A huge advantage for LTE in this space is that solutions of this type can piggy-back on top of existing and widely available infrastructure which can only continue to improve and expand. For this reason, the cellular-based IoT market is expected to grow at a healthy clip, 34% annual CAGR over the next 5 years. And those growth rates don't include applications using short-range solutions today which may choose to switch to cellular.
Also interesting is how operator response to these new standards varies by region. In the US, AT&T and Verizon are deploying CAT M1 now. They apparently will also support NB-IoT but not for 1 to 2 years. China and Korea are aggressively going after NB-IoT and will introduce next quarter, with no plans yet for CAT M1. Meanwhile Europe is all over the map 😎, some operators starting with NB-IoT deployment, others with CAT M1. Overall this gives interesting insight into who/what is driving priorities in each region.
Particularly in these narrow-band applications, IoT nodes must be ultra-low power and ultra-low cost. For example NB-IoT targets a 10+ year battery life in user equipment. And cost pressure for IoT devices is always acute, perhaps no better exemplified than in asset-tracking applications where a device may sometimes need to compete with an RFID alternative. This suggests for this class of applications that it may not be enough to simply adapt SoC architectures but that architects should be looking at more aggressive ways to reduce cost, complexity and time to market.
CEVA has an interesting and aggressive proposal, based on their CEVA-X1 solution. First, you need a processing core for the radio, so why do you need another core for general processing? Certainly, a challenge to the dominant paradigm but you can’t argue with the area logic – one core will be smaller than two. They also say you can remove hardware accelerators thanks to a few dedicated IoT instructions they have added to cover operations which are usually not efficient in a DSP. They also have some interesting ideas for reducing RAM footprint through more extensive use of embedded flash (reducing need for RAM) along with clever caching.
In general for these devices, where latency is not critical they suggest heavier use of software rather than hardware in the spirit of minimizing area and complexity. They point out that typical functions - modem, GNSS and sensor fusion - can easily multiplex on one processor core. Also, compressing a data payload prior to transmission (to further reduce RF power) is a function readily provided on a DSP platform.
CEVA has partnered with NextG-Com Limited to offer packaged solutions to simplify the development of CAT M1 and Cat-NB1 platforms. These incorporate the CEVA-X1 single-core IoT processor together with NextG-Com's ALPSLite-M CAT M1 or ALPSLite-NB Cat-NB1 protocol stacks, in a highly compact and power-efficient manner. The solutions have also been carefully optimized to ensure smallest memory footprint, minimizing the requirements for embedded flash and static RAM.
CEVA still have some interesting questions to answer, particularly regarding security and provisioning for devices built along these lines. But no question they are thinking outside the box; with this class of solution they are offering a very low-cost and power-effective cellular IoT solution.
To learn more about CAT M1, CAT-NB and the CEVA-X1 solution, click HERE. You can also download the CEVA-X1 product note HERE.
More articles by Bernard...