Monday, January 7, 2019
After an unusual two-year delay, silicon for a new Wi-Fi standard is starting to emerge. Over the next few months, a handful of startups will sample chips for 802.11ah, a 900-MHz version of Wi-Fi targeting long-range links especially for the internet of things.
The so-called HaLow products promise delivery of up to Mbits/s over distances of tens of meters to a kilometer and support for thousands of nodes on an access point. They will occupy a space between ultra-low-power and -cost LoRa and Sigfox networks and below more power-hungry LTE Cat-M and Narrowband-IoT networks that come with data plans.
Some of the initial products will be single chips made in 40-nm processes, not likely to carry large price premiums. They represent a new breed of IoT options for unlicensed bands that support both IP networking and OFDM modulation familiar to the broad PC ecosystem. Long term, they will eke out a position as a fourth band extending the range of links for smartphones and PCs.
“HaLow stands out for its versatility in enabling new business models as well as capabilities in capacity, range, and battery operation, all of the attributes the market needs — it just needs an ecosystem to emerge,” said one wireless veteran who asked not to be named.
The Wi-Fi Alliance announced HaLow at CES 2016 as the outcome of the 802.11ah process kicked off with significant help from Qualcomm engineers in 2014. Usually, by the time an IEEE 802.11 spec is finished, several top Wi-Fi chip vendors are already shipping silicon in a race to the market — but not this time.
Two years since the standard was finished, only one company, Korean startup Newracom, is said to have shipped a chip seen in a gateway from Korean Telecom. The startup, formed by a Wi-Fi expert from Korea’s ETRI research center, got its start selling IP blocks for existing standards like 802.11n.
Four other startups around the world aim to sample HaLow silicon next year. Morse Micro, a Sydney startup formed by Wi-Fi experts from Broadcom and Radiata, is one of the most impressive of the group. Established Wi-Fi chip vendors such as Broadcom, Cypress, Huawei, Mediatek, Qualcomm, and Realtek may follow rather than lead this time.
Despite early enthusiasm for .11ah, the initial timing for the HaLow release was bad. In 2016, the next big version of mainstream 2.4-/5-GHz Wi-Fi, 802.11ax, was also ready. “There were serious design decisions companies had to make — no one could afford to miss the .11ax cycle,” said the wireless veteran.
The choice was an easy one. The .11ax products would clearly rise quickly to sell millions of units. Companies such as Qualcomm and Mediatek were also ramping LTE IoT and 5G products, leaving them no resources for HaLow. In addition, bringing Wi-Fi to the unlicensed ISM bands would require some missionary work.
you go to a new band, it takes time,” said the wireless veteran. “Five gigahertz was a 10-year process, and it took .11ac to move it to widespread use,” while 60 GHz Wi-Fi is still getting on its feet.
“It’s a tough road to grow a new ecosystem around a new band … HaLow is in the same spot, so it will take time,” he added.
Others note that in 2016, a lot of startup energy was getting pumped into an emerging class of low-power, wide-area networks. LoRa, Sigfox, and other IoT networks for the ISM bands were on the rise, and the LTE IoT standards — Narrowband IoT and Cat-M — were in the wings. There weren’t investment dollars or entrepreneurial energy left for HaLow.
Today, the big players are expected to get into the market around August when the Wi-Fi Alliance launches its HaLow certification program. The trade group will lead up to that milestone with a series of plugfests starting this month (January).
The case for 802.11ah in a crowded LPWA space
In two white papers, engineers at Morse Micro made the case for HaLow.
Although it has gained backers, including Comcast, LoRa is limited to data rates typically below a few kilobits/second — not enough, they say, to support over-the-air updates needed to keep up with changing security standards. LoRa’s relatively “low network capacity limits the number of sensors [that] a base station can service,” it said.
More established 802.15.4 networks such as Wi-Sun and Zigbee require more power but still don’t support large networks of sensors on a single base station, added the white paper.
For its part, HaLow should deliver at least 160 Kbits/s over at least 100 meters and longer with line of sight. What’s more, it can accommodate more than 8,000 nodes on an access point, said Michael De Nil, co-founder of Morse.
So Morse targets a range of IoT uses ranging from farm fields to utility meters, warehouses, supermarkets, and remote video surveillance cameras. “We could be selling millions of units a year by the end of 2019,” said De Nil.
accelerate its ramp, Morse is opening an office in China.
“It’s a great market, with very large OEMs who move fast … we will sample chips in July, and within six to eight months, they will be in volume production,” said De Nil. “For U.S. or European OEMs, the process will take until 2020 or even 2021.”
HaLow is not too late for the IoT market, argue backers. “The IoT market size and growth was over-hyped five to 10 years ago,” said the wireless veteran. “We’re still in the early stages of IoT market development, which is good news for HaLow.”
“LoRa and Sigfox are very low throughput, they have their space, and they have traction, but for TCP/IP and internet-like security, discovery, and upgrades, you need more throughput — for the real IoT, you need a technology like HaLow.”
IoT is just the start. Long term, HaLow will become a block in mainstream chips for access points and clients, bringing them new distance and low-power capabilities, say proponents. It’s a typical chicken-and-egg scenario, familiar in semiconductors, of new standards going from nowhere to everywhere.
Three more HaLow startups in the wings
Besides Morse and Korea’s Newracom, said to be the furthest along of the HaLow startups, three other startups aim to sample chips in the next year or so.
Methods2Business has IP for a HaLow media access controller based on a Tensilica DSP. Founded in 2010 in Eindhoven, it also has an engineering team in Serbia.
In Silicon Valley, Adapt-IP has developed an FPGA version of a HaLow baseband. It is considering a partnership to design a complete HaLow chip by early 2020. The startup has been seeking its way forward in the slow-to-emerge IoT market since 2014.
Another startup, Palma Ceia SemiDesign (PCS) got its start doing RF contract design. That led to selling some of its own IP blocks for NB-IoT and HaLow transceivers. Now, the company aims to raise funds and roll out a HaLow device in 2019 as its first chip-level product.
PCS is already profitable from its design and IP business but is raising funds to support its first chip tapeouts. It aims to sample before June chips, probably made in a 40-nm node, for a simple HaLow access point supporting a limited number of end nodes as well as a single-stream end-node chip.
“We may not be first, but we will be fast followers … we look to a future where HaLow and NB-IoT chips may be integrated,” said Kevin M. Walsh, vice president of strategic marketing for PCS.
Like Morse, PCS has set up a China office in the Shenzhen area where its chips are being designed, close to module makers who will be among its first customers. It will make its chips at TSMC, the foundry for which it has historically designed its IP.
On Morse, RISC-V, and Sydney’s Wi-Fi enclave
The Morse design incorporates 1 to 8 Mbytes flash, uses private keys for secure boot and cloud authentication, and fits in a 6 × 6 QFN package. De Nil boasts that it was created from a clean sheet of paper with novel filters and accelerators and no third-party IP blocks.
“The .11ah standard has stringent radio requirements with relatively wide bandwidth, so wide dynamic range is needed to deal with interference and OFDM requirements to get the best performance and lowest power,” he said.
The design uses multiple 32-bit RISC-V cores, customized to act as wireless signal processors. “Using open-source cores has really accelerated our design … [without them,] we would have needed Arm processors and negotiated royalties — instead, we took open-source RTL and added accelerators,” he said.
The design takes a “processor-driven approach,” said De Nil. “Traditional Wi-Fi designs tune hardened IP to reach the highest bandwidth tuned. We’ve done a few generations of test chips, so now we believe we can scale this approach to other IoT protocols such as Zigbee and NB-IoT long term.”
As for RISC-V, “the quality of the open-source code that’s available is really high, it’s surprisingly mature … now we need more students who know how to use RISC-V tools to generate systems,” he said.
The startup’s headquarters is in Sydney, home to two early Wi-Fi pioneers, Neil Weste and John O’Sullivan, who now work for Morse. De Nil and co-founder Andrew Terry worked in a local Broadcom office as digital and RF design leads, respectively, for multiple generations of Wi-Fi chips for the Apple iPhone.
Several engineers on the Morse team were behind startup Radiata, which built in the late 1990s one of the first 5-GHz Wi-Fi chips for the .11a standard. Cisco later acquired Radiata, but much of the Wi-Fi talent remained in Sydney.
“Broadcom is a good company and culture, but it gets boring cranking the handle on the next-gen iPhone chip every year,” said De Nil. “Eventually, the company stopped innovating and was just generating next-gen IP for large customers.”
“We saw they weren’t really addressing IoT that would be massive … six months after the Avago acquisition, it [looked like a] good opportunity, and we met [Radiata founder] Neal Weste, who joined us,” said Terry.
“We were up for the challenge of putting what we knew to good use and creating some interesting circuits.”
Morse started as a self-funded company with six people in August 2016 and later got some angel funding from semiconductor experts. It’s now kicking off a $15 million A-series fundraising to fuel plans to hire about 30 people and take its chips to production before the end of 2019. Its name refers to Morse code, the pioneer of long-range wireless communications.
Copyright © 2018 CST, Inc. All Rights Reserved