Design and verification standards are critical if we want to get a new generation of Internet of Things (IoT) devices into the market, according to panelists at an Accellera Systems Initiative breakfast at the Design Automation Conference (DAC 2015) June 9. However, IoT devices for different vertical markets pose very different challenges and requirements, making the standards picture extremely complicated.
The panel was titled “Design and Verification Standards in the Era of IoT.” It was moderated by industry editor John Blyler, CEO of JB Systems Media and Technology. Panelists were as follows, shown left to right in the photo below:
- Lu Dai, director of engineering, Qualcomm
- Wael William Diab, senior director for strategy marketing, industry development and standardization, Huawei
- Chris Rowen, CTO, IP Group, Cadence Design Systems, Inc.
In opening remarks, Blyler recalled a conversation from the recent IEEE International Microwave Symposium in which a panelist pointed to the networking and application layers as the key problem areas for RF and wireless standardization. Similarly, in the IoT space, we need to look “higher up” at the systems level and consider both software and hardware development, Blyler said.
Rowen helped set some context for the discussion by noting three important points about IoT:
- IoT is not a product segment. Vertical product segments such as automotive, medical devices, and home automation all have very different characteristics.
- IoT “devices” are components within a hierarchy of systems that includes sensors, applications, user interface, gateway application (such as cell phone), and finally the cloud, where all data is aggregated.
- A bifurcation is taking place in design. We are going from extreme scale SoCs to “extreme fit” SoCs that are specialized, low energy, and very low cost.
Here are some of the questions and answers that were addressed during the panel discussion.
Q: The claim was recently made that given the level of interaction between sensors and gateways, 50X more verification nodes would have to be checked for IoT. What standards need to be enhanced or changed to accomplish that?
Rowen: That’s a huge number of design dimensions, and the way you attack a problem of that scale is by modularization. You define areas that are protected and encapsulated by standards, and you prove that individual elements will be compliant with that interface. We will see that many interesting problems will be in the software layers.
Q: Why is standardization so important for IoT?
Dai: A company that is trying to make a lot of chips has to deal with a variety of standards. If you have to deal with hundreds of standards, it’s a big bottleneck for bringing your products to market. If you have good standardization within the development process of the IC, that helps time to market.
When I first joined Qualcomm a few years ago, there was no internal verification methodology. When we had a new hire, it took months to ramp up on our internal methodology to become effective. Then came UVM [Universal Verification Methodology], and as UVM became standard, we reduced our ramp-up time tremendously. We’ve seen good engineers ramp up within days.
Diab: When we start to look at standards, we have to do a better job of understanding how they’re all going to play with each other. I don’t think one set of standards can solve the IoT problem. Some standards can grow vertically in markets like industrial, and other standards are getting more horizontal. Security is very important and is probably one thing that goes horizontally.
Requirements for verticals may be different, but processing capability, latency, bandwidth, and messaging capability are common [horizontal] concerns. I think a lot of standards organizations this year will work on horizontal slices [of IoT].
Q: IoT interoperability is important. Any suggestions for getting that done and moving forward?
Rowen: The interoperability problem is that many of these [IoT] devices are wireless. Wireless is interesting because it is really hard – it’s not like a USB plug. Wireless lacks the infrastructure that exists today around wired standards. If we do things in a heavily wireless way, there will be major barriers to overcome.
Dai: There are different standards for 4G LTE technology for different [geographical] markets. We have to make a chip that can work for 20 or 30 wireless technologies, and the cost for that is tremendous. The U.S., Europe, and China all have different tweaks. A good standard that works across the globe would reduce the cost a lot.
Q: If we’re talking about the need to define requirements, a good example to look at is power. Certainly you have UPF [Unified Power Format] for the chip, board, and module.
Rowen: There is certainly a big role for standards about power management. But there is also a domain in which we’re woefully under-equipped, and that is the ability to accurately model the different power usage scenarios at the applications level. Too often power devolves into something that runs over thousands of cycles to confirm that you can switch between power management levels successfully. That’s important, but it tells you very little about how much power your system is going to dissipate.
Dai: There are products that claim to be UPF compliant, but my biggest problem with my most recent chip was still with UPF. These tools are not necessarily 100% UPF compliant.
One other concern I have is that I cannot get one simulator to pass my Verilog code and then go to another that will pass. Even though we have a lot of tools, there is no certification process for a language standard.
Q: When we create a standard, does there need to be a companion compliance test?
Rowen: I think compliance is important. Compliance is being able to prove that you followed what you said you would follow. It also plays into functional safety requirements, where you need to prove you adhered to the flow.
Dai: When we [Qualcomm] sell our 4G chips, we have to go through a lot of certifications. It’s often a differentiating factor.
Q: For IoT you need power management and verification that includes analog. Comments?
Rowen: Small, cheap sensor nodes tend to be very analog-rich, lower scale in terms of digital content, and have lots of software. Part of understanding what’s different about standardization is built on understanding what’s different about the design process, and what does it mean to have a software-rich and analog-rich world.
Dai: Analog is important in this era of IoT. Analog needs to come into the standards community.
Richard Goering
Cadence Blog Posts About DAC 2015
Gary Smith at DAC 2015: How EDA Can Expand Into New Directions
DAC 2015: Google Smart Contact Lens Project Stretches Limits of IC Design
DAC 2015: Lip-Bu Tan, Cadence CEO, Sees Profound Changes in Semiconductors and EDA
DAC 2015: “Level of Compute in Vision Processing Extraordinary” – Chris Rowen
DAC 2015: Can We Build a Virtual Silicon Valley?
DAC 2015: Cadence Vision-Design Presentation Wins Best Paper Honors