I’d like to avoid a semantics battle here. Historically, M2M has been mostly cellular, and IoT mostly other shorter-range wireless area networks. We could talk for a long time about topologies, structured versus unstructured data, coverage and mobility, traffic predictability, and other system-level differences between the two. What I want to do here is focus on one piece – low-end LTE-enabled radio technology and an overlap between the two camps.
Measuring just cellular M2M connections, we get something like 243M devices in 2014, and still growing – expected to be around 2B devices within 5 years. Most of that today is 2G or 3G modules from vendors such as Huawei, Gemalto, Sierra Wireless, Telit, u-blox, and ZTE. M2M is absolutely exploding in China driving much of the growth. 2G is sunsetting at AT&T and other carriers around the globe, which will set off a wave of migration to newer technology.
The reality is most M2M applications don’t need the massive bandwidth available at the upper end of the LTE range. They are perfectly happy on the mature 2G and 3G networks, and can usually absorb a moderate module cost and size. IoT edge devices usually don’t need a lot of bandwidth either, and can’t afford a lot of cost or power consumption.
One could certainly design a 4G M2M module with higher performance chipsets – something like digital signage comes to mind, where content is downloaded frequently – but that would be at the top of the pyramid and would miss almost all IoT applications. To hit a broader range of devices, lower power LTE chipsets are needed.
Fortunately, the LTE specification created a space at the low end where both M2M and IoT applications have a path to smaller, less expensive, lower power chips. Emmanuel Gresset at CEVA has assembled one of the more interesting pieces of slideware I’ve seen lately.
A couple of points make the demarcation even clearer. Voice support stops at Cat 1. Moving from full-duplex to half-duplex at Cat 0 reduces the RF front-end BOM cost considerably. Dropping the transmit power for Cat M makes CMOS die integration of the radio possible. While NB-IoT is 18 months or more away from initial rollouts, the opportunity for devices to jump right into a high-90s national coverage situation with a cost effective data plan sans voice would be huge.
Our readers noted the recent acquisition of Altair Semiconductor by Sony. Roaming around the Altair website and M2M vendors’ sites shows a lot of energy going into Cat 1 chips and modules right now. Not surprising since Cat 1 has network support today, and support for a voice channel helps many applications, particularly automotive.
What about IP, for companies designing their own parts? Last November we shared CEVA’s new CEVA-XC5 and CEVA-XC8 DSP cores. Now, CEVA is introducing Dragonfly, a reference platform adding software support for multiple standards including LTE Cat 1, Cat 0, and Cat M, plus Wi-Fi 802.11n/ah, ZigBee, LoRa, SIGFOX, Ingenu, NB-IoT, GNSS, and LTE-OTDOA (more on that in a moment). Dragonfly also supports wired PLC (power line communication) interfaces which often show up as requirement in many regions. CEVA says this is the first LTE Cat 0 IP platform, and the first multi-mode M2M platform combining these capabilities. There are actually two versions of Dragonfly, one with voice enabled LTE Cat 1 support, and the other targeting the Cat 0 and Cat M (expected to need a software upgrade when the specification is ratified) applications.
One of the big pluses of Dragonfly is concurrency. Requirements may call for running LTE Cat 0, Wi-Fi 802.11n, and a PLC interface simultaneously. That capability is a function of both the power of the underlying low-power DSP and the software stack. It also points to a potential issue for chipset developers: regional or carrier differences. The advantages of a DSP-based soft implementation may prove very advantageous in this age where standards adoption and carrier support is in flux.
LTE-OTDOA stands for observed time difference of arrival, a new approach for indoor and urban canyon positioning – use cases that give GNSS fits. By grabbing three timing measurements from dispersed eNodeBs with “good” geometry (for example, the user is not very close to any of the towers), hyperbolae of the time difference can be plotted and intersected. (One guess who is behind this standard.) Combine that idea with a drone use case, and you get the demonstration CEVA is about to run at MWC 2016 on an emulated LTE Cat 0 network:
Being able to provide both a data uplink and downlink plus dual-mode positioning for a drone could help with the autonomous package delivery scenario.
The point here is CEVA has combined low-power DSP IP with software for these emerging M2M and IoT specifications to jumpstart development and protect designs if specifications and adoption are a bit tenuous in the early going. More in the CEVA Dragonfly press release.
I’ll say this again: it is starting to look as though LTE Cat M and NB-IoT could drive all the other IoT WAN standards into a niche, just like what high data rate LTE did to WiMAX. Spectrum availability and data plan pricing have to cooperate for this to happen. And, there is that question some well-intended reader asked me the other day: “What does 5G have to do with the IoT?” When these low data rate LTE categories solidify, the 5G specification will embrace them, and things will get a lot clearer down the road for M2M and IoT applications.
More articles from Don...