Potential Power Picks
Choosing power options for an embedded system is a bit like ordering from a Chinese menu: “One from column A, one from column B, and one from column C.” Instead of being labeled “appetizer,” “soup,” and “main course,” the columns are labeled “primary,” “recharge,” and “backup.
Choosing power options for an embedded system is a bit like ordering from a Chinese menu: “One from column A, one from column B, and one from column C.” Instead of being labeled “appetizer,” “soup,” and “main course,” the columns are labeled “primary,” “recharge,” and “backup.”
For primary power, choose the source that will power your embedded system the vast majority of the time. For most embedded applications, this is good ol’ utility mains power—most likely 120 V ac. Second choice is probably some form of rechargable dry cell, such as Lithium-ion (Li-ion). A growing number of network-connected applications are able to derive primary power directly from the network (such as USB). The number of embedded systems that rely on plain-old non-rechargable dry cells is shrinking rapidly.
The recharge column indicates what is used to keep the primary going indefinitiely. Many high-reliabilty systems, such as medical patient monitoring systems, have built in uninterruptible power sources. System power for vehicles invariably consists of a primary battery recharged by an alternator whenever the engine is operating.
Backup sources are there when all else fails. Typically, backup is designed to provide power for minimum necessary functions, such as maintaining volatile memory in mobile devices during main battery replacement. Most personal computers, for example, have a non-rechargable lithium battery (as opposed to rechargable Li-ion) to maintain BIOS firmware when the computer is disconnected from main power.
Which power source is right for which category depends entirely on application characteristics. For example, wireless sensors deployed in remote or hard-to-access locations is one of the few growing applications where long-lived, but non-rechargable, lithium batteries are the first choice. Hot environments are problematical for Li-ion rechargables because high heat drastically shortens their service lives.
A typical embedded system linear supply schema uses commercially available transformer/rectifier inputs to couple electric power from utility mains to the embedded system.ing ground, and a +/- 15 V double-ended supply.
Applications where battery power is the best choice fall into three categories: low power requirements, high power requirements, and backup. Wireless sensors deployed in remote locations require low power consumption. Hybrid vehicles need lots of electric power, but the primary source really needs to be a battery for other reasons. Medical monitoring equipment is an example where the application’s high-reliability requirement calls for on-board battery backup even though the primary source is utility mains.
When power availability is low, the embedded system designer must incorporate system features that minimize the system’s power requirements and maximize the efficiency with which power is used.
Choosing the correct devices and circuit forms is the first step in minimizing power usage. Microprocessor clock speed, for example, has a large impact on a device’s power consumption, with power requirements rising nonlinearly with clock rate for the same processor. Similarly, “wider” processors (such as, 32-bit vs. 16-bit) require more power than “narrower” ones. Similarly, some circuit forms are inherently more efficient than others. For example, switching power supplies are much more efficient than linear power supplies. Of course, running directly off battery voltage is more efficient than either. So, the best choice for minimizing power requirements is to use the lowest clock speed, narrowest microprocessor, and most efficient circuit forms consistent with operational needs.
After optimizing the circuit for minimum operating power requirement, the next most effective—perhaps the most effective overall—strategy for reducing power consumption is providing a sleep mode and watchdog timer. In sleep mode, all non-essential functions are shut off for an extended period of time and power goes only to a separate low-power timing circuit. The timer’s purpose is to automatically wake the system up for a burst of activity.
Relatively few high-power applications use batteries for primary power. It’s not that batteries can’t deliver heavy currents—they can—the problem is that they generally can’t do it for long. Usually, engineers opt for primary battery power in these applications only when other considerations make it the best choice.
For example, I often build small devices for short-term experimental applications. Some years ago I built an extremely-low-frequency (ELF) oscillator based on a dualμa741 op-amp for road testing a digital storage oscilloscope. The thing’s planned operational life was a couple of hours, and I didn’t happen to have an appropriate benchtop supply to power it with.
A switched-mode power supply or "switcher" provides regulated dc at much higher efficiency than a linear supply.
What I did have was a pair of 9 V transistor batteries and holders. So, I soldered the positive lead of one holder to the negative lead of the other, and both of them to the circuit ground. I then soldered the free positive lead to the op-amp’s positive supply terminal and the free negative lead to the negative terminal. The resulting
A more common application for primary battery power when lots of “juice” is needed is vehicle propulsion. Hybrid and all-electric vehicles typically use battery-powered electric motors for propulsion, and some other source for recharge.
A relatively new wrinkle on these systems is the availability of ultracapacitors. Batteries store energy in chemical bonds. They release this energy through chemical reactions. The main factor limiting their maximum current is the speed at which the chemical reactions can run. Ultracapacitors, on the other hand, actually store energy in the form of electric fields. Their peak current is limited only by Ohmic heating of the capacitor plates and internal wiring. They also can stand more charge/recharge cycles.
Embedded systems using batteries for backup incorporate an internal uninterruptible power source (UPS). I’ve already mentioned medical patient monitoring equipment, which often uses such technology to assure service should utility power be lost. It also makes it possible to maintain continuous monitoring when a patient has to be moved: technicians just pull the plug and move the monitor with the patient without shutting it off. The equipment runs primarily on battery power, while mains provide a trickle charge whenever available.
The pulse-width modulator at the heart of every switched-mode power supply uses digital switching technology to control the power output with minimal dissipative loss.
Most household appliances derive their power from 120 V ac utility power. As the use of embedded control systems for these devices has grown, the number of systems operating directly from utility power has grown as well.
Unlike the timer-and-relay-based control systems they replace, microprocessor-based automated control systems do not run on 60 Hz ac power. They use carefully regulated dc power at relatively low voltages (often 5 V dc).
There are two basic classifications of power supply circuits for converting utility-grade ac power to electronic-grade dc power: linear and switching.
Linear power supplies start with a transformer to reduce the powerline voltage to approximately the supply’s output voltage. Next, a bridge rectifier converts the low-voltage ac to pulsating dc. A filter section—most often just a large electrolytic capacitor followed by a relatively low value power resistor—smooths out the pulsations.
At that point, we have dc power at a voltage that is whatever it is based on the utility voltage, transformer turns ratio, load current, and resistances—with a 60 Hz ripple superimposed on it whose amplitude depends on both the filter’s RC constant and the load current. In other words, the output current and voltage are whatever they happen to be at the time, not what is required!
To improve the dc quality, power-supply designers add a linear regulator. The linear regulator is a linear power amplifier powered by the filter output that amplifies a small reference voltage up to the desired value. Feedback elements make it possible to change the final output voltage, or to make the circuit regulate current rather than voltage.
Linear power supplies are rather inefficient, typically dissipating as much power as they deliver. Always a waste, and sometimes a very expensive waste, this power loss can be mitigated by using a switching power supply.
A switching power supply has the same transformer, rectifier, and filter sections as the linear supply. Improvements come from replacing the linear regulator with a switching regulator. Instead of using a linear amplifier, where the transistors are never turned all the way off or all the way on, transistors in a PWM are alternately driven into saturation (effectively a short circuit), then cut off (effectively infinite resistance). In either condition, power dissipation is quite small.
The PWM puts out a train of pulses with variable duty cycle. A feedback system modifies the duty cycle to provide exactly the output value needed for the load. A regenerating filter (essentially a large inductor) averages the current to smooth out the pulses.
Switching power supplies are complex to design and build. Luckily, there are vendors who manufacture the things in huge quantities at very low cost, so there is seldom any need to design one for your special application. Embedded system designers now look at them as black boxes that take “dirty” utility power in and pump “clean” dc out. Just go to a catalog, pick out the block you want based on the output voltage(s) and current(s) you need, and order it. The hard part is making a space on your system layout to install it.
Often, the power available to run embedded systems does not come from utility mains or batteries. System power in an automobile, for example, comes from an engine-driven alternator/voltage regulator system and provides 12 V dc at often quite substantial current levels. Trains, boats, cars, spacecraft, wind turbines, and solar power plants all have power sources that have nothing to do with utility mains. In most cases, these sources provide dc power at some standard voltage with highly variable power quality.
To run embedded electronic systems in these situations, the designer typically needs to change the voltage and regulate the output. The only option is the dc-to-dc convertor.
A dc-to-dc convertor uses the available dc supply to run a square-wave oscillator to make ac power. A transformer steps this ac voltage up or down as needed. A full-wave bridge rectifier then recovers the dc. Since a square-wave duty cycle is 100%, the rectifier’s output is relatively pure dc with only high-frequency imperfections to filter out. If further regulation is needed, a switching regulator will do the job quite efficiently, thank you.
Other systems make electronic-grade power available as a matter of course. Products connecting to USB ports, for example, have 5 V dc available from the USB connector. Similarly, clause 33 of the IEEE 802.3-2005 Ethernet standard (IEEE 802.3af, commonly referred to as “power over Ethernet”) recommends providing 48 V dc over two of the four conductor pairs in standard CAT-3 or CAT-5e Ethernet cable to supply power for terminal equipment. The standard recommends a current limit of 400 mA.
Mixing and matching these available power sources for primary, recharge, and backup duty gives embedded system designers a wide range of options to power their creations. Choosing the best scheme is a matter of comparing what is available in a particular application with what the application needs.
C.G. Masi is a senior editor at Control Engineering . Reach him at firstname.lastname@example.org