Memory flexibility key to FPGA-based designs
Bob O'Donnell, President and Chief Analyst of TECHnalysis Research, LLC for Lattice Semiconductor, discusses the importance that memory technologies, particularly FPGAs, have in embedded designs.
While much of the tech world is typically focused on all the latest cutting-edge advancements, the truth is that a huge portion of the devices powered by semiconductor technology often use components that can be several years old. The recent challenges that the automotive industry faced, for example, were primarily focused on what some refer to as 'lagging edge' or 'trailing edge' chips, based on older manufacturing processes.
The world of embedded devices powered by FPGAs is even broader and deeper than the automotive industry, so it’s particularly important to be able to work with an extremely wide range of different components. In some instances, FPGA-powered devices use some of the newest technologies available, while in others, they can leverage chips that are as much as 20 years old.
One area that’s particularly important for embedded device designers is the choice of memory technologies. Many modern FPGAs incorporate some amount of SRAM (Static Random Access Memory) within their designs, allowing certain algorithms or other embedded software to run within the FPGA, but for many applications, engineers need to combine them with various types of external DRAM (Dynamic RAM) to be able to achieve the functionality that a particular device or application requires.
In a modern image sensor-based device, for example, the ability to do real-time AI-driven processing of high-resolution video signals to do something like identifying different types of motion, for example, requires the use of DRAM for things like frame buffering of the incoming signals.
The trick with external memory is that there are an enormous range of different types, different performance characteristics, capacities, power draws, and so on. As a result, it’s essential to be able to connect to as many of these different memory types as possible. One of the most common types in current use is LPDDR4 (Low Power Double Data Rate, version 4), which provides device designers with a great range of different capacities/densities, speeds and power requirements that can be matched to specific applications on which they may be working.
Because of its low-power nature, it’s particularly well-suited for mobile and other battery-powered devices but can also be used to build low energy-consumption-focused devices in industrial/automotive, as well as in applications where thermal management is challenging.
Even though the initial spec for LPDDR4 first came out in 2014 (and the LPDDR5 spec was introduced in 2019), it’s still extremely popular among many device designers because of its flexibility and lower cost. LPDDR4 is now the most popular DRAM in industrial and embedded systems due to optimal power, cost and supply longevity. Plus, it offers a big improvement in power efficiency over the older but still commonly used LPDDR2 and LPDDR3-type memories.
As a result, memory companies like Micron offer an enormous range of different LPDDR4 components to meet the varied demands of the market.
As part of its strategy to modernise the small capacity FPGA market, Lattice recently introduced the Lattice CertusPro-NX family, which is currently the only FPGA family in its class to support LPDDR4. As a result, engineers are now able to leverage not only the faster speeds, but the lower power and lower frequency options that can be critically important for embedded devices.
What’s exciting about this combination is that it brings the ability to achieve powerful capabilities such as machine vision and other smart camera-type applications to a much broader range of devices. In particular, it’s now possible to bring AI functionality to smaller, cheaper, and lower power devices, extending the range of applications where this exciting, powerful new technology can be used.
From a memory maker’s perspective, like Micron, broader support for LPDDR4 in more host/processing devices like CertusPro-NX FPGAs, means the company has an even broader range of customers to which it can sell its memory components. For device designers, the new LPDDR4 support means they can leverage the enormous range (and longevity) of these popular and well-established memory technologies, knowing that they will have a stable, long-term supply for devices that have long manufacturing lifetimes.
Ultimately, it’s critical to remember that the world of embedded devices continues to grow and diversify at an impressive rate. As a result, it’s true win-win-win for all parties involved when the range of component choices and means of interconnecting them continue to evolve and expand. The combination of LPDDR4 memories with FPGAs can clearly enable a powerful range of applications that can be specifically developed for a wide variety of different environments. For device designers looking for flexibility, it’s an exciting new time.