The Massive MIMO Question (technical)

I don’t want people to think I make a hobby out of trying to rain on the Open RAN parade. But I do have questions. And ORAN has plenty of boosters; I can’t add anything there.

Massive MIMO is interesting for many reasons. One of these days it might be worth writing a post about how the actual implementation of mMIMO is so completely different from the theoretical work that birthed it. (Unlikely to go viral.) But a more immediate reason is the question of whether ORAN mMIMO can truly compete with a closed proprietary version in the sweet spot: mid-band dense urban networks.

In the mid-band (2.5-5 GHz, let’s say), we want to use active array antennas (AAS) to overcome the less-favorable propagation, using beam-forming. Each antenna in the array (or small group of antennas) is driven by its own radio chain. So a 64T64R AAS has 64 radio chains, thus 64 transmitters and 64 receivers.

Now, the physics says that, for a given transmitter power, when you assemble transmitters into a beam-forming array, the effective power in the beam goes up with the square of the number of elements. So 64 transmitters deliver four times the power in the main beam direction as 32. Visualize it as the result of adding power while also narrowing the beam.

On the other hand, the physics also says that the effective sensitivity of the combined receiver only goes up in proportion to the number of elements, 64 receivers being twice as sensitive as 32. Visualize this as the result of adding received signal power while also adding more receiver noise.

The result of this difference between transmit and receive gains is a gap between uplink and downlink. As the array increases in size, from 16T16R all the way to 128T128R, for a given data rate the downlink reach exceeds the uplink reach, all else being equal. Or, looking at it differently, the usable cell size is controlled by the uplink, while the downlink (transmit beam) may extend deep into the neighboring cell.

This creates a number of problems, which I won’t go into. But it leads me to something Ericsson began promoting this summer: their “Uplink Booster”.

Briefly, Ericsson claims that, by moving a number of processing functions out of the DU server and into the Radio Unit, they can improve the massive MIMO uplink by a factor of 10 (10dB). The result, they claim, is better performance and reduced interference through improved beam-forming, as well as better spectral efficiency through improved multi-user MIMO.

Impressive claims. And I’m inclined to credit them, if for no other reason than the effort they must have put into the actual engineering.

So let’s say for the sake of argument that Ericsson’s “Uplink Booster” does what they say it does. Where does that leave ORAN?

It seems to leave ORAN far behind, at least in mid-band massive MIMO, because the most important open API (the fronthaul specification based on the 7.2 split) does not allow these processing functions to move into the RU.

Consequently, it seems that mobile operators using ORAN and massive MIMO will have to supplement with low-band radios for coverage, or else locate their cell sites closer together and turn down the transmit power. This may be acceptable for greenfield deployments, like the Dish 5G network, or for rural Huawei replacement, but it strikes me as a non-starter for upgrading dense urban 4G networks to 5G — unless the cost of ORAN-compliant hardware is so much cheaper than what a tier-1 like Ericsson charges that it doesn’t matter. And maybe it will be, but considering that Ericsson’s operating margin is probably around 4%, it just sounds like a losing proposition.

Maybe the ORAN Alliance will develop a new API for massive MIMO. (That could take a couple of years.) Maybe vendors will switch to a different functional split, moving more functions into the RU and leveraging 3GPP-defined interfaces. That will be a major change, and if it means moving the layer 2 MAC and scheduler, could change the ORAN business model entirely.

Or maybe the ORAN parade will rethink where it wants to march.

I’m very much interested in comments.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: