Adapt, Galvanizing IoT Edge Device Deployment

The semiconductor industry today is filled with inspired ideas for useful IoT edge devices, a ready-made opportunity for new approaches and strategic changes to building cost-sensitive, low-power SoCs.

Despite commendable intentions, IoT is an immature market with little of significance created yet. Some of that is because designers are experimenting. Others with domain experience don’t have the expertise, resources or interest to attract new “Things” to the internet and implement the necessary hardware devices. On the other end of the spectrum, clever entrepreneurs have too little design expertise or resources to build an economical SoC to deploy at a large scale. They need savvy engineers, as well as investors to foot foundry costs. They have other hurdles. In some cases, good ideas are too expensive because an IoT edge device often is only viable as a single-chip implementation. The cost to produce an SoC limits any experimentation, especially in the IoT space where networks must be built using many small devices.

What’s needed is a practical way to make a single or small number of these devices and have them connect to each other naturally and effectively.

Adapt, a Lanza techVentures portfolio company, may have a solution. Its vision is to develop SoCs efficiently enough to make it practical to implement low-cost, minimum viable devices to support applications with flexible, non-rigid specifications before real-world experiences.

At the center of the Adapt vision is WiFi HaLow, otherwise known as the IEEE standard 802.11ah wireless networking protocol. Through the eyes of Lucio Lanza, managing partner of Lanza techVentures, HaLow will enable computer signals to become citizens of the Internet by connecting cost-effective and economical processors.

Any profile of Adapt must include its original creator John Sanguinetti, Adapt’s CTO. John founded Chronologic Simulation, developer of VCS, the fast Verilog simulator. From there, he founded and served as CTO of Forte Design Systems, a high-level synthesis (HLS) tool company, acquired by Cadence Design Systems in 2012. John’s goal for Forte was to bring in the system-level view to make design as easy to implement as possible. He believed Verilog was too low-level and proposed SystemC for design entry as a way to raise the level of abstraction for the designer.

According to John, “Adapt’s goal is to create generic SoC platforms that can be easily modified to accommodate new, experimental applications.” Adapt’s first IoT platform implements WiFi HaLow along with a base level of processing power. The platform can be augmented with additional logic to produce custom SoCs at a low design cost.

Michael McNamara, CEO

John Sanguinetti, CTO

Lanza techVentures is an investor because of John Sanguinetti, says Lucio who he knew from being Forte’s board chair. “Adapt is almost the next generation of Forte –– high level, consumer-style description and automatic implementation that can be moved to different implementations including custom ICs when success requires and justifies it.”

Indeed, Adapt uses a high-level methodology to reduce difficulty and costs for design creation and verification with results that dramatically improve design efficiency over traditional methods. The cost and effort of producing variations of existing designs using high-level design tools usually is reduced by more than an order of magnitude.

Adapt isn’t a traditional service provider that needs a set of deliverables, such as Verilog source code. More Internet innovators are not experienced enough to do something like this, nor do they want to learn such a complex methodology.

Instead, Adapt provides a higher level of design so the user can specify what the chip does without concern for how it’s implemented, not the traditional EDA-centric environment. The user knows how he or she wants to make the device a citizen of the Internet and Adapt can reduce the barrier between them and the chip implementation.

Adapt could be considered the interface to getting the chip done at the user level and not the foundry’s level through its high-level design flow. It uses SystemC, a programmatic description to implement a generic platform built around a HaLow communications model. HaLow, power efficient, long-range WiFi that can go for more than a kilometer, is the latest in a succession of wireless protocols. Because the chip is generic, devices can be produced cost effectively, and it can be the core of an even cheaper custom chip that integrates sensors and a communications platform. The Adapt HaLow platform offers enough processing power to make it a viable edge device by itself, and the basis of a range of efficient IoT edge devices.

Adapt’s design methodology was refined on a succession of standard communication devices over a period of years. It started life in 2009 as HighIP Design, an IP company building a USB3 device controller using HLS, a novel approach 10 years ago. The founder visited John who was intrigued by the proposition and started helping while continuing to work for Forte. He provided funding, as did Lucio.

Lanza techVentures is an investor because of John Sanguinetti, says Lucio who he knew from being Forte’s board chair. “Adapt is almost the next generation of Forte –– high level, consumer-style description and automatic implementation that can be moved to different implementations including custom ICs when success requires and justifies it.”

The USB3 project took longer than expected. The small group of engineers had to pipe clean the design by starting with USB2 with the plan to reimplement the USB2 design in SystemC to USB3. The design was executed in an FPGA and the demo worked and was demonstrated at the 2010 DAC.

Meanwhile, Cadence acquired Forte and John was able to devote his attention full time getting the USB3 IP fully operational. After two years and confident of the engineering efforts, John recruited Michael (Mac) McNamara, a colleague from Ardent and Chronologic who was a key technologist. Mac founded SureFire Verification that was acquired by Verisity, then folded into Cadence.

Adapt HSoM Family

At that point, Mac was managing the Cadence C-to-Silicon group off the Cadence campus with a group of former Cadence Berkeley Labs engineers and experienced the resistance to HLS adoption. However, both he and John found that IP designers used HLS to great effect. In fact, they believe IP design may be the perfect application for HLS. In their experience with users, HLS’ primary value is the ability to create a variation of an already implemented IP design. One large Japanese company, for example, uses High Level Synthesis for all its IP designs.

This revelation convinced Mac as CEO that the company should be called Adapt, not HighIP Design. Mac brought other changes as well, including two other technical contributors who liked what Adapt was doing and wanted in.

Along about the same time, the Internet of Things was catching on an and one paradigm anticipated sensors spread out across the world with a collector that periodically gathered the data. It would need to be controlled viably, cheaply and maintenance free. The device would have a button battery that would last several years so it could be thrown away or abandoned when it no longer worked. The ideal device would be deployed by the thousands and made from a single SoC. It’s a problem unless there’s an application that’s going to succeed. Designers with the idea don’t know how to make the chip. They would rather spend their time and resources addressing larger problems they are trying to solve than attempting to develop the expertise required to do the implementation.

By chance, John and Mac met an entrepreneurial technologist looking at the IoT market who believed HaLow was a great opportunity. He joined Adapt and Adapt provided the environment to develop the IP.

Adapt devoted all its efforts on the wireless protocol, a big job to be sure, John noted, but one that could fulfill the IoT paradigm. Initially, the work was done through a partnership between Adapt and two other companies. Adapt worked on the new baseband. One partner focused on the radio, the other on the MAC layer. After a while, the partnership ended and Adapt went it alone.

The first IoT platform SoC for custom IoT edge devices –– the Wi-Fi HaLow Development Platform (HDP) –– was launched earlier this year. It delivers data more reliably, securely, over longer distances and more cost effectively than any other electronic data transmission to meet edge device requirements. HDP is meant for system integrators and application developers targeting the industrial IoT market. The HDP includes a radio from Analog Devices and an FPGA which contains all of the custom logic. Adapt is also making the move to become a fabless semiconductor company to be able to provide ICs and an IP module for SoCs. The plan is to tape out a 22nm WiFi HaLow chip at 925megahertz signal frequency as more advanced nodes aren’t necessary for this type of application. Tape-out of the first chip will occur after market experience with the HDP.

Excitement is growing around Adapt’s HaLow design. Its FPGA-based system enables WiFi data across long distances, addressing key requirements for large-scale industrial IoT deployments, including long range, low power, low cost and excellent data rate. A huge differentiator is its ability to identify edge devices within one meter of accuracy. This capability is especially interesting to the construction industry where logistics are all important, and could be used in agriculture, logistics, manufacturing and retail also.

Adapt HaLow Development Platform

Of course, HLS is a large part of the equation. Adapt engineers are writing less code than they otherwise would and are able to do the same kind of designs with four engineers, not 40. The advantage of HLS is the derivatives business, as John and Mac noted all those years ago.

Adapt and another Lanza techVentures portfolio company Efabless share a similar vision to accelerate SoC design while reducing cost. Adapt is creating the base platforms for SoCs as Efabless reduces the design costs through its online marketplace with open source tools and IP bundles.

Once Adapt and Efabless reach their goals, they will enable millions of IoT edge device deployment that otherwise would not be economically viable …

… and the Internet will expand to become the Internet of Things. Finally.