The short version: Before you pick an EDFA, do the link budget on paper. Add up the fiber, the splitters, and the connectors on your actual route, subtract that from your launch power, and check what’s left at the far end. On Premlink’s WDM PON EDFA (typical +18 dBm out, NF ≤5.5 dB, input window -10 to +10 dBm) that math has decided more than one of our customers’ projects before they ever placed an order. The rule we hand every engineer: +3 dB on output roughly doubles what you can split or reach. Do the budget first; the datasheet number is only useful after you’ve computed it.
Why I keep telling people to stop looking at output power first
Nine out of ten selection conversations we get start the same way: “what’s the highest output you have?” And nine out of ten times, that’s the wrong first question.
Output power only matters relative to what your route actually eats. A 40 dBm-capable unit won’t save you if your splitter plan and fiber length add up to more loss than your far-end receiver can tolerate. I’ve watched a project team buy a high-power EYDFA and then discover their real problem was a 1×32 splitter running off a 1550 transmitter with a dirty connector eating 3 dB before the signal even hit the node. The amplifier was never the bottleneck. The budget check would have caught it in ten minutes.
So: the EDFA exists to add a block of gain in the middle of the link. Choosing its output tier means calculating how large that block has to be. Not the other way around.
The one equation that runs this whole article
Received power (dBm) = Launch power (dBm) − Total link loss (dB)
Где total link loss = fiber + splitters + connectors + splices + a design margin. The link works if:
Received power ≥ Receiver sensitivity + ≥3 dB margin
For an ONU that’s roughly -27 to -30 dBm. For an optical node it depends on whether it has an integrated receiver—often -2 to +0 dBm. Those two cases change the answer completely, which is why “what’s at the far end” is a required input, not a nice-to-have.
Premlink WDM PON EDFA: the real numbers
These come straight off the product page, not from marketing copy:
| Параметр | Ценить | Почему это важно |
|---|---|---|
| Input window | -10 to +10 dBm | You can feed it from a weaker transmitter without extra stages |
| Typical output | ≥18 dBm | The practical reference point for most designs |
| Max output | ≤40 dBm | A ceiling, not a target |
| Noise figure | ≤5.5 dB | Sets how clean the signal stays, especially cascaded |
| Input wavelength | 1545–1565 nm | The 1550 nm CATV video window |
| GPON passthrough | ≤1 dB | The 1310/1490 data path is nearly lossless |
The three segments, done properly
Segment 1 — what reaches the EDFA input.
Transmitter output, minus the short fiber and connectors between it and the EDFA. Example: a CATV transmitter doing 8 dBm, through two APC connectors (~-0.5 to -1 dB), lands at ≈7–7.5 dBm at the EDFA input. Inside the -10/+10 window, fine.
Segment 2 — what the route eats after the EDFA.
Here’s a table I keep pinned in our office, because it’s the part people round away and regret later:
| Артикул | Typical (1550 nm) |
|---|---|
| Fiber | ~0.35–0.4 dB/km |
| 1×8 splitter | ~10 dB |
| 1×16 splitter | ~13 dB |
| 1×32 splitter | ~16-17 дБ |
| APC connector | ~0.3–0.5 dB each |
| Splicing point | ~0.05–0.1 dB each |
Segment 3 — the far end’s minimum.
ONU at -27 dBm, or an optical node at -2 to +0 dBm. Two different worlds.
A worked example that shows the trap
Say an EDFA at 18 dBm feeds a 1×32 splitter (-17 dB), over 8 km (-3.2 dB), with two splices (-0.2 dB) and two connectors (-0.8 dB):
Total loss ≈ 17 + 3.2 + 0.2 + 0.8 ≈ 21.2 dB
Received ≈ 18 − 21.2 ≈ -3.2 dBm
- Far end is an ONU (-27 dBm): margin ≈ +23.8 dB. Fine. Almost embarrassingly fine.
- Far end is a passive optical node (needs ≥0 dBm): margin ≈ -3.2 dB. Not enough.
Same amplifier, same route, two completely different verdicts. This is the single most common place I see selection mistakes—people size the EDFA without pinning down what’s actually at the far end.
The 3 dB law and how to use it
Two anchors: 0 dBm ≈ 1 mW, и doubling power = +3 dBm.
So moving from 18 to 21 dBm is literally doubling the power. In practice that buys you about double the split scale or a meaningful reach increase. When a design is short by 2–3 dB, try these before reaching for a bigger amplifier:
1. Drop the splitter grade (1×16 saves ~4 dB over 1×32).
2. Check your connectors—APC and clean end-faces. A dirty face costs 3 dB and nobody budgets for it.
3. If truly needed, step the EDFA output up by 3 dB.
Cascading two EDFAs: watch the first stage
Long reach sometimes means two stages. The rule of thumb:
System NF ≈ NF₁ + (NF₂ − 1) / G₁
G₁ is the first stage’s gain. The first stage’s NF and gain dominate the whole chain. Premlink’s WDM PON EDFA at NF ≤5.5 dB is a solid first stage. And don’t let that first stage run on a weak input—keep it near 0 dBm or the real-world NF climbs.
Why we tell some customers to dial it back
More output isn’t automatically better. A unit that can reach 40 dBm doesn’t mean you should run it there. Beyond a certain point you’re just adding SBS risk (stimulated Brillouin scattering) and noise for no benefit.
The Premlink unit has an LCD and SNMP/WEB management, and output can be trimmed in the -10 to +0 range. When a customer’s budget works at, say, 16 dBm, we’d rather have them set 16 and keep the link clean than run 20 because the box can. The right number is “just enough plus 3 dB margin.”
Where the numbers and rules come from
- Fiber and splitter loss figures above are standard operating values used across the industry, consistent with FOA reference tables and typical ITU-T G.651/G.652 application guidance for single-mode 1550 nm links.
- EDFA NF cascade math is the standard Friis-formula for noise in cascaded optical amplifiers, covered in ITU-T G.661 and general EDFA engineering references.
- Premlink product parameters are quoted directly from the WDM PON EDFA product page.
Our position, stated plainly: for most FTTH triple-play builds, a well-budgeted 18 dBm EDFA beats an oversized high-power unit every time. Buy the amplifier your budget says you need, keep 3 dB of margin, and clean your connectors.
Recommended reading
- WDM PON EDFA / EYDFA series — the full family and three amplification modes
- ВДМ ЭДФА и XGSPON EDFA — specific models
- CATV + PON EDFA Combiner — if you’re combining video and data on one fiber
ЧАСТО ЗАДАВАЕМЫЕ ВОПРОСЫ
Q: Can too much input power damage the EDFA?
A: As long as the input stays inside the -10 to +10 dBm window you’re safe. Too high pushes the amplifier into saturation and distortion; too low degrades the noise figure. Confirm your transmitter output minus the short upstream losses lands inside that window.
Q: What’s the difference between 1×32 and 1×64 for the EDFA?
A: Roughly one 3 dB tier (1×32 ≈ -17 dB, 1×64 ≈ -20 dB). If 18 dBm comfortably drove 1×32, plan on about 21 dBm for 1×64 to keep the same margin.
Q: Does the far-end device change the choice?
A: Significantly. An ONU at -27 dBm gives you huge margin. A passive optical node needing ≥0 dBm is far more demanding. Identify the far-end device type before you size anything.
Q: Can this EDFA carry GPON data on the same fiber?
A: Yes. It’s an amplifier plus WDM combiner, merging 1550 nm CATV video with GPON/XGSPON/50G-PON data on one fiber (GPON pass-through loss ≤1 dB). That’s the single-fiber triple-play play.
*Target keywords: edfa amplifier / optical budget. Written for FTTH/HFC network engineers and anyone budgeting an EDFA link.*