In sum, what to know:
-Signals Research Group recently conducted what is almost certainly the first independent testing of 5G massive and Multi-User MIMO in FDD spectrum, in a live network.
-The tests were conducted across multiple cells in Verizon’s network in St. Paul, Minnesota, where Ericsson is the equipment vendor.
-SRG found that Verizon had performance gains from deploying 32T32R massive MIMO, and the addition of MU-MIMO enhanced the uplink even further. However, downlink improvements were spottier.
As mobile network operators seek to squeeze ever more performance and capacity out of their existing spectrum, they have turned to massive Multiple-Input Multiple Output (MIMO) systems, and also to the use of Multi-User MIMO (MU-MIMO), which enables devices to share the same network resources, serving up data transmissions to multiple devices at the same time. MU-MIMO is used in Wi-Fi systems, and in a cellular context, it has largely been deployed in Time Division Duplex (TDD) spectrum.
Verizon has already embraced the use of Massive MIMO in TDD spectrum (including some notable speed achievements), but it has also been one of the few operators working to advance massive MIMO systems in FDD spectrum. That work seems to be paying off: Signals Research Group has published the results of recent testing where both Massive MIMO and MU-MIMO were in play in Verizon’s network, and found that the resulting improvement in uplink throughput was “simply outstanding.”
For Mike Thelander, founder of SRG, the results were notable because FDD presents a tougher challenge for Massive MIMO and MU-MIMO implementation than TDD spectrum. The technology is also still a rarity in real-world networks. “You can probably count on using the fingers on a couple of hands the number of operators that have commercial traffic running over 5G FDD massive MIMO, and even then we believe the number of massive MIMO FDD cell sites is probably fairly modest, especially outside of China,” he wrote in the SRG report.
Why is FDD, as Thelander put it, “an entirely new animal” when it comes to MIMO-related advances? Because while TDD operation uses the same channel for uplink and downlink transmissions (meaning, the network can use channel information from one direction to optimize transmissions in the other), FDD relies on different frequencies for the uplink and the downlink—which makes it more challenging to have a good grasp of channel information. To Thelander, that left an open question as to whether MU-MIMO could be effectively implemented in FDD.
To answer that question, he revisited an area in St. Paul, Minnesota where SRG had previously tested the performance differences between 4T4R MIMO and 32T32R MIMO on Verizon’s network. At the time of that testing, however, Verizon hadn’t fully implemented the MU-MIMO capabilities in the uplink and downlink for Massive MIMO. What Thelander found in the latest testing is that performance improvements from moving to Massive MIMO were further enhanced by the new addition of MU-MIMO—although the effects were more consistent in the uplink.
“When Verizon upgraded from 4T4R to 32T32R, they got significant, significant gains in performance in both the downlink and the uplink, even before you turn on MU-MIMO,” Thelander said. “Now you turn on MU-MIMO, and in the uplink, it’s another big, big boost in terms of performance.”
The tests involved both stationary and walk-testing with two Motorola razr fold 2026 smartphones locked to 5G Standalone operations in Band n2 (1.9 GHz), with capture and analysis tools from Accuver Americas and Keysight Technologies. SRG found that uplink MU-MIMO use “occurred nearly all the time, delivering close to four uplink MIMO layers on a consistent basis from the network perspective, even when the two smartphones were collocated” and serving up “very high double-digit gains on a percentage basis” when compared to Single-User MIMO (SU-MIMO) with 32T32R. In short: Massive MIMO had already outperformed 4T4R MIMO, and MU-MIMO improved the uplink performance even more.
The downlink, however, was a different story. On one hand, SRG concluded that its earlier testing had underestimated how well Massive MIMO could perform in the downlink with SU-MIMO alone. But Thelander also characterized the gains from downlink MU-MIMO as “hit or miss, relative to what you already got with the Massive MIMO.” The two UEs shared the same network resources with MU-MIMO (known as UE pairing) on some walk-testing routes and produced “solid double-digit gains in spectral efficiency” over 32T32R with SU-MIMO, according to the SRG report. But on other routes, UE pairing was less consistent, leaving “opportunity for improvement,” as Thelander put it.
GET A LITTLE CLOSER
SRG’s testing also offered some intriguing perspective on how much physical separation is actually required between devices, for uplink MU-MIMO to be useful.
“What was surprising is that when I first tried to test it, I literally had two phones sitting next to each other in my vehicle, and they were using MU-MIMO. … They were sharing the same network resources—literally, 18 inches apart,” Thelander said. “Now, I’ve seen that in TDD. I didn’t expect to see that in FDD.”
Generally, the more physical separation between UEs, the better that MU-MIMO would be expected to work: More space between devices means that it’s easier for the network to figure out there are multiple devices to be served. But in this case, there wasn’t much spatial separation between the two phones. Yet within 10 seconds, according to SRG, the two phones were sharing the same network resources.
And while Thelander adds the caveat that he was testing in an unloaded cell with only a couple of devices, the smooth MU-MIMO use by two devices so close together bodes well for FDD MU-MIMO on a larger scale. “What that tells me is that … even if you had lots of UEs out there and a lot of them are co-located, MU-MIMO in the uplink should still work reasonably well,” he said.
You can access a preview of SRG’s report here.
Facts Only
* Signals Research Group conducted testing on 5G massive and Multi-User MIMO in FDD spectrum in a live network.
* Tests were conducted across multiple cells in Verizon’s network in St. Paul, Minnesota, utilizing Ericsson equipment.
* Deploying 32T32R massive MIMO provided performance gains in both uplink and downlink before MU-MIMO was added.
* The addition of MU-MIMO further enhanced the uplink performance.
* Downlink improvements were inconsistent.
* Uplink MU-MIMO use delivered close to four uplink MIMO layers consistently across routes with two smartphones collocated.
* When compared to Single-User MIMO (SU-MIMO) with 32T32R, uplink MU-MIMO resulted in very high double-digit gains.
* Downlink MU-MIMO gains were variable, showing solid double-digit gains on some routes but less consistent performance on others relative to SU-MIMO.
* Testing involved two Motorola razr fold 2026 smartphones operating in Band n2 (1.9 GHz) in 5G Standalone.
Executive Summary
Independent testing by Signals Research Group examined the performance of 5G massive and Multi-User MIMO in Frequency Division Duplex (FDD) spectrum within Verizon's live network in St. Paul, Minnesota, with Ericsson as the equipment vendor. The tests revealed that deploying 32T32R massive MIMO, combined with Multi-User MIMO (MU-MIMO), resulted in significant performance gains in the uplink throughput. Specifically, the addition of MU-MIMO enhanced the uplink performance further. However, improvements in downlink performance were less consistent.
The testing highlighted a key difference between FDD and Time Division Duplex (TDD) spectrum; FDD presents a greater challenge for implementing Massive MIMO and MU-MIMO due to differences in channel information availability. While Massive MIMO alone provided substantial gains over 4T4R MIMO, the introduction of MU-MIMO offered further uplink improvements. Downlink performance gains from MU-MIMO were described as variable, showing solid double-digit spectral efficiency gains on some routes but less consistent results on others.
The spatial separation between devices showed that even phones placed close together could effectively share network resources via uplink MU-MIMO within a short timeframe. This suggests potential viability for FDD MU-MIMO implementation at scale, especially for the uplink, despite the inherent challenges associated with FDD spectrum characteristics.
Full Take
The central tension in this findings lies between the observed, highly favorable uplink results and the more ambiguous downlink performance and the inherent difficulty of implementing MU-MIMO in FDD relative to TDD. The observation that MU-MIMO provided a "big, big boost" to the uplink performance when layered on top of Massive MIMO suggests that channel sharing mechanisms might be significantly more effective or less constrained in the uplink than the downlink under FDD conditions. This implies that the fundamental challenge of FDD—the lack of simultaneous feedback for channel optimization across directions—may be less prohibitive for uplink-focused spatial multiplexing than for downlink strategies.
The finding regarding physical separation, where near-co-located devices still achieved shared resource utilization in the uplink, challenges the assumption that spatial separation is a mandatory prerequisite for effective MU-MIMO. This shifts the focus from macro-level antenna placement constraints to network signaling and resource allocation protocols as the primary limitations. For future development, the lack of consistent downlink gains suggests that whatever mechanisms govern spectral efficiency in FDD transmission might require specific architectural adaptations beyond simple MIMO layer aggregation. The pattern suggests that performance improvements are dependent not just on the physical radio link geometry but critically on how channel state information is managed across disparate frequency bands, pointing toward a need for standardized cross-link awareness in FDD Massive MIMO deployment.
Bridge Questions: How do the differing requirements for channel information in FDD versus TDD spectrum fundamentally alter the necessary protocols for successful MU-MIMO implementation? What specific architectural or signaling adjustments are required to ensure consistent downlink spectral efficiency gains across varied FDD scenarios? If uplink performance is significantly enhanced while downlink is hit-or-miss, what does this reveal about the asymmetry in how these two directions handle channel estimation in FDD networks?
Sentinel — Human
The text reads like an analysis or summary of a technical report, successfully synthesizing complex engineering findings while weaving in expert interpretation regarding the implications for 5G network architecture.