Marginal GPON Power Budget Ahead of an XGS-PON Overlay
Helder Telecom serves 480 subscribers from one GPON OLT port through a 1:64 splitter over 19 km of G.652D feeder and 1.5 km of drop, using Class B+ optics (28 dB nominal budget at 1490 nm). At the ONT bulkhead, 12 of the 64 ONTs read -28.6 dBm downstream; all 12 sit on the longest feeder branches. Those ONTs re-range 5-20 times per day and 40% of them lost their provisioning twice last month. The physical path includes 9 fusion splices at 0.1 dB each, 6 connectors at 0.4 dB each, and the design assumed 2.0 dB of margin for ageing and future repairs. The operator plans to overlay XGS-PON on the same fibre with a WDM combiner (1.2 dB insertion loss) within 12 months. Subscriber growth is flat, but marketing wants to sell 1 Gbps packages on this footprint next quarter.
1. Build the link budget and show quantitatively why these ONTs are failing.
Read the worked answer
Loss build-up, 1490 nm downstream (longest branch):
| Element | Loss |
|---|---|
| 1:64 splitter | 20.5 dB |
| Feeder 19 km + drop 1.5 km, G.652D at 0.25 dB/km (cable-inclusive) | 5.1 dB |
| 9 fusion splices at 0.1 dB | 0.9 dB |
| 6 connectors at 0.4 dB | 2.4 dB |
| Installed loss | 28.9 dB |
Against the Class B+ budget of 28 dB: the link is already 0.9 dB over budget with zero margin, before the 2.0 dB design margin and before the 1.2 dB XGS-PON combiner. With an OLT launch power of +1.5 dBm (Class B+ minimum) and ONT sensitivity of -27 dBm, the worst-case received power is 1.5 - 28.9 = -27.4 dBm, i.e. below sensitivity. The -28.6 dBm measured is consistent with that: the ONTs are operating outside their specified range.
Why they flap rather than simply fail: an ONT below sensitivity still occasionally decodes a frame at the noise margin, so it ranges, loses the ranging grant, and re-ranges. Each loss of frame triggers re-ranging and, on some vendors, a configuration re-download. The 5-20 re-ranges per day and the provisioning losses are therefore a symptom of optical margin, not a provisioning-server fault or a faulty batch of ONTs.
Read-across: the 12 affected ONTs are exactly the longest branches, which is the signature of a distributed loss problem, not a card fault.
2. Rank the remediation options by cost and benefit, and answer marketing's request to sell 1 Gbps packages next quarter.
Read the worked answer
Ranked remediation (cheapest useful action first):
- Inspect and clean every connector on the 12 branches; re-terminate any with poor polish. A dirty SC/APC connector typically costs 1-3 dB and is often the single biggest recoverable loss. Cost: hours of technician time; benefit: up to several dB.
- Re-splice or replace the worst feeder sections if OTDR shows a lossy splice above 0.1 dB.
- Re-split for the affected branches. Move 16 subscribers onto a second OLT port with a 1:32 splitter: 20.5 dB becomes 17.5 dB for the moved subscribers, a 3.0 dB gain, and the remaining 1:64 port gains headroom as users leave. Cost: one OLT port and one splitter; benefit: 3.0 dB plus future headroom.
- Upgrade to Class C+ optics (32 dB budget). This must be done at both OLT and ONT - a C+ OLT with B+ ONTs buys nothing on the downstream. Cost: fleet-wide ONT replacement on the long branches; benefit: 4 dB.
On 1 Gbps next quarter: no. This is GPON: 2.5 Gbps downstream shared across 64 subscribers, roughly 38 Mbps of committed average per subscriber at a 1:1 contention assumption, and the upstream is 1.25 Gbps shared - a 1 Gbps package cannot be delivered on a 1:64 GPON port even with perfect optics. Fix the optical budget now (items 1-3), and sell 1 Gbps only after the XGS-PON overlay, on branches designed for it.
3. What changes when XGS-PON is overlaid, and what would you insist on before the combiner goes in?
Read the worked answer
XGS-PON uses 1270 nm upstream and 1577 nm downstream, so it coexists with GPON 1310/1490 through a WDM combiner - but coexistence is optical, not free.
Budget impact:
- The combiner adds about 1.2 dB at each end. That takes this link from already 0.9 dB over budget to roughly 3-4 dB over budget for GPON. The GPON power problem must be fixed before the overlay, not after - the combiner will push the 12 marginal ONTs fully out of service.
- XGS-PON does not have the same 28 dB budget. Class N1 optics give 29 dB, N2 about 31 dB and E2 about 33 dB. At 20.5 km with a 1:64 splitter (20.5 dB loss) plus splices and connectors, N1 is marginal and N2/E2 is the realistic choice. Design XGS-PON as 1:32 on the long branches and 1:64 only where the fibre is short.
Insist on before installation:
- A full OTDR trace and bi-directional power reading on every branch, with the 12 marginal ONTs remediated and re-measured.
- A documented loss budget per branch that includes the combiner and the 2 dB ageing margin, signed off against the chosen XGS-PON optics class.
- 1650 nm test-port filtering or a plan for live OTDR testing, since the combiner blocks the out-of-band test wavelength unless a filter port is provisioned.
- A migration plan that keeps GPON ONTs on the same splitter: XGS-PON ONTs and GPON ONTs on one combiner port require the combiner to be wavelength-selective per direction, and a mis-cabled port takes out both services.
- A per-port subscriber cap for the 1 Gbps tier, since XGS-PON is 10G/10G shared and 64 x 1 Gbps is an oversubscription of 6:1 before any contention.