Location Technology

Network Centric

Z-Axis Accuracy for 5G/6G

How Carriers Can Improve Vertical Location

by Simply Adding Vertical Antennas

Public-safety rules in the United States now expect mobile network operators (MNOs) to deliver floor-level vertical location for emergency 911 calls in major US cities.

This white paper proposes a network‑centric vertical positioning. The core idea is simple: attach a vertically separated stack of additional 5G FR1 antenna panels at selected base‑station sites and use the carrier‑phase differences between these panels to estimate user height. The method fits into existing 5G positioning architectures as a Z‑axis refinement step on top of today’s TDoA/OTDOA engines.

This works because adding vertical antennas breaks the co-planar assumption that has haunted legacy positioning solutions. The positioning Jacobian gains a strong vertical component. In controlled simulations with 5G FR1, PRS and a handful of sites equipped with vertical interferometry, 3D error at the 80 percentile drops from tens of meters to around a meter or below when the interferometric phase terms are added ..

Download PDF

1. The Vertical Location Challenge

Radio networks and GNSS constellations have historically been optimised to tell us where a device is on the ground plane. In that plane, modern techniques routinely achieve metre‑ or even decimetre‑level accuracy. The moment we care about height – for example, whether a 9‑1‑1 caller is on the tenth floor or the fifteenth – the geometry of today’s networks reveals its limitations.

In a typical LTE or 5G deployment, most base stations sit at roughly the same elevation, on towers and rooftops spread across a city. From the handset’s perspective, the network is almost entirely co‑planar (Figure 1). Timing‑based methods such as OTDOA infer position from range differences to those sites. If all the transmitters lie nearly in a plane, the geometry matrix is well behaved in the horizontal dimensions but weak in the vertical dimension. Small timing errors translate into large uncertainty in height. GNSS, which might help outdoors, is heavily attenuated or unavailable indoors, exactly where most dense urban emergencies occur.

2. Vertical Interferometry Antennas

Locaila’s network-centric solution is, at heart, very simple: at a site that already has a 5G NR FR1 antenna module, the operator installs another antenna of the same type a few meters above or below the existing panel. We call this a Vertical Interferometry Antenna.

These two antennas share the same oscillator and transmit carefully designed PRS patterns. In the air, their waves overlap and create an interference pattern like bright and dark “fringes” in space. This is exactly the same physical principle as the interference pattern in Young’s double-slit experiment. If you could somehow paint these RF fringes onto the side of a building, you would see a set of horizontal stripes. Each stripe corresponds to a particular combination of the two antenna signals – a particular phase relationship.

These stripes are fixed in space: they do not drift with weather or temperature, and they are laid out at precise intervals set by the carrier wavelength.

A UE at a given height experiences a specific phase difference between the signals coming from the two antennas. If the network measures that phase difference accurately, it can tell which “stripe” the UE is sitting on. When we add a second or third vertical spacing – for example, a third panel at a different height – each spacing produces its own fringe pattern. Taken together, these patterns behave like an RF barcode along the height axis: the combination of phase differences across the different baselines uniquely identifies the user’s absolute altitude. Locaila’s algorithms use these vertical interference patterns to generate a Z-axis signal that remains usable even when the UE is deep inside a building.

It is noted that this method is not a beamforming-based AoD scheme, because in most practical 5G deployments, the elevation beam resolution is relatively poor. Further, it is not like an uplink based AoA method.

3. Architecture and Standards Integration

The proposed solution fits into the existing 5G positioning architecture (for example, as defined in TS 38.211 and TS 38.215) without disruptive changes. At the radio layer, selected RU/DU instances expose additional RF chains for the vertically separated panels. PRS resources are configured in the usual way, but the network now knows which physical panels form each vertical baseline.

Modern 5G antenna modules already contain dense grids of patch elements. By choosing different pairs of elements, and assigning them distinct PRS patterns, vendors can create multiple virtual baselines inside a single or dual-panel structure without redesigning the hardware. Because FR1 carriers such as 3.5 GHz (with a wavelength of roughly 8.6 cm) are already widely deployed, even a 10 cm vertical spacing between elements or panels is enough to produce distinct phase pattern over a ±3 m height range in the far field. Using multiple different spacings enriches the interference pattern and makes the RF “barcode” more robust.

Crucially, the panels share a common clock, so the PRS sources are phase-coherent. When the network computes phase differences between panels, most of the common oscillator phase noise and static offset cancels out. What remains is a relatively clean, geometry-driven signal that is sensitive to height but tolerant of hardware imperfections.

On the UE side, nothing special is required. The device does not need to know about fringes or antenna geometry. It simply receives PRS as defined in TS 38.215, measures carrier-phase differences between those PRS signals and reports one or more RSCPD(Received Signal Carrier Phase Difference)-style observables back to the network, using mechanisms that are already being standardised for NR positioning.

On the network side, gNB software and the LMF (Location Management Function) are extended so they understand which PRS resources correspond to which vertical panels or element groups. The LMF fuses the reported RSCPD measurements with existing timing and, where available, angle measurements to estimate a full 3D position with a very tight Z component. The resulting 3D coordinates can be fed into the emergency call flow (E9-1-1 / NG9-1-1) or exposed via location APIs to public-safety and commercial applications.

4. Simulations and Field Trials

To judge whether vertical arrays are worth the effort, we look at two layers of evidence: system‑level simulations in realistic urban geometries and early outdoor trials with prototype antenna stacks.

In simulations, a typical scenario assumes FR1 carrier frequencies around 3.5 GHz, five or six gNB sites with inter‑site distances of a few hundred metres, and two or three vertically separated panels per selected site. TDoA noise levels are set so that a legacy 5G positioning engine delivers several metres of vertical RMSE under indoor penetration and partial NLOS conditions. Phase noise is modelled to reflect practical receiver performance, for example a few degrees of standard deviation in carrier‑phase estimates.

In this baseline, a TDoA‑only solution often yields 3D error distributions with tens of metres of spread and Z‑axis errors clustering around five metres or more at the 80th percentile (Labeled Legacy TDoA)

When vertical refinement is added using interferometric phase measurements from the new antenna stacks, the picture changes (Labeled 3D coarse, Figure 3). Median Z‑axis RMSE moves into the sub‑meter regime in the refined proprietary algorithm, and 80th–90th percentile errors can be brought comfortably within the ±3 m regulatory threshold. Horizontal error also tightens slightly because the overall geometry matrix becomes better conditioned (Labeled 3D refined).

These results are not tied to a single geometry. Across multiple layouts and random user drops, the same pattern appears: the combination of a modest number of vertical baselines and existing TDoA observables moves the operating point from “barely compliant” towards “robustly within ±3 m”, with a substantial fraction of calls enjoying much finer height resolution.

Early outdoor trials support this trend. In one example, three vertically separated antennas were installed on a building at roughly 0.3 miles from the test area, transmitting an emulated 5G‑like waveform in the low GHz range. A receiver moved floor‑to‑floor inside nearby buildings. The measured vertical phase differences tracked height changes cleanly.

When fused with timing information in a simple two‑step solver, the system consistently identified the correct floor in multi‑storey buildings at high probability, even in the presence of realistic indoor penetration loss.

While these prototypes do not yet represent fully integrated 5G macro deployments, they demonstrate that the interferometric concept survives real‑world impairments such as multipath, hardware offsets and environmental noise, and can be engineered into carrier‑grade equipment.

Page updated

Google Sites

Report abuse