5G Infrastructure Needs Chips!
September 16, 2020
Open RAN looks like our best bet for a competitive 5G infrastructure market. But existing implementations are too expensive. We need more chips and suppliers to build DU and RU units for 5G at scale.
5G won’t take off without significant infrastructure. And it’s no secret that the dominant supplier of that infrastructure is now politically unpopular.
What other options do we have?
Even before politics became a factor, network operators were looking for better solutions. The classical radio access network (RAN) – radio units and a lot of baseband electronics at the foot of each cell tower, connecting through backhaul to central stations – is already hard to scale. This requires huge investment per tower, each servicing highly variable loads throughout the day.
To scale up 5G, operators need to drive down costs through better utilization and more competition between equipment providers. This has motivated new standards, particularly Open RAN (O-RAN), which should stimulate off-the-shelf hardware and open interfaces. While O-RAN is a very appealing option, there’s a gap in scaling this market – we need more chips to support hardware builders and encourage new builders to enter the market.
An Irresistible Force
There’s a lot of business momentum behind making O-RAN scale. The network operators are certainly motivated, but there are new players who are also pushing. Rakuten, the company best known for their e-commerce ads, is nearing a full O-RAN launch in Japan in the hope that it will channel more e-commerce traffic directly to them. Meantime Dish Network has committed to an O-RAN network covering 70% of U.S. users by 2023. Facebook is pushing hard through their Telecom Infra Project, no doubt expecting that, like Rakuten, the more traffic that can be pushed through the RAN, the more their advertising revenues will rise.
Naturally all the established players are active as well, Nokia, Ericsson, Samsung and others.
Just as important is the political momentum. The U.S. Government, represented by the FCC, is very bullish on O-RAN, for obvious reasons: it stimulates competition (which is good for everyone) and it avoids having to use the dominant supplier (which we are told is bad for national security).
Plugging a Hole in the Supply Chain
This all sounds perfect. So why isn’t it already up and running? Because there’s a gap in the supply chain – a shortage of innovative chips on which those infrastructure equipment solutions can be built at scale.
Without getting too deep into O-RAN technology, in a simple view you have an antenna, which connects to a radio unit (RU) that handles the front end of the physical-layer functions. That connects to a distributed unit (DU), which performs the rest of the baseband functions. This connects through mid-haul to a centralized unit (CU), which then connects through backhaul to the core network (I should add that this is one of several possible topologies).
With O-RAN, clearly specifying the interfaces between the RU, DU, and CU units (until now proprietary for each of the incumbent RAN vendors) will allow operators to mix and match units from different vendors. The interfaces are open, but the hardware and software running on each can still be vendor specific, allowing vendors to compete and differentiate.
How can that hardware be built? Dedicated chips provide the best price/performance and have been the option of preference for the big equipment providers so far, but custom chips require a lot of investment of capital and expertise. Or you can use standard server platforms for the CU and DU, but those are expensive and difficult to scale to the demanding massive MIMO dimensions we expect from modern 5G networks. FPGAs are a possibility in the RU modem, but are also expensive and burn a lot of power.
The obvious answer to this problem is to develop more custom chips to support O-RAN. To create more competition around features, price, and performance.
There’s a healthy ecosystem in this area, from secure virtualization platform suppliers to CU and DU hardware suppliers, RU hardware and software suppliers, virtualized RAN software, and system integrators. We just need more chip platform options.
Those chip makers must build on proven infrastructure in the O-RAN space, such as DSP-based platforms that can be a cost-efficient alternative to current COTS-based solutions. Such platforms, building on a long legacy of expertise and success in wireless communications, have already been adopted by leading infrastructure equipment OEMs. It is time to move to the next phase in the O-RAN evolution by using state-of-the art DSPs optimized for baseband processing and allowing them to compete head-to-head with traditional RAN solutions.
Whether you plan to build DUs or RUs, you might be surprised how far DSPs like the CEVA-XC16 baseband processor can take O-RAN infrastructure while offering a smoother, more cost-effective migration path to 5G baseband IP platform scalability than standard COTS platforms.