5G Private Networks: Production Reality vs. Sales Pitch
7 min read
The Ground-Level Reality
- The Operational Pain: Legacy factory connectivity drops packets when autonomous robots traverse metal-heavy corridors, stalling automation.
- The Architectural Fix: Deploying private cellular spectrum to decouple industrial communications from congested local Wi-Fi.
- The First Step: Audit your physical RF environment and map out your spectrum access strategy before buying any hardware.
Why Spot Stalls at the Warehouse Threshold
A four-legged robot freezes at the threshold of an industrial Amsterdam processing plant because its ping spiked past 1,200ms during an access point handoff. This is the reality of legacy factory connectivity in heavy industrial settings, like those food production giant Cargill operates across its global network of 1,100 facilities. When a multi-million dollar automation initiative depends on real-time telemetry, a dropped packet is not just a minor lag; it is a physical safety hazard that halts the line.
Sales decks promise gigabit speeds, sub-millisecond latency, and effortless coverage. In production, radio frequency propagation behaves like water. It bounces off corrugated steel, gets absorbed by wet raw materials, and dies inside concrete silos. 5G private networks for factories are frequently sold as a magic blanket of wireless coverage, but the engineering reality is a complex exercise in physics, spectrum coordination, and hardware configuration.
This gap between the brochure and the physical floor is why deployments frequently stall. When Cargill launched its factory connectivity strategy in March 2025, it targeted 50 of its facilities by February 2026, with plans to scale to more than 100 sites per year. To make this work, they had to move past the marketing hype and address the ground-level physics of industrial radio waves.
The Physics of Shared Spectrum: CBRS and the SAS Engine
To understand why private cellular works where Wi-Fi fails, you have to look at how the spectrum is managed. In the United States, the rise of the Citizens Broadband Radio Service (CBRS) has changed the economics of industrial wireless by offering shared access to 150 MHz of spectrum in the 3.5 GHz band. This is not the wild west of unlicensed 2.4 GHz or 5 GHz Wi-Fi, where your neighbor's microwave can degrade your robotic telemetry.
Instead, CBRS relies on a three-tiered access model coordinated by a cloud-based Spectrum Access System (SAS). The top tier belongs to Incumbents, such as military radar. The second tier consists of Priority Access Licenses (PAL), which are auctioned off. The third tier, General Authorized Access (GAA), is open to anyone but must be coordinated to prevent interference. When you deploy a private network on GAA spectrum, your local radios must constantly check in with the SAS to ensure they are not stepping on higher-priority users.
How Spectrum Access Systems Dictate Your Uptime
Every small cell, known technically as a Citizens Broadband Radio Service Device (CBSD), must register its precise GPS coordinates with a SAS provider like Google or Federated Wireless. Once registered, the CBSD requests a grant to transmit on a specific channel. If the SAS detects a military radar sweep nearby, it can instantly revoke that grant and force your radios to change channels or power down. This introduces an external dependency that traditional enterprise network engineers are rarely prepared to manage.
"If your local edge core loses its WAN backhaul to the cloud-hosted SAS for more than 240 seconds, your entire factory floor goes dark."
This 240-second window is a hard regulatory limit. Under FCC rules, if a CBSD cannot refresh its heartbeat with the SAS within four minutes, it must cease transmission. If your WAN link drops, your local robots lose connectivity even if your local edge servers are running perfectly. This is the kind of operational detail that rarely makes it into the vendor's slide deck.
A Production-Grade Roadmap for Private Wireless
Moving from a vendor's proof of concept to an active, multi-site deployment requires a systematic approach to hardware, spectrum, and software configuration.
- Map the physical RF topology: Conduct a passive and active RF sweep of the facility to identify dead zones caused by heavy machinery, steel gantries, and liquid storage tanks.
- Select your core architecture: Deploy a local User Plane Function (UPF) on-premises to keep data traffic local, avoiding the latency penalty of routing packets to a public cloud.
- Provision the SIM profiles: Configure custom Quality of Service Class Identifiers (QCIs) to prioritize robotic telemetry over worker tablet traffic, ensuring critical control loops always have bandwidth.
- Verify SAS heartbeat resiliency: Simulate a WAN outage during a scheduled maintenance window to confirm that your local Evolved Packet Core (EPC) can gracefully handle a temporary loss of SAS connectivity.
Illustrative figures for explanation — representative, not measured.
The Fork in the Road: Managed NTT vs. DIY GAA-CBRS
When deploying 5G private networks for factories, systems architects face a fundamental choice between two valid but structurally different approaches. Neither is universally superior; each requires accepting a specific set of operational trade-offs.
- The Managed Service Model (The Cargill/NTT DATA Approach): This path involves outsourcing the entire stack to a global integrator. NTT DATA manages the RF planning, hardware procurement, SAS coordination, and carrier-grade SLA monitoring. This model suits organizations scaling across dozens of sites annually, but it comes with high recurring OpEx and slows down internal developers who must submit support tickets for simple configuration changes.
- The DIY GAA-CBRS Model: This approach leverages open-source cores like Open5GS running on local commodity hardware, paired with off-the-shelf CBSD radios. It eliminates subscription fees per SIM and gives your systems architects complete control over routing tables. However, you become your own carrier; when handoffs fail between gNodeB radios, your internal engineering team must debug the packet loss using specialized spectrum analyzers.
The deciding variable is not technology; it is your organization's engineering depth. If you do not have staff who understand the difference between an RSRP signal metric and a SINR noise metric, the DIY approach will result in expensive, unresolvable downtime.
How Private Cellular Deployments Quietly Fail
Many systems architects make the mistake of treating private cellular as if it were simply a more expensive version of Wi-Fi. This assumption leads to several common anti-patterns in production.
- Treating 5G like Super Wi-Fi: Assuming you can swap access points for small cells on a 1:1 basis without calculating link budgets or understanding how antenna polarization affects moving vehicles.
- Ignoring the SIM card lifecycle: Believing provisioning is a set-and-forget task, only to have automated robots stall when physical SIM profiles expire or fail to authenticate with the local Home Subscriber Server (HSS).
- Over-provisioning the downlink: Allocating too much bandwidth to downlink traffic when factory AI applications, such as Boston Dynamics' Spot checking for overheating equipment, demand an asymmetric, uplink-heavy configuration to stream high-definition thermal video.
Rule of Thumb: If your factory floor does not have moving parts traversing more than 50,000 square feet of metal-obstructed space, you do not need private 5G; a properly engineered, multi-channel Wi-Fi 6E network will cost 80% less and deploy in a fraction of the time.
Private cellular shines when assets are moving quickly across large physical distances, or where the sheer density of metal structures creates a multipath interference nightmare for Wi-Fi. If your assets are stationary, or if their movement is confined to a small, clear line of sight, the complexity of managing SIMs, gNodeBs, and SAS heartbeats is rarely worth the investment.
Frequently Asked Questions
What happens to our automated robots when the local WAN link to the cloud-hosted SAS provider drops?
If your local Evolved Packet Core (EPC) is designed with local survivability, it will continue to route on-premise traffic between your devices and your local edge servers. However, under FCC rules, if the CBSD radios cannot refresh their grant with the Spectrum Access System (SAS) within 240 seconds, they must stop transmitting, immediately halting all wireless operations on the factory floor.
Why does our robotic inspection video feed stutter when the overall signal strength (RSRP) reads as excellent?
Excellent Reference Signal Received Power (RSRP) only means the radio can hear the transmitter; it does not guarantee throughput. In industrial settings, Signal-to-Interference-plus-Noise Ratio (SINR) is the critical metric. Reflections from metallic structures create multi-path interference, which degrades the SINR and causes massive packet retransmissions, even if the raw signal is strong.
How do we handle handoffs between different gNodeB small cells for fast-moving automated guided vehicles (AGVs)?
Handoffs require careful tuning of the A3 event parameters (time-to-trigger and hysteresis) in your core network. If tuned too aggressively, the AGV will "ping-pong" between adjacent cells, dropping packets; if tuned too slowly, the connection will drop before the handoff completes. This is a manual optimization process that must be performed on-site with active telemetry logging.
The Architect's Verdict: Do not buy into the vendor promise of a single, unified wireless fabric without first auditing your team's ability to run a cellular core. If you choose the managed route, negotiate strict SLAs on configuration change latency, not just raw uptime. If you choose the DIY route, start by deploying a single gNodeB and a local UPF in your most challenging RF environment to find where the signal actually breaks.
How many times did your automated guided vehicles drop their Wi-Fi sessions last week, and what would it actually cost your operations team to migrate those specific roaming routes to a dedicated CBRS band?
Engineering References & Signals
This guide is synthesized directly from active engineering signals and the reporting within the Source Data above.
- Cargill's multi-site private 5G rollout and deployment of Boston Dynamics' Spot at its Amsterdam facility [2].
- The role of Citizens Broadband Radio Service (CBRS) and shared spectrum frameworks in U.S. manufacturing [1].
Related from this blog
- Edge Computing Hardware: Rugged IPCs vs. Plant Servers
- Can AGVs in Manufacturing Safely Abandon the Floor Tape?
- Edge computing hardware: US custom vs global modular
- SCADA modernization and the $50M water technology split
- Industrial IoT Cybersecurity: The Reality of OT Discovery
