Don Dingee Published on 01-12-2016 11:00 AM
Kids have a tendency to put things in their mouths. Any parent can relate to the statement, "Put that down! You don't know where it's been!" After the first child, concern usually relaxes quite a bit. People joke about a 5-second rule on the premise if an object was just dropped on the floor, it may not be contaminated yet.
Should there be a 5-second rule for IoT chips? OK, maybe it's a 5-microsecond rule, but the point is we don't know where those chips are going to be or how they are going to be contaminated. Unsuspecting users are going to pick them up, put them in their living spaces and use them for highly precious personal information, with no idea what they are actually doing.
I see a big difference between functional verification of IoT chips and truly validating them. The oversimplified question of:" Are we building the right product?" is becoming increasingly difficult to answer as products become interconnected and ingrained in personalized use. The standard of care is also changing; an IoT product only has to mess up once for a consumer to be permanently harmed. That same consumer can then go out on social media and really do some damage to a company's prospects.
Frightened yet? Designers should be. We went through years of the dreaded blue-screen behavior; if it rolls over and stops working, reboot it. We accepted that, strangely, but we didn't really have a lot of choice. We also got drilled by all the technology market window theory that we had to build complex products more quickly, or some competitor would come in and take our share. Late to market was one of the worst labels a product could get, and it got managers fired.
Your odds of being burnt go up dramatically if you're not testing IoT chips very, very thoroughly. Beckwith, who works for Progressive, isn't as well-known as Flo, but he’s just as sharp and spot on. Allowing consumers to put these IoT chips that we just slapped together in their "mouths" or other sensitive places and find out what happens is asking for trouble.
How do we avoid this fate? There are no guarantees, of course, but FPGA-based prototyping is a cost-effective tool that enables earlier and more extensive testing and exploration of both hardware and software in IoT chip design. Where many firms have been driving hard for bigger FPGA-based prototyping systems with multi-FPGA partitioning technology, most of the IoT chip designs are small and fit in a single FPGA.
S2C has gone back to this low end with IoT parts in mind in introducing the Single KU115 Prodigy Logic Module. With one Xilinx Kintex UltraScale XCKU115 FPGA, it is a lower cost solution that fits with the framework of S2C tools for standalone or cloud use with up to 16 modules. It supports addition of the S2C daughter cards for I/O, and the ProtoBridge AXI software for transaction-level testing.
Wouldn't a non-UltraScale device be even cheaper? There are several advantages to using the XCKU115 on a prototyping platform for IoT parts. First is the advanced clock management, which allows designs to be carved into more domains. Second are 48 high speed transceivers and 624 I/O pins, spread out on 8 high-speed connectors on this S2C module for connectivity.
Third are the DSP resources, with 5520 slices. Many IoT chips aren't going to be jellybean control parts wiggling a few I/O pins and an A/D, but instead a class of more capable parts with algorithms that speed up data, particularly for real-time analytics and localized modeling. S2C is making a bet that will likely pay off later in enabling these concurrent connectivity and algorithms to be explored deeply by IoT chip design teams pre-silicon.
The companies that figure out the right IoT algorithms and connectivity and get it in highly reliable silicon will win, especially if competitors who rush to market with sketchy, less tested parts get clobbered. The next 5 seconds on the IoT may prove really, really critical.