Supercapacitor Backup Power Design Guide
A supercapacitor sits between a battery and a conventional capacitor in the way it stores energy. It holds far more charge than an electrolytic of the same size, delivers it far faster than a battery, and tolerates hundreds of thousands of cycles. Those properties make it the right choice for a backup that lasts seconds or minutes, and the wrong choice for one that lasts hours.
What a Supercapacitor Is Good At
The device stores charge in the electric double layer at the surface of a porous carbon electrode, so there is no chemical reaction and no wear mechanism of the kind that limits a battery. Charge and discharge are fast, the internal resistance is low, and the cycle life is measured in hundreds of thousands rather than thousands.
The trade is energy density. A supercapacitor stores perhaps a twentieth of the energy of a lithium cell of the same mass, and its voltage falls linearly as it discharges rather than staying nearly constant. A backup design therefore has to work with a falling input voltage and has to use a converter that can extract energy down to a low input level.
Temperature behaviour is better than a battery in the cold and worse in the heat. The capacitance falls and the internal resistance rises at low temperature, which reduces the usable energy, while high temperature accelerates the ageing of the electrolyte. Both effects should be checked against the application rather than assumed to be negligible.
Energy and the Usable Voltage Window
The energy stored is one half of the capacitance times the square of the voltage, and the energy that can actually be used is the difference between that value at the top of the window and at the bottom. Because the voltage appears squared, the last part of the discharge contributes very little energy, and stopping at half the rated voltage already gives up three quarters of the stored energy.
This is why the input range of the downstream converter matters so much. A converter that operates down to two volts can extract far more from a five volt supercapacitor than one that stops at four volts. Choosing the converter for a wide input range is often more effective than choosing a larger capacitor.
The rated voltage of a single cell is low, typically between two and three volts, so a stack is needed for any useful rail. Cells in series add their voltages but divide the capacitance, and they introduce the balancing problem that dominates the practical design of a stack.
Inrush Current and Charging
An uncharged supercapacitor looks like a short circuit to the supply, and the inrush current is limited only by the internal resistance of the capacitor, the resistance of the wiring and the impedance of the source. A large cell connected directly to a low impedance rail can draw hundreds of amperes for a few milliseconds.
The remedy is a controlled charge. A series resistor bypassed by a switch limits the initial current and is shorted out once the capacitor has reached a defined voltage, a current limited source charges the stack at a defined rate, or an active circuit regulates the charge current throughout. The choice depends on how quickly the backup has to be ready after power is applied.
Charging current also heats the capacitor through its internal resistance. The energy dissipated is significant during a fast charge, and the heat has to leave through the leads and the board. Rapid charging shortens the life of the cell, so a charge rate that keeps the temperature rise modest is a better engineering choice than one that fills the stack in the shortest possible time.
Cell Balancing in Series Stacks
cell balancing keeps the voltage across each cell in a series stack within its rating. Manufacturing spread in capacitance and leakage means that the cells charge at slightly different rates, and the weakest cell reaches its limit first. Without balancing, the stack is limited by that cell and its life is shortened.
Passive balancing uses a resistor across each cell to draw a small current that equalises the voltages. It is simple, cheap and adequate where the charge current is modest, at the cost of a continuous drain that matters in a low power application. The resistor value is chosen so that the balancing current is larger than the spread in leakage between the cells.
Active balancing moves charge from the higher cells to the lower ones using a small converter, which is more efficient and more complex. It suits large stacks and high charge currents, where the loss of a passive network would be significant. Whichever method is chosen, the balancing circuit is part of the backup system and its own current consumption has to be included in the standby budget.
Hold Up Time and the Load Profile
hold up time is how long the backup has to carry the load, and it is calculated from the usable energy and the load power. Because the load is rarely constant, the calculation has to use the actual profile rather than an average, and a load that draws a large current for a short period will dominate the result.
The conversion efficiency of the downstream regulator also enters the calculation. A backup that has to supply a rail at constant power draws a current from the capacitor that rises as the capacitor voltage falls, so the useful energy is less than the simple capacitor formula suggests. A boost converter with a wide input range recovers most of that energy.
Where the backup must also signal that it is running, the monitoring threshold should be set with margin above the point at which the converter drops out. A warning issued at the last possible moment is not useful, and a warning issued too early wastes the available energy. Measuring the actual hold up time on a prototype is the only reliable way to set the threshold.
Leakage and Self Discharge
leakage current flows through the capacitor continuously and discharges it even when no load is connected. The value is quoted after a period of stabilisation, because the initial leakage is much higher than the long term figure, and a design that uses the initial value will overestimate the standby performance.
Self discharge also comes from the balancing resistors and from any monitor circuit connected across the stack. In a low power application, those currents can easily exceed the leakage of the cells themselves, and the standby time is then set by the design rather than by the component. Totalling all of the currents that flow when the system is idle is a necessary step.
The practical consequence is that a supercapacitor keeps its charge for days or weeks, not months. Where the backup has to be available after a long period without power, a battery is a better choice, while a supercapacitor suits a system that is powered most of the time and needs a short ride through.
Charging Circuit Topologies
The simplest topology is a current limited supply that charges the stack directly to its rated voltage. It works well when the main rail is already at the stack voltage, and it has the merit of simplicity. The charge current has to be limited and the voltage has to be regulated to avoid overcharging the stack.
A boost converter charging the stack from a lower rail allows the stack voltage to be chosen independently of the system rail. The converter operates in a mode that limits the charge current and then transitions to voltage regulation as the stack fills, which requires a control loop that remains stable across both modes.
Where the stack supplies the load during a power failure, a bidirectional converter can charge the stack and then discharge it into the rail, using the same inductor and switches. That approach saves components and board area, at the cost of a control scheme that has to switch modes cleanly. Whatever the topology, the return path for the charging current should be separate from the sensitive measurement ground, as described in our guide to power integrity.
Layout and Protection
A supercapacitor delivers a large current in a short time, so the connections have to be sized for the peak rather than the average. Wide traces, multiple vias and a connector rated for the current are all necessary, and the general placement rules for the bulk energy storage that supports a rail are described in bulk capacitor placement.
Place the capacitor as close to the load as the mechanical design allows, because the resistance and inductance of the connections limit the current that can be delivered quickly. Where the backup supplies a switching converter, the input capacitor of that converter and the supercapacitor should be connected by a low inductance path with nothing else sharing it.
Protection is mostly about the discharge path. A reverse connected cell, a shorted stack or a fault that discharges the stack through an unintended path can all damage the board, and a fuse or a current limit in series with the stack limits the damage. The partition between the power path and the signal ground follows the same rules used for any mixed signal board, as described in mixed signal board design.
FAQ
How long can a supercapacitor power a load? Seconds to minutes for most practical stacks. The hold up time is set by the usable energy above the converter drop out voltage and by the load profile, not by the capacitance alone.
Do I need balancing for two cells in series? Usually yes. Even a small spread in leakage causes the cells to diverge, and the weaker cell will reach its limit first. Two resistors are enough for most low power designs.
Why does my stack voltage fall faster than calculated? Include the balancing resistors, the monitoring circuit and the rising input current of the downstream converter. Those three items account for most of the difference in practice.
