Your wall socket is 50Hz or 60Hz, 120 or 240 volts rms ac, but most importantly it is a voltage source. That is, the voltage remains constant at, say, 235 volts, regardless of whether you’re drawing 0 amps or 20 amps from it. Actually, this is a little bit of a lie; the voltage does reduce somewhat as you draw current from it. Maybe the resistance of the wire in the wall is 1Ω, so if you’re drawing 20 amps, the voltage falls to 215 volts. And as your house draws more current, the voltage on the whole circuit to your house from the transformer falls a bit, though probably not as much as the voltage drop from the wires in your wall. But the voltage is relatively constant.
This configuration gives rise to a number of phenomena we’re familiar with, not all agreeable:
None of these are inherent to electricity in general. If we used an RF current source rather than a extremely-low-frequency voltage source, all of the above items change.
First, what is a voltage source?
A voltage source is an energy source with a constant voltage across the load, regardless of the current. Its output impedance is zero. The voltage of your wall socket is a close approximation of a voltage source; its impedance is on the order of 1Ω. It has a limit to how much current it can supply — even leaving aside the fuses and circuit breakers, to draw 240 amps from your 1Ω 240V source, you would need a short-circuit load, which means that none of the power actually reaches the load, and also you cannot reduce the impedance any further in order to draw more current.
A current source, by contrast, is an energy source with a constant current, regardless of the voltage. Its output impedance is infinite. A photovoltaic cell in constant light at a voltage well below its maximum output voltage is a close approximation of a current source, because each photon it converts generates an electron-hole pair regardless of the voltage; I think its impedance is on the order of 1MΩ. Any current source has a limit to how much voltage it can supply — if you open-circuit the photovoltaic cell, its voltage will rise to about 0.7 volts and stay there. At that point you cannot reduce the conductance any further in order to get higher voltages.
This is one of many applications of a certain duality in electronics in which we interchange serial and parallel, voltage and current, resistance and conductance, and I think capacitance and inductance, and maybe frequency and period, and everything comes out the same.
The study of electrical currents began with batteries, which are pretty close approximations to voltage sources. Current sources, on the other hand, are generally thought of as artificial or theoretical constructs. But in today’s world of power electronics, our voltage sources are usually complicated circuits, and we can build current sources just as easily. Stick-welding machines and TIG welding machines are commonly designed as high-power current sources, under the name “constant-current source”, or “CC”, for example.
It turns out that if you use a current source rather than a voltage source for your mains power, and run it at a higher frequency, there are a number of interesting results. In particular, all of the bullet points in the first section above become false, and there are a number of potential safety, cost, and complexity benefits — although of course we won’t know if these work out in practice until we have real experience with them. Current sources defy the expectations we have built up over centuries of working with voltage sources — among other things, about what is safe and what is dangerous. To take the simplest example, short-circuiting a high-power current source is totally safe and actually the right way to turn off whatever it’s powering. Open-circuiting it, by contrast, can produce a hazardous voltage.
The extremely low frequencies we’re accustomed to using date from the late 19th century, when they derived from the frequencies of rotation of the steam-driven generators used to produce the power. This approach is still used today, with generators’ rotation synchronized to the grid frequency before putting them online. But supplying power at higher frequencies has a number of substantial advantages in safety, electronic noise, and equipment size, cost, and complexity.
In addition to potential uses in houses and industry, this system should also be useful in spacecraft, where wiring harnesses commonly contribute an alarming amount both of very expensive mass and of very expensive power consumption.
Consider a system supplying one ampere at 32768 Hz, the frequency of a watch crystal, to every wall outlet, with current-source compliance up to 1000 VAC.
The outlets on a circuit are wired up in series — the current flows first through one appliance, then another, then another. When no appliance is plugged in, a short-circuit shunt is mechanically imposed across the two terminals of the outlet, so current flows through the outlet with no significant voltage drop. When all the outlets on a circuit are short-circuited, the total resistance of the circuit might be 1Ω, so the 1A will produce 1V of voltage drop and 1W of power consumption.
If a 10W appliance is plugged in, once its contacts have made contact with those of the outlet, its plug mechanically removes the short-circuit shunt, introducing the appliance into the circuit. The appliance’s impedance of 10Ω produces a voltage drop of 10V across the plug (once the current source’s output voltage rises to match), delivering 10W to the appliance.
A switched outlet or appliance has the switch in parallel to it. When the switch is closed, there is no voltage across the appliance, and consequently no current through it, or a tiny residual current due to the tiny voltage drop across the closed switch. Opening the switch forces the current to flow through the appliance instead.
A short circuit between an outlet and the neutral wire removes power from the appliances downstream from it, but causes no other problems. An open circuit, however, could cause a hazardous voltage, and therefore safety measures must be taken to shunt current past the open circuit, just as fuses or circuit breakers must create an open circuit in a voltage-source system to interrupt hazardous currents.
All the normal wires are the same tiny size, 24 AWG, 510 μm in diameter, which is enough to carry 1A without getting warm, but not so thick as to waste any significant copper due to the 400-micron skin depth in copper at 32768 Hz. Higher-current internal ac wires in some appliances need to be made of copper tape of less than 1 mm thick. The thickness of the wire does not depend on the length of the run, because the voltage drop is inconsequential; the current source will automatically raise the voltage to compensate, so that the load will receive the full 1A anyway, as long as the run isn’t so long as to approach the current source’s compliance limit (1000 V, thus 1000 Ω at 1 A.)
Physically, the outlets are strange. The two prongs are each 20 mm long, with their first 15 mm insulated, as in France; they are positioned a rather large 40 mm apart, with a ground prong in the middle. Behind the outlet, the attachments for the tiny wires are on strange stalks that curl back around. All of this is for high-voltage safety in the unlikely case that a single outlet is called upon to supply the full 1000 V limit of a circuit; it needs the creepage allowance not to form a conductive, but high-resistance, path between the electrodes.
Despite the possibility of using a single wire from one outlet to the next, in fact the return wire runs alongside the hot wire as a lightly twisted pair for the whole circuit in order to reduce EMI. It isn’t connected to anything until the last outlet in the string.
Each circuit is fed from a separate transformer, a bit smaller than the usual circuit breaker, on a higher-voltage-compliance constant-current “bus”, which is wired in series, not in parallel as usual. The transformers are center-tapped, and the center tap is grounded, so the net voltage surrounding the twisted pair of wires is zero, and of course their net current is also zero. This reduces EMI and keeps the voltage from any point to ground to a minimum. Each transformer also has a safety shunt attached to it.
First, in at least a couple of ways, the system I’m describing here is potentially more dangerous than the traditional system. 240V is often not enough to kill you, depending on how dry your skin is. 120V almost never is. But one ampere through your arms is traditionally considered very likely to kill you, and the current source will happily apply 1000 volts to you if you are suddenly the only path through its circuit and that is what is needed to drive an ampere through you. 1000 watts applied to your body is also very likely to kill you.
Also, kilovolts can do surprising things that the more usual voltages do not do, jumping through air and burning tracks through dust on surfaces and whatnot.
However, there are a variety of ways this system is inherently safer, as well as a number of safety features that can be added.
First, the vast majority of circuits will not have hazardous voltages present on them at all, because they will not have hundreds of watts of load, so they won’t need hundreds of ohms of impedance. And even those circuits that do have hazardous voltages will generally not have hazardous voltage differences within a single appliance.
??? I’m conflicted about whether grounding the center taps is a good idea. Without that current path to ground, there would never be any hazardous voltages relative to ground in the system, only potentially hazardous voltage differences within the circuit. But without it, grounding an enclosure will not produce a detectable ground fault if a hot wire comes in contact with the enclosure.
As explained above, short circuits are not hazardous in a current-source system — the current source will only supply its usual current to them, rather than an unbounded current, and they will dissipate no energy. Open circuits produce a hazardous voltage but no immediate fire. The real danger is near-open circuits of a few kΩ, which could potentially dissipate a few kW.
I think one amp is not enough to sustain an arc in air (???), which would eliminate the usual risk of arcing, despite the high voltages.
Using RF is a safety advantage because currents at these frequencies cannot (???) penetrate deeply into the human body, instead staying on the skin, and so they cannot cause muscular contractions (???) or cardiac fibrillation. They can still cause burns, and RF electrical burns are notorious, but that requires a lot more energy. Unfortunately, this system is capable of supplying that energy.
While it’s harmless and probably useful to include traditional series overcurrent fuses in the system — if the current deviates significantly above 1 A, you have a bad problem in your current source and it would be a good idea to disconnect from it — they won’t detect hazardous voltage soars or open circuits, which are the kinds of problems that could arise from electrical faults inside your house.
Grounding the cases of appliances provides a more effective safety measure against hot cases than in the traditional voltage-source system. In the standard system, a hot wire contacting the grounded case produces a ground fault, which fires the circuit breaker. In this system, by contrast, it shorts out half of the transformer and the upstream appliances, effectively dividing the circuit into two circuits of 500 mA each, joined at a shared ground point. The enclosure remains grounded and thus safe to touch; there are no sparks and no danger of electrical fire. However, nothing on the circuit will work properly until the faulty appliance is removed, because it will be getting a quarter of its usual power.
The simplest safety shunt against open circuits is just a buzzer-type SPST relay; its normally-closed contacts in series with a normally-closed pushbutton short the load, and its winding in series with the load holds those contacts open. If at some point the load goes open-circuit, the current through the load and the coil will cease, the contacts will spring closed, and the load will remain shunted out of the circuit until someone pushes the pushbutton†. This fails safe in case of coil failure and in case of power failure; it may be inconvenient to have to reset all your circuits after a power failure, but it’s better than having to replace all your appliances. It may be possible to tune this circuit to reliably detect the ground-fault half-current case, too.
The other problem the above circuit has is that it doesn’t actually limit the voltage; it just responds to the source’s inability to sustain its current in the face of overwhelming resistance. In the case where the source is actually capable of hitting 10kV, it might fail to activate because the deadly overvoltage has burned a track across the outlet and is efficiently heating it to incandescence.
A simple truly-overvoltage-driven alternative would be a large gas tube. A fluorescent lamp tube would sort of work, but its discharge maintenance voltage is high enough (in the range of 100V) that it would probably cause major damage to the faulty load.
A fully-solid-state alternative might be a diac in series with a small inductor and a NC pushbutton, driven from a bridge rectifier in parallel with the load. If the diac goes into conduction due to overvoltage, it should crowbar the circuit rather effectively until someone presses the button; the inductor keeps it in conduction as the voltage crosses through zero 65536 times per second. This should give you under 10V across the circuit. (e.g. the Littelfuse K1400GURP SIDAC can handle an amp steady-state and will crowbar down to 1.2 volts, with a breakover voltage of 130–146V; you could use 7 of them in series, giving you a breakover of 910–1022 volts, and a voltage of 8.4 volts, plus the bridge and inductor voltages. It costs 37¢ in bulk.)
An alternative relay circuit would use a latching relay activated by current through, say, a calibrated spark gap, MOV, diac, or gas tube. It could latch either mechanically or, by virtue of energizing its own coil, electromechanically. This last has the dubious advantage of resetting automatically after power failures, including after being shunted out of the circuit by an upstream shunt.
† At which point the circuit potentially goes really open-circuit if the load still isn’t repaired, thus causing further safeties to trip further upstream, so maybe this design is inadequate. Or maybe you just need two of them, maybe controlled with a make-before-break pushbutton that opens one circuit after closing the other.
Everything is topsy-turvy in the current-source world, but it’s overall simpler. It’s fine for things to short out when they fail; it’s problematic and possibly dangerous for them to create open circuits, for example by breaking the filament of an incandescent lightbulb. If your lightbulb filament breaks, it will trip the safety shunt, and you will probably want to turn off all the lightbulbs on the circuit, reset the safety, and then turn them on one at a time until you figure out which one is tripping the safety.
Resistive heating is simple. 1Ω is 1W. To get a given temperature under given cooling conditions, you use a given filament; you get the same temperature no matter how much of it you use, but more filament gives you more power. You should probably mount it on some kind of conductive backing with light insulation in between so that it becomes a short circuit if it melts, not an open circuit.
More resistance gives you more power; this is precisely backwards from the situation with voltage sources, where load resistance and power are inversely proportional.
3V 1.5W LEDs can be driven directly from the regulated 1A with no resistor. Smaller LEDs will burn out; larger LEDs will only shine 1.5W. Large fluorescent tubes of about 100W can, too, with no ballast or starter, as long as they can cold-cathode start at a low enough voltage, and fast enough to keep the safety shunt from tripping. 50W tubes should work if half-wave rectified — with one diode in series and another opposite diode shunting the combination. But most LEDs and fluorescent lights will require some kind of power supply.
Voltage-step-up transformers can be used to step down the current for some of these devices, such as small LEDs. Voltage-step-down transformers are probably useful to step up the current for larger resistive elements, so that they can use thicker wire.
Low-voltage low-power power supplies are really easy; a tiny 1A 47-microhenry inductor has an impedance of ωL = 4.8Ω and thus a voltage across it of 4.8V (ac rms), which is 6.8V peak. Stick a silicon bridge rectifier across that, dropping 1.4V, and you can charge a capacitor to 5.4V dc at any current up to 100mA average or so (at which point the inductor voltage starts to drop because you’re stealing its current). For 10% ripple at 100mA, the capacitor itself only needs to bear the 100mA for 15.3 microseconds without dropping more than 0.5V, which means 3 μF is adequate; an LC filter would allow you to use a smaller capacitor and also get better stability. If you want a regulated voltage, you might want to use a slightly larger inductor to get more like 9V peak, then drive a 7805 off it. Then it would be okay to steal more current from the inductor.
In any case, you won’t have any 50Hz or 60Hz ripple. You’ll have 32768Hz ripple, plus some harmonics like 65536Hz and 98304Hz, which are a lot easier to filter out.
Combining the two above, you can quite reasonably use a voltage-step-down transformer to feed a 22μH inductor 5A, then rectify, filter, and regulate, to get a 60% efficient 1A 5V regulated USB power supply in five components.
In the simplest case, of course, you can get a voltage by using a resistor rather than an inductor, but it will dissipate power. Or you could use an unpolarized capacitor of a few μF.
Inrush currents from hooking up capacitors to line voltage no longer exist. Instead you have inductive voltage spikes when you switch an inductor into the 1A circuit. You may want a snubber network to tame this, but that may be unnecessary — you aren’t going to design in an inductor with more than 1kΩ of reactance, which means more than 10mH. ½LI² = 5 mJ, which may not be enough energy to cause any real problems.
Switching power supplies are still a thing, despite the greatly increased convenience of old-style transformer power supplies, and they need no inductors. If you were using dc instead of ac, a switching power supply amounts to switching between a dead short and a capacitor-diode series. You run your voltage-regulated circuit off the capacitor, and the diode keeps the capacitor from discharging during the time that the power is turned off with a short. You can do this with ac if you use a bridge rectifier or just a reverse-protection diode. For lower-power supplies, you could just use a rectifier diode to generate a voltage dissipatively and switch that voltage onto a capacitor at times.
Squirrel-cage induction motors should be barely feasible: 800 poles around the stator should give you 164 revolutions per second, which is 9840 RPM. The “squirrel cage” should probably be a millimeter-thick copper or aluminum sheet. I don’t think you can use laminated electrical steel in the rotor at these frequencies. On the plus side, the starter capacitor can be quite small.
Universal motors are not feasible at all without rectification.
The higher peak voltages allowed by a current-source system, as well as the automatic compensation for wiring losses, allow us to use much lower maximum currents and therefore thinner wire than in the traditional system. Instead of building the copper for the worst case and the insulation for the average case, it allows us to build the copper for the average case and the insulation for the worst case. Insulation is much cheaper than copper.
24AWG copper wire is 510μm in diameter and 84Ω/km, and it’s recommended for currents up to 3.5 A. 84Ω/km means 84mΩ/m and thus, at 1A, 84 mW/m, which is not totally insignificant in terms of heating, but not really dangerous either. It means you can go over 5 km round trip with this wire before you hit the source’s 1000 V compliance limit.
A bigger issue at these frequencies is likely to be stray inductance. 1Ω is only 9.7 μH. The twisted pairs should reduce this problem, but they won’t eliminate it.
Copper’s density of 9.0 g/cc means this works out to 1.8 grams per meter of wire, or 3.6 g if you run the neutral wire next to the hot one. But that doesn’t count the insulation, which needs to be safe at 1000 V.
100 turns of this wire on an inductor work out to a coil, say, 5.1 mm long and 5.1 mm thick, for maybe 15 mm of diameter on the whole inductor. Transformers might be a little larger. None of that takes into account the insulation, though, so probably the reality will be several times that.
1000 V probably needs over 16 mm clearance between exposed conductors and 64 mm creepage distance, although I should look up the standards.
As explained in the “appliance design” section, the higher frequencies allow the use of much smaller capacitors, inductors, and transformers, which dramatically reduces cost and weight. The current-mode design allows us to get whatever voltage we want by adjusting a reactance; the use of rf ac allows us to get whatever current we want with a tiny transformer.
are too high for humans to hear, so they cause less problems with audio equipment; and they