Freescale Semiconductor has succeeded in cramming an entire cellular base station onto a single chip. It’s a claim many chipmakers have made, but other “base-station-on-a-chip” designs have focused primarily on small cells and femtos. But at Mobile World Congress on Monday, Freescale revealed it has reduced the baseband capacity of a big honking tower-based macrocell to a system-on-a-chip (SoC) design.
That’s not only an impressive feat of miniaturization and integration, it could kick off the next-generation of LTE deployments, lower the costs of building mobile networks and reduce the energy required to run them. When all those factors are taken into account, base station SoCs could cut the cost of delivering a bit of data, which ultimately could lead to cheaper mobile data plans for the consumer.
The brains of a base station typically reside in a channel card, a sort of hopped-up motherboard designed to perform the extremely complex task of encoding and decoding radio signals. Usually wireless infrastructure vendors like Ericsson and Alcatel-Lucent design those channel cards as a bunch of discrete components: digital signal processors, applications processors, and a variety of hardware accelerators. The new QorIQ Qonverge design stamps all of those discrete components onto a single piece of silicon.
So why is this significant? SoCs are much cheaper to manufacture than the sum cost of all of those separate components, and also drain far less power. According to Scott Aylor, director and GM of Freescale’s wireless access division, a single QorIQ chip can support the capacity of a three-sector 20 MHz LTE cell site – the same configurations Verizon and AT&T are using in their new 4G networks – for one quarter of the cost. Aylor also said the highly integrated platform also drains three times less power, which will help operators design more energy-efficient networks. If operators can build cheaper networks and cut their operating costs, they could theoretically offer mobile broadband at cheaper prices.
That performance and power efficiency will make QorIQ a building block for future network technologies such as LTE-Advanced, Aylor said. LTE-Advanced will require enormous processing resources as the bandwidth pumped out by each cell grows well beyond 100 Mbps. “We’re well ahead of the LTE-Advanced curve,” Aylor said.
Furthermore, Freescale customers like Alcatel-Lucent and Nokia Siemens Networks are already exploring a radical shift in network design. Known as cloud radio access network, or cloud-RAN, it seeks to divorce the base station from the cell site. Instead, operators could build signal-processing farms in a private cloud, in essence virtualizing their base stations. Whenever capacity is needed, cell sites – which are little more than radio heads at this point – would reach into the cloud and grab it. Again, Aylor said SoCs would be ideal for such a scenario.
“We can build farms of 64 of these things on a single card, then daisy-chain them together,” Aylor said. “There’s not a significant limitation on our side as to how far we can scale.”
Aylor said the key to developing the SoC was Freescale’s utilizing new 28-nanometer process technology, allowing it to condense a lot more performance in much less space. That means Freescale’s competitors can’t be far behind as most of them are already going down the 28-nm path.
Digital signal processing giant Texas Instruments has already developed powerful SoCs for use in Cloud-RAN platforms and is using them to power smaller-sized cells. Aylor, however, said Freescale believes it has significant advantage over its competitors as it can supply a single-chip solution across the board, from the lowliest femtocell to the most powerful macrocell.
Not all network vendors subscribe to the SoC approach, Aylor admits, but Fujitsu and Alcatel-Lucent are already converts. The Franco-American networking giant is already using QorIQ chips in its new lightRadio architecture, initially in its Cube small cells, but it plans to begin designing its macro base stations around the new SoCs as well.