Skip to content
ISP Operations

Access & Last-Mile Networks

The last mile is where subscribers actually experience your network — and where most African ISPs bleed capacity, spectrum, and reputation. We design, upgrade, and operate fibre and fixed-wireless access networks built to survive growth.

Your core can be perfect. If the access layer is improvised, the business is improvised.

01 — Why Access

The Access Layer Is Where ISPs Win or Lose

Backbones are the easy part. The hard engineering lives between your PoP and the subscriber: splitter budgets, sector loading, interference floors, CPE quality, and the pole-by-pole decisions nobody documents.

Most access networks in our markets were not designed — they accreted. A tower here, a splice closure there, whatever radio was in stock that month. It works at 300 subscribers. At 3,000 it produces nightly congestion, unexplainable outages, and support queues that eat margin.

Our practice covers both physical media honestly: fibre where density and payback justify the civils, fixed wireless where they don’t — and engineered handover between the two, because in African deployment practice the answer is almost always a hybrid.

02 — Fibre Access

Fibre Access: GPON, FTTH & FTTB

PON economics reward disciplined outside-plant design — the fibre you bury outlives three generations of electronics

OLT & Headend Planning

  • OLT placement, sizing, and port economics
  • GPON today, XGS-PON-ready ODN by default
  • Redundant uplinks and power engineering
  • VLAN and service-model design per PON port
  • ONT/ONU provisioning and profile discipline

ODN & Splitter Design

  • Optical budget planning (Class B+/C+)
  • Centralized vs cascaded splitter topologies
  • Split ratios matched to take-rate reality, not hope
  • Closure, FDH, and FAT placement for maintainability
  • As-built documentation and fibre records that survive staff turnover

FTTH / FTTB Deployment Practice

  • Homes-passed vs homes-connected cost modelling
  • FTTB with in-building Ethernet or GPON risers
  • Aerial vs buried trade-offs on real African civils costs
  • Wayleave, pole-attachment, and make-ready planning
  • Acceptance testing: OTDR, power meter, and light-level records per drop

Capacity & Upgrade Path

  • Per-PON oversubscription modelling against actual usage data
  • Coexistence planning: GPON and XGS-PON on one ODN via WDM
  • Upgrade triggers defined before you need them
  • Business-service overlays (dedicated fibre, active Ethernet) on shared plant

The 2026 economics are settled: XGS-PON electronics carry roughly a 20–30% premium over GPON, which dilutes to a few percent of total cost per home passed. Build the ODN once, to XGS-PON optical standards, and let the electronics follow demand.

03 — Fixed Wireless

Fixed Wireless Access & Distribution

Wireless is not the cheap option — it is a different engineering discipline, and it punishes operators who treat it casually

PtMP Access

  • Sector planning: azimuth, tilt, and frequency reuse
  • GPS-synchronized platforms to enable dense reuse
  • Realistic per-sector subscriber and throughput budgets
  • Cambium, MikroTik, and Ubiquiti-class platform selection by role, not brand loyalty
  • Modulation and SNR targets enforced at install time

PtP Backhaul & Distribution

  • Licensed microwave for trunk links that must not fail
  • Unlicensed 5/6 GHz and 60–80 GHz for metro distribution
  • Path profiling, Fresnel clearance, and rain-fade margins
  • Ring and mesh topologies over pure hub-and-spoke
  • Capacity headroom rules tied to sector growth

Licensed vs Unlicensed Strategy

  • When a licensed band assignment pays for itself
  • Regulator engagement: CAK-class licensing and frequency assignment processes
  • Unlicensed-band survival tactics in congested urban RF
  • 6 GHz outdoor use tracked as regulators open the band
  • Migration sequencing from unlicensed to licensed without subscriber downtime

Tower & Site Engineering

  • Structural loading, grounding, and lightning protection
  • Power autonomy: solar, battery, and generator sizing for grid-poor sites
  • Colocation negotiation and interference coordination with neighbours
  • Site documentation, labelling, and spares strategy
Every unplanned frequency is a future outage with a date you don’t know yet.

05 — RF Discipline

RF & Spectrum Discipline — Proven in the Field

Interference is the silent killer of wireless ISPs. It rarely takes a network down outright; it degrades modulation, inflates retransmissions, and shows up in your business as churn nobody can explain. Fixing it takes spectrum analysis, a frequency plan, and the discipline to enforce it — not bigger antennas.

We have done this work where it is hardest. For DSI (DATCO Service Internet) in Eastern DRC, we upgraded the RF and core network from Ubiquiti to carrier-grade Cambium Networks and Cisco equipment across licensed and unlicensed bands — ending the chronic interference-driven outages that had defined the network — and delivered the IP and RF plan for residential expansion, including VLAN segmentation and spectrum optimisation.

At SkyTrend Networks, a CAK-licensed NFP ISP in Kenya, we re-designed the fibre-optic and wireless network infrastructure and systems architecture for nationwide expansion. Our frequency-planning practice is grounded in formal Cambium training (2016) and twenty-plus years of operating networks where spectrum is contested and power is unreliable.

A frequency plan that lives in one engineer’s head is not a plan. We leave you with documented channel assignments, reuse maps, and change control.

06 — Migration Path

The WISP-to-Carrier-Grade Migration Path

  1. 01

    Audit the RF Reality

    Spectrum sweeps, sector loading analysis, link budgets recomputed from measured data — not the install-day assumptions.

  2. 02

    Stabilize

    Frequency re-plan, sync enforcement, and elimination of the worst self-interference before any new hardware is bought.

  3. 03

    Re-Platform Deliberately

    Carrier-grade radios and routing where they earn their cost, staged sector by sector with rollback paths and no subscriber flag days.

  4. 04

    Layer In Fibre

    GPON into the dense clusters wireless struggles to serve, freeing spectrum for the areas where wireless remains the right answer.

  5. 05

    Operationalize

    Monitoring, capacity thresholds, spares, and documented change control — so the network stays carrier-grade after we leave.

The upgrade is a sequence, not a purchase order. Buying carrier-grade radios into an unplanned RF environment reproduces the old problems at a higher price.

07 — CPE

CPE & the In-Home Experience

Subscribers judge you by their Wi-Fi. Most "network complaints" end at the customer’s router.

CPE Strategy

Standardize on a short, tested CPE list per access technology and buy against a lifecycle plan, not the cheapest carton this quarter. Operator-owned, centrally managed CPE costs more per unit and less per subscriber once truck rolls are counted.

Remote Management

TR-069/TR-369-class management, zero-touch provisioning, and firmware discipline turn the CPE fleet from a support liability into a diagnostic sensor at every subscriber premises.

In-Home Wi-Fi Quality

Placement guidance at install, dual-band or Wi-Fi 6 units where plans exceed what 2.4 GHz can carry, and mesh only where the building demands it. The goal is that the speed you sell is the speed the phone sees.

Support Economics

Every CPE decision is a support-cost decision. Visibility to the CPE — signal levels, retrains, Wi-Fi congestion — lets first-line support resolve calls without dispatching a technician to discover a router behind a fridge.

08 — Capacity

Capacity Planning at the Access Layer

Access capacity fails quietly, then all at once. We put numbers and thresholds on the layer most operators run on feel:

  • Per-sector and per-PON utilization tracked against defined upgrade triggers
  • Oversubscription ratios derived from your measured peak-hour data
  • Busy-hour forecasting tied to sales pipeline, not last year’s average
  • Spectrum headroom treated as inventory — audited, allocated, and reserved
  • Backhaul sized ahead of access, so an upgrade at the edge is never throttled one hop up

Growth should be a provisioning task, not an engineering emergency.

Field note — 2026

The 2026 access-layer picture is unusually clear: GPON still carries the volume, but with XGS-PON electronics settling at a modest premium the only defensible new ODN is one built to 10G optical standards — the fibre outlives the electronics by decades. On the wireless side, the interesting fight is spectrum, not speed: 5 GHz is saturated in every African city we work in, regulators are moving at very different paces on outdoor 6 GHz, and the WISPs that thrive are the ones treating frequency plans and GPS sync as operational discipline rather than install-day settings. The vendors will sell you gigabit radios all day; nobody sells clean spectrum.

Assess Your Access Network

Start with a structured review of your fibre plant, RF environment, and last-mile capacity — before growth turns quiet weaknesses into outages.