This is an old revision of the document!

Capacitor Selection

In theory, a batteryless device can completely forgo energy storage and rely entirely on ambient energy harvested from the environment. In practice, most devices require some form of energy storage buffer between the harvester and the device, if only to smooth out the overall power curve and provide somewhat more predictable operation.

So, why capacitors over batteries?

Batteries vs. Capacitors: A Brief Introduction

Broadly speaking, batteries and capacitors both function as power storage devices: both store a potential difference (or voltage), which can be discharged to power a device or circuit. How they do so varies, however, with significant implications on capabilities and behavior.

Batteries store their potential difference via electrochemical reaction. One portion of the battery (the anode) releases electrons that are accepted by the other portion (the cathode): this movement of electrons generates current (and by extension, power). Chemical operations limit how quickly batteries can discharge (or charge) their energy, but results in a comparatively energy dense storage method with a constant voltage over most of its charge/discharge cycle.

Depending on the materials used for the anode and cathode a battery will be either disposable (the chemical reaction is irreversible) or rechargeable (the chemical reaction can be reversed by an outside power source). However, even rechargeable batteries have a finite lifetime, losing effectiveness over multiple recharge cycles: chemical composition plays a significant role in base lifetime, but environmental conditions and how often the battery is charged and discharged also limit a rechargeable battery’s practical lifetime.

Capacitors store power via electromagnetism, rather than a chemical reaction. Most capacitors contain two conductive plates separated by a non-conductive layer (called the dielectric). When connected to a power source, the dielectric blocks the normal flow of electrons between the plates, leading to one plate accumulating electrons while the other loses them: this results in the polarization of the dielectric as the plates exert equal but opposite electrostatic force on the dielectric, storing a potential difference in the field. Once disconnected from a power source, the plates will seek to equalize charges, and the resulting electron flow from one plate to another creates a current (and power). This behavior allows for rapid charge and discharge times relative to a battery, though with the caveat that voltage will fluctuate to a greater extent.

Storing power via electromagnetism has limitations, however. All dielectric materials have a breakdown voltage, or a point where the resulting electric field becomes strong enough to act as a conductor: lighting is a visible example of this behavior in nature, occurring when the potential difference between clouds and the earth exceeds the breakdown voltage of air. This breakdown voltage places a practical ceiling on the maximum voltage of a capacitor, and so the only other way to increase power storage is to increase the size of the plates: due to the physics involved, a capacitor will usually be significantly larger than a battery that can store a comparable amount of energy.

Capacitors are also more limited in terms of long-term storage compared to a battery. In theory, if an ideal capacitor is charged and then disconnected from a circuit, it will retain a charge indefinitely: in practice, the physics of real-world capacitors results in some amount of discharge even when disconnected from a circuit or otherwise inactive. Combined with their generally lower overall capacity, energy stored in a capacitor tends to be “use it or lose it” in a way that it is not with batteries.

Why Capacitors?

Most IoT devices (and electronic devices more generally) are built around the assumption of a constant, steady supply of power, which has traditionally been the domain of batteries. They cannot be casually swapped for capacitors, which have on average lower power density and significantly different I-V characteristics over their discharge cycles: if using capacitors as the primary energy storage, a device will need to be designed around these limitations. So if batteries are tried and tested and the “default” for most applications, why bother with capacitors?

Sustainability: in most electronic devices, the battery is typically the component with the lowest working life (around a few years at most). Most batteries also contain heavy metals that are environmentally toxic: even if replacement is viable, disposal is a constant concern. In a world with millions (or billions or even trillions) of small IoT devices, the quantity of potential battery e-waste goes from an inconvenience to a significant problem. Capacitors, by contrast, generally have significantly longer active lives than capacitors do, reducing the need for replacement: the composition of most capacitors also makes e-waste less of a concern than with batteries (though not entirely irrelevant).
Maintenance: Most batteryless devices are intended to be deployed in conditions where access may be difficult (remote wilderness) or impossible (microsatellites). Even when accessible, the human effort required to regularly replace and dispose of batteries may make certain deployments impractical. The goal of batteryless devices is to not only reduce the need to dispose of e-waste from batteries, but the human labor required to maintain them as well.
Quick charging: Many UAV/robotics platforms rely on batteries for power storage, but the batteries used in most platforms take significantly longer to charge than discharge: this results in a relatively low duty cycle (period where a robot/UAV is active vs. recharging) as recharging becomes a primary bottleneck. As capacitors can be charged significantly quicker (seconds vs. minutes or hours) than batteries, they are being explored as an alternative energy storage method to reduce downtime.
Batteries may be infeasible: Some applications may prevent the use of batteries at all. Devices deployed to monitor conditions in the Mithraetum of Circus Maximus in Rome were forced to forgo batteries, as environmental conditions ran the risk of causing battery leakage and damage to the site. Capacitors can also be made at scales far smaller than any battery, and (barring any breakthroughs in battery technology) will likely be the only practical method of energy storage for micro-scale systems like smart dust.

Capacitor Sizing: Why Bigger is Not (always) Better

If using a capacitor in place of a battery, it might be tempting to use a capacitor with the highest capacitance possible to minimize the difference in storage capacity. However, a larger capacitor may end up being detrimental overall if two main factors are not taken into account:

Charge Time and Voltage: Charge on a capacitor is expressed as:

Q = CV

Where Q is charge, C is the capacitance, and V is voltage. By extension:

V = Q/C

The larger the capacitor, the higher the charge accumulation necessary to reach a specific voltage value (and the longer it takes to accumulate it).

All electrically-powered devices have minimum voltage thresholds necessary for operation (and may have restricted operations at those minimums): the longer it takes for the capacitor to reach those thresholds, the longer the downtime. In some applications this may be acceptable if a larger amount of power is needed for specific work: for other applications (e.g. sensors trying to detect specific events), longer downtimes may interfere with the device’s primary tasks. Leakage and Resistance: As mentioned previously, real world capacitors must contend with some level of leakage. There are multiple contributing factors, from minor flaws in the plates from manufacturing to some small level of current passing through the dielectric, which larger capacitors exacerbate.

Real world capacitors also have an equivalent series resistance (ESR) which is a product of the resistance of their material composition: larger capacitors will often have a higher ESR simply due to having more material (and thus, resistance) for current to travel through.

Because of leakage and ESR, it should not be taken as given that the extra power gained from a larger capacitor will be converted entirely to useful work.

Capacitor Selection Guidelines

Given the above, we can come up with some general guidelines when selecting capacitors for a batteryless device:

Assume short bursts of energy in most cases: it is difficult to match the steady power of a comparable battery device short of some supercapacitors or a source of harvested energy capable of providing more power than is actively consumed. If either option is not available, then selecting purely for high capacitance may provide no benefit or even be counterproductive, and many batteryless devices will need to be designed with an appropriate checkpointing strategy to best make use of these short bursts. Have enough energy to complete the largest discrete task: At minimum, the capacitor should be able to provide enough charge to complete the largest discrete (i.e. must be done in one power cycle) task: for example, if the device has a radio that it uses to transmit data, it should have enough power to complete the transmission without encountering a power failure. The smaller the capacitor, the more reactive the device: the less time a device spends charging, the more often it will be active (and capable of detecting/reacting to events, if necessary). This is often in tension with the previous point, as many devices have a small number of tasks or peripherals with significantly higher energy demands than the others. Be aware of non-ideal behaviors: like batteries, a capacitor’s paper performance can be degraded by various factors such as temperature, humidity, and frequent charge/discharge cycles. What may work on a test bench or simulation may fail in a practical deployment if these factors are not properly accounted for.

Capacitor Setups

A number of capacitor setups have been explored, ranging from simple single capacitor setups to more complex power storage and management. The most common setups and examples are explored below.

Single Capacitor Storage

The simplest implementation merely replaces a device’s battery with a capacitor (or a single fixed bank of capacitors in series and/or parallel with each other): this capacitor then serves as the primary energy storage for the device. This design has an advantage in relative simplicity, and energy management is likewise relatively straightforward as there is only one source to manage. However, using only one capacitor limits the device to, at minimum, the capacitance required for its most intensive task: this can potentially be an issue if the device’s other tasks would have more optimal performance with a smaller capacitor.

Federated Storage

Batteryless devices often include multiple peripherals with their own individual energy requirements, and these conflicting requirements often mean that selecting an ideal capacitor for one means compromising the performance of others.

Federated storage arises from the observation that if capacitors cannot provide the steady power of a traditional power source, then there is no need to restrict energy storage to a single capacitor or bank: separate individual capacitors can be used to store energy and power for individual peripherals or tasks, allowing each capacitor to be optimally sized for what it is powering. Separate energy banks also allow for more flexible operation: a sensor can gather readings without impacting the processor that parses them, which in turn does not use energy that the radio needs to send the readings to another node.

With this flexibility comes increased complexity. A separate energy management circuit is usually required to measure and distribute energy to the individual capacitors, which leads to increased overhead and circuit size. The device itself must also be able to account for the current energy state of each individual peripheral or task: using the previous example, if a device wishes to transmit data it must be able to recognize whether or not the radio has sufficient power for operation and adjust its behavior accordingly.

Examples United Federation of Peripherals (UFOP) http://dx.doi.org/10.1145/2809695.2809707 Stash https://doi.org/10.1145/3641511

Reconfigurable Storage

Reconfigurable storage setups approach the capacitance issue from a different angle. Capacitors connected in parallel have increased capacitance (and reduced capacitance when in series): by using switches to (dis)connect or adjust connections between an arbitrary number of capacitors connected in parallel (or series) with each other, the overall capacitance can be dynamically adjusted to best meet the current demands on the device. This provides (from the circuit’s perspective) a single power source, but with the flexibility of being able to adjust capacitance to best match current demands.

Like federated energy, there is a price in both circuit size and complexity, requiring multiple capacitors (even when some may be rarely active) and some control circuitry and logic to switch configurations as needed: there is also some energy loss when the configuration changes, due to equalization of charges across the capacitors. While capable of a wider range of capacitances than single or federated storage methods, the range of capacitances is ultimately limited by the number of possible combinations, so the energy profile of the device should still be well understood to ensure proper capacitor sizing.

Examples Capybara https://doi.org/10.1145/3173162.3173210