SANTA CLARA, CA — While some of the biggest news out of last week's ARM TechCon centered on security for smaller IoT devices, plenty of other interesting technology was on display. Everything from SoC design IP to software agents for location services were represented. Here are ten of the interesting technologies I encountered while at the show.
Ada development coming to the IoT
AdaCore brought its development tools for the Ada programming language to the show, with a message for IoT developers: Ada is coming. “Our development tools have traditionally been used in the aerospace and defense industries,” AdaCore's lead of business development Quentin Ochem told EE Times at the conference. “But we are expanding into IoT and automotive.” Ochem explained that Ada may still have a poor reputation among many developers that stemmed from the early implementations, but these early problems are long resolved. Ada 2012, the most recent of three revisions since the original introduction of Ada, is a mature, standards-based language, Ochem noted.
Ada was specifically designed to address issues such as testing, quality assurance, porting to new platforms, and functionality upgrades to deployed software in an efficient manner, simplifying development and long-term maintenance of code. Because Ada tools verify software while it is being developed, Ada projects tend to be those where software quality and reliability are paramount. Now that avoiding software weaknesses that can compromise security is a rising concern for the IoT, Ochem said, Ada could increasingly find a home there, as well.
To that end, AdaCore is in the midst of developing a new product line crafted for the IoT market, Ochem said. He expects the official launch of those tools in late 2017.
Comtech Telecommunications demonstrated its Location Studio product at the conference, showing how IoT devices can gain access to location services without compromising BoM and constraints. Without a GPS system or external beacons, the IoT Location Platform (ILP) is able to provide devices with information on their place in the world, and the location-based behaviors like tracking and geofencing that the information enables.
The key to keeping the system's impact on an IoT device design is a divide-and-conquer approach, Comtech's Sameer Vuyuuru told EE Times. Rather than performing the location calculations on the device, he explained, a 2kB software agent simply captures information from the device's wireless links, such as signal strength and server ID codes, and sends a short (~200 byte) message to Comtech servers, which perform the calculation needed to turn that information into a location fix. From there Comtech can provide a variety of location-based services to IoT device users, including real-time mapping of device locations.
The approach works with whatever radio and other resources the IoT device has available, Vuyuuru explained, including providing dead reckoning if accelerometer data is available. Accuracy, of course, depends on those resources. A location fix based on a snapshot of GPS signals, for instance, would be more accurate than those based solely on Wi-Fi. Vuyuuru also noted that the software agent needs to be customized to the operating system the device uses in order to gain access to the data it needs.
Resistive RAM speeds past flash
Crossbar has been developing a non-volatile RAM technology based on the anti-fuse behavior of amorphous silicon that it calls resistive RAM (RRAM). That development is now paying off, according to company vice president Sylvain Dubois, with the technology reaching commercialization. Dubois told EE Times that the company has signed its first manufacturing partnership for its technology in a 40nm process and they are now licensing their technology to chip developers.
The RRAM cell is a vertical resistive layer of amorphous silicon between two metal pads. The device stores data by imposing a voltage across that layer to create controlled growth of a conductive “filament” through it. The filament does not fully short between the metal layers, however, which allows the process to be reversed to eliminate the filament. The difference between programmed and unprogrammed states is a 100:1 increase in current through the cell, making it easy to sense the cell's state. As this is a physical change in the cell, it remains intact when power is absent, making the memory non-volatile, and Dubois estimates that the technology will be able to achieve wear tolerance up to 1M cycles.
The RRAM's read/write circuit design allows an array to access a cell in around 20 nsec, and to write in 1-2 µsec with byte-level access, Dubois said. This speed provides a considerable advantage over traditional Flash in terms of access times as well as providing energy savings by eliminating the need to block erase in order to alter a memory location. The cell's structure allows it to fit entirely between metal layers, which provides a density advantage, including an ability to create 3D arrays.
To read the rest of this article, visit EBN sister site EE Times.