Battery selection

Have you ever had this conversation?

Marketing: The battery has to be readily available to the public.
Engineer: Okay, car batteries are available. This'll make power consumption easy.
Marketing: It has to be lightweight.
Engineer: Scratch the car battery. How about a D cell battery? Decent power storage, but now I need to start watching power consumption.
Marketing: It has to be the size of a pager.
Engineer: Okay, maybe AA batteries and we swap them out every month or so.
Marketing: It has to run for 10 years without recharge.
Engineer: Get a life!

Agreeing on battery size and technology is often a battle between marketing and engineering. Everyone wants a device that is no larger than the display and keypad, light as a feather, and runs forever on a single (low cost, widely available) battery. While the battery affects the size and weight of the unit, it is often overlooked until the end of the design or chosen arbitrarily by marketing.

These questions can help narrow the selection:

• What voltages are needed in the system?
• Does the battery need to be rechargeable?
• What do the typical use/charge cycles look like for this product?
• What are the standard and peak currents used by the system?
• Does the battery need to be easily replaced by a consumer?

Once these questions are answered, you can specify the battery type and size required to achieve the desired battery life. But, even with these questions answered, you must still choose from a number of battery chemistries. Table 2 is not comprehensive, but it lists the more common varieties of batteries.

Table 2 Battery chemistry comparison
Battery Type	Vnominal	Maximum Load Current*	Energy by Wgt. (Wh/Kg)	Energy by Vol. (Wh/L)	Operating Temperature	Recharge	Maintenance**	Life Cycle (Recharge)	Shelf life/ Self discharge
Alkaline	1.5V		150	375	-20 to 55C				5 yrs to 80%
NiCd	1.25V	>2C	45-80	125	-40 to 60C	t	1/Mo	1500	20% per Mo
NiMH	1.25V	0.5-1C	60-120	180	-20 to 60C	t	1/3Mo	500	30% per Mo
Lithium ion	3.6V	1C	100	270-325	-20 to 60C	t		50-100	10% per Mo
Lithium polymer	3.7V	0.2C	120-160	230-270	-20 to 60C	t		500	10+ years
Sealed lead acid	2V	0.2C	30	80	-20 to 60C	t	1/6Mo	200-500	5% per Mo
Zinc air	1.4V		300	1150	-20 to 60C				3-12 weeks to 50%
Silver oxide	1.55V		130	500	-20 to 60C				2 yrs. to 84%
Lithium poly	3.0V				-40 to 85C				10+ years
carbon monofluoride
Lithium manganese	3.0V		225	550	-20 to 60C				10+ years
diode
Lithium thionyl	3.6V		710	1300	-55 to 100C				9 yrs to 80%
chloride
Lithium sulfur	3.0V		290	500	-60 to 85C				10 yrs to 80%
* C = Amp-Hour rating of battery divided by 1 Hour, e.g. for a 1500mAH battery, C = 1500mA
** Required interval between deep cycles

Before selecting a battery type, additional research is required. Each battery chemistry has it own nuances. For instance, a Zinc Air battery cannot be sealed. It must be exposed to air for it to work properly. Lithium batteries prefer a slow discharge and require low voltage protection circuitry.

To generate the working voltages, use either linear regulators or switchers. If the voltage coming onto the board is close to the needed voltage, a low drop-out (LDO) linear regulator might be the best choice:

• Doesn't generate noise on the board
• Low parts count
• Inexpensive
• Can't create a voltage higher than the input

But linear regulators generate heat.

On the other hand, a switcher:

• Usually has a better efficiency
• Doesn't generate as much heat
• Can generally generate any output voltage from any input voltage

But switchers are noisy and have higher parts counts.

Many designers are still afraid to use switching regulators because they are perceived as "touchy" to design and most digital designers naturally stray away from any math that is not boolean. Fear not, many companies offer excellent application notes to go with their switching regulator products (Maxim and Linear Technology are two of our favorites). If you follow their instructions, it works every time.

Another issue to consider is a battery voltage that crosses the generated board voltage. For example, if the required working voltage is 3.3V and the power source is a lithium ion battery that will vary from 2.9V to 4.0V during its useful life, you have several options. One is a buck-boost configuration, where the voltage is generated through two stages of power supplies. The problem with this is that the efficiencies of the two stages are multiplied, reducing the combined efficiency.

Another option is a SEPIC configuration. This is a switcher configured with two coupled inductors (possibly a small transformer). This type of power supply does not have the same efficiency as a typical buck or boost but does better than the two-stage approach.

A third and, hopefully, obvious solution is to arrange the batteries in series or parallel so that the battery voltage no longer crosses any of the board voltages during the course of discharging the battery.

Sleeping and reduced speed clocking

Many embedded applications spend most of their time waiting for something to happen: receiving data on a serial port, watching an I/O pin change state, or waiting for a time delay to expire. If the processor is still running at full speed when it is just waiting for something to happen, we are burning up battery life while accomplishing little.

Microcontroller designers dealwith inactivity by providing sleep modes and adjustable clock speeds. When a processor is put into a sleep mode most (if not all) of the device is turned off. The oscillator is either turned off or switched to a lower frequency and the CPU stops executing instructions. Many processors have different levels of sleeping, each with a different power consumption based on which peripherals are left on.

Microcontrollers are typically awakened from a sleep mode when an interrupt is generated. When using a sleep mode that stops the processor clock, stabilizing the clock before executing instructions typically takes a long time. This time is often measured in tens to hundreds of milliseconds. When you go into sleep mode, keep in mind that the device cannot react to the stimulus for however long the clock takes to stabilize. Also, look at the expected frequency of events. When an event takes place 10 times per second and 50 milliseconds of warm-up is required to come out of sleep mode, you may save substantially more power by running the processor at a lower clock speed instead of sleeping.

General considerations

Now your design is complete. You can proceed with layout, fabrication, and assembly, verify the board, and start production. That would be perfect, but perfection is something you strive for yet rarely achieve on the first try. So, say it with me now: "There will be a second spin of the PCB." Repeat this mantra until you are saying it in your sleep.

Once you admit that a second spin will take place, life gets easier. The design can always be reviewed again, but there is a point of diminishing returns. By no means is this a license to frivolously spin boards. Test everything on paper to a reasonable level, then spin the board. Now you have a platform to verify the design and measure current values. The following list of features should be considered for the first spin, though they may be removed for subsequent versions:

• Test points connected to unused I/O-this will help ease software development.
• Test points and 0W resistors to allow for current measurement.-primarily this would be used for measuring battery current coming onto the board and current from each of the on-board power supplies. Additionally, they should be added to examine current consumption for various sections of the design.
• Build it larger than the final form factor. You know there is going to be a second spin (you already had to say it). So, give yourself room to work.

Ultimately, designing for low power does not require black magic, but it does require extreme attention to detail.

Mike Willey is VP of development for Paragon Innovations. He has over 23 years of experience in embedded systems design. He holds a BS in electrical engineering from Texas A&M University. His e-mail address is willey@paragoninnovations.com

Kris Stafford is director of operations for Paragon Innovations. He has over 13 years of experience in embedded systems design. He holds an ME in electrical engineering from Texas A&M University. His e-mail address is kris@paragoninnovations.com.

Return to January 2002 Table of Contents