN AC ISN'T ON, somehow. This would let me >feed in 12VDC straight. The power supply SHOULD (this is one of my >questions) have regulators to ensure proper voltage. (I would try to apply >this power before the system regulators). I would then only have to build >an AC convertor to power my monitor. My A500/A2000 Tech Reference manual doesn't cover the power supply, but the Amiga's "unregulated DC" power supply is almost certain to be at quite a high voltage - 250 to 300V, I'd expect. Switching power supplies like the Amiga's are most efficient in high input voltage situations (High voltage power FETS are cheaper than high-current ones at equivalent controlled-power levels, transformer windings are smaller, and input energy-storage caps are smaller). The 250V is usually produced by running the AC line power into either a voltage-doubler (for 110VAC operation), or a simple rectifier/capacitor filter (for 220VAC). It's very easy to jumper-configure a voltage doubler versus fullwave bridge selection, so the power supply is comfortable pretty nearly anywhere in the world; the only difficulty would be its fan, which would have to be matched to the line voltage, unless it's DC powered. Anyway, the DC/DC switching converter/regulator probably generates its multiple output voltages (+/- 12V, +5V) with multiple windings on its converter transformer. Typically, the 5V winding's output is used as the feedback signal to the switching converter, and the "+/- 12V" supplies either fall where they may, or the switcher generates slightly higher voltages and then uses a linear post-regulator. In any case, I don't expect you'll be able to find anywhere in the standard Amiga power supply to feed 12VDC in directly. If you want to run the Amiga from 12V, I think your best shot would be to do the reverse of what the Amiga's supply does: use a high-frequency, high-efficiency DC/DC converter to generate a 150VDC, reasonably regulated output at the requisite 2-3A, and then use some high-voltage power FETS to chop this into a 300Vpk-pk 60Hz pulsetrain. The reason for such a high voltage is that the power supplies in the bulk of electronic equipment generate their unregulated DC power using simple capacitor-input filters, which "peak detect" the AC voltage fed into their rectifiers. So, they're counting on seeing a peak line voltage of SQRT(2) times the RMS value. You can't just go applying a squarewave of that amplitude to the equipment, though, because it MAY also have transformers in it, whose primaries would then see much larger volt-time integrals than the sinewaves they were designed for would provide. Their primaries would very likely saturate. You could combat this by goosing the frequency up to 85Hz, but the equipment might be deriving timing from its supply. Besides, resistive loads powered by such a UPS would overdissipate by 42%. You could be really fanatic and try to synthesize a sinewave, but I don't think there's any need to. All that should be necessary is to come up with a waveshape that has the requisite peak voltage of 150V, and yet an RMS value of only 120V. Recalling that the RMS value of a square pulsetrain is given by the amplitude of the pulses divided by the square root of their duty factor, we can quickly calculate that a 150V, 50%, 60Hz pulsetrain ought to do nicely, and be quite easy to synthesize with simple switching elements: +150 _____ 0V ___/ \_____ ____ -150 \____/ In fact, the supplies in both the monitor and the Amiga should prefer such a source, as long as the edge speeds are tamed just a little (I'd put a little damped (resistor in parallel) inductor in series with your "60Hz" power). The right values will help ease life for the chopping transistors, as well, by allowing them to switch fully ON at low-current. If you use a full H-bridge chopper (four FETs), you have the opportunity of cleverly cycling the gate drives of the four transistors so that one pair of transistors handles the turn-ON transients, looking into a series inductor, and the other pair handles the turn-OFF transients, leaning into a capacitor. All this fanciness results from trying to compensate for output-voltage variation in the storage battery (Don't automotive supplies vary between 11 and 14V depending on the battery's charge and load?). If the battery voltage is more stable than this, a person might be able to get away with simply chopping the battery into a step-up transformer at a high enough multiple of 60Hz to minimize the transformer's size, and then using synchronous rectification techniques to down-convert the stepped-up Of course, if you can get yourself a big enough 12V-120V stepup transformer, you can do the chopping directly. > > I'm looking for suggestions for different ways to wire this system, >information on the workings of the power supply, any success and failure >stories, which design is most practical, fast ways of switching (if I have >to), would a capacitor in the computer cover-up switching time (I assume a >very large one would, and any other general suggestions. > A larger stack of batteries might be worth considering. If you had 150V worth of batteries, you could get away with very little magnetics; you'd only need a couple snubber inductors. And, at 150V, the current drain would be down below 4A, meaning that you could probably use quite small cells - maybe even those D-cell-sized gell cells. I, too, have always wondered why UPSes don't continually operate from battery, and use the line power to constantly charge. I can only come up with two guesses: 1) UPS designers are daunted by the prospect of a charging circuit capable of supplying a shade over the full-load current requirement for the system going amok and stuffing that much current into the battery *when the load is off*. I grant that that's a concern, but I think it ought to be fairly easy to sense the net current into the battery, and regulate it to the "trickle charge" value. 2) People worry about the accuracy of the oscillator that generates the simulated 60Hz. This one's simple: Phase-lock the rascal to the line, as long as the line is hot. Brian Rhodefer