Samsung and KDDI Show Off 3.6 Gbps at 7 GHz with Extreme MIMO for 6G

KDDI, KDDI Research and Samsung Research reported another interesting result from the increasingly active 7 GHz 6G research space. In an outdoor trial at Samsung Electronics' R&D campus in Seoul, the companies achieved a 3.6 Gbps downlink using 100 MHz of spectrum, combining Extreme Massive MIMO, eight-layer single-user MIMO and 1024QAM modulation.

The measurements were carried out from 18 to 20 August 2026. According to KDDI, the Extreme Massive MIMO base station used more than 1,000 antenna elements, allowing the transmission direction to be controlled very precisely and radio energy to be concentrated towards the terminal. KDDI describes the 3.6 Gbps result as being at a world-leading level for outdoor 7 GHz transmission using 100 MHz bandwidth, based on research by Samsung Research and KDDI Research as of 2 September 2026.

The headline speed is interesting, but perhaps the more useful part of the experiment is understanding how it was achieved. Under the same test conditions and using eight spatial layers, 256QAM delivered 3.0 Gbps while 1024QAM increased this to 3.6 Gbps, a measured improvement of 20%. 256QAM carries eight bits per modulation symbol, while 1024QAM carries ten. In theory that increases the number of bits per symbol by 25%, although practical throughput also depends on coding, overhead and other aspects of the radio link.

Moving to 1024QAM is not simply a matter of changing the modulation setting. As the modulation order increases, the constellation points become more closely spaced and the receiver needs a cleaner signal to distinguish between them reliably. In other words, 1024QAM needs sufficiently good signal quality and Signal-to-Interference-plus-Noise Ratio, or SINR. This is where the very large antenna array becomes important. By concentrating more energy towards the terminal and providing much stronger beamforming capability, Extreme MIMO can create the radio conditions needed for higher-order modulation.

1024QAM itself should not be regarded as a new 6G technology. Downlink 1024QAM for NR FR1 was introduced by 3GPP in Release 17. What is interesting here is achieving it in an outdoor 7 GHz Extreme MIMO system with eight downlink spatial layers, rather than the modulation scheme in isolation.

There is also an important distinction between antenna elements, digital ports and spatial layers. Thousands of antenna elements do not mean thousands of independent data streams. The antenna elements are combined to create highly directional beams. Digital antenna ports provide the baseband system with degrees of freedom for beamforming and spatial processing, while spatial layers are the separate data streams actually transmitted to the user.

Samsung's previously described 7 GHz X-MMU, or eXtremely large Massive MIMO Unit, provides a useful example. It has 1,024 antenna elements and 256 digital ports, and Samsung's associated UE emulator has eight receive antennas mapped to eight digital ports for eight-layer downlink reception. Samsung's prototype operates in the 7.125–7.325 GHz range with a 100 MHz channel.

The latest KDDI announcement says only that the trial used more than 1,000 antenna elements, so we should not automatically assume that every hardware detail was identical to Samsung's earlier prototype. Nevertheless, the overall architecture clearly points in the same direction.

This is also where the terminology becomes interesting. We have discussed Giga-MIMO and Gigantic MIMO, or gMIMO, several times on this blog. The basic idea is to take the principles that made Massive MIMO successful in 5G and scale them considerably further, particularly in the upper-mid-band where the shorter wavelength allows far more antenna elements to be fitted within a practical physical aperture.

Samsung generally refers to its approach as eXtreme MIMO or X-MIMO, while KDDI uses the term Extreme Massive MIMO. Extreme MIMO is not simply a Samsung name. ITU-R Report M.2516 described Extreme MIMO, or E-MIMO, as an evolution of Massive MIMO using much larger-scale antenna arrays, and the IMT-2030 Framework Recommendation M.2160 continues to use the term. China's IMT-2030 (6G) Promotion Group was also discussing Extreme-MIMO as a key 6G technology several years ago.

At the same time, Qualcomm increasingly uses Giga-MIMO for its upper-mid-band approach. Interestingly, Qualcomm's current 7 GHz prototype also has 1,024 antenna elements and 256 digital ports, with a prototype device supporting up to eight downlink layers.

I therefore increasingly think of Extreme MIMO, Giga-MIMO and Gigantic MIMO as overlapping concepts rather than fundamentally different technologies. There are differences in how individual companies and researchers define the details, but the basic direction is becoming clear: hundreds of digital antenna ports, potentially thousands of antenna elements, much greater beamforming gain and significantly more spatial processing capability, especially in the upper-mid-band.

We previously looked at this evolution in The Rise of 'giga MIMO' or 'gigantic MIMO' (gMIMO) and The Journey from mMIMO in 5G towards gMIMO in 6G. We also looked at Samsung's earlier 6G MIMO prototypes, including its 128TR 6.9–7.3 GHz radio, in Samsung Pushes Boundaries of 6G MIMO.

The move towards these much denser arrays is particularly important because 7 GHz brings a fundamental propagation challenge. Compared with today's commonly deployed 3–4 GHz mid-band systems, the higher frequency results in greater path loss. Without additional antenna gain, an operator could therefore end up with a smaller cell footprint and might need additional sites to obtain equivalent coverage. That would significantly weaken the economic case for using the new spectrum.

The shorter wavelength, however, is both the problem and part of the solution. It allows far more antenna elements to be packed into a panel of practical dimensions. With suitable radio and baseband processing, these elements can create narrower, higher-gain beams that help compensate for the additional path loss.

This is why the second result reported by KDDI may ultimately be at least as important as the 3.6 Gbps headline.

KDDI says that the 7 GHz system achieved a communication area comparable with a Sub-6 base station installed at the same location. More precisely, within the measurement area, the proportion of locations meeting the received-power level considered sufficient for stable communications reached 95% of the corresponding Sub-6 result.

That distinction is important. This does not mean that 7 GHz will always reproduce a 3.5 GHz cell footprint, nor does it prove universal coverage parity. Propagation environment, transmit power, antenna configuration, clutter, building penetration and many other factors will determine the result in a commercial deployment. It does, however, add another useful field measurement supporting the idea that very large antenna arrays could allow upper-mid-band spectrum to be deployed using existing macro-site grids rather than requiring wholesale site densification.

The latest result is also part of a continuing Samsung research programme rather than an isolated speed demonstration. In February 2026, Samsung, KT and Keysight announced outdoor tests at the same Seoul R&D campus, achieving up to 3 Gbps at 7 GHz. That system simultaneously transmitted eight data streams from a prototype base station with 256 digital ports to a single-user Keysight 6G terminal testbed.

Samsung subsequently published considerably more technical detail, while a March 2026 paper, Extreme-MIMO Field Trials in 7 GHz Band: Unlocking the Potential of New Spectrum for 6G, documented a 256-port architecture and outdoor urban measurements. The researchers demonstrated eight-layer downlink SU-MIMO over 100 MHz with more than 3 Gbps to a single user.

The September KDDI result therefore looks like another step along the same development path. The particularly interesting change is the successful application of 1024QAM, moving measured single-user throughput from 3.0 to 3.6 Gbps while retaining 100 MHz bandwidth and eight spatial layers.

There is a wider spectrum story as well. Under WRC-27 Agenda Item 1.7, ITU-R is studying the 7.125–8.4 GHz range, or portions of it depending on ITU Region, for possible IMT identification. The work includes extensive sharing and compatibility studies with the fixed, satellite, Earth exploration and other incumbent services already using parts of this spectrum.

This is important because the attraction of the 7 GHz region is not merely higher frequency. Operators need additional contiguous spectrum if 6G is to provide another meaningful capacity layer. At the same time, going too high in frequency makes wide-area coverage increasingly difficult. Upper-mid-band spectrum is therefore being investigated as something of a compromise between the coverage characteristics of today's mid-band and the very large bandwidths available at millimetre-wave frequencies.

The standards work is also becoming much more concrete. 3GPP's Release 20 Study on 6G Radio is already under way, while Release 21 is intended to contain the first normative 3GPP 6G specifications for the IMT-2030 submission. Recent RAN work is already addressing areas including higher-order modulation, constellation shaping and other aspects of the future 6G radio interface.

It is still too early to take a prototype result such as 3.6 Gbps and extrapolate directly to future commercial 6G performance. The experiment used a research base station and test terminal rather than a normal handset. Eight-layer reception places substantial demands on UE antenna and RF capability, and scaling very large arrays introduces challenges around power consumption, signal processing, calibration, cost and thermal management.

Nevertheless, the progression is becoming increasingly interesting. The industry is not simply experimenting with another new frequency band. We are beginning to see a fairly consistent architecture emerge for upper-mid-band 6G: denser antenna arrays, hundreds of digital ports, stronger beamforming, more spatial layers and higher-order modulation, all intended to extract considerably more capacity from spectrum around 7 GHz while retaining a practical macro-network footprint.

For KDDI, the work also connects with its Digital Belt Initiative, which aims to combine high-capacity, low-latency communications infrastructure with AI computing infrastructure for an AI-centric society. KDDI sees the large-capacity communications demonstrated in this trial as one of the technologies that could contribute to that vision.

The 3.6 Gbps number will naturally receive most of the attention, but it may not be the most important result. The more significant question for 6G is whether upper-mid-band spectrum can provide substantially more capacity without forcing operators to build a much denser macro network.

Extreme MIMO, Giga-MIMO or Gigantic MIMO, whatever name the industry eventually settles on, is increasingly looking like one of the key technologies being developed to make that possible.

Related Posts: 

Posted by on

Comments

Post a Comment