capacitive filtering of power line transients without power loss

Question

I've been researching so-called "surge capacitors", which are claimed to augment surge protection devices (SPDs). The claim is that SPDs usually have at least a 5-10 ns response time, and require a voltage increase of 10% or more, whereas surge capacitors are "always on" and can react to any increase in voltage immediately. The trade-off, of course, is that surge capacitors cannot handle very large currents, which is why they are marketed to only augment (and not replace) SPDs.

One brand of commercially available surge capacitors is Delta, and on their webpage explaining these devices, they say:

The continuous current flowing to a surge capacitor does not cause a waste of electricity.

They claim that an "internal automatic discharge circuit" will "absorb the voltage spike, then return it to the circuit as harmless low voltage current". They have a simple graphic on the previously noted webpage that shows the transient energy being added back to the load at close to the expected 60 Hz power frequency.

I am familiar with how RC or LRC filtering circuits can shunt unwanted high frequencies to ground, but I thought this would result in some power loss (or "waste" of electricity), due to some non-zero current bleeding to ground. Can anyone provide a circuit diagram for how maybe this capacitive filtering circuit could work without wasting any power?

Answer 1

Well, without any detail of what device is in question, or applicable patents if any, there are two typical possibilities.

First, consider the equivalent circuit used to test surge:

^{simulate this circuit – Schematic created using CircuitLab}

C1 roughly corresponds to the spare charge leftover from a lightning direct or induced strike. R1-R3 and L1 set the rise time, duration and discharge of the waveform, as well as the effective impedance of around 2Ω.

The mains distribution system uses SPDs (big stacks of MOVs, fused with an explosive bolt to rapidly separate the protector if it overheats or otherwise fails), which being up on poles, have significant stray inductance to ground; ground itself is also lifted by such a strike, and nearby loops (hence: induced lightning) pick up the EMP anyway, so the whole process is very "squishy" to analyze in the vicinity of an actual strike. At some distance, enough SPDs have been encountered that the peak voltage is within expected limits (maybe 3-5 times nominal line peak voltage), within the ratings of distribution transformers, and the output is simply a transformed-down surge. Or the low-voltage side experiences induced lightning directly, but in either case a few kV is a typical maximum limit for differential surge.

As there's some distance between the source (or last low-impedance point anyway, SPD or otherwise) and load, there is some inductance, which also sets the approximate impedance as this is a dynamic event. Thus we arrive at a circuit something like the above (it's called "combined" because it has a 8/20us response for current into a short circuit, and 1.2/50us response for voltage into an open circuit; both extremes of waveforms, and the intermediate cases, find use in EMC testing).

Note that let-through is not generally concerned about nanoseconds, but microseconds. ns's are fairly trivial to filter: such signals (100s MHz) "don't like to stay in wires", i.e. they are easily lost to radiation, dielectric absorption, mutual inductance, etc., so, say an arcing process on a pole some blocks away, doesn't transmit all that much to a given customer, at these frequencies. Granted, it might still be significant voltage as RF goes -- tens of volts perhaps, but just not full-amplitude surge voltage. After going through distribution cables, transformers and wiring, bandwidth is limited to some kHz, and maybe a strong surge still comes through in the low-µs time scale, and that's about it.

Mind, nearby strikes can do all sorts of nasty things, in terms of bandwidth or risetime as well as amplitude. But, that's very much getting into a different regime altogether, and demanding very high reliability (or any reliability on very unfortunate sites e.g. antenna/tower, skyscraper, mountain top, etc.) requires different solutions (yard grounding grid, facility shielding, more and bigger SPDs and filtering, etc.). So we tend to consider surge protection within the extents of commercial equipment and IEC and related standards.

ns scale let-through is also produced by discharge type SPDs, like GDTs (gas discharge tube) and spark gap structures. In this case, it takes some time for the gas to ionize and become conductive. I suppose a facility might be protected by such structures as a first-line defense, and then the ~ns scale overshoot would be easily found within the facility itself. In any case, it's very easy to filter frequencies this high (little capacitance or inductance is required), it's harder to shield it (again, because it "doesn't like to stay in wires", which means equipment near such (unfiltered) wiring can experience problems too).

Anyway, back to the circuit: MOV1 is a diode, crudely modified to give a MOV-like characteristic; CircuitLab doesn't have a MOV component and I can't paste in a SPICE model. If you run the simulation with MOV1 or C3 deleted, or replaced with a say 1mΩ resistor, you can see the above waveforms as claimed. As shown, we get waveforms such as:

Compared to the raw 1kV peak, both seem to have significant effect. 10uF obviously isn't very large in relation to the 8uF source, but we can make it much larger, 100uF say, without running into much problem at mains frequency.

And so we come to one possibility: just load it down with enough capacitance to quash the surge.

The other possibility is a more subtle approach: clamp the surge into a large capacitor, but use diodes to do it. This draws a small idle current from the line (costing some power that the plain shunt cap might not), but can keep the peak voltage quite low: maybe some 10s of V above nominal peak. The capacitor normally stays charged to nominal peak voltage, and gets charged up a couple volts more after an event. This doesn't really need to discharge back into the line (and also is supposed to skate over multiple cycles, so can't exactly return energy in sync at any given point anyway), so I suspect isn't what they're getting at here.

As for power dissipation, that's simple: a direct capacitor (and nothing else; their description would seem to be applicable to this, where the "return energy back to the load" circuit is a trivial wire link) always draws and returns some energy from the mains. When the draw and return are phased very accurately (close to 90°), the average loss is near zero. An ideal capacitance does this at exactly 90°; real capacitors have some ESR, which reduces the phase shift and dissipates power.

As for power or energy of the surge, note that since surge has a finite source impedance, it can be momentarily shorted out, reflecting power back towards the source. A passive capacitor can do this by superposition, and a TVS or MOV can do it in part, depending on relative amplitude and clamping voltage of course.

Capacitors actually can have extremely high current ratings: comparable values, at much higher voltages, are used in rail guns and other high-energy apparatus, for example.

It's entirely possible that the referenced sheet is pure marketing waffle for an utterly conventional power factor correction capacitor: these types are typically a simple and cheap polypropylene film and aluminum foil, rolled up together, and impregnated with dielectric oil. Connections are made with a couple tabs crimped/welded to the foils, and brought out to terminals. These connections are the weakness of the oil-can capacitor: they're more or less dimensioned for adequate ratings at mains frequency, so aren't too useful at high frequencies (pulsed or otherwise). They also serve a protective purpose: if arcing occurs in the capacitor, the gas generated causes the casing to bulge, lifting the lid (with terminals) away from the roll, breaking the tabs off. A sort of pressure fuse. (At least, in applicable types of course!)

Put another way: they may simply be trying to pawn off regular PFC caps as surge protection. This isn't a special function, it's just a natural consequence of how it works.

Answer 2

An ideal capacitor is has only reactance and no resistance. So the ideal capacitor is only an energy storage device, rather than energy dissipative. So when considering "active power" (aka "real power"), a pure capacitor load on an AC supply draws current 90 degrees out of phase with the voltage, which results in positive instantaneous power for 1/4 cycle as capacitor charges and then negative instantaneous power (outputting power back to the supply) for a 1/4 cycle as the capacity discharges. Best article I found for some nice diagrams.

I like your idea of comparing this product to RC filters, since placing a low ESR capacitor across the AC supply line is really an RC filter with very low R. "Surge Capacitor" seems like just a marketing term, because what they are suggesting is simply adding capacitor as a filter to suppress high frequency noise, which may or may not have originated from to any type of voltage surge/spike event. Read this for another take on capacitor application more appropriately called "AC Filter Capacitors".

I believe adding such caps will provide less noisy voltage to the load, at the cost of increased out-of-phase current and tiny power dissipation from the ESR of the capacitor. But if the building voltage temporarily jumps to 480Vac due to distribution transformer fault, these capacitors won't protect anything, unlike the transient voltage clamp devices (SPDs per above terminology).