A circuit turn-up is a repeatable sequence, not improvisation. Do it the same way every time and the failure modes become obvious. This page is the runbook: four ordered phases — pre-check, physical turn-up, service turn-up, acceptance. Every line is a checkbox with the reason it matters and the value you should see or record beside it; work top to bottom. Your progress is saved locally so you can turn up over two days and pick up where you left off.
Mental model: A turn-up is a build-then-verify pipeline. Pre-check is design validation on paper. Physical turn-up gets light on the glass. Service turn-up gets a customer signal across that light. Acceptance is the "definition of done" — the record that proves it works and gives future-you a baseline. Never skip a phase because the previous one "looked fine."
Common mistake: Patching first and thinking later. The number one cause of a turn-up that "won't come up" is a design or path detail that was never confirmed — wrong wavelength assigned, single-fiber design patched as a pair, loss budget never actually calculated. The pre-check phase is where you catch these on paper, for free, before a truck roll.
Product note: The Ekinops/Celestis provisioning screens, menu paths, and field names referenced here are conceptual — verify exact steps and labels against current Ekinops / Celestis documentation. The sequence and the checks below are platform-independent; the specific clicks are not.
Phase 1 — Pre-Check (validate the design on paper)
Nothing physical happens here. You are confirming that the circuit can work before anyone touches a patch panel. Each item lists why it matters and what to confirm/record.
Field checklist — the two that bite hardest: APC vs UPC mismatch (an APC connector mated to a UPC ferrule looks seated but wrecks return loss and can damage endfaces), and single-fiber-vs-pair (a single-fiber bidirectional design uses different wavelengths each direction on one strand — patch it as a pair and you get nothing). Confirm both before the truck rolls.
Phase 2 — Physical Turn-Up (get light on the glass)
Now you are on site with a fiber scope and a power meter. Order matters: clean first, patch client then line, prove orientation, then read power against the budget you calculated in Phase 1.
Read the diagram as: your local Tx must land on the far end's Rx. A "lines up locally but dark at the far end" symptom is almost always a swapped Tx/Rx somewhere in the patch path.
CORRECT CLIENT-SIDE PATCH CORRECT LINE-SIDE PATCH
(router <--> Ekinops client port) (Ekinops line port <--> mux/fiber)
router Tx ───────────► client Rx line Tx ───────────► mux/span (add)
router Rx ◄─────────── client Tx line Rx ◄─────────── mux/span (drop)
(a "cross": Tx meets Rx) (a "cross": Tx meets Rx)
SWAP SYMPTOM: Tx wired to Tx / Rx to Rx → local link may seem to try,
far end sees NO LIGHT / LOS. Fix = swap the strand, re-scope, re-read Rx.
Every mated pair is a "cross" — a Tx on one side always meets an Rx on the other. If you ever wire Tx-to-Tx, one direction goes dark: the give-away is a clean local transmit reading with LOS at the far end. Swap the strand, clean/inspect again, and re-read Rx.
Phase 3 — Service Turn-Up (get a customer signal across the light)
Light is on the glass and clean. Now build the service in the NMS and prove real traffic quality end to end.
Key idea: "Link is up" is not "service is good." A span can show a healthy link while quietly running high pre-FEC error rates that eat your entire correction margin. Always confirm signal quality (FEC / BER / errored seconds), not just signal presence, before you call a turn-up complete.
Phase 4 — Acceptance (record the definition of done)
The circuit works. Now capture the evidence and the baseline. This record is what a future outage investigation, an audit, or the next engineer will lean on. Incomplete acceptance data is a slow tax you pay every time something breaks later.
Acceptance record template
Fill one of these per circuit and store it with the service. This is the baseline that turns a future "seems slow" into a measured "we lost 5 dB." Populate every cell — a blank is a question someone asks during the next outage.
Field
Value
Field
Value
Service ID
__________
Line rate
__________
A-site (rack/shelf/port)
__________
Wavelength / channel
__________
Z-site (rack/shelf/port)
__________
Local Tx / Rx (dBm)
______ / ______
Module / slot / port (A)
__________
Remote Tx / Rx (dBm)
______ / ______
Module / slot / port (Z)
__________
Pre-FEC BER (+ margin)
__________
Client speed / optic
__________
Span loss vs budget (dB)
______ / ______
Protection scheme
__________
OTDR / OLTS result
__________
Date turned up
__________
Technician
__________
Common mistake: Turning up under time pressure and promising to "capture the baseline later." Later never comes, and six months on you have a degrading span with nothing to compare against. The baseline is worth the most on the exact day the circuit is healthy — capture it before you leave the site.
Top turn-up mistakes (learn them so you never make them):
Skipping the cleaning/scope step. One dirty endface = high loss and rising pre-FEC, and it can burn under power. Inspect–clean–re-inspect every mate, every time.
No baseline saved. "Link up, moving on" leaves future-you with nothing to compare against. Record all four powers + pre-FEC before you leave.
APC ↔ UPC mismatch. Seats fine, wrecks return loss, can damage ferrules. Match polish at every junction.
Wrong channel / wavelength. Collides with a neighbour on the same glass and takes down two services. Confirm the channel plan before you light the line.
Unverified far end. Reading only your local side hides a Tx/Rx swap or a one-direction fault. Always confirm remote Rx within budget before calling it done.