Super-capacitors are emerging as a possible alternative to batteriesfor energy-storage in some applications. However, the major advantagesthat super-capacitors offer must be balanced against some significantdisadvantages.
On the plus side, super-capacitors have a virtually unlimitedlifetime of around 10,000,000 charge/discharge cycles and they cancharge and discharge at phenomenal currents in excess of 1,000 Amperes.They are also largely immune to temperature variations. However, theycannot compete with batteries in energy density or cost: typicallysuper-capacitors offer just 3-5% of the energy density of Li-Ionbatteries and cost 10 to 15 times more.
There are, however, some applications for which the advantagesoutweigh even these limitations. But super-capacitors also present twosignificant design challenges in how they charge and retrieve energy.With charging, the challenge is to transfer energy to the capacitorwhen it is completely discharged (effectively presenting a shortcircuit), while retrieving energy also becomes progressively moredifficult as the capacitor voltage approaches 0V. Overcoming these twochallenges is the main hurdle for the efficient use of super-capacitorsas replacements to battery storage.
|Figure1: SMPS-based constant-current charger|
The charging challenge
Linear chargers dissipate a large percentage of energy when charging acapacitor which is completely discharged. Then, as the capacitorcharges, a smaller percentage of the energy is lost, and more energymakes it into the capacitor. Adding the power absorbed by the capacitorand the power dissipated in the charger, the charger will actuallydissipate more than half of the available energy as heat, over a fullcharge cycle. In fact, a linear charger throws away almost 58% of theavailable charging energy as heat.
The other charging option is to use a system based on a Switch-ModePower Supply (SMPS), where the difference between the output capacitorvoltage and the source voltage is dropped across an inductor. In avoltage-regulated SMPS design (Figure1, above ) the inductor current is driven by the differencebetween the voltage across the output capacitor and a fixed referencevoltage. This difference voltage is then amplified, integrated, andphase-shifted, before it is fed back into the Pulse-Width-Modulation(PWM) comparator.
The PWM comparator then uses that voltage to determine how muchcurrent to pump through the inductor on the next cycle. Often, SMPScircuits can achieve conversion efficiencies of greater than 80-90%,with careful design.
In the charger circuit, very little time is spent operating with aconstant output voltage. By definition, the charger circuit is designedto do most of its work while ramping up the capacitor voltage from zeroto the final voltage. It is during this charge-up period that energytransfer needs to be optimised.
The charging circuit requires a system that will regulate thecharging current of the capacitor, independent of the output voltage,and only use the voltage feedback as the means of determining when thecharge is complete. Figure 1 shows how this can be accomplished using avariation on the typical SMPS design. Here, the current in the inductoris regulated by comparing the current in the inductor against two fixedlevels; one at the maximum desired current, and the other at theminimum.
Initially, it will take the inductor very little time to ramp upfrom the minimum to maximum current, as the voltage across the inductoris at its maximum. The discharge time will be correspondingly longer,as the inductor has to discharge into a relatively small voltage. Asthe charge in the capacitor increases, however, the voltage differencewill drop, increasing the ramp-up time, and the capacitor voltage willrise, shortening the discharge time.
While something similar can be implemented with a traditionaltime-base driven PWM, the selection of the inductor becomes critical tomaintaining the minimum current level. Additionally, instability canoccur when the duty cycle is greater than 50%. A simple solution toavoid this instability is to use a relaxation-oscillator, 555Timer-style system, using two comparators and a SR flip-flop, so thatthe inductor component values set the frequency.
Getting energy from the super-capacitor also presents challenges. It iseasier to retrieve energy from a battery because it maintains arelatively flat discharge voltage as its charge is diminished. Acapacitor, on the other hand, has a steep discharge slope that dropslinearly from the full-charge voltage down to zero.
As there are few circuits that can operate effectively over thisvoltage range, it naturally follows that a SMPS-style boost circuit isrequired to convert the variable capacitor voltage into a reasonablyconstant load voltage.
As the capacitor voltage is now the source voltage for the SMPS,when it drops, the inductor current ramp-up time must increase if theoutput voltage is to remain constant. This seemingly trivial problemhas one very serious consequence:
If the pulse of current delivered to the output remains constant,but the time between deliveries increases due to longer ramp-up times,the output voltage will begin to sag between deliveries. This meansthat, as the capacitor's charge is diminished, the output-voltageripple will increase.
Given this behaviour, using a fixed inductor current is no longerpossible, if the circuit is to maintain a reasonably constant outputvoltage. There are, however, three possible solutions. The first is tostack super-capacitors together, to extend the usable range of theretrieval circuitry, so that the lost capacity is a smaller percentageof the capacitor's total storage capacity.
Alternatively, declare a minimum operating charge voltage for thesuper-capacitor and shut down when the charge drops below this level,which effectively discards part of the super-capacitor's capacity asunusable. The third option is to limit the inductor current, so thatthe output of the retrieval circuit becomes increasinglycurrent-limited as the charge diminishes.
Unfortunately, there are two problems with the series-stackedapproach: First, there is the problem of balancing the charge in thecapacitors, and the second concerns the SMPS design of the circuitretrieving energy from the capacitor. There are charge-balancingtechniques that shunt charge current around the individual capacitors,based upon their charge voltage. One such technique uses a Zener diodein parallel with each capacitor. The second technique uses voltagecomparators and MOSFET transistors to shunt the current, based upon amonitoring circuit.
The Zener diode circuit is by far the simplest solution. The problemis that Zener diodes do not have a perfectly sharp turn-on knee andactually start conducting below their Zener voltage. As a result, evenwhen the capacitors are balanced, there will still be some conductionaround the capacitors that is bleeding-off charge.
The active MOSFET charge balancer solves this problem through a morecomplex switching system. It monitors the voltage across the variouscapacitors in the string and when any capacitor reaches its maximumworking voltage, the monitoring logic disconnects the capacitor fromthe string and shunts the charging current to the other capacitors tocontinue charging.
This eliminates the leakage-current problem, but at the cost of amore complex system for monitoring the voltages across all thecapacitors. This additional circuitry also burns some of the chargecurrent to power itself during the charging cycle, reducing thecharging efficiency of the system.
The second problem with a series-stacked super-capacitors is thetotal voltage of the capacitor string. Initially, the stacked voltagewill typically be higher than the required load voltage, necessitatinga buck topology SMPS design. However, as the charge in the capacitorsis depleted, the stacked voltage will eventually drop below the loadvoltage, as it ramps down to zero.
This means that, at some point in the discharge curve, the SMPSdesign will have to switch gears and become a boost topology SMPSdesign. An alternative solution is to put multiple super-capacitors inparallel, reducing the slope of the discharge line. There are severaladvantages to this method, including the fact that the capacitors willautomatically charge-balance to a common voltage.
The charge will then route to those capacitors with greatercapacitance without active direction, and the total voltage for thebanked system remains low. This allows the use of a simple boost SMPSto retrieve power from the capacitors. The boost SMPS retrieval circuitcan also pull more current from a parallel configuration, whilereducing the I2R losses in the individual capacitors, because thecurrent load is shared by all the devices.
While banking multiple super-capacitors in parallel reduces theslope of the capacitor voltage, it does not eliminate the basicproblem, but merely delays it. When the capacitor voltage falls, thetime to charge a fixed inductance increases. The only two solutionsthat are both simple and practical are to shut down the boost when thesuper-capacitor voltage drops below a reasonable minimum voltage, or tolimit the inductor current as the charge in the super-capacitor isdepleted.
The super-capacitor boost circuit does not need a fixed-currentlimit. Instead, what is needed, is a current limit that decreases asthe capacitor voltage drops. This will correspondingly drop the maximuminductor current and maintain a reasonable ramp-up time.
To do this, the circuit can be modified so that the current-limitinput is driven with feedback from the super-capacitor. This willreduce the maximum current linearly, with the reduction in voltage.This keeps the equation balanced and the maximum charge time reasonablyconstant, as the capacitor voltage drops to zero.
|Figure2: Composite charge-retrieval SMPS circuit|
There is a certain similarity between the buck topology circuit used tocharge the super-capacitor, and the boost topology circuit used toretrieve energy: Replacing the Flyback diode in the buck circuit with asynchronously switched MOSFET, creates the same power-chain circuit asreplacing the Flyback diode in the boost circuit.
Figure 2 above shows whatcan be achieved with a little embedded intelligence. Re-routing the PWMsignals to both MOSFETs, and re-routing the feedback and current-sensesignals, creates a circuit that can, with some embedded intelligence,handle buck charging as well as boost retrieval. The intelligence canalso be tasked with monitoring the capacitor, output and sourcevoltages for charging, to determine which topology is required.
This circuit can either use a microcontroller with the necessarymixed-signal peripherals to build the conversion circuitry, or a simplemicrocontroller combined with an array of sufficiently programmableexternal mixed-signal devices to allow the required switching andcontrol. The PIC family of microcontrollers includes bothlow-cost controllers, as well as versions which integrate mixed-signalperipherals optimised for power applications.
The use of super-capacitors for energy storage does have significantadvantages, as well as disadvantages, compared to chemical, batterystorage. Its extended life and immunity from temperature effects canmake the super-capacitor the preferred storage medium, despite its costand limited energy-density. The challenges associated with charging andretrieving energy from a super-capacitor system, however, also have tobe factored into the cost/benefit analysis.
These challenges can be significant, but they can be handled withcareful design and the inclusion of some simple embedded intelligence.Recent developments in the PIC microcontroller family includecontrollers which integrate many of the mixed-signal peripherals neededto implement SMPS-style charging and retrieval.
The addition of embedded intelligence can, therefore, help to reducethe size and cost of using super-capacitor storage as an alternative toconventional batteries.
Keith Curtis is PrincipalApplications Engineer in the Security, Microcontroller and TechnologyDevelopment Division at MicrochipTechnology Inc.