Cell-free massive MIMO Channels in an Urban Environment -- Measurements and Channel Statistics

Cell-free massive MIMO (CF-mMIMO), where each user equipment (UE) is connected to multiple access points (APs), is emerging as an important component for 5G and 6G cellular systems. Accurate channel models based on measurements are required to optimize their design and deployment. This paper presents an extensive measurement campaign for CF-mMIMO in an urban environment. A new “virtual AP” technique measures channels between 80 UE locations and more than 20,000 possible microcellular AP locations. Measurements are done at 3.5 GHz carrier frequency with 350 MHz bandwidth (BW). The paper describes the measurement setup and data processing, shows sample results and their physical interpretation, and provides statistics for key quantities such as pathloss, shadowing, delay spread (DS), and delay window. We find pathloss coefficients of 2.9 and 10.4 for line-of-sight (LOS) and non line-of-sight (NLOS), respectively, where the high LOS coefficient is mainly because larger distance leads to more grazing angle of incidence and thus lower antenna gain in our setup. Shadowing standard deviations are 5.1/16.6 dB, and root mean squared (RMS) DSs of -80.6/-72.6 dBs. The measurements can also be used for parameterizing a CUNEC-type model, which will be reported in future work.

💡 Research Summary

This paper presents an extensive measurement campaign aimed at characterizing cell‑free massive MIMO (CF‑mMIMO) channels in a realistic urban environment, a prerequisite for the design and optimization of upcoming 5G and 6G networks. While prior works on CF‑mMIMO have been limited to a handful of access points (APs) or indoor virtual arrays, this study overcomes those constraints by introducing an improved “virtual AP” methodology. A cherry‑picker vehicle lifts a single patch antenna to a typical AP height of 13 m and drives at 0.4–0.6 m/s, creating a dense virtual linear array with element spacing of 4–6 cm. Over two city blocks (≈200 m × 200 m) on the USC campus, 80 distinct user equipment (UE) locations were selected, and for each UE the channel was measured to more than 20 000 potential AP positions, yielding a dataset of unprecedented spatial density for outdoor CF‑mMIMO.

The channel sounder operates at a 3.5 GHz carrier with a 350 MHz bandwidth, employing a multi‑tone waveform (2801 sub‑carriers, 125 kHz spacing) that provides a delay resolution of 2.857 ns (≈0.857 m) and a maximum non‑aliased delay of 8 µs. The transmitter chain consists of a Keysight AWG, up‑conversion to 3.5 GHz, dual band‑pass filtering, and two power amplifiers delivering 42 dBm through a vertically polarized patch antenna. The receiver comprises an 8‑element omnidirectional dipole array, low‑noise amplifiers, an 8‑to‑1 RF switch, and an automatic gain control (AGC) unit that maintains output power variation within 1.2 dB despite a dynamic range spanning more than 120 dB. Both ends are synchronized with GPS‑disciplined Rubidium clocks providing a 10 MHz reference and a 1 PPS UTC‑aligned pulse, ensuring sub‑microsecond timing accuracy for the 10 Hz switching cycle (80 µs per SISO capture, 640 µs per Rx switch).

Calibration is performed in two stages: a back‑to‑back (B2B) measurement of the complete Tx‑Rx chain for all 64 Tx/Rx port combinations, and a frequency‑dependent antenna gain calibration in an anechoic chamber using a precision rotor. This yields accurate transfer functions and antenna patterns for subsequent data processing.

Post‑processing includes frequency‑domain compensation, windowing, and extraction of multipath components. LOS/NLOS classification is based on visual inspection of the impulse responses and geometric knowledge of the environment. Path‑loss fitting reveals a surprisingly high LOS exponent of 2.9 (greater than the free‑space value of 2) which the authors attribute to the increasing grazing angle at larger distances, reducing the effective gain of the vertically oriented patch antenna. In NLOS conditions the exponent rises to 10.4, reflecting severe diffraction and reflection losses in the dense urban canyon. Shadow fading standard deviations are 5.1 dB for LOS and 16.6 dB for NLOS, indicating much larger variability when the direct path is blocked. RMS delay spreads are –80.6 dB (LOS) and –72.6 dB (NLOS), both very low, suggesting that even with a 350 MHz bandwidth the multipath richness is limited in this environment, which is beneficial for wideband modulation schemes. The authors also analyze the delay window (maximum observable delay) and relate it to the physical layout of streets, alleys, and building facades, providing a physical interpretation of the observed impulse‑response evolution.

Statistically, the dataset offers a comprehensive view of the joint UE‑AP channel characteristics required for CF‑mMIMO system modeling, which differs fundamentally from conventional single‑BS models. The authors propose to use these measurements to parameterize a CUNEC‑type distance‑based path‑loss model in future work, enabling more accurate system‑level simulations, AP placement optimization, and power‑control algorithms.

In conclusion, the paper delivers the most extensive wideband outdoor CF‑mMIMO measurement campaign to date, demonstrating a practical and precise virtual‑AP approach, detailed hardware design, rigorous calibration, and thorough statistical analysis. While the study is confined to a specific downtown Los Angeles area and employs a single vertically polarized patch antenna, it establishes a solid foundation for extending the methodology to diverse urban morphologies, other frequency bands, and multi‑polarized AP configurations. The resulting channel models will be instrumental in guiding the deployment of next‑generation cell‑free networks.