Reference designs reduce design risk, says Sondrel
A family of reference designs, released by ASIC company, Sondrel, could reduce design costs, risk and time by up to 30 per cent, compared to starting from scratch, claims the company.
“In a given application area, there is always considerable duplication in the design of each ASIC as there is a lot of communality in the interconnections and supporting IP that forms the architecture of the device,” explained Sondrel CEO, Graham Curren. “Rather than start from scratch with each new design, we have created reference designs that distils our experience of designing the architecture for such chips to create reusable IP platforms,” he continued.
“Onto this, we add the customer’s IP with some customisation to create a bespoke solution . . . This reduces the overall design costs and risk as our IP platform is tested and ready to use,” Curren said.
The company calls this approach Architecting the future.
“Potential customers for ASIC are often concerned that a custom ASIC will be very expensive,” admitted Ian Walsh, Sondrel’s vice president of worldwide sales. “Our semi-custom, reusable IP platforms not only reduce costs but also makes it much easier to give indicative costs for customers . . . Our experience with each reference design means that we can estimate the ballpark costs for design, IP licensing, foundry, test, qualification and packaging right through to the total cost per unit,” Walsh said.
Sondrel will be releasing more than five IP platforms in this family this year, which target advanced nodes to provide cost effective devices. Two will be targeted at advanced driver assistance systems (ADAS) and three will have scalable amounts of processing power to address different application areas.
The first design architecture to be announced is the SFA 200. It is targeted at fixed and mobile (battery powered) applications, such as smart homes, smart metering, sensor fusion and other applications where a compact chip can add processing intelligence with local end-point data processing for data collection and analysis as well as inference processing. The design can be arrayed to implement scalable processing applications if required.
Sondrel also offers a full turnkey service that turns designs into fully tested, shipping silicon.