GPON shelf upgrade: 1:64 re-split versus XGS-PON optics
An operator runs a 7360 ISAM FX in a suburb with four GPON line cards, 1:32 splitters and 3,800 residential subscribers. Two cards reach 85–92% downstream utilisation between 20:00 and 22:30, driven by 4K streaming and work-from-home uploads; upstream peaks at 70%. The business case is to move 2,000 subscribers to 1 Gbit/s symmetric service within 12 months. The ODN is single-stage: exchange to street cabinet, then passive to the home. The longest path is 15 km with 18 fusion splices and four connectors; planning figures are 0.35 dB/km at 1310 nm and 0.25 dB/km at 1550 nm, 0.10 dB per splice and 0.30 dB per connector. The field team proposes re-splitting the two hot cards from 1:32 to 1:64 to free ports. Deployed GPON optics are class B+ (28 dB); the cabinet has one free 2U and limited power headroom.
1. Calculate whether XGS-PON over 1:64 at 15 km fits a 29 dB (N1) optical budget in both directions. Which direction binds, and what would you actually deploy?
Read the worked answer
Use a planning figure of 3.5 dB per 1:2 stage, so 1:64 = 6 × 3.5 ≈ 21.0 dB.
Downstream, 1577 nm
- Splitter 21.0
- Fibre 15 × 0.25 = 3.75
- Splices 18 × 0.10 = 1.80
- Connectors 4 × 0.30 = 1.20
- Total 27.75 dB → 1.25 dB margin against N1
Upstream, 1270 nm
- Splitter 21.0
- Fibre 15 × 0.35 = 5.25
- Splices and connectors 3.00
- Total 29.25 dB → negative margin against N1 (29 dB)
Upstream binds because 1270 nm has higher fibre loss and the splitter loss is the same in both directions. A 1.25 dB downstream margin is also below the 3 dB normally reserved for ageing, temperature and repair splices.
Deploy N2 optics (31 dB) on the 1:64 ports, which restores roughly 1.75 dB upstream margin, or keep 1:32 for upgraded ports (23.25 dB upstream, about 5.75 dB margin). Given the 2U and power constraints, using more 1:32 ports on additional cards is the lower-risk choice unless port density is the binding constraint.
2. Compare combo PON against an external WDM overlay for this shelf. Would you accept the 1:64 re-split on the shared ODN?
Read the worked answer
Combo PON puts GPON (1490/1310 nm) and XGS-PON (1577/1270 nm) on one port with an integrated coexistence filter, so it reuses one fibre and one ODN and saves the cabinet space you do not have. That is the main argument in its favour here.
But the splitter is shared, so both budgets must close. At 1:64, class B+ GPON is 21.0 + 3.75 + 1.80 + 1.20 = 27.75 dB against a 28 dB budget - 0.25 dB margin, which fails as soon as you add an ageing allowance or a repair splice. So the 1:64 re-split on the shared ODN is not acceptable.
Practical design: stay at 1:32 and use class C+ GPON optics (32 dB) plus N2 XGS-PON. If a 1:64 split is required for density, apply it only to XGS-PON-only ports where no GPON wavelength is present.
An external WDM1r overlay is the fallback if the installed GPON card cannot host combo optics, but it consumes the free 2U and adds a powered element - the wrong trade in this cabinet.
3. Size the upgrade for 2,000 subscribers at 1 Gbit/s symmetric. What split ratio and uplink capacity would you specify, and what must the SLA say?
Read the worked answer
Sold rate is 2,000 × 1 Gbit/s = 2 Tbit/s, which is clearly statistical, not committed.
Port count. At 1:32, 2,000 subscribers need about 63 PON ports. With 16-port XGS-PON cards that is four cards, matching the existing four-card shelf footprint.
Concurrency. If 30% of subscribers on a fully loaded port demand 1 Gbit/s simultaneously, that port needs 0.30 × 32 × 1 Gbit/s = 9.6 Gbit/s - 96% of a 10 Gbit/s PON. That is not a design point. Either cap the design at about 20% concurrency (6.4 Gbit/s, 64% utilisation) with per-subscriber shaping and DBA, or use 1:16 for premium tiers.
Uplink. Each 16-port card needs 2 × 25G (or at minimum 2 × 10G) toward the aggregation layer; verify the shelf fabric and uplink before ordering cards.
SLA wording. Because upstream is also 10 Gbit/s shared, a 1 Gbit/s upstream rate must be written as a peak rate with a committed information rate, plus contention and DBA guarantees. Otherwise the first busy hour produces a commercial dispute.