Verizon tech boss says open RAN still fails at massive MIMO

Ever since Ericsson effectively gave a Eurovision-like "nul points" to the open RAN (or O-RAN) standard, describing it as poorly suited to advanced 5G technology, critics have made the Swedish vendor out to be a typical spoilsport. Ericsson would say that, they argue, because open RAN could badly upset its own performance. But dismissing similar complaints from one of the world's biggest mobile operators – a clear target for open RAN wannabes – is not so easy.

The gloomy take on the telco side comes from none other than Joseph Russo, Verizon's president of global networks and technology (a role less verbose telcos would describe as CTO). Russo was this week being quizzed by a Wells Fargo analyst at the bank's TMT Summit when he gave his downbeat assessment of open RAN's suitability for massive MIMO, an advanced 5G technology that crams transmitters and receivers into radio units (RUs).

More basic radios come in configurations that go up to eight transmitters and receivers (so-called 8T/8R) per unit. Massive MIMO, in commercial networks, begins at 16 and goes all the way up to 64 (even higher numbers are found in labs). "And we're very supportive, and we're doing a lot of work to develop that ecosystem," Russo started out on the topic of open RAN, according to Verizon's own transcript.

"But if I look at look at the performance of the O-RAN at this point, it can't do the kind of things like massive MIMO at 16T/16R [to] 64T/64R, those kind of performance measures, we're just not there yet," he continued. "And until we are, it doesn't have a place necessarily in the kind of performance that I want to deliver to my customers."

Progress, not perfection

The problem, as far as Ericsson and others are concerned, is with the original 7.2x specification developed by the O-RAN Alliance, the telco-led body behind open RAN standards. In the interests of simplicity, it pushes most of the functionality into the distributed unit (DU), a server box responsible for baseband processing, and avoids overcomplicating radios. The goal, partly, was to foment competition to big kit vendors on the radio side – a priority for some telcos. But that doesn't seem to work for massive MIMO.

It wasn't just Ericsson complaining. Chipmaker Qualcomm and French telco Orange also had concerns that uplink performance would suffer if lots of data had to be sent from the RU to the DU for processing. Others joined the campaign and two rival solutions finally emerged. Class or "operation mode" A, backed by Ericsson, proposed moving important uplink functions, including the interference-addressing equalizer, back into the RU. The class B camp wanted to keep the equalizer in the DU, as the original 7.2x intended, but move channel estimation and beamforming, which are other uplink functions, to the RU.

A summertime agreement was reached in Osaka. While nothing is mandatory (meaning no one is forced to abandon plain-vanilla 7.2x), vendors now have the option of producing these two other flavors of O-RAN-compatible radios. The compromise is that any DU must include an equalizer to ensure it works with the different RU options. This obviously duplicates features across the RU and DU for the class A camp and could drive up costs.

But that doesn't mean the alternative is trouble-free. Without an equalizer, a class B radio would still have to send a large volume of data over the fronthaul link for DU processing. The fix is to do channel estimation in both the RU and the DU and reduce the amount of data that needs to be sent between them. Again, this sounds duplicative. Worse, say critics, it risks interoperability problems if the RU vendor differs from the DU vendor and their channel-estimation algorithms do not neatly align.

Arguably the prevailing concern for the open RAN community is that its open fronthaul specification is already starting to fragment. Ericsson has publicly said it will not be developing any radios compatible with the class B option, even though this is preferred by Orange, one of the Swedish vendor's big customers. Smaller vendors like Mavenir seem opposed to any tinkering whatsoever. They have insisted the original 7.2x is good enough.

Nokia, which is understood to have sided with Ericsson during the debate and backed class A, has not revealed its intentions. In recent days, it claims to have carried out interoperability tests with Mavenir, proving that Nokia's DUs will sync with Mavenir's RUs. But these appear to be more basic radios, not massive MIMO. On a recent press trip to its research facility in Oulu, Nokia told Light Reading that developing O-RAN-compatible massive MIMO equipment is not a current priority.

Going mainstream

Much of this could be overlooked if massive MIMO were a relatively niche technology. Originally, it appeared to be of interest only in urban hotspots. But this no longer seems to be the case. Some operators in Denmark and the Netherlands reportedly have deployed it across most sites. "Hutchison in Denmark has 32T/32R on 80% of its 3.5GHz sites," said John Strand, CEO of Danish advisory firm Strand Consult, referring to a two-year-old story in the Danish press.

Telcos generally are using 2.5GHz and C-band spectrum (which covers 3.5GHz) to deploy "active MIMO," meaning 16T/16R to 64T/64R radios, rather than less advanced "passive MIMO" equipment, says Kim Larsen, a former CTO at T-Mobile Netherlands who now runs an advisory firm called Techneconomy. The move is "not terribly economical" and might not be strictly necessary outside cities, he told Light Reading. But it gives telcos a network they can market as "the best," he said.

The question posed in a LinkedIn post by Alok Tripathi, an independent market analyst, is why Verizon has not used open RAN for the areas where 8T/8R or even 4T/4R radios are up to the job. Rather than mixing and matching different RAN vendors, it has instead focused on RAN virtualization. That has meant combining Intel's general-purpose processors with Samsung's software in DUs. And on that front, Russo seems to have no performance complaints.