A fiber route can reach the right cabinet and still fail the carrier's transport requirements. The missing input is often the processing boundary: which radio functions stay at the site, and which move to a remote baseband location. We establish that boundary before treating a route as suitable for a wireless connection.
For an ISP or neutral host, the practical question is what the offered circuit must support. A tower company needs a clear handoff. Fiber designers and builders need an approved route basis that survives a change in equipment location without silently changing the service being delivered.
For outside-plant design, we compare these transport segments by asking what each requires of the route. The useful output is a route with documented assumptions, rather than a generic reach figure attached to the word fronthaul. The carrier's architecture controls the starting point.
Backhaul vs fronthaul starts with the RAN split
The distinction in backhaul vs fronthaul is where traffic travels relative to baseband processing: fronthaul connects the radio unit to baseband functions, while backhaul connects the baseband or aggregation hub toward the core. A split 5G architecture can also require midhaul between distributed and centralized baseband functions, changing the transport requirements.
Fronthaul: radio unit to baseband
Fronthaul carries information exchanged across the radio-to-baseband boundary. Its payload depends on where processing is divided. Traditional CPRI transports digitized radio information with a different traffic model from packet-based eCPRI, so we ask for the actual interface specification before assigning fiber or accepting a leased connection as suitable.
Whether the radio unit and its serving baseband functions occupy the same site or separate locations determines which connection the outside plant must support. A drawing labeled simply "fiber to tower" does not tell us whether the incoming OSP cable carries fronthaul or whether fronthaul begins after the ground-level baseband equipment. We identify both endpoints in the route schedule.
The physical run up the tower is the fiber-to-the-antenna path, with cable support and tower-entry details governed by its installation requirements even when the logical traffic is fronthaul. That physical description does not establish the RAN split. We keep the route description separate from the interface name.
Midhaul: distributed unit to centralized unit
Midhaul connects separated baseband functions, commonly a distributed unit (DU) and a centralized unit (CU). In 3GPP TS 38.401, NG-RAN Architecture Description, the gNB-DU connects to the gNB-CU over F1. F1 carries control-plane and user-plane communication; it occupies a different processing boundary from the lower-layer radio connection.
With the DU and CU in one facility, that interface may need no separate outside-plant route, but a remote CU makes its location part of the transport design. We mark the F1 endpoints explicitly rather than assuming every baseband link has the same allowable delay.
A fiber designer given only the label "Centralized RAN" still needs the named split to understand the assignment: moving a DU changes the fronthaul route while moving only a CU changes the midhaul route. Our drawings should let the carrier review that distinction without decoding a cabinet label or tracing an undocumented patch.
Backhaul: baseband or aggregation toward the core
Backhaul carries processed traffic toward the core through the operator's transport network. In the CU/DU architecture described by 3GPP TS 38.401, the CU provides connectivity toward the core. A physical aggregation hub can serve that path, but a hub that only combines radio-to-DU circuits still aggregates fronthaul traffic.
Following the logical endpoints through intermediate facilities lets us identify the segment even where traffic is aggregated. We also record the owner of each handoff, especially where a neutral host supplies passive infrastructure to several carriers. Shared duct space can contain circuits with different transport requirements even when every cable terminates in the same building.
General backhaul options include fiber, point-to-point microwave, millimeter-wave radio and satellite, with suitability determined by the operator's service requirements. This article concerns the fiber route implications. The network owner selects its transport architecture, and we use that approved basis to define the civil and optical pathway.
Backhaul vs fronthaul: compare the transport requirements
Use the comparison below alongside the interface schedule, reading each column against the carrier's selected split and the circuit's actual endpoints. We keep latency and synchronization separate because an acceptable packet transit time does not establish acceptable phase alignment at a radio.
| Design question | Fronthaul | Midhaul | Backhaul |
|---|---|---|---|
| Purpose | Radio unit to baseband functions | Separated baseband functions, commonly DU to CU | Baseband or aggregation hub toward core |
| Latency | Split-specific transport budget; lower-layer processing is sensitive to delay | F1 and operator performance budget | Service and end-to-end application budget |
| Timing | Frequency and phase/time requirements follow the radio use case | May distribute timing according to the synchronization architecture | May carry timing toward access sites; verify the selected profile |
| Capacity | Radio configuration and functional split determine transport load | Traffic across the DU/CU boundary and growth allowance | Aggregated traffic and capacity during failures |
| Medium | Often fiber, dedicated or shared through approved optical transport | Usually packet transport over the selected physical network | Fiber or a qualified wireless transport alternative |
| Resilience | Alternate path must preserve the split's transport budget | Protection must retain DU/CU connectivity within requirements | Alternate path must carry surviving demand within service limits |
IEEE 1914.1-2019, IEEE Standard for Packet-based Fronthaul Transport Networks, distinguishes NGFI-I low-layer splits from NGFI-II high-layer splits. It defines architecture and transport-node requirements. Those terms reinforce why a label on a fiber map cannot replace the functional split, and why we require the operator's interface definitions before final design.
Match capacity to the traffic at that boundary
When reviewing the operator's traffic model for capacity planning, we ask whether the supplied figure represents an interface line rate or an expected traffic load, because those values support different decisions about sharing a path. The fiber designer should not apply a backhaul oversubscription assumption to a fronthaul circuit merely because both arrive on Ethernet interfaces.
If the carrier expects another optical interface at a node, we give that growth a physical assignment showing how it reaches the serving hub and where it terminates; if growth stays within an existing managed service, the provider must confirm the available transport capacity. Those cases can produce different cable and termination requirements for the same forecast.
Traffic moving onto a surviving link creates a loading condition that needs its own review; the link's normal utilization no longer describes the demand it must carry. We preserve the protection-path assumptions in the route package so the transport team can evaluate its capacity without reconstructing the passive network from disconnected drawing notes.
Build separate latency and timing budgets
Allocate delay before measuring route suitability
Before filling in the latency worksheet, we establish its measurement boundaries and obtain the permitted one-way transport delay and the portion allocated to the passive route, with active-network allowances supplied by the carrier or its transport designer. Round-trip and one-way figures cannot be substituted for each other without knowing what was measured.
Measuring a straight line between facilities would leave slack and building entry routes out of the cable path. Even a GIS centerline may omit a diversion around a crossing that has not received approval. We reconcile the route length with the installation drawings before calculating the propagation contribution to the allocated transport delay.
The eCPRI Transport Network Requirements V1.2, Table 1A, defines High25, High100, High200 and High500 classes with maximum one-way frame delays of 25, 100, 200 and 500 microseconds. The document includes propagation and switching delay within that transport measure. Those classes describe requirements for specified flows, not universal allowable fiber distances.
To keep the remaining delay allowance explicit, we check the carrier's allocation. If it gives us a total transport allowance without reserving active-network delay, we return that unresolved allocation before accepting the route; otherwise a favorable fiber calculation can conceal a path that fails once intermediate equipment is included. The normal and protection paths need separate calculations.
Route-review tip: Put the delay measurement boundaries beside the route-length assumption. This lets the carrier identify a missing allowance before a civil drawing is mistaken for approval of the complete transport path.
Preserve the synchronization design
For the separate synchronization limits, ITU-T G.8271.1 addresses network limits for time synchronization with full timing support, while G.8275.1 defines the associated Precision Time Protocol telecom profile for phase/time synchronization. G.8275.2 addresses partial timing support. The selected profile and network arrangement determine which timing assumptions belong in the design basis.
Our OSP record supports that work by preserving actual paths and handoff locations. A reroute can change directional delay asymmetry, and a protection event can move traffic onto a different path with different timing behavior. The carrier must evaluate those effects against its synchronization requirements rather than infer timing performance from optical continuity.
Physical changes that affect the carrier's analysis go into our route documentation so the equipment team has the path information it needs when configuring the clocks. For route teams that need the passive-path calculations coordinated with the drawings, our fronthaul and backhaul fiber route design covers that interface between latency inputs and the proposed outside plant.
Translate interfaces into fiber and optical loss budgets
Count assigned fibers before selecting a cable
To turn a radio count into a fiber count, we need to know the optical arrangement. A duplex optical link uses 2 fibers, while a supported bidirectional link can use one; the approved interface and optical arrangement decide which assumption applies. We separate working assignments from reserved capacity so a spare strand never gets mistaken for a second protected route.
With wavelength multiplexing, the allocation changes because it can carry several channels over a common fiber path, but it introduces optical components and dependencies that the carrier must approve before our fiber schedule treats those channels as available. A dark-fiber handoff has a different assignment basis from a managed packet service.
Tracing each required handoff through the splice schedule to its destination gives us a way to check those assignments. Our small-cell fiber backhaul engineering guide develops the per-node planning question further. The count should remain understandable after a node is added, rather than relying on an unexplained reserve percentage in the cable label.
Keep loss and delay in different worksheets
In the optical power budget, we account for the selected fiber attenuation and each planned splice or connector pair, using the approved optical component specifications and the owner's acceptance criteria. A route can meet its loss allowance and still exceed its latency allocation; neither result substitutes for the other.
Because passive joints consume optical margin and affect maintainability, we record their locations in the loss worksheet and route documents. Their count cannot establish a packet-delay limit, so we evaluate latency from the actual cable path and the carrier's active-network assumptions without inventing a universal delay penalty per fusion splice.
During the original design, we allow for future cable repairs that may add joints and replacement cable. We reserve the optical margin required by the owner and make the underlying assumptions visible. Our limitation is that a passive-route calculation cannot certify packet behavior through equipment whose configuration and loading we do not control.
Hub-location check: Review the building entry before accepting the route endpoint. An approved street-side connection leaves an unresolved design gap when the carrier's handoff sits elsewhere inside the facility.
Specify what the acceptance evidence must prove
For acceptance of the passive path, optical continuity and insertion-loss results answer different questions from OTDR testing: its traces help document its events and length under the agreed test method. We ask the carrier to identify the separate transport-performance evidence needed before service acceptance, including the conditions under which the protected path will be evaluated.
The acceptance package should preserve fiber identities through every handoff. If the test report names a port differently from the splice drawing, the owner cannot confidently associate a passing result with the circuit under review. We resolve that naming mismatch in the records before treating the package as complete.
Locate hubs around feasible and resilient routes
Compare locations using the serving paths
When comparing hub locations, we evaluate every connected branch using feasible cable routes and the carrier's processing placement, rather than choosing the geometric center of a coverage area and assuming the fronthaul allowance will follow. Existing fiber helps only when its path and usable assignments meet the selected interfaces.
At a building entrance, even a favorable route may lack a usable pathway to the agreed handoff. That access problem leaves the nominal endpoint different from the point used in the delay worksheet. We carry the design through that entrance and identify where the customer's inside-plant responsibility begins.
Check common failures and the alternate path
Separate fibers in the same cable remain exposed to the same cable cut. Even with separate cables, a shared duct bank can leave both circuits vulnerable to the same excavation, so the review must consider how they could fail together. We show common structures on the route review so the owner can decide whether the proposed protection matches its resilience requirement.
On a ring, protection may send traffic along the longer side, changing the relevant delay and increasing demand on the surviving segment. Calculating that alternate path establishes whether the operating state remains acceptable; the ring symbol supplies none of those results. Our cell-tower fiber backhaul design discussion explains the ring-planning context behind that review.
The route comparison should carry each alternative far enough to expose shared crossings and constrained entrances. We use the approach described in fiber route design for 5G backhaul to connect corridor decisions with the network requirement. An alternate alignment only helps when the owner can build and use it under the required operating conditions.
Choose the design inputs before releasing the route
Confirm the design basis: Freeze the functional split and handoff requirements before releasing the fiber design. Provide the transport and timing allocations that the passive route must meet, then require a separate review when a DU or CU moves. We recommend treating that relocation as a design change even if the civil corridor remains available.
Document the shared infrastructure: Define each tenant's handoff and identify shared physical failures. Reserve assignable fiber with a documented purpose, rather than selling protection based only on separate strand numbers. Our engineering and delivery team can help establish the route scope while the carrier retains responsibility for its radio and transport equipment.
Reconcile the release package: Release drawings only when the fiber assignments agree with the approved endpoints. Carry field deviations back into the length and loss calculations, and retain the revised records in closeout. A changed entrance can affect the calculation even when the outside cable still follows the permitted corridor.
Draftech's engineering is 100% in-house, and full turnkey construction is delivered by Draftech-managed subcontract crews under Draftech's QA/QC. We coordinate the passive route from the carrier's requirements through the construction documents, keeping equipment configuration with the responsible provider. Send the proposed endpoints and available interface requirements through our project contact page or email info@draftech.com.
For a route that qualifies for our free engineering offer, Draftech engineers the first 20,000 linear feet at no cost, from feasibility and field survey through permit approval. The owner reviews each request before Draftech commits the package. That review lets us establish the route's scope and the carrier inputs needed to begin.

