OSI-ORB-TR1 · Version 0.9 draft · August 2026
The specification suite defines what the platform is. This report explains why it can exist at all, and why a closed architecture will struggle to answer it.
Section 2
Temporary industrial sites were underserved by every generation of network technology for structural reasons, not accidental ones. The architecture is best read as a point-by-point answer to each.
Constraint 1
Fixed networks amortize over decades; a jobsite exists for months. Trenching, towers, and macro siting never clear the cost hurdle on an 18-month project, so nobody builds, and the site inherits a grid engineered for highways and rooftops rather than pits and corridors.
Constraint 2
Interference protection meant an exclusive license, auctioned nationally and held by carriers, unavailable to a site operator at any sensible price. The unlicensed alternatives were short-range bands with no coordination mechanism, so they could not carry infrastructure.
Constraint 3
Multi-band deployment historically demanded frequency planning, interference studies, link budgets, and antenna engineering per site. That labor does not exist in the construction trades and cannot be imported economically for every job.
Constraint 4
A perfect site network is useless without a WAN. Remote sites had no path to one: no fiber, thin or absent macro coverage, and satellite that was narrowband, high-latency, and expensive.
Constraint 5
Every radio technology meant a different chip vendor, firmware stack, and management plane. Five radio technologies meant five vendor roadmaps, and the client side was worse: bespoke devices for every non-mainstream band.
The shape of the answer
Each constraint fell separately, for unrelated reasons, over roughly fifteen years. The platform is a composition across all of them, which is precisely why it has not been assembled before.
Section 3
Every enabler below is individually available to any competitor. The value concentrates in the layer that binds them: the coordination and control software, and the published standard.
| Enabler | Matured | What it is | Constraint dissolved |
|---|---|---|---|
| Database-coordinated spectrum | 2012 to 2024 | PAWS for TVWS, SAS for CBRS, AFC for 6 GHz | Licensed-spectrum model: interference management became an API call rather than an auction |
| CBRS three-tier framework | 2020 | GAA open access, optional PALs | Licensed-spectrum model: carrier-grade mid-band for anyone, protection purchasable per county |
| 6 GHz unlicensed opening | 2020 to 2023 | 1,200 MHz under AFC, standard power outdoor | Capacity ceiling: more open mid-band than most carriers hold licensed |
| Programmable ASIP basebands | 2019 to 2026 | One silicon platform, many waveforms, openly licensable | Silicon fragmentation: LTE, NR, and future waveforms as firmware, on both ends of the link |
| Merchant module maturity | 2021 to 2026 | Wi-Fi 7, HaLow, NR+, multiprotocol 802.15.4 and BLE | Silicon fragmentation: certified modules with mainline open-source drivers |
| Open-source network software | 2015 to 2026 | OpenWrt lineage, open 5G cores, SD-WAN steering | Integration cost: the control plane is assembled from production-hardened parts |
| LEO broadband | 2021 to 2026 | Flat-panel terminals, hundreds of Mbps, minutes to acquire | Backhaul dependency: a WAN exists everywhere on Earth, portable, on day one |
| Multiband antenna integration | 2015 to 2025 | Stacked-aperture macro panels, high-gain wide-angle arrays | RF expertise wall: filtering and gain engineering move from the field into the factory |
| Edge compute economics | 2020 to 2026 | Server-class ARM and accelerators at site power budgets | Backhaul dependency: process locally, ship results; the WAN leaves the experience path |
| Neutral-host frameworks | 2019 to 2026 | OnGo shared HNI, MOCN, Passpoint and OpenRoaming | Device fragmentation: every phone on every carrier becomes a native client |
Section 4
Path loss grows with the square of frequency, so every octave upward costs roughly 6 dB of link budget, while the wide allocations that carry tonnage only exist higher up. No band is both wide and far.
4.1 The ladder
TVWS at 470 to 608 MHz buys the lowest path loss and the best diffraction and foliage penetration in exchange for 6 MHz channels. HaLow at 900 MHz trades a little reach for a cheap standardized IoT ecosystem. CBRS at 3.5 GHz and Wi-Fi at 5 to 7 GHz trade reach for the channels that carry the site's tonnage. The three-zone model is this physics made into a product promise.
4.2 The uplink
Carrier TDD frames are fixed for download-dominant consumer traffic, typically a quarter of airtime or less for uplink. A private CBRS carrier under the site's own scheduler inverts that: the same 100 MHz grant becomes 300 to 500 Mbps of sustained upload, the direction jobsite traffic actually flows. A policy freedom, not a radio improvement.
4.3 Diversity
Seven bands span 470 MHz to 7.125 GHz, four regulatory regimes, and three coordination mechanisms; their interference environments, propagation failure modes, and legal statuses are largely uncorrelated. Weaker than a licensed guarantee on one band, stronger than any single band against the full set of real failure modes, and it requires owning nothing. PALs remain a purchasable hardening.
Section 5
The deepest enabler is regulatory, not technical. Over twelve years the FCC replaced exclusive licensing with database coordination across three successive regimes, and all three share one transaction shape: a device at a location asks a database what it may transmit, and receives a grant.
5.1 The inversion
PAWS made television white space queryable in the early 2010s. The CBRS SAS framework proved in 2020 that a dynamically coordinated three-tier band could carry carrier-grade infrastructure nationally. The 6 GHz AFC extended the pattern to 1,200 MHz of prime mid-band. An API can be automated, which is what makes zero-touch multi-band deployment lawful rather than merely convenient.
5.2 The ledger
One client architecture with per-regime adapters, one geolocation and device-identity substrate, one grant lifecycle, one record store. Every grant event is written to an append-only, tamper-evident ledger keyed by device, location, band, power, and time, and the same pattern extends to attachments, neutral-host sessions, and WAN path assignments. This is the form of assurance that autonomy liability, insurance underwriting, and audit actually consume.
5.3 The hard case
Database coordination was designed for fixed infrastructure, and its hardest edge is a radiator that moves: a CBRS relay on a haul truck holds a grant bound to a changing geolocation. A ledger-native architecture treats it as a data problem: continuous position integrity, grant renewal on movement per policy, automatic radiation suspension when authorization lapses, and a complete lifecycle record. Fleet-as-network is lawful and auditable because this case was designed in from the start.
Section 6
Classical baseband silicon hard-wires one waveform family. The ORB cellular complex specifies a programmable signal-processing fabric on which LTE and 5G NR are firmware loads and future waveforms are ports.
6.1 The ASIP baseband
The TVWS carrier is not a separate radio product but the same LTE stack retargeted to UHF through a different RF chain, so the tail tier costs an antenna section rather than a vendor relationship. The band plan evolves by firmware. And a full-specification DECT NR+ physical layer, which commercial silicon only partly implements today, becomes engineering backlog on owned tooling rather than a request to a chip vendor's product committee.
6.2 The mezzanine boundary
The baseband is a removable mezzanine with a normative interface: PCIe and multi-gigabit Ethernet for data, a defined O-RAN functional split for fronthaul, PTP timing ingress, and a published mechanical envelope. Baseband silicon is the one layer where geopolitical provenance can matter to a buyer, and the mezzanine turns that from an architectural commitment into a sourcing decision.
6.3 Symmetry
With the same programmable baseband on both ends, the link becomes a designed system: uplink-heavy frames are honored end to end because both schedulers agree; a machine crossing from CBRS into the TVWS tail keeps its session because both ends run the same transition logic; deterministic traffic is scheduled cooperatively rather than negotiated through standard QoS. Symmetry is also what makes the dual-role CPE economical, since the marginal cost of a coverage point approaches a second RF chain rather than a second product.
6.4 The merchant boundary
Everything outside the cellular complex stays on merchant silicon: Wi-Fi 7, HaLow, NR+, and multiprotocol 802.15.4 and BLE modules are certified, cheap, mainline-supported, and independently replaceable. Custom engineering concentrates where differentiation lives. The one deliberate asymmetry is the NR+ roadmap: merchant parts serve the deterministic tier today, and the ASIP platform holds the option on a full implementation tomorrow.
Section 7
A single shared aperture cannot span 470 MHz to 7.125 GHz: a structure resonant at the bottom of that range is electrically enormous and couples destructively into everything above it. The integrated panel solves it the way multiband macro panels solved their own 5:1 versions, then extends it.
Consequence one
Isolation becomes a designed, chamber-verified property, which is a precondition for zero-touch deployment: the mast assembles by connector map and torque value because every hard RF decision was made once, in the factory.
Consequence two
Per-band and per-beam gain patterns and isolation matrices are consumed directly by AFC registrations, SAS filings, and white space database entries. The antenna documentation and the coordination plane are two views of the same engineering, and a second-source manufacturer inherits both from the specification alone.
Section 8
Safety-relevant coordination cannot share fate with bulk data, and the platform separates them twice over: by regulation and by physics.
By law
DECT NR+ operates in spectrum reserved for DECT technology, so the command-and-control layer faces no Wi-Fi, no ISM traffic, and no site data by law rather than by luck.
By physics
At 1.9 GHz the tier sits nearly a gigahertz from the congested access bands in both directions, so even pathological saturation of the data tiers is invisible to it.
By protocol
Deterministic scheduling, mesh self-healing, and direct device-to-device operation mean machines coordinate with machines with the infrastructure removed, and the mesh rejoins the ledger when connectivity returns.
Stated in the language a safety case uses: the probability that fleet command and control degrades due to load, interference, or failure elsewhere in the site network is engineered toward zero by construction, not managed toward zero by prioritization. QoS on shared spectrum manages contention; a reserved band with a frequency gap removes it.
Section 9
The control plane is deliberately an assembly of production-hardened open components. Original engineering concentrates in exactly three places.
Assembled
Owned
Everything else is integration, a cost advantage that compounds: the open foundations are maintained by their communities, patched at ecosystem scale, and familiar to every engineer the program will ever hire.
The edge-compute inversion
Even the best portable WAN is an order of magnitude thinner than the site fabric, so Class A units process capture workloads locally and ship results. The site feels LAN-fast on any WAN because the WAN was demoted from the experience path to the synchronization path.
Section 10
Graceful degradation is a doctrinal requirement, so it is worth stating exactly what the architecture guarantees under each failure class.
| Failure | Cause | What degrades | What remains |
|---|---|---|---|
| One access band lost | Interference, database outage, or hardware | Aggregate capacity on that tier | All other bands; steering rebalances; deterministic tier unaffected |
| All WAN paths lost | Every uplink down at once | Cloud sync, remote access, offload interconnect | Entire local fabric; edge compute keeps workflows running; C2 mesh unaffected |
| Serving ORB lost | The base station itself goes dark | Site-wide capacity tiers and coordination services | CPE Oases continue locally; NR+ C2 mesh continues machine to machine; records queue for the ledger |
| Coordination database unreachable | PAWS, SAS, or AFC unreachable | New grants on the affected regime | Existing grants per regime policy; all other regimes; unlicensed tiers unaffected |
| GNSS degraded | Jamming, canyon, or antenna fault | New location-bound grants; timing holdover clock starts | Operation on holdover per Part 96 limits; alarm raised; non-cellular tiers unaffected |
| Site power lost | Generator or mains failure | Mains-fed infrastructure | Machine-mounted Oases on vehicle power; portable units on battery; C2 mesh rides the fleet |
The pattern across every row is the architecture's deepest property: no single failure severs connectivity or the deterministic tier, because no single element is load-bearing for either. The fleet itself is the redundancy of last resort.
Section 11
A jobsite network must serve three device populations it does not control, and the architecture answers each with a standards body's work rather than a proprietary bridge.
The phones
The universal population is served to kilometer-plus range with no app, no provisioning, and no client hardware, because receive-side array gain closes the uplink from phone-class transmit power.
The commodity devices
A border router on every unit admits the commodity IoT population over Wi-Fi and Thread, and multiprotocol silicon adds BLE mesh for person-and-asset scale at zero hardware cost.
The subscriptions
OnGo shared-HNI with MOCN on CBRS, and Passpoint with OpenRoaming on Wi-Fi, let licensed carriers' subscribers ride open-spectrum infrastructure natively. That closes the site's emergency-reachability gap and inverts the historical relationship: the licensed world pays to ride the open one.
Section 12
Every hardware element is procurable: merchant modules, antenna manufacturing, LEO terminals, even programmable basebands. The defense rests on three layers that do not commoditize, and on the shape of the problem itself.
Section 13
Spectrum became an API, silicon became programmable and licensable, radios became certified commodities, backhaul became portable, and network software became an open assembly. The architecture is the disciplined composition of those facts.
ORB-TR1 is the companion technical report to the specification suite. We share the complete document with operators, integrators, and second-source manufacturers under evaluation.