Switcheroo Power Supply

Synopsis:
See link for electronic schematic. It is largely correct and fairly complete, but a couple details are probably missing (the transistor symbols are fairly crude, and the Controller needs to be connected to the main AC input, to make measurements). "Broken" lines indicate wires that cross over or under other wires. The part marked "Controller" is not magic; I just don't know enough electronics to design the stuff that goes in there. Basically a fairly low frequency oscillator (likely less than 20 kilohertz) is used to create timing pulses that activate either individual switching transistors, or groups of switching-transistors. They work so that a connected-in-series of group of capacitors is allowed to become charged during one whole cycle of Alternating Current. Then, while a second series of capacitors is being charged, the first group is being discharged in parallel. Switching the capacitors from series to parallel allows a low-voltage Direct Current output (to the squggly-line Load symbol at the lower-left of the schematic). The exact output voltage depends on such things as the number of capacitors in each group, and how long each capacitor is allowed to be charged.

Details:
The type of power supply commonly found in ordinary personal computers is known as a Switching Power Supply. While I am interested in lots of different things, I managed for many years to just accept that, and never bothered to study the details of how they worked. Well, one day, in a burst of wild imagination, I came up with something that could FIT THE LABEL of "swtiching power supply", even if it had no similarity to the standard gadget -- and it doesn't. So, I call it a Switcheroo Power Supply, because switching flows of power is all it does! Do note that I am aware that because the power goes through so many transistors (all needing heat-sinks for cooling), this is likely not a hugely efficient way of converting AC line voltage to low-voltage DC. But at the HalfBakery, things like efficiency are less important than uniqueness...and this Idea does have interesting aspects to it, like no "voltage regulator" components, no transformer, no radio-frequency emissions...and possible usefulness for international travelers.

In the Beginning, Everything Is Off. The Controller will have to have some sort of connection to the input power, so that it can measure such things as total voltage and current-frequency -- and use that information. This will make the overall Switcheroo Power Supply suitable for use all over the world, where some places have 120V house-power and others have 220V, and some places use 60 hertz current-frequency and some use 50Hz -- and it will work just fine even in places suffering from brown-outs (as will be explained below)! Protection from power spikes, however, will have to be external to this device (but that's already normal for ordinary switching power supplies, so is not an objection to this Idea).

Now consider only Side I of the schematic (see link), as an alternating-current wave-form begins to flow into it. Note at the lower-left where the letter B is located near the "base wire" of a crude transistor icon. There is a vertical (broken) wire connecting all the base-wires of a stack of transistors; since they are the activation-wires for transistors, one signal on that connecting wire can cause all of these transistors to be Off at this time. (But this is an oversimplification; RELATIVE voltages among the capacitors may require, and dealing with brown-outs DOES require, individual control signals, all activated together.) Meanwhile, the letter A is located near the base-wire of another transistor, at one end of another (oversimplified) connected stack; all of these are to be switched On right at the moment when the wave-form of the Alternating Current Cycle is at its lowest possible voltage-level. These will remain On for one full cycle, and then be switched Off. Next, there are several other transistors marked with 1, 2, 3, and N, which will be activated individually, for short moments of time.

Note that in the middle of the stack of capacitor-icons are a couple of diagonal lines. They are NOT wires; they are only there to indicate that the total number of capacitors is arbitrary. The precise number will depend on the total line-voltage that the Switcheroo Power Supply is expected to convert, and how low a DC output voltage is desired. For the moment, let us pretend we are converting 120V to 12V; the ten-to-one ratio means we need at least ten capacitors (not counting the caveats below).

At the beginning of a single alternating-current cycle, with all the "A" transistor-group activated, we are dealing with a stack of capacitors that is connected in series. If each capacitor is rated for 12 volts, then obviously a connected-in-series group-of-ten of them can accommodate 120 volts. The problem, of course, is that Alternating Current has different voltage-levels at different moments, during the course of a full cycle of power. The solution is the sequence of transistors marked 1, 2, 3, ... N. They are to be activated (and deactivated) in sequence as the voltage curve rises from the minimum to the maximum. Here is where we have to deal with the caveat mentioned above, because the maximum voltage of a household 120V AC cycle is actually rather higher than 120 volts!

In a 120V system, the 120V is more formally and accurately known as "120 volts RMS", with RMS standing for "root mean square". This is a special kind of AVERAGE, such that we can pretend the AC voltage is comparable to a Direct Current voltage. That is, for DC, if we have ten amperes of current being pushed by 120 volts, then BY DEFINITION that is 10x120=1200 watts of power -- and for AC, if we have ten amperes of current being pushed by 120V RMS, then that also just happens to be 10x120=1200 watts of power. The actual momentary peak voltage of a 120V AC cycle is about 170 volts -- and this is why we actually need more than the previously-mentioned ten capacitors in the example series being discussed here (fourteen for 170V). However, do note that while a large region such as the United States may claim to use 120V RMS and 60Hz for power, there are local variations (almost always in the voltage, not the frequency). During a "brown-out", for example, the local voltage may only be 100V RMS (a lower-than-standard voltage causes incandescent bulbs to glow more dimly, thus the label). This is why I specified that the Controller for a Switcheroo Power Supply needs to be able to study the power that it supplied to it -- let IT decide how many capacitors to use in series! Finally, as the next part of the description begins, keep in mind that during the first half of an overall AC cycle, voltage rises from minimum to maximum, and during the second half of the cycle, voltage falls from maximum to minimum.

So, as our Alternating-Current cycle of power rises from its lowest voltage level, Transistor 1 is activated. This allows line-current to flow and begin charging the first capacitor in the series. After a short time, the cycle of power has risen 12 volts, so Transistor 1 is switched Off, and Transistor 2 is switched On. Two capacitors rated at 12V each can, in series, withstand a total of 24V; three in series can withstand 36V, and so on. The quantity of that "short time" of each numbered transistor's activation depends on the power cycle; let's pretend we are dealing with 50Hz here (50 cycles per second, or "50 Hertz", is normal in Japan). So, if in half of 1/50th of a second the power rises from its minimum to its maximum (a range of 170 volts), then we want 1/14th of 1/2 of 1/50th of a second, and as a rough first estimate, Transistor 1 would only be activated for 1/1400th of a second. That "fairly low frequency oscillator" mentioned in the Synopsis is thus pulsating at 1400 Hz -- but this would need to be rather faster if we wanted to do the Switcheroo for, say, 220V household power (common in Europe), or for power at 60Hz. Note that by comparison, the ordinary already-existing average switching power supply has perhaps a 150,000Hz oscillator.

And now for three different and rather important finicky details. First, because I am describing a system in which power flows through a transistor into a capacitor, the Electronic Engineer who wants to construct a Switcheroo Power Supply has to take into account something known as the "voltage drop" caused by the resistance of a transistor. See, transistors are not a perfect switches, the resistance that one offers to a current flowing through it means that 12V going in might, when it comes out, have the effect of only being at 11V. So, to compensate (not only because transistors are in the circuit as the capacitors are charged, but are also in the circuit as the capacitors are discharged), it could be that the actual voltage-incrementation by which we seek to charge the capacitors will be more like 14V than 12V. Yet that value depends on the quality of the transistors used (some offer less resistance than others), and leads to messy calculations and fractional voltage-values that are beyond the scope of this description. I am going to ignore it completely here, and stick with simple 12V increments, although in practice it cannot be ignored.

The second finicky detail is that because the wave-form of AC power is a special curve known as a "sine wave", it is an unfortunate fact that a simple division of that wave into equal time slices is not going to give us equal increments of 12 volts. For this reason, among others, it might be concluded that the Controller of a Switcheroo Power Supply is going to be a rather sophisticated device. For example, one way to deal with the time-slice problem is to use finer units of time (a faster oscillator), and then use different quantities of those slices for different portions of the wave-form, when sequentially/temporarily activiating Transistors 1, 2, 3, ... N. Of course, if the Controller is going to be fancy enough to accommodate a wide range of power sources, and variations on those sources such as brown-out conditions, then also handling different quantities of small time-slices is not such a big deal.

The third finicky detail regards the great likelihood that whatever desired output voltage we want, there won't be an exact way to divide that into the supplied input voltage (14x12V=168V, not the 170V mentioned earlier). This is OK; we simply ensure that when Transistor N is activated, it accommodates the final possible WHOLE desired voltage increment -- and after Transistor N is switched Off, the final fractional/momentary rising voltage is simply ignored.

OK, when the power wave has reached its maximum voltage level, the Controller now has to properly handle the Switcheroo as the voltage level starts to drop. Consider this sequence:
=
Z. 84v-96v
=
Y. 72v-84v
=
X. 60v-72v
=
This represents three adjacent 12-volt ranges, with a crude capacitor-symbol on either side of each range. When the AC voltage level RISES in Range Y, 72v-84v, we want the capacitor between ranges Y and Z to be getting charged (to 84 volts above the bottom of the connected series of capacitors). But when the AC voltage FALLS in Range Y, 84v-72v, we want the capacitor between ranges Y and X to be getting charged. See, that second capacitor originally received some charge during the voltage-rise up to 72 volts -- but there is no guarantee that it was fully charged during the short time the appropriate switching transistor was activated. So, when the overall voltage curve is dropping, as long as it is still ABOVE 72 volts, that second capacitor can receive charge UP TO 72 volts, just as we want it to have. That is, if a capacitor actually does get momentarily excessively charged, then because AC power systems are quite forgiving in terms of current flows, we can be sure that as long as the appropriate switching transistor is still connecting that capacitor to the Line WHILE the voltage is dropping to the desired level, the capacitor will discharge back to the Line almost all the excess, and end up close-enough to the desired level, when the transistor is switched Off. (Yes, I know there are ways of pre-computing the "capacitance" of a capacitor, such that we know it will become fully charged during a certain period of time. But remember these capacitors are part of a system that is expected to accommodate different voltages and frequencies, and a particular capacitance that is perfect for one such combination will either be too much or too little at another. Also, because we have two opportunities to charge a given capacitor during one full AC cycle, there is no reason not to take advantage of both those opportunities.)

What all that means is, although we turned Off Transistor N while the voltage level was rising that last fraction to its maximum level, we will begin the falling-voltage portion of the cycle by turning On Transistor N right away. And, during the final 12volt-drop of Line Voltage, down to the minimum, NO transistor in the 1, 2, 3, ... N series is connecting any capacitor to the Line. At this time we can now switch the entire "A" group of transistors Off, thereby breaking all the connections that gave us a series of capacitors. This means we are now free to activate the "B" group. However, before we get to the consequences of having just done that, let's consider Side II of the Switcheroo Power Supply, which is basically a duplicate of Side I (merely mirror-imaged in the schematic). Although not marked, it also has an "A" group and a "B" group of switching transistors. After switching On the Side I "B" group, we can now switch OFF the Side II "B" group, and we can follow that by switching On the Side II "A" group, in preparation for letting the next AC cycle charge the capacitors in Side II, in exactly the same way as has been described for Side I.

The Switcheroo complete, we now examine the power flowing out of the capacitors in Side I, courtesy of its "B" transistor group. Individually, each capacitor possesses 12 volts of charge, along with a reasonable quantity of stored current. Like a small battery cell, we could hook up a gadget that needs 12V of power, and each capacitor could run that gadget for a short time. However, because of the way the "B" transistor group connects the capacitors, all of them are now working together (in "parallel"), to provide 12 volts of power. Note that they only need to provide this power for 1/50th of a second, because after that, the capacitors in Side II will be charged! This means that quite a large amount of current can be drawn from a Switcheroo Power Supply, just as if a transformer had been used to change 120V to 12V. But the advantage is that here the output is pure Direct Current, while the output of a transformer is still Alternating Current.

Now, I do know that AS power is drawn from a capacitor, the voltage-level of that capacitor drops. This effect can be minimized by using fairly high-capacity capacitors -- and of course we are aided by the fact that power is only drawn for 1/50th of a second, before the capacitors get re-charged. That is, once sufficiently charged (consider 50 charging pulses per second, before any load is attached), only a portion of that charge would be used to supply the load, and so the voltage level of the linked capacitors would only drop slightly, before getting re-charged. Thus I am fairly sure that a Switcheroo Power Supply can directly provide a sufficiently stable output for most any device that needs low-voltage DC, without additional voltage-regulation circuitry. Not to mention that in places like the USA, where power is supplied at 60Hz, the capacitors only have to supply power for 1/60th of a second, before getting re-charged.

In closing I need to mention that it is not necessary to restrict a Switcheroo Power Supply to a single type of output voltage. Depending on how long each capacitor is charged, and how they are connected together when being discharged, SEVERAL groups of capacitors can provide output-power at different voltages. For running a computer, one group could output 12V, another group could provide 5V, still another could provide 3.3V, and so on, so that a computer's motherboard need not have any power conversion circuitry of its own. Indeed, depending on how the capacitors are connected RELATIVE TO EACH OTHER, a group could provide -12V, another could provide -5V, and so on!