This is an original planning guide. It is not a vendor manual, an official standard, or legal advice. Example addresses use documentation ranges. Review every command before you paste it.
Copper, fibre, and the thing you are actually buying
A cable design chooses a medium, a path, a label, and a test. Category copper and fibre grades are families, not moods. The connector, the patch, and the installed cable are one channel. A perfect horizontal cable with a crushed patch still fails. Plan the channel, then label both ends, then test, then record the test with the label.
Single mode fibre is the long-term default for backbones because the fibre can outlive several generations of optics. Multimode is a short-reach choice with a grade and a distance that must match the optic. Mixing a multimode patch on a single-mode run does not produce a slower link. It produces no link. The colour and the label exist so you do not do that at 1 a.m.
Copper categories are about the channel you need, not about buying the largest number. A run that is too long for the category does not become compliant because the jacket says a higher grade. Stay inside the length the design assumed, including the patches at both ends. If you cannot, move the closet or use fibre.
The loss calculator on this site adds the numbers you type. It does not see your connectors. Use it to reject a drawing that cannot work, then use a meter to accept the drawing that might.
Planning budgets for a few common runs
a rack-to-rack patch of 30 metres. multimode is unnecessary. A copper patch or a short single-mode jumper both work if the optic matches. Do not buy a long-haul optic for a patch. A planning budget, not a standard quotation, might use 0.4 dB per kilometre of fibre, 0.3 dB per connector, and 0.1 dB per splice. That is 0.01 dB of fibre, 0.60 dB of connectors, and 0.00 dB of splices, about 0.61 dB before you add margin. If the optic's budget is 6 dB and this sum is already near it, you do not have margin for a dirty connector. Clean, test, and write the measured loss next to this estimate. The estimate is how you choose the optic. The measurement is how you accept the run. This page is not the TIA or ISO text. It is a way to see whether the drawing is even plausible.
a floor run under 100 metres of 80 metres. copper category cable can still be the horizontal run if the channel, including patches, stays inside the category limit you are designing to. Fibre is for the backbone and for runs you will not want to replace. A planning budget, not a standard quotation, might use 0.4 dB per kilometre of fibre, 0.3 dB per connector, and 0.1 dB per splice. That is 0.03 dB of fibre, 0.60 dB of connectors, and 0.00 dB of splices, about 0.63 dB before you add margin. If the optic's budget is 6 dB and this sum is already near it, you do not have margin for a dirty connector. Clean, test, and write the measured loss next to this estimate. The estimate is how you choose the optic. The measurement is how you accept the run. This page is not the TIA or ISO text. It is a way to see whether the drawing is even plausible.
a building riser of 150 metres. this is past a normal copper channel. Use fibre. Two connectors and one splice is a modest riser. Budget the loss before you pick the optic. A planning budget, not a standard quotation, might use 0.4 dB per kilometre of fibre, 0.3 dB per connector, and 0.1 dB per splice. That is 0.06 dB of fibre, 0.60 dB of connectors, and 0.10 dB of splices, about 0.76 dB before you add margin. If the optic's budget is 6 dB and this sum is already near it, you do not have margin for a dirty connector. Clean, test, and write the measured loss next to this estimate. The estimate is how you choose the optic. The measurement is how you accept the run. This page is not the TIA or ISO text. It is a way to see whether the drawing is even plausible.
a campus hop of 400 metres. single mode is the default. Multimode distances depend on the grade and the optic, and a campus is where a cheap multimode choice becomes a second project. A planning budget, not a standard quotation, might use 0.4 dB per kilometre of fibre, 0.3 dB per connector, and 0.1 dB per splice. That is 0.16 dB of fibre, 0.60 dB of connectors, and 0.20 dB of splices, about 0.96 dB before you add margin. If the optic's budget is 6 dB and this sum is already near it, you do not have margin for a dirty connector. Clean, test, and write the measured loss next to this estimate. The estimate is how you choose the optic. The measurement is how you accept the run. This page is not the TIA or ISO text. It is a way to see whether the drawing is even plausible.
a longer campus or a metro handoff you light yourself of 2000 metres. loss, not just length, decides. Four splices are easy to forget if the drawing shows one line. A planning budget, not a standard quotation, might use 0.4 dB per kilometre of fibre, 0.3 dB per connector, and 0.1 dB per splice. That is 0.80 dB of fibre, 0.60 dB of connectors, and 0.40 dB of splices, about 1.80 dB before you add margin. If the optic's budget is 6 dB and this sum is already near it, you do not have margin for a dirty connector. Clean, test, and write the measured loss next to this estimate. The estimate is how you choose the optic. The measurement is how you accept the run. This page is not the TIA or ISO text. It is a way to see whether the drawing is even plausible.
a long run of 10000 metres. this needs a loss budget and an optic chosen for that budget. A diagram with a 10 kilometre line and no decibels is not a design. A planning budget, not a standard quotation, might use 0.4 dB per kilometre of fibre, 0.3 dB per connector, and 0.1 dB per splice. That is 4.00 dB of fibre, 0.60 dB of connectors, and 0.80 dB of splices, about 5.40 dB before you add margin. If the optic's budget is 6 dB and this sum is already near it, you do not have margin for a dirty connector. Clean, test, and write the measured loss next to this estimate. The estimate is how you choose the optic. The measurement is how you accept the run. This page is not the TIA or ISO text. It is a way to see whether the drawing is even plausible.
Labels and what a test must carry
A label should tell a stranger the far end without a phone call. Building, floor, closet, rack, and port is enough. A label that says uplink is not enough when there are four uplinks. Both ends get the same identifier. If one end is relabelled, the other end is wrong until you fix it. The test report uses that identifier, not a memory of which strand was the blue one.
Test copper as a channel or as a permanent link, and say which. Test fibre with a loss figure in each direction if the method you use requires it, and keep the meter calibration in mind. An OTDR trace is how you find a break or a bad splice. Reading one is its own skill. This pillar does not replace that skill. It tells you to attach the trace to the label of the strand it measured.
Colour codes help if the whole site uses one code. Inventing a code per closet helps nobody. Write the code on one page and put that page in the closet. A strand used for the core should not be the same colour as a strand left dark for a customer, unless the label carries the difference. The label wins over the colour when they disagree, and then you fix the colour.
Optics, pairs, and polarity
An optic has a reach and a fibre type. A short-reach single-mode optic and a long-reach single-mode optic are not interchangeable just because both are blue or both are grey. Read the part number against the loss and the distance. A long-reach optic on a short patch can overload the receiver. An attenuator is a planned part, not a surprise. A short-reach optic on a 10 kilometre run will not become long-reach because you wanted it to.
Duplex fibre has a transmit strand and a receive strand. If both ends transmit on the same strand, the link stays dark. Polarity is the crossing that makes one end's transmit meet the other end's receive. The scheme can be in the cassette, the patch, or the trunk. Pick one scheme for the site and write it down. A closet that crosses and a closet that does not cross will produce a link that works only when someone swaps a patch at random.
Bi-directional optics put both directions on one strand with two wavelengths. They save strands and they fail in pairs of mismatched wavelengths. Label them as such. Do not patch a bi-directional optic into a duplex pair and expect the spare strand to save you. The spare strand is spare only if the optic uses two strands.
Dirty connectors dominate real loss. The estimate on this page does not include a fingerprint. Clean before you test, and clean before you blame the fibre. A test that passes after a clean is a pass. A test that fails after a clean is a real fault. Write which one you saw, because the next person will otherwise replace a good strand. Keep a known-good patch in the tester's kit and use it when a strand fails, so you can tell a bad patch from a bad trunk. Replace the patch first if the known-good patch clears the fault. Replace the trunk only when the known-good patch still fails. That order is cheaper than pulling fibre because a patch was dusty. Note the patch part number in the result when the patch was the fault, so purchasing does not send the same bad batch back to the closet. File that note with the strand label the same day.
Before the ceiling closes
Count the strands you need, then add spares, then add the spares you will waste on mistakes. A 12-strand run that uses 8 on day one has 4 left, which is one optic pair and one spare pair. That is tight. A 24-strand run costs more now and less than a second installation later. The expensive part is the path, not the extra glass.
Keep copper and fibre bend radius in the plan, not as a hope. A kinked patch fails a test you could have avoided. Power separation and fire stopping are part of the path. They are not optional decoration. If you do not know the rule that applies to the building, stop and ask before the cable is in the wall.
Hand the tester a list of labels, not a drawing with no names. When a strand fails, the list tells you which one. When a strand passes, the list is the record you will need when someone asks what the building contains.
- Choose single mode for backbones unless a short run has a reason to be multimode, and write that reason.
- Estimate loss with connectors and splices, then compare it to the optic.
- Add spare strands.
- Label both ends with the same identifier.
- Test against that identifier.
- File the result where the next person can find it without opening the ceiling.
Mistakes that fail a test or a later optic
A multimode patch on a single-mode trunk.
A copper channel that is over length once the patches are included.
A loss budget with no connectors because the drawing was a single line.
Labels that say uplink.
No spare strands.
An OTDR trace with no strand identifier.
Buying a higher copper category to excuse a run that is too long.
Leaving the test result in a technician's inbox.
Cable checks before the ceiling closes
Confirm this before you call the change done: Every backbone run is identified as single mode or multimode, and the patches match. The loss estimate includes connectors and splices and still fits the optic with margin. Copper channels, including patches, fit the length the category requires. Spare strands are in the count. Both ends share one label scheme. The test report uses those labels. Bend radius and fire stopping are in the work order, not discovered after the pull. A failed strand has a named replacement, not a shrug. Documentation addresses are not labelled onto live fibre as if they were circuit IDs. Someone other than the installer can find the test from the label. And a known-good patch was used before a failed strand was condemned. If one of those fails, stop.
Tools for this guide
These tools run in the browser on this site. They do not log into your machines.
Related pillar guides
Guides in this cluster
Rack, power, and room pages are in the server room guide. This guide is for cable and fiber design.