Functional splits are proving to be very divisive in ways that were never intended. The general idea is to determine which elements go into a radio unit (RU) at a mobile site and which are included in a box called the distributed unit (DU). This is the basic setup in an open radio access network (RAN), where the RU comes from one supplier and the DU from another. But proposals on the technical options have led to clashes and discontent inside the O-RAN Alliance, the group responsible for developing industry specs.
Much of the current unhappiness is blamed on Ericsson, the big Swedish manufacturer of mobile network equipment. After several years of activity, the open RAN community has settled on a functional split known as 7.2x for its open fronthaul spec (fronthaul describes the link between the RU and the DU). This creates a very simple RU, shifting most of the clever stuff into the DU. For less advanced and lower-cost radios, it seems to work just fine. Where it falls short, as far as Ericsson and several other stakeholders are concerned, is on massive MIMO, a 5G technology that bristles with antennas.
Simplifying the RU to the extent that 7.2x does would cause performance problems in massive MIMO networks, according to Ericsson. It is not the only company with this view, but its proposed fixes have been controversial. Critics believe they could hinder open RAN, and a few have now gone on record with their concerns. Some think Ericsson is deliberately trying to upset a technology it sees as a threat.
Its latest proposal, discussed in some detail at Informa's recent Big 5G Event in Austin, was for an uplink-only modification that would move some features from the DU to the RU, including the equalizer used to mitigate interference. Ericsson had backing from AMD, a chipmaker, as well as Nokia, its main Nordic rival, and ZTE, a Chinese competitor.
Yet the proposal in its original form was evidently unacceptable to other members of the O-RAN Alliance. Instead, the group has approved a compromise between Ericsson's proposal and another one put forward by US telco giant AT&T, chipmaker Qualcomm and French service provider Orange. Despite that approval, not everyone was delighted by the move.
South Korea's Samsung, the biggest of the Asian kit makers after Huawei and ZTE, is known to have abstained from voting out of concern that too many flavors of open fronthaul could lead to fragmentation of the ecosystem. Voting against the proposal would have been awkward because of the support it had drawn from AT&T and Orange, both of which are Samsung customers.
Another supplier that remained neutral during the vote was the much smaller Parallel Wireless, a US developer of RAN software. Steve Papa, its CEO, told Light Reading via email that his team's concern was over the backward compatibility of the new spec, although this is supposed to be guaranteed.
Japan's NEC made guarded comments when approached by Light Reading. "NEC is proceeding with the development of vRAN [virtual RAN] and considering a wide range of virtualization options in order to flexibly support the functions of future radio units," said the Japanese company by email.
Mavenir, another US software developer, declined to comment. But John Baker, its senior vice president of business development, has been highly critical of Ericsson's original proposal on LinkedIn, describing it as "delay tactics" in a previous post.
7.3 shenanigans
The backstory adds considerable intrigue. Sources including Nokia say Ericsson has previously been keen on more sweeping changes under the label of 7.3. "We have the O-RAN Alliance-compliant interface that is 7.2x on both uplink and downlink, and Nokia is in the camp, as are NEC, Fujitsu, Mavenir, Rakuten, Samsung and so on," said Mark Atkinson, Nokia's head of RAN. "Then there is the 7.3 camp, which is not an O-RAN-compliant interface, which is Ericsson, Huawei and ZTE, and maybe some others as well."
Ericsson denies any involvement with a 7.3 proposal, referring to it as 7.3x in emails sent to Light Reading. "It is a common misunderstanding in the industry – there has never been a 7.3x," said a company spokesperson. "Ericsson has never proposed a 7.3x split to the O-RAN Alliance. Ericson has only always proposed UL [uplink] improvements to 7.2x."
But Nokia is not the only one that links Ericsson to 7.3 and a desire for more dramatic changes. A source who requested anonymity vouched for Atkinson's comment about an Ericsson-backed 7.3 split, saying it would entail both uplink and downlink changes. And earlier this year, the O-RAN Alliance's Working Group 4 (WG4) produced a technical report comparing 7.2x and 7.3, according to that source.
Its conclusions were that 7.3 would increase the complexity of the radio interface and be more power-hungry in all deployment scenarios, said the source. Light Reading approached the O-RAN Alliance about obtaining a copy of that report but was told "there is currently no published Technical Report from O-RAN WG4" by a spokesperson, who failed to respond to other questions.
The big concern about making the radio interface more complex is twofold. First, it could drive up development costs for suppliers, especially if a new spec does not come with backward compatibility. This would be far worse for smaller companies that lack the research-and-development budgets of the big kit vendors. Second, as noted by Atkinson, it could make building a genuinely multivendor network much harder.
"The more intelligence that moves into the radio, the more technically complex it becomes," he said. "Separation of intelligence between baseband and RF needs to be carefully managed because the features need to be aligned."
Rakuten, a Japanese ecommerce company that has recently set up a business selling telco network products, echoed some of those concerns in emailed comments. "While there is a DL [downlink] challenge with respect to bandwidth for centralized DU deployments as radios increase in capacity (massive MIMO), there have been discussions around the downsides of making the O-RU [open radio unit] more complex with additional active electronic processing demands on O&M, field maintenance costs, etc.," said a company spokesperson.
Multivendor mayhem
People still uneasy about the 7.2x uplink modification the O-RAN Alliance has approved see the latest version as a watered-down 7.3, including two flavors of that split as optional extras (one outlined by the Ericsson proposal and the other by the Qualcomm group) while mandating support for 7.2x. The key technical differences between 7.2x and what the industry refers to as 7.3 were recently described in a detailed blog by Paul Rhodes, the head of RAN for AtlasEdge, a UK provider of data centers.
"The difference is that functions such as channel estimation and Interference Rejection Combining (IRC), which are implemented inside the DU in option 7.2x, are implemented inside the RU in option 7.3," he wrote. "This brings lower fronthaul BW [bandwidth] requirements as more functions are done inside the RU and higher performance compared to option 7.2x due to channel estimation being done inside the RU."
The downside of all this is clearly spelled out by Rhodes. "Unfortunately, 7.3 requires powerful chips in the RU resulting in a higher cost RU relative to option 7.2x, a more complex interface than option 7.2x and therefore more difficulties in multi-vendor integration." To many in the sector, none of that will sound very open.
Update: The original version of this story mistakenly said NEC had declined to comment. This has now been corrected and its remarks have been included. Apologies.
Related posts:
— Iain Morris, International Editor, Light Reading