TL; DR: For more than three decades, Lattice Semiconductor has been building innovative technologies for consumers and an array of industries worldwide. Today, the company is leading the field-programmable gate array (FPGA) manufacturing space, developing products with a focus on cost savings, low power consumption, and small form factor. With these budget-friendly, energy-efficient solutions and the company’s supplemental programming software, Lattice is pioneering edge computing and making it easier than ever for engineers to turn their ideas into reality with tiny — yet powerful — silicon chips. All signs point to machine learning and artificial intelligence (AI) as the technologies that will drive the Internet of Things (IoT), and Lattice’s solutions are providing a roadmap to more quickly bring this future to the world.
When topics of artificial intelligence (AI) and machine learning come up in conversation, many people envision post-apocalyptic landscapes, like scenes out of The Terminator or The Matrix, where malicious robots have taken over the planet. But, if you ask most computer scientists and engineers, these visions are best left to the realm of science fiction.
Studies show that nearly half of modern businesses plan to leverage machine learning to gain better competitive advantages. In the most probable future use cases, the technology will deal with predictive analytics that help save maintenance costs associated with running equipment. Machine learning is also changing how we live in relation to the growing Internet of Things (IoT), with computers being employed in our homes to run various appliances and devices.
Sitting right at the cutting edge of all this technology is Lattice Semiconductor, which for 30 years has been building products that are helping shape our future. The company’s line of FPGAs enables flexible, secure connectivity and computing for a variety of IoT applications at the Edge.
This kind of edge computing relies on real-time sensor data processing — and FPGAs are perfect for the job, given the inherent parallelism and programmability benefits. Lattice’s FPGAs are optimized for the highest operations per second at the lowest possible power consumption levels, meaning these devices are ideal for use not only for traditional algorithm-based processing but also for the artificial neural network architectures that will power the smart machines of the future.
“Our strategy has always been to focus on low power and small form factor FPGAs,” said Deepak Boppana, Lattice’s Director of Product Marketing. “Lattice FPGAs are optimized for connectivity and computing at the Edge.”
Lattice offers a wide range of programmable arrays with varying lookup tables and speeds suitable for any electrical engineering need. Lookup tables significantly reduce computational time due to a more simplified structure with no need for a complete input/output (I/O) operation. The bottom line: Lattice’s solutions save space, optimize power consumption, and, most importantly, cut costs.
How Lattice Optimizes Power & Cost in Connectivity & Acceleration Apps
Lattice offers programmable logic with its ECP, iCE, CrossLink, and MachXO product lines. The company is the only manufacturer to provide FPGAs at such a small form factor combined with low cost and power consumption. These are commonly used in connectivity and acceleration apps because of their versatility and efficiency.
The capabilities of FPGAs lead to excessive value in the long run. They can be reprogrammed with ease, whereas many arrays can only be hard-wired once without some kind of significant hardware change. CrossLink is the first programmable bridging solution for digital cameras and provides support for 4k Ultra HD for speeds up to an astounding 12 Gbps — typically running at less than 100 MW. The MachXO Family trades lower costs for higher I/O densities. Many of these chips offer glue logic and bus bridging, making it possible to lighten processor workloads with logical filtering.
The iCE40 Ultra line is the world’s smallest FPGA category. Between its lookup table range of 640 to 5,280, flexible signal positioning, and outputs for RGB and IR LED lights, these are commonly used in barcode scanners, infrared-powered remotes, and light controllers.
“Lattice’s product portfolio includes a range of low-density FPGAs optimized for use in mobile and Edge applications, including smart homes, smart cities, smart factories, and smart cars,” Deepak said. “In 2015, we expanded our portfolio to include wired and wireless ASSPs.”
Unlike the more versatile FPGAs, ASICs and ASSPs are programmed to be more application-specific. These products complement one another to accelerate computation. ECP5 chips cost an average of 40% less than competing FPGAs and include an improved architecture that features dual channel SERDES (serializer/deserializer) to compensate for low I/O. The Lattice ECP5 family effectively breaks the rules of power, size, and cost for applications such as automotive systems and micro-servers.
Focus on Low Density Helps Leverage Mainstream Process Technologies
Lattice provides the most comprehensive portfolio of low-density FPGAs that run on extremely low power for an efficient price. A focus on low density means Lattice can manage a great deal of power at a very low cost of entry.
“Given our focus on low-density FPGAs, we can leverage some of the mainstream process technologies out there,” Deepak said.
Smaller arrays save room for sensors and other input and output features, which is crucial for handheld devices like smartphones. Specifically, the iCE40 FPGAs slash power consumption up to 100 times, while making always-on sensor integration a reality for context-aware devices.
Powerful Design Software to Help Bring Ideas to Market Faster
Lattice’s design software and IP help simplify the design process for FPGA programmers. Loaded with pre-tested functions, Lattice’s IP lets engineers test the programming in a forgiving environment.
The IP suite comes loaded with nearly 300 blueprints for anything from hot swap controllers to video interface technology.
Lattice also offers supplemental programs that give engineers an edge over the competition, including Lattice Diamond, iCEcube2, and ispLEVER Classic. Lattice Diamond Software features a user-friendly interface and Tcl scripting to tailor to the most intuitive design flows. The iCEcube 2 Design Software helps optimize ultra-low density FPGAs for power and cost using simulation and synthesis inferencing. Meanwhile, ispLEVER Classic is used for Lattice CPLDs (complex programmable logic devices) and allows for an in-depth walkthrough of the complete design process.
Embedded design software, such as LatticeMico System, which is used with microcontrollers, is available, too. Mixed signal devices can be simulated in PAC-Designer, while ORCAstra is used to debug Lattice FPGAs.
Lattice Moves Forward with a Continued Focus on Edge Applications
Ever focused on the future, Deepak told us he envisions a world where Lattice’s FPGAs drive AI and machine learning. It’s not the scary world of The Terminator or The Matrix; rather, it’s more reminiscent of the utopian scenes of The Jetsons or Star Trek, where machines work to make our lives easier.
“We see a trend of intelligence moving from the cloud to the Edge devices due to privacy, latency, and network bandwidth limitations,” he said. “Lattice FPGAs are ideally suited to implement energy-efficient machine learning solutions at the Edge, while also providing the flexibility to adapt to the rapidly evolving algorithms and architectures. It’s very exciting.”
Edge computing pushes computation to the literal edges of a network, meaning smart devices must work independently and yet cooperatively with one another within the Internet of Things. This will give us greater interconnectivity of devices and embed connections in everyday objects, including vehicles, phones, sensors, and much more.
Since 1983, Lattice has been providing the world with low-density programmable logic devices. Now, more than three decades later, engineers are programming these devices to do what was once unimaginable — stretching the limits of exploration and ingenuity to create machines that think, cars that drive themselves, and homes, factories, and cities capable of self-regulation with smart sensors.