Energy harvesting is the art of extracting useful power from energy sources normally deemed too weak, too dilute, or too impractical to be usable. I have already presented an example of such a “harvester” (“Trick A BJT-Based Converter Into Starting At Only 250 mV DC”). This time, I will deal with the other extreme—the circuit presented here can convert voltages in the 1- to 30-kV range into “normal” electronic voltages, at power levels ranging from milliwatts to watts.
This presents interesting design challenges regarding the choice of the architecture and active element, and the method to get the circuit started. It’s obvious that none of the traditional solutions can be adapted here. Transistors don’t go much higher than 2 kV, valves require heating power, and multilevel thyristor systems are suitable only for big industrial or railway applications.
The active element selected for this design is a very old classic: a spark-gap. How can this icon of 19th century technology be put to good use in a modern circuit?
Let’s forget any preconception and examine the facts about spark-gaps:
• Their voltage can be readily modified by changing the spacing of the electrodes.
• They require no standby power, are self-triggering (when the breakdown voltage is exceeded), and self-extinguishing (when the current falls below the holding current).
• They have good efficiency—the off-state current is negligible, and the on-state voltage could be anywhere between 100 and 500 V. This may look large, but with a supply voltage 20 or 40 times larger, this is negligible.
• They are robust (no electrostatic-discharge precautions required).
• They are inexpensive.
The properties of the spark-gap make it an ideal candidate for a relaxation oscillator (Fig. 1). The high-voltage source is symbolized by a current generator. This approximates most practical high-voltage generators, which have a large internal resistance.
The role of R1 is to isolate the oscillator from the external world. It minimizes the egress of interference, and avoids the generation of high peak currents in the cabling that may arise if the power source is capacitive.
The input current charges C1 until the spark-gap reaches its breakdown and discharges it into the primary of the high-voltage transformer. The resulting pulse is transferred to the secondary, and then to the storage capacitor C2, via a Schottky diode.
An additional element, the leakage inductance of the transformer (not visible on the schematic), has a very important role. This inductance spreads the current pulse in time, and reduces its peak value to levels acceptable to the diode and the capacitor. This is excellent news, because it also simplifies the transformer’s design. Even with this inductance, the peak currents are rather hefty for such a low-power circuit, typically between 10 and 100 A. This is why such a large diode is used, together with an organic semiconductor (OSCON) cap. The slightest ohmic losses have a dramatic impact on the efficiency.
Here, only one diode is used, because the input polarity is assumed to be known and constant. If this isn’t the case, a full-wave rectifier must be added, either at the primary or at the secondary, depending on the nature of the power source. If it has a low frequency and a current high enough to charge the oscillator’s capacitor in one cycle, the bridge can be placed at the secondary. If the current has a higher frequency and requires a number of cycles to charge the capacitor, a high-voltage bridge must be used upstream from the converter.
To maximize power and efficiency, the output must be adapted, i.e., the loaded voltage must be half of the open-circuit voltage. In fact, it should be a little less, to account for the losses, imperfect coupling, etc. In this case, the transformer ratio is 1/100 and the spark-gap voltage is adjusted to 4 kV, yielding a theoretical 20-V output. The actual optimum is reached at 17.5 V.
The capacitor must not only be able to withstand the raw input voltage, but also the pulsed mode of operation. General-purpose film or ceramic multilayer caps are unable to withstand repetitive pulse currents, and their metallization will fail rapidly at the junction with the terminals.
Traditional disk capacitors have a much more robust electrode construction, and are thus suitable. They tend to have rather high dielectric losses, due to the nature of their material (typically Y5U or something similar), but considering their low cost, the slight reduction in efficiency is acceptable.
The transformer was based on a RM6 core, made of generalpurpose 3B ferrite material. Good high-voltage practices have to be adhered to for the primary winding. The voltage limits of 1 to 30 kV are only indicative: Below 1 kV, the spark-gap has to be adjusted with micometric accuracy; over 30 kV, the problems caused by corona and similar effects become intractable.
The pulsing part of the converter is a powerful interference generator, which requires careful shielding to avoid disruption to other equipment. The construction must not be airtight, though. The spark-gap generates ozone and corrosive nitrogen compounds, and if they’re allowed to build up inside an enclosure, they will corrode metals and degrade insulating materials. A pure nitrogen environment might be a solution.
The spark-gap must not conform to the stereotype of the two pointed tips. Ideally, the electrodes should be rather massive, smooth, and slightly convex. If a continuous operation is envisioned, their material should be hard and refractory enough to resist erosion, such as tungsten.
On a 1-k load, the prototype delivered 306 mW from an input power of 480 mW. This power was estimated by looking at the frequency of oscillation, which is 153 Hz in this case (P = 0.5 CV2F). This mode of calculation was adopted in order to remove the characteristics of the source from the equation. These figures put the efficiency at 63%, which is quite respectable for such a lowtech circuit.
Possible power sources include primarily piezoelectric or pyroelectric elements, but there are other possibilities. For example, you could build a “Power Thief ” (Fig. 2a). An antenna collects the stray electrostatic field from a high-voltage power line and feeds the high-voltage rectifier. This is similar to the fluorescent-tube trick, but gives out usable electrical power.
In theory, it should also be possible to exploit the atmospheric gradient (Fig. 2b). This hasn’t been tested, and if attempted, great care must be exercised to avoid being struck by lightning.
The circuit could also be powered by an electrostatic machine, but only for demonstration purposes. Such machines are rather inefficient when compared to electromagnetic generators, and their output current doesn’t exceed tens of microamperes.