mmWave is the 24-100 GHz radio band that buys huge bandwidth at the cost of severe path loss and blockage, forcing every system built on it to be a beamforming system, and the current research frontier is not whether it works but how much overhead the beamforming, channel estimation and hardware cost.
Summary
Millimetre-wave denotes carrier frequencies roughly between 24 and 100 GHz (the sources also reach upward into sub-THz work at 111 GHz and 300 GHz). The attraction is arithmetic: bandwidth scales with carrier frequency, so mmWave channels offer the data rates and low latency that 6G and industrial IoT programmes are specified around. The cost is equally arithmetic: free-space path loss rises with frequency, the channel becomes sparse and directional, and any hand, wall or body between transmitter and receiver removes the link. Consequently mmWave is not really a band, it is an architecture. Every practical system uses large antenna arrays with beamforming to recover the lost link budget and to exploit channel sparsity.
That architecture is where all the engineering difficulty sits, and the supplied literature is essentially a catalogue of its overheads. Fully digital arrays need one RF chain per antenna, which is impractical at scale on cost and power grounds, so systems use hybrid analog-digital precoding, which in turn needs iterative optimisation and channel state feedback that both scale badly with array size and subcarrier count. Local oscillators for coherent downconversion are themselves named as a major source of cost, power and complexity in mmWave and sub-THz receivers. As apertures grow, users fall inside the near field and the planar-wave assumption underlying conventional codebooks breaks, requiring distance-aware channel models and codebooks. Under high mobility the time-invariance assumption behind OFDM, the waveform of the 4G and 5G eras, also becomes hard to maintain at mmWave and sub-THz, with Doppler-induced degradation prompting a return to spread-spectrum and chirp-based waveforms.
The parameters that decide mmWave in practice are therefore: beam management latency and overhead; blockage probability and the mitigations available (reconfigurable surfaces, obstacle-bending beams, antenna placement); channel estimation pilot cost; RF chain and LO count per array; and mobility, which sets the Doppler regime. Proposed answers in the sources include radar-assisted narrowing of the candidate beam set for non-terrestrial links, delay-phased arrays that split control and data beams without losing per-beam gain, large reconfigurable surfaces to route around blockages, Airy beams that curve around obstacles, and transparent on-screen antenna arrays to escape the limited aperture and ‘death grip’ of edge-mounted phone antennas.
A second, quieter use of the same band is sensing. mmWave radar keeps working in fog, smoke, dust and low light, which makes it attractive for size-, weight- and power-constrained robots, drones and wearables. Because comms arrays and radar arrays are the same hardware, integrated sensing and communication (ISAC) is a headline 6G feature, with the side effect that networks may infer location, movement, activity and even breathing or heart-rate-related information about people and bystanders.
Evidence base
- A validated 60 GHz testbed pairing a Xilinx Zynq UltraScale+ RFSoC with a Sivers beamforming front-end scanned 63 transmit directions to build a full 63x63 beamspace across 350 points on a 1.95 m x 3.60 m indoor grid, published 8 May 2026.
- Transparent on-screen antenna arrays are simulated to raise uplink spectral efficiency by over 50% at 28 GHz and more than 150% at 300 GHz versus edge-mounted antennas, while resisting user-induced blockage, 3 June 2026.
- A prototyped delay-phased array at 4-7 GHz decoupled control and data beams for the first time, achieving nearly double the spectral efficiency of conventional phased arrays without extra power, 2 June 2026.
- Radar-assisted beam selection for mmWave non-terrestrial networks suppressed an unintended user’s received power below -135 dBm while adding roughly 2 dB of beamforming gain for legitimate users, 2 June 2026.
- Dual-laser self-injection locking to a micro-Fabry-Perot reached about -148 dBc/Hz at 1 MHz offset on a 111 GHz carrier, and an integrated photonic mmWave radar chip was reported in November 2025; the source states the remaining question is commercial, not scientific ref.
- ISAC in mmWave and sub-THz 6G deployments may collect or infer location and environment data, behavioural profiles such as movement and activity, and physiological data including breathing frequency and heart-rate-related information, from users and bystanders alike, 28 May 2026.
Open questions
- Does the delay-phased array that decouples control and data beams reproduce its near-2x spectral efficiency gain when built at 28 GHz or above, rather than on the 4-7 GHz testbed where it was prototyped?
- Is Doppler degradation at mmWave under high mobility severe enough in measured channels to justify abandoning OFDM for delay-Doppler or chirp waveforms, or do MIMO, coding and scheduling remedies remain sufficient?
- Will privacy regulation permit the spatial sensing that ISAC’s value case depends on, given that mmWave and sub-THz deployments can infer behavioural and physiological data about non-consenting bystanders?
- Can the combination of reduced-feedback precoding, closed-form analog precoding, LO-free receivers and learned channel estimation actually bring per-array cost and power down enough for dense deployment, or do the individual savings fail to compose?
Assessment drafted 2026-08-31 from up to 18 KB sources using the technology-scorecard framework; scores are a draft read pending review.