Evolving energy harvesting with FRAM

In the past few years, energy harvesting solutions have evolved to become more efficient and cost-effective. They now feature ultralow-power microcontrollers and wireless transceivers that consume less power, as well as transducers that generate more energy. The latest development to advance energy harvesting applications is embedded ferroelectric RAM for ultralow-power MCUs.

The combination of ultralow-power MCUs and embedded nonvolatile FRAM is a revolutionary advancement for applications in which low power consumption is a key concern. Such apps include wireless sensor networking and energy harvesting. The memory reduces the industry’s best active power by up to 50 percent, enabling operation at 80 µA/MHz in active mode and 1.5 µA in real-time-clock mode. When operating at 2 volts, devices consume less than 200 µW when running at 1 MHz and as little as 3 µW in standby.

In addition, many low-power MCUs have integrated the power management circuitry needed to manage the unstable output of even the best energy harvesting environment. The required level of energy can be easily supplied by small, cost-effective vibration, thermal and solar energy harvesting modules that are smaller than a human palm.

Faster writes at lower power

The main advantage of FRAM is not in replacing SRAM but in replacing flash or another nonvolatile memory such as E²PROM. Not only do those memories require high programming voltages and charge pumps, but they also require slow ramps and long write cycles at high currents, sometimes above 10 mA. 

Figure 1. Model node using an ultralow-power MCU. Click on image to enlarge.

 
 

In energy harvesting applications, peak currents exceeding 10 mA can be a serious challenge since the incoming power levels are frequently several orders of magnitude lower, as shown in Figure 2.  By reducing the peak power numbers, it’s possible to reduce the size, cost and complexity of the energy harvesting system, thereby removing obstacles that have prevented some designers from using energy harvesting. 

Figure 2. Average energy harvesting results.

Additionally, FRAM can write 1,000 times faster than traditional memory at similar current levels, so it can use up to 1,000 times less energy per byte. Combine this with the robust power system in microcontroller families such as the MSP430 MCU from Texas Instruments, and there is now a low-cost, simple way to add energy harvesting to your system. In an average usage example, described below, the FRAM device writes at 2 Mbytes/second, while the flash device of the same architecture is limited to 13 kbytes/s. Combine this with the current usage at those given speeds, and more than 450 times less energy is used.

Record more data, longer

A common application for energy harvesting systems is sensor monitoring and logging. The relatively low write endurance (only 10,000 to 100,000 write cycles) of traditional memories such as flash or E²PROM can present a challenge in applications with a long intended lifetime. FRAM, on the other hand, has nearly infinite write endurance.

The amount of data a designer must accommodate is determined by the accuracy, longevity and redundancy needs of the application, but a general rule is that more information is better. If the cost or capability of the memory limits the data size, then the designer will have to sacrifice recording frequency, accuracy or device life span.

FRAM eliminates this concern and frees the designer to focus on obtaining the best data possible.

The designer can even add another sensor, record more often, use a higher-resolution A/D converter or record continuously without fear of wearing out the memory. To better understand the difference FRAM can make, consider the following:

Data = Memory Size x Endurance

Dflash = 16 kbytes x 100,000 Cycles = 1.6 Gbytes

DFRAM = 16 kbytes x 10¹⁵ Cycles = 16 exabytes

In other words, you can record 10 billion times more data over the device lifetime in FRAM than in flash or E²PROM. (How often to you get to talk about exabytes of data?)