Web site created by Donato Borello, Amsal Karic, and Randal C. Nelson



Power System Details

The project specified an electronic artifact that would operate for 1000 years. The sytem did not necessarily have to do something continuously, but should operate on demand during those 1000 years. Specifically, it was not to be a "time capsule" intended to be opened once at the end of the period. An electronic device by definition requires a source of electric power. It was explicitly specified that this be intrinsic to the object, and not dependent on external infrastructure such as a replaceable battery or a power outlet.

The electronic mechanisms we proposed required at least a few tens of microwatts to operate. This was scaled down after we failed to find a power souce that could supply the few milliwatts our initial designs required. Coming up with a durable electric power supply that could provide even this proved to be one of the more challenging aspects of the project.

Energy Source

We considered a number of possible sources of electrical energy.

Single use batteries seemed, prima facie, out of the question. They just don't seem to store enough energy to do anything interesting for long enough. Also, they tend to degrade chemically over a period of decades at most. We were thus fascinated to discover that there is a battery-powered device, the Oxford Electric Bell that has been running for over 170 years. This experiment is an electrostatically powered ocsillator where a small metal ball swings as a pendulum between two oppositely charged metal bells. The charge is maintained by a pair of high voltage dry cells, probably Zamboni piles where a stack of paper-separated discs of zinc and silver, produces power via slow reaction with oxygen in the atmosphere. The output is in the kilovolt range with a current of a few nanoamps. The cells that have kept the Oxford bell ringing are far too large to use in our pocket-size device, do not provide enough power to do much, and require specialized circuitry to bring the voltage down to a useful range. Nevertheless, a most curious device.

Another source of energy that has been used to power mechanical devices long-term is the expansion of air or some other substance, either through variation of barometric pressure, or variation in temperature. Devices of this sort power both the Beverly Clock and the proposed 10,000 year clock (currently under construction). Unfortunately, these mechanisms are intrinsically bulky (to provide sufficient expansion volume), and both complex and delicate mechanically. We would also need a way to convert mechanical power into electrical energy. On a different scale, thermally driven expansion of some solid, coupled with piezoelectric crystals, a high-voltage reduction circuit, and storage of electric energy might do the trick, but bulk for sufficient power would not fit in our space, and building such a device promised to be difficult given our skill set.

Radioactive decay is another potential source of long-term power. In principle large amounts of energy can be stored in a small volume. Getting at this energy is a problem. Thermo-electric generators using radioactive heat sources are used to power deep-space probes. Unfortunately the operating temperatures needed to generate usable power are much higher than we would want in our compact device, and producing them requires significant amounts of highly radioactive materials. We can't get them, (and don't have the ability to work safely with them if we did).

There are also designs for making nuclear batteries that convert radiation from radioactive isotopes more directly into electricity. There are commercially available devices based on betavoltaic technology. Unfortunately for us, these use relatively short half-life, low-energy beta emmitters (e.g. tritium) and would supply power for at most a few decades. Technology using longer-lived, higher energy alpha emitters does not seem well developed. (The 1600 year half-life of Radium-226, famously discovered by Marie Curie, would be ideal for our device, but as it produces radon and nasty gammas it is not so ideal in other ways. A better bet all-round would be Americium-241, half-life 432y.)

Simple mechanical generators are sometimes used in self-winding watches and other devices such as "battery-free" flashlights and radios. The simplest design is a permanent magnet, usually contained by a tube, that slides back and forth inside a coil when the device is shaken, or more gently rocked in some designs. The resulting (alternating) current can be rectified and used to charge a capacitor or rechargable battery. Most of the mass-market devices of this sort contain a rechargable battery, so "battery-free" is a bit of a misnomer. There are also wind-up designs. We rejected this approach mainly because having to shake or wind the artifact to get it to work seemed counter to the intrinsically powered philosophy. But also, a winder, or a magnet rattling back and forth over a millennium would be likely to wear out.

The approach we ultimately settled on was to use solar panels to convert ambient light energy to electricity, and store electrical energy in an ultra-capacitor for use when no light was present. The energy source is not internal, but since light is present in most human-occupied environments, neither is active effort (such as shaking or winding) required. At most, the artifact would need to be placed in a better lit location. Solar panel technology is well developed, and increasingly mass-marketed, so we had that working in our favor. Long-lasting energy storage is less developed. Most commercial solar devices needing energy storage rely on rechargable batteries, which typically function for only a few years before suffering serious chemical degradation. We used electric-double-layer ultracapacitors, which are newer, less-well-tested technology, and have somewhat lower capacity, but have the potential to significantly outlast batteries.

Solar Panels

Pholtovoltaic power has been the subject of intensive research and commercial development in the last decade, and high quality components are available at reasonable cost. Direct sunlight at noon has a power density of about 1000 watts per square meter, or 100mW per square centimeter. Current silicon photovoltaics easily achieve 10%-15% efficiency, so direct sun should be able to provide 10mW per square centimeter of electrical power. A pocket sized object would have few square inches, 10 or more square centimeters, easily available for solar cell, which would provide 100+ mW in direct sunlight. So it seemed that even in diminished light, the few mW needed to run the proposed designs would be easily available.

Finding sufficiently small solar panels proved to be a bit tricky. The housing design allowed a maximum panel size of about 1.5 by 3.0 inches. Moreover, to avoid incorporating voltage-boosting circuitry, we wanted the panels to provide at least 5 volts. Most commercially available panels, since they are designed to produce power, are considerably larger than the space we had available. Small, individual cells, which generate about .5V, were easier to find, but we wanted to avoid assembling and encapsulating our own panel arrays if possible. We finally found some small, (37mm x 33mm) epoxy-encapsulated panels, rated at 6.7V and 31mA from Solarbotics, a small company that supports the micro solar-powered-robot hobbyist community. The panels were pre-mounted on fiberglass circuit board, and thus quite robust to start with. Two of them would fit on our design, and the specs suggested that plenty of power would be available.

We hooked up the panels and discovered that under bright office lighting, they would not power any of our test circuits. The voltage rating on solar panels refers to the open circuit voltage, and the current to the short-circuit current. We ran some tests on our panels and discovered that the open-circuit voltage under bright indoor lighting was only 2.5V, and the short-circuit current about .1mA. Definitely way out of the spec range. We had expected somewhat lower current under office lights, but not the dramatic drop in open-circuit voltage.

What we missed initially was the huge quantitative difference between bright office lighting and sunlight. Light power available to a solar panel is approximately proportional to the illumination measured in lux (lumens per square meter). Ordinary indoor lighting is about 100 lux, bright office lighting about 500, and direct sunlight close to 100,000. There is a factor of 200-1000 difference between indoor lighting and sunlight. Solar panel ratings are based on a model of direct sunlight, and when we took ours outside at noon, they met the advertised specifications. Indoors, the low open-circuit voltage indicated we were outside the range of the voltage-source plus series resistance model that can often be used for solar panels operated below their rated power.

A more accurate model of a solar cell is a light-dependent current source in parallel with a diode and a shunt resistor. Under usual, high illumination operating conditions, the voltage is set by the forward operating potential of the diode (which can be thought of as "turning on" at the rated voltage). At low illumination however, the diode is "off" and the voltage is determined by the current flowing through the shunt resistor. Indoors, our panels were, unfortunately, operating in this range.

We "solved" the problem by putting two cells in series, which gave us about 5V open circuit under bright office lights, and limiting the current drain to a few tens of microamps. This current limit seriously impacted our options for electronic functionality, as discuss in the electronics section. Also, if the lighting was very much below our bright office test conditions, the output voltage fell below the minimum needed for our circuit (about 4V). Equivalent solar cells with higher internal shunt resistances would be the best solution, but it is not clear they exist, or if they do, what other limits they might have. Another solution would be some sort of "store and go" charging circuit that would accumulate charge in a capacitor and then fire off a voltage-boosting circuit when a threshold was reached, to pump up the primary energy storage unit. Even this would not work below about 2 volts, and the quiescent drain needs to be below a microamp or two, so it hardly seems worth it.

Our ultimate "solution" also involved placing the artifact on display near a north-facing window, which admits enough light to power the system even if the office lights are left off. We have gained a new appreciation of $10 solar calculators. Maybe an armoured version will be the next project.

Energy Storage

We wanted to be able to keep the system running at least overnight, since artifacts in human environments will generally be subject to such periods of darkness. A day contains 86,400 seconds, 100,000 in round figures. 10 microamps for 100,000 seconds is 1 Coulomb, so we need capacitance in the farad range. The only current devices that can provide this in a pocket-size package are electric double-layer ultracapacitors. Some low-voltage electrolytics come close, but electrolytics have a tendency to fail catastrophically after a decade or two, especially if the device is not activated for long periods. Such periods of neglect seem likely with anything that stays around for 1000 years.

A bare ultracapacitor is a low-voltage device, limited to just over 1V. We needed to be able to store energy at a higher potential, since our circuit needs 4V just to run, and we wanted to avoid the complexity of voltage-boosting circuitry (although that would certainly be an option in future devices). To store higher voltages, several ultracaps need to be connected in series. There are well known problems associated with doing this. Differences from the rated capacitance mean that the voltage will not be distributed exactly equally. More serious, in a system that is repeatedly charged and discharged, small differences in leakage rates between individual caps can cause excessive charge to accumulate on individual elements of the chain, eventually exceeding the breakdown voltage and destroying the capacitor. If the failure is a short, a chain reaction can destroy all the other caps in the chain.

To prevent such problems, some sort of "balancing" circuitry is generally employed when capacitors are used in seried. The simplest solution is just to put a resistance in parallel with each capacitor. This is called passive balancing, and works well as long as the leakage introduced can be tolerated. If the leakage is excessive, more complex active systems that dynamically monitor and gate excess voltage can be used. Constructing all this, of course, is extra work for the circuit designers and makers.

Fortunately, ultracapacitor vendors supply pre-assembled and balanced units for higher voltage use. We obtained 1F units rated to 10V from the Tecate Group. These fit within a one inch cube and are available in actively and passively balanced versions The specific device numbers are the PowerBurst PB5-10 and PB5-10P. We initially used the passively balanced version with the idea that less complexity meant less that could go wrong. Unfortunately, the internal resistances used discharged the capacitors with a time constant of about 3 hours, which meant they would not store energy overnight. Replacing them with the actively balanced units solved the problem, increasing the self-discharge time constant by an order of magnitude to about a day.

Circuitry

A basic power circuit is simply to connect the output of the solar panels to the capacitor. For good performance two other elements are needed. The first is a diode in series with the photovoltaics. The solar panels have an effective resistance, and when the illumination is not high enough to produce a voltage greater than what is already on the capacitor, it will discharge through the panels. We used a Schottky diode to minimize the forward voltage drop. The second is a Zener diode to protect the ultracapacitor from excessive voltage. Ultracaps are distinctive in that application of voltages only slightly higher than the rating will produce current flow and rapid degradation of the device. Our series pair of solar panels will produce over 13V in direct sunlight, which exceeds the 10V rating of the ultracapacitor. To prevent damage, we placed a 9.1V zener diode (in reverse bias) across the terminals of the storage capacitor. The basic circuit is shown below.

Prospects

Silcon solar panels are currently available that are rated to last 40 years in full sunlight with 20% or less reduction in output. this is projected based on accelerated exposure experiments; There are few panels that have actually been around that long, and none of the current high-efficiency models. Most of the projected degradation is in the housing and encapsulation materials. The basic photo-electric mechanism appears to be extremely durable, though some studies indicate a slow deterioration in photo-electric efficiency, correlated mostly with exposure to radiation in the blue and ultraviolet. Solar panels on satellites have provided useful power for a decade or more in the very harsh radiation environment of space, which also suggests good intrinsic durability.

Our panels are epoxy-embedded behind quarter-inch plate glass. We originally intended to add an ultraviolet filter, but did not find a cheap source of ultraviolet filter glass. The components available from e.g., Edmund and other optical manufacturers are rather pricey, probably because they meet other high optical standards (flatness etc.) rather than intrinsic cost of the glass material. Since the solar panels in our artifact are in a fairly deep recess, a filter could be easily added. It certainly seems possible that they could provide power for several centuries if they remain physically undamaged.

The electric double-layer ultracapacitor is the greatest unknown. These have been around less than a decade, and experience is limited. The unit we have is rated for 500,000 charge-discharge cycles, which at one per day, would give us our 1000 years. The time rating is 10 years, but this is a conservative guess by the manufacturer. The basic materials are porous carbon and an organic electrolyte sealed in a stainless steel canister. These substances are intrinsically quite stable. The weak point is probably the canister seals, which as noted in the housing section, we have done our best to protect from air, water, and light. We have also tried to protect against leakage by secondary embedding in paraffin.