Arm’s mission is to grow the entire IoT ecosystem by providing the secure architecture that underpins end-devices, and as such, we are in a unique position in the industry insofar as being broadly connectivity agnostic.
Having recently returned from meeting and talking with global silicon partners, infrastructure vendors, carriers and the end device users of LPWA technology, I have been reflecting on what I have learnt, and contrasting it with some of the commentary by industry pundits of the state of LPWAN technology.
Here are 6 common themes that have surfaced about Cellular IoT:
I have read several articles claiming that Cat-M1 (more correctly “LTE-M”) and NB-IoT are proving slow to be adopted, while the non-cellular technologies such as Sigfox and LoRa are establishing themselves as the de facto LPWA deployment standard. There are multiple elements to consider here:
I would reflect that with few exceptions, the technology industry has invariably set over-optimistic expectations of how quickly wireless technologies are adopted, but over the long term always under-forecast the eventual volume shipments of those technologies. What has been a constant theme is that it is the open wireless standards that eventually reach commercial success. Wi-Fi, Bluetooth and 3G are tangible examples of open standards where the early hype was met with a resounding lack of actual commercial traction, but once the ecosystems developed and commercial uses cases became clearer, the respective successes now speak for themselves.
I also believe that any open LPWA technology will need to achieve greater than 100M individual node deployments to robustly establish itself as a self-sustaining critical mass, and so enable economies of scale. There is a parallel here with digital ecosystems, and analyst Richard Windsor asserts that a digital ecosystem (illustrated using Spotify as an example) needs a minimum of 100M daily active users to have a fighting chance to achieve critical mass, and additionally it is only when 300M active users are achieved that the ecosystems assets can be profitably monetized. While this is not a like-for-like comparison with LPWA, I believe the issue of critical mass and subsequent profitable monetization are equally applicable to LPWA technology, and the 100M deployed nodes threshold is broadly in line with the experiences of other wireless technologies.
From my recent China visit, I met the national carriers, chipset vendors, end vertical users of LPWA technology, and also had the opportunity to attend the GSMA’s MWC Shanghai. It was clear to me that NB-IoT and Cat-M1 are firmly establishing themselves as key LPWA technologies, and LoRa is also enjoying early commercial traction. The GSMA noted at MWC Shanghai that globally 37 carriers and 27 vendors are now supporting cellular LPWA, and there are 56 active cellular LPWA pilots covering applications such as meters, parking, bike sharing and security panels. There are now globally 9 commercially live Cat-M1 or NB-IoT networks from 7 major carriers.
While these GSMA statistics and other recent industry announcements are encouraging, what brought home the reality of cellular IoT’s prospects to me was the recent news that the Chinese government’s ministry responsible for telecommunications (Ministry of Information and Information Technology - MIIT) issued a policy that states that 400K cellular base stations are to be NB-IoT capable by the end of 2017, and 1.5M by 2020. These infrastructure enablement numbers translate to an MIIT stated target of 20M and 600M NB-IoT nodes in 2017 and 2020, respectively.
As we have seen previously in China, ministry policies have a galvanizing effect upon the domestic ecosystem, and it would be a brave person to wager such scale of deployment numbers is not possible.
This is ‘fake news’ to coin a topical phrase. While some industry commentators have been keen to create a VHS vs Betamax narrative, the two technologies have attributes that address different use cases and are not fundamentally overlapping.
NB-IoT is suited to simple sensor node applications, where the connected ‘thing’ is typically stationary and the use case does not require more than a few 10’s of bytes of data transferred intermittently throughout the day. Smart parking, smart utility meters, agricultural sensors and smart municipal lighting are use cases that are well suited to NB-IoT.
Conversely, IoT applications requiring higher data rates, mobility or voice, such as digital signage, real time asset tracking and alarm panels, are Cat-M1’s sweet spot. While there will always be use cases that fall in the grey area, these two technologies are very complementary, and this was also the very clear message I heard from carriers and infrastructure vendors in my recent discussions.
Arm therefore believes that to address the full spectrum of IoT applications, single mode and multimode cellular LPWA solutions together with non-cellular LPWA will need to coexist.
Most of the cellular LPWA silicon solutions available today (or in development) have their engineering roots in pre-existing higher category LTE modems (Cat-1, Cat-4), and have not been designed from the ground-up as single mode Cat-M1 or NB-IoT solutions. Additionally, the Cat-M1 standard supports enhanced functional capability relative to NB-IoT (i.e. higher data rate, voice and mobility), and these features necessitate higher processing and greater memory requirements than NB-IoT.
There has been a narrative developing in some industry quarters that NB-IoT ‘comes for free’ when part of a Cat-M1 solution i.e. the incremental impact to silicon die area by adding NB-IoT to a multimode Cat-M1+NB-IoT solution is modest. This may be a broadly true observation, but conversely therefore, NB-IoT must be a more cost effective standard relative to Cat-M1 due to its lower processing needs and smaller subset of features it supports which translates as a smaller silicon die area.
Certainly, this is Arm’s assertion, and we will revisit and expand upon this perspective in a future article.
While this headline makes for exciting reading, it is extremely disingenuous. I have spoken with multiple global silicon vendors, carriers and infrastructure vendors to explore this reported interoperability ‘issue’.
Consistently I have heard from those at the coal face of testing the technology that the interoperability ‘issues’ are a result of the natural time offset between the device and infrastructure vendors’ software development cycles. This development cycle offset between the ecosystem was exposed when some non-backward compatible change requests (CRs) were introduced into 3GPP's Release 13 March 2017 specification update. This issue will disappear over the coming months as the device and infrastructure vendor software deliveries synchronise, and is a great example of multiple stakeholders co-working to ensure global interoperability, and is not indicative of any inherent lack of robustness of the NB-IoT standard.
Incidentally, vendor development cycle misalignment for newly introduced or updated standards is not limited to cellular LPWA; one major infrastructure told me that this is a natural occurrence for all new cellular standards evolution, and cited recent interoperability issues related to LTE Category 16 smartphone platform testing which have not made headline news (because those in the industry actually working on this understand this is a natural phenomenon in any complex multi-vendor technology development).
Are Cat-M1 and NB-IoT power efficient enough to enable 10 year battery life? There is a 3GPP technical note (TR45.820) that proposes a methodology to calculate battery life for Cat-M1 and NB-IoT. This makes certain well-informed assumptions for receiver, transmitter standby and deep sleep power consumption (for a given power amplifier and efficiency, and using a 5Whr battery).
The analysis also considers the impact of three different coverage scenarios (i.e. link budget) , which I simplistically refer to as ‘Good’, ‘Typical’ and ‘Bad’ coverage. The 3GPP analysis also considers example IoT applications requiring either once daily or 2 hourly sensor data updates of 500 bytes or 200 bytes to the network (for reference, a residential smart meter application would typically update the network once daily and require no more than around 100 bytes of data upload).
The 3GPP analysis demonstrates that in ‘Good’ coverage and low sensor update rate use cases, the node does not need to frequently transmit (or retransmit) data, and in fact the battery life is dominated by the standby and deep sleep power consumption. In these cases, a 5Whr battery life comfortably exceeding 10 years is theoretically possible for NB-IoT.
Conversely, if the node is deployed in a ‘Poor’ coverage scenario (e.g. deep indoors or at a cell edge), or, if the application mandates sensor data updates every few minutes, say, the repeated data transmissions may not allow for much greater than 1 year life of a 5Whr battery for NB-IoT. In these instances of ‘Poor’ coverage or multiple sensor updates use cases, the node’s receive and transmit power consumption will be the dominant contributors to battery life.
For reference, a typical battery for these applications could be a AA or D-type lithium cell, and commercially available cells readily achieve greater than 3GPP's assumed 5Whr. Small lithium ‘button’ cells could in theory be used, but around a dozen connected in parallel would be needed to deliver the required peak transmit current for NB-IoT or Cat-M1. It should be noted that the 3GPP calculations do not consider self-discharge (around 1%/year for lithium cells) or temperature effects (self-discharge doubles for each 10 degrees centigrade increase in temperature). These real-life battery effects will serve to incrementally erode the overall battery lifetime. Incidentally, while NB-IoT is a more power efficient LPWA technology than Cat-M1, in certain cases of poor coverage, this benefit can be eroded.
In conclusion, cellular LPWA technologies can enable battery life exceeding 10 years, but in deployment this will be a function of several variables, and the discussed challenges of poor coverage or frequent sensor data update rates will place similar constraints upon non-cellular technologies also. Since the differing LPWA technologies are each better suited to different use cases, it is not insightful to take a broad-brush comparison approach to battery life.
Given enough scale and volume, economies of scale can be leveraged in the deployment of any technology and drive down cost. In the case of ‘traditional’ cellular M2M, 2G GPRS modules are available today for well under USD$5.
But the cellular industry has had over 25 years and several billion unit shipments of 2G chipsets for the silicon and module pricing to hit maturity. However, the industry cannot wait another 25 years to start benefiting from the attributes of cellular LPWAN-connected ‘things’, so it is my assertion that vertical integration of the LPWA module supply chain is needed to deliver a revolutionary – and still profitable - step change in pricing sooner.
The traditional higher category and higher data rate 3G or LTE modules have not led to vertical integration of the supply chain because the end use cases such as laptop computers, surveillance cameras or automotive connectivity, have typically been able to sustain higher module pricing. For the aggressive price targets for LPWAN, ‘stacking’ of silicon and module gross margins, the need for external SIM components and RF front end passive components are all areas that I believe the industry must collectively focus to aggressively drive down the pricing – hopefully some visionaries will step up to this challenge, and Arm is actively exploring how it can support innovators within its partnership in the pursuit of this goal.
Incidentally, if you are wondering why 2G GPRS isn’t the answer to LPWA given that it addresses many key attributes of IoT use cases (e.g. voice capability, mobility and highly mature costs), there are two key reasons: 2G spectrum is extremely valuable to carriers, and it is being re-farmed globally for higher data rate applications, and secondly, 2G’s power consumption is not as competitive as the cellular LPWA technologies, which have been specified at inception with dedicated power saving and discontinuous reception modes.
IoT is not really about a specific technology, it is about using the appropriate technology to enable viable commercial business models that solve real world problems. There is no one-size-fits-all connectivity solution for IoT, and the industry needs to accept and embrace this.
PAN, LAN, WAN, licensed and unlicensed connectivity will all need to coexist and will complement each other, so the “my connectivity technology is best” argument is not only false, it is actually misleading and confuses the very end verticals our industry is targeting to help (i.e. utilities, agriculture, logistics, building management, insurance etc).
LPWA connectivity for sure has an important role to play in the wider context of IoT, and Arm will continue to ensure it enables secure, power efficient processor and connectivity solutions to help enable the broader IoT ecosystem.
Learn more at www.arm.com/cordio