CEVA unveils Dragonfly NB2, but now we have to learn more acronyms


CEVA has revealed the successor to its Dragonfly NB1 silicon design, supporting the new 3GPP Release 14 standards. The followup includes a GNSS hardware package, for adding satellite positioning to your designs, as well as CEVA’s ClearVox voice-recognition IP. So while CEVA is claiming that this is the world’s first licensable Release 14 offering, it does mean that we’ve got some more names to learn.

The chip designer isn’t to blame here, for the problem of the nomenclature. CEVA is hoping to snare more customers for the silicon, looking to build on the success of the Dragonfly NB1, which it says has been widely licensed for use in smart cities, transport and logistics, and consumer electronics. However, when asked, CEVA is not saying how many chips that actually comprises. ZTE’s Rosefinch line is a confirmed customer.

So then – the new terms. While the two variants were being worked on for Release 13, there was LTE Cat-M1 and LTE Cat-M2 – occasionally referred to using the umbrella term LTE Cat-M (meaning both), and sometimes using MTC (Machine-Type Communications). Respectively, after Release 13 was published, they were then badged as LTE-M and NB-IoT, although many still used Cat-M or Cat-M1 to denote LTE-M.

Now with Release 14 out in the wild, there are new names to learn, thanks to the enhancements made to both LTE-M and NB-IoT. We appear to have reverted to using ‘Cat,’ with Cat-NB1 (Release 13 NB-IoT), Cat-NB2 (Release 14 NB-IoT), Cat-M1 (Release 13 LTE-M), and Cat-M2 (Release 14 LTE-M). The Release 14 versions incorporate the enhancements made in the standards, but rather annoyingly for those who will inevitably use a search engine, Cat-M2 now has a double meaning.

There are also catch-all terms to be aware of now. Starting at the top, ‘Cellular IoT’ refers to all types, but specifically, ‘Massive MTC’ refers exclusively to Cat-NB1 and Cat-NB2. Frustratingly, there is a further delineation to accommodate. Cat-M1 is now apparently also called eMTC (Enhanced MTC), and Cat-M2 is FeMTC (Further Enhanced MTC). Seriously. Numeric versioning would have been so much easier. Can we at least stick with this scheme now? So that when Release 15 fiddles with the specs they just become Cat-M3 and Cat-NB3? And why weren’t we doing that from the off?

Moving on – the Dragonfly NB2 centers around CEVA’s X1 processor, which handles both the DSP functions (turning analog signals into digitally readable ones, and vice versa), includes the aforementioned GNSS block, the RF transceiver, the power amplifier (PA), and on-chip memory and flash controller. CEVA reckons this combination, which doesn’t need another dedicated application processor and also supports eSIM and USIM, ensures the lowest possible BOM costs. It’s a similar package to Altair’s new ALT-1210, which also houses a GNSS element.

CEVA says the design is proven in both 55nm and 40nm production processes, and includes new power management and sleep mechanisms, for ultra-low-power sleep modes – a few micro amps (µA). The faster speed of Cat-NB2 also helps with power consumption, as the devices can spend less time on air sending messages.

Emmanuel Gresset, Head of Business Development in CEVA’s Wireless unit, spoke to us about the launch. He explained that this extra speed, up to 160Kbps compared to Cat-NB1’s 60Kbps, was a noticeable improvement, but that the extra speed hasn’t really changed the use cases – just improved the performance.

The Massive MTC section, comprising Cat-NB1 and Cat-NB2, remains most comparable to the unlicensed LPWAN (U-LPWAN) like LoRa and Sigfox. With that target ten-year battery life and duty cycles of around 1% (spending almost all their time asleep), Gresset said that Smart Grid was a big volume play today, but that other like asset tracking and environmental monitoring were emerging – very different from Cat-M1 and Cat-M2.

Pointing to Ericsson’s recent Mobility Index, Gresset noted that it was a bit odd that there was apparently no growth of the Cellular IoT (licensed LPWAN, L-LPWAN) volume in India. Compared to the rest of the world, which Ericsson thinks is going to grow to 3.5bn L-LPWAN devices by 2023, India doesn’t seem interested – perhaps because of relatively poor LTE network penetration, which might favor the U-LPWAN options. Nonetheless, Gresset said that the huge deployment of NB-IoT in China could possibly be replicated in India.

Gresset also noted that there is significant upheaval in the chip supply chain, as chip makers move closer to becoming module makers. He pointed to Nordic Semiconductor’s first non-Bluetooth chip, a Cat-M1 and Bluetooth combo, as evidence of this – adding that this is a clear move to target the likes of Telit and Sierra Wireless. Gresset, and CEVA, is also seeing OEMs and MNOs exploring developing their own modules and chips, mostly out of pricing. For the likes of CEVA, these companies, which don’t have any experience in silicon, are great opportunities.

CEVA believes that NB-IoT will soon be a drop-in option for chip designs, akin to how Bluetooth and WiFi are often paired in a combo chip. Industrial designs are apparently a potentially lucrative avenue, with the likes of Microchip and Infineon turning to a company like CEVA to easily add a cellular option to their portfolios. In addition, the company thinks that once prices for NB+GNSS silicon gets low enough, there will be a sharp rise in volumes.

Gresset adds that Cat-NB1 and Cat-NB2 is around half the cost of Cat-M1 and Cat-M2 currently, and uses around a third of the power. He notes that only dual-mode (M + NB) chips are available today, and that these are not yet low-cost enough. In terms of competition with U-LPWAN, Gresset stressed that Quality of Service (QoS) and OTA updates are key, and that customers are especially insistent on the latter. Lower latencies are also much desired, which is driving interest in L-LPWAN.

CEVA believes that most MNOs will upgrade to support Cat-NB2. Gresset said that the speedy update to Release 14 came about because of the problems with power draw in Cat-NB1, where sleep-state power usage was too high, compounded by the relative high consumption of the PA when using such a slow data rate. Gresset said this is why no operators are going to properly commercialize Release 13, and will move very swiftly up to Rel. 14.