FRAM Tears Down Traditional Microcontroller Design Barriers
FRAM is reshaping the landscape of microcontroller applications due to its ability to provide the best benefits of SRAM and flash architectures.
Microcontroller memory has seen expansive change since the first microcontrollers were introduced. Memory technologies transitioning from PROM and EPROM to EEPROM and flash were significant factors in the steady growth of microcontrollers in recent products. Microcontrollers with a partition of SRAM and flash have been a staple in the industry for a long time, but even these standard memory technologies must evolve for microcontrollers to become ubiquitous. Modern memory technologies aim to bridge the gap between SRAM and flash, and also eliminate other shortcomings like limited endurance and high-power memory accesses. Ferroelectric random access memory (FRAM or FeRAM) is one of the emerging memory technologies today. There are many advantages of FRAM microcontrollers over traditional RAM and flash microcontrollers; in certain applications these advantages significantly advance the application.
FRAM provides many of the benefits of SRAM, but with the added benefit of being non-volatile. This is achieved by a fundamentally different approach to how the memory state is stored in the cell. An FRAM cell uses the polarization of a ferroelectric crystal to store memory content as opposed to a SRAM cell, which uses a stored charge within transistors. After power is removed, the state of the FRAM crystal is retained while the charge in SRAM depletes.
A microcontroller application needs non-volatile memory to store its program content, and fast, low-power memory for data storage and manipulation. These requirements led modern microcontrollers to consist of two types of memory: most commonly, SRAM and flash.
Due to its unique properties, FRAM-based microcontrollers can use only one memory type instead of a combination of SRAM and flash. This is a significant step in the right direction for flexible memory architectures because FRAM eliminates the partition between SRAM and flash on traditional microcontroller devices. FRAM provides unified memory with dynamic partitioning, allowing for full customization of the memory to fit the requirements of any application. Instead of being confined by a fixed amount of SRAM and flash on a device, the FRAM can be partitioned into any ratio of data and program size.
|Figure 1: FRAM unified memory with dynamic partitioning allows for full customization to fit the requirements of any application.|
Unified memory provides many advantages, particularly for applications that use a significant amount of traditional RAM. In a traditional microcontroller, SRAM and flash sizes typically scale larger together. RAM-intensive applications usually require a microcontroller that has both a large amount of SRAM and flash, causing a large chunk of flash to go to waste. Wasted flash is a severe drawback as it increases the physical size of the die as well as the cost. These large memory devices are also larger in size with a higher pin count. Many applications need a large amount of data memory without the additional scaling of other aspects of the microcontroller.
These applications include audio and video, LCD, signal processing and math-intensive designs. Common to these applications are large data buffers that consume traditional RAM memory, but it doesn’t need to be one straightforward operation using RAM. Having several moderate RAM users can also push an application into this realm. For instance, one application that consists of driving an LCD display, running a real-time operating system (RTOS) and running a USB or wireless stack can quickly consume large amounts of data space.
FRAM enables flexibility of design by no longer restricting memory options to the small sample of predetermined RAM and flash sizes of traditional microcontrollers. As an application evolves, so does the memory footprint by simply sliding the divider between data and program memory. This flexibility during the design phase allows designers to get the product to market at a much faster pace. Additionally, the device can accommodate future applications. The process of adding functionality to an existing product, or even major application overhauls during a field firmware update, no longer faces the limitations imposed by memory boundaries.
FRAM memory enables more than just flexible, unified memory. Its characteristics enable drastic improvements in low-power applications. When comparing FRAM memory to flash, memory access requires 2.5 times less power. This is a significant savings, but more importantly, FRAM memory access is 100 times faster! To see how both of these characteristics truly affect an application, total energy instead of power needs to be evaluated. From an energy standpoint, FRAM requires 250 times less energy than flash does!
This ultra-low-energy technology enables a wide range of improved applications from portable, battery-powered applications to data-logging and wireless applications. An energy-harvesting application incorporates several features that make FRAM an ideal candidate.
|Figure 2: FRAM memory cuts power to reduce batteries in an application.|
Energy harvesting applications will typically log data and transmit the data via wireless protocol. As shown above, writing equivalent blocks of data to FRAM requires 250 times less energy than flash. So why not just use the SRAM of the traditional microcontroller to avoid this energy penalty? Most of the power used in an energy-harvesting application comes from the wireless transmission of data. Because SRAM is volatile memory, data would have to be transmitted each time it was received to avoid data loss in the event of power loss. This isn’t a practical use case for an energy-harvesting application.
FRAM also provides several other features that stand above flash. It takes so little energy to write to FRAM memory that in the event of power loss, the data write is guaranteed to finish. A flash-based architecture would require a very large external capacitor or battery to finish a data write. FRAM also has nearly unlimited write endurance compared to 10,000 writes for flash. Flash often requires wear-leveling algorithms to avoid reaching its endurance limit. If that is not enough, FRAM memory is bit addressable like SRAM allowing for added manageability. Flash memory is not bit or byte addressable; it is required to be written in segments. To avoid data loss, redundant blocks of data are often used in case of power loss during a segment write. These redundant data blocks and wear-leveling algorithms require additional flash memory for the same amount of stored data. FRAM suffers from none of these shortcomings.
These advantages of FRAM allow for reduced bill-of-materials, smaller memory footprint and significant power savings. In mobile applications, the battery life can be extended or replaced with a smaller and cheaper battery. These system simplifications aid the developer and allow for a product that is more robust, sustainable and lower cost to build.
FRAM memory is reshaping the landscape of microcontroller applications due to its makeup as one memory that provides the best benefits of SRAM and flash architectures. FRAM is such a revolutionary technology that microcontrollers are now used in applications where they were previously never found. With countless ways for FRAM to be leveraged, this technology is ushering in an exciting future for the world of microcontrollers.
Michael Stein is an embedded applications engineer focusing on TI’s MSP430 MCU customer applications. His responsibilities include designing and engineering system-level hardware and software using MSP430 MCUs, providing training and support to TI’s customers and field teams as well as providing applications expertise during development of next generation MCU products such as silicon, tools, software and more. He holds a bachelor of science and master of science degrees in electrical engineering from the University of Colorado.