Verification Trends Enabling A 5G Future

Applications have driven requirements for verification for quite some time now, as I have written previously regarding Aero & Defense, AI and Machine Learning and the Internet of Things. In wireless communication, we are just at the brink of the transition to Fifth Generation Networks, or 5G. This transition will not only lead to new applications and use models that will impact our day-to-day lives, but also lead to a new set of verification challenges.

Since mobile phones were just that in the 1990s—phones that we could take with us—we have experienced significant advances in communication networks from 1G analog all the way to 4G networks with up to 50Mbit/s data rates with less than 100ms latency. The next-generation 5G networks promise huge advances in the three areas of data rates, latency/reliability, and the number of devices per square kilometer. It’s often referred to as the 5G Triangle, as shown below.

The 5G Triangle (Source: Bigstockphoto, Cadence)

Specifically, 5G will enable the following capabilities, as compared to the existing 4G networks:

• 10X more bandwidth for enhanced mobile broadband with peak data rates of up to 20 Gbit/s
• 50X better latency, down to 1ms combined with high reliability of 10^–9 error rates
• Up to 1 million device connections per/km² with high-energy efficiency

The 5G network has multiple layers. Endpoints like mobile devices and IoT “things” must take multiple radios into consideration and require ultra-low power consumption in cheaper and smaller form factors. They connect to base stations, for which besides throughput, low power is a major challenge and leads to important architecture considerations. Continuing the path of data, at the edge of the network, the requirements include low latency, high throughput, and low power. Given that the targeted low latency requirement is data transmission from the user device to the cloud—and back—in 1ms or less, many devices need to interact efficiently and especially the location of processing becomes hugely important. In the core network and in the cloud, the key requirements are to allow the flexibility of the software to perform specialized applications such as data manipulation and to provide software management to enable field updates, all with as low power as possible.

While each layer of the network has its unique challenges, three main trends for verification of the 5G future emerge:

• Low power verification and optimization both for mobile components living on battery power as well as plugged-in devices for which thermal effects are important
• “Systems of systems” verification ensuring that devices, networks, cloud edges and clouds all work together correctly
• Architecture analysis and performance verification needs to answer questions way beyond just pure single participants in the layered network, like handling of high-bandwidth broadband video and thousands of new device connections that may suddenly appear within a coverage area

I have written on Systems of System Emulation before, and it is really no surprise that, for example, low power verification and optimization and mixed-abstraction level execution combining simulation and emulation are gaining more and more importance, as was shown at the recent CDNLive China. For instance, HiSilicon reported on “Accelerating Performance Exploration with a Hybrid Platform,” combining Palladium Z1 Emulation and transaction-level models. They described how performance and low power become the two biggest challenges in consumer chips. To address the requirement of performance exploration, they develop transaction-level models (TLM) to simulate a variety of scenarios. Given that it is hard to develop all TLM models, HiSilicon built a hybrid platform combining TLM simulation and RTL in emulation to cover more scenarios. Many software issues were resolved during the bring-up.

Both HiSilicon (“The Emulator DPA’s Using in the GPU’s Power Big Data”) and Phytium Technology (“Application and Expansion of DPA in Power Consumption Assessment”) described their experiences with low power using Palladium Z1 and using Cadence Joules RTL Power Solution to connect activity data gathered in emulation with power information gathered from .lib files describing the silicon technology. Phytium won the best paper award, concluding that DPA combined with the Joules solution and Cadence Palladium Dynamic RTL (DRTL) capabilities not only were crucial to their assessment of power consumption, but also enabled power-driven design optimization, power-driven placement, and power network design based on IR drop analysis.

Both companies, of course, target 5G and the cloud quite heavily. CDNLive China confirmed how low power, system of systems and architecture/performance are the key drivers for verification of devices targeting that application domain.