Both ends take an MPO connector, so the hard part is not the connector. It is deciding which combination of fiber count, polarity and gender the link needs, and whether the two cords in that link fight each other.
This page covers configuration. If you are still working out what each field means on its own, the MPO patch cord selection guide covers the individual specifications first.
Patch Cord or Trunk — Two Different Jobs
An MPO to MPO assembly does one of two jobs, and the construction differs accordingly.
| Patch Cord | Trunk | |
|---|---|---|
| Role | Equipment patching inside a cabinet | Permanent backbone between cabinets or zones |
| Typical run | Short, within a rack or between adjacent racks | Long, spanning MDA, HDA and EDA zones |
| Jacket | Single jacket, flexible for tight bends | Heavier construction, built for pulling and long-term strain |
| Fiber count | 8 to 24 | 12 to 144 |
| Handled how | Plugged and re-plugged during moves and changes | Installed once, left in place |
| Pulling eye | Not applicable | Available on End A, End B or both |
What the product name does not tell you
Order against the physical requirement rather than the product name. Three fields settle it:
- Jacket construction — flexible single jacket for patching, heavier build for a permanent run
- Length — measured along the routed path, not cabinet to cabinet
- Pulling eye — required if the assembly gets pulled through conduit or tray
A pulling eye protects the connector during installation. Fibconet supplies it on End A, End B or both ends, and it is specified at order time rather than added later. On a long trunk pulled through a duct, dragging unprotected MPO connectors is how end faces get destroyed before the link is ever tested.
Length tolerance, and why it matters more on trunks
Fibconet builds to a graduated tolerance, always positive:
| Ordered length | Tolerance |
|---|---|
| Under 1 m | +5 / −0 cm |
| 1 to 10 m | +10 / −0 cm |
| 10 to 40 m | +15 / −0 cm |
| Over 40 m | +0.5% of length / −0 |
Every band runs positive-only, so an assembly never arrives short. On a 100 m trunk that means up to 50 cm of extra cable, which coils into the manager. A cord built to a symmetrical tolerance could arrive 50 cm short, and a trunk that will not reach the patch panel is scrap.
Fibconet builds these as LSZH MPO trunk assemblies and as shorter jumpers from the same MPO/MTP patch cord range.
Fiber Count and Optical Grade for a Two-Ended MPO Link
Matching count across the channel
Fiber count follows the transceiver: 8F for 40G and 100G SR4 or 400G SR8 breakouts, 12F for general parallel optics and structured backbones, 16F for 400G QSFP-DD and OSFP, 24F where density matters most.
On a two-ended MPO link the count has to hold across every segment. A 12F trunk feeding a 12F jumper works.
Cable diameter tracks the count. Runs at 8, 12, 16 and 24 fibers use a 3.0 mm jacket; 3.8 mm and 4.5 mm builds are available, along with micro-distribution, oval ribbon and armoured constructions for higher counts or harsher pathways.
Reading insertion loss and return loss against your link
Two loss grades are available on Fibconet MPO assemblies:
| Standard loss | Low loss | |
|---|---|---|
| Single-mode (APC) | ≤ 0.75 dB max, 0.4–0.5 dB typical | ≤ 0.35 dB max, 0.2–0.25 dB typical |
| Multimode (PC) | ≤ 0.6 dB max, 0.35–0.5 dB typical | ≤ 0.35 dB max, 0.2 dB typical |
Return loss is specified by end face rather than as a single figure, because the polish geometry determines the reflection:
| End face | Single-mode | Multimode |
|---|---|---|
| PC | ≥ 50 dB | ≥ 20 dB |
| APC | ≥ 60 dB | ≥ 40 dB |
Insertion loss and return loss are measured per IEC 61300-3-4 (insertion method B), at 1310/1550 nm for single-mode and 850/1300 nm for multimode.
Ask for the test method alongside the number when you compare quotations, since the same assembly measures differently under different methods.
A published maximum only becomes useful once you count mating points and add them up:
Worked example. A structured channel running jumper → trunk → cassette → jumper presents four MPO mating points. At multimode standard grade the worst case stacks to 4 × 0.6 dB, or 2.4 dB of connector loss before fiber attenuation. At low-loss grade the same four points stack to 1.4 dB. Compare each figure against the link budget on your transceiver datasheet.
One mating point on a legacy 40G or 100G link generally runs on standard grade. Two or more mating points, or any 400G and 800G parallel link, calls for low loss.
Assemblies built with imported ferrules measure around 0.15 dB per core on single-mode in typical production testing.
That is a typical value from measurement, not the specification limit — the committed limit stays at 0.35 dB for the low-loss grade.
When a supplier quotes a single impressive number, ask whether it is a maximum or a typical, because only the maximum is contractual.
Polarity Across the Whole Link, Not One Cord
Polarity describes how fiber positions map end to end. On a single cord the rule is simple. On a channel with two or three cords in series, the mappings combine, and that is where links fail.
Quick recap: Type A maps position 1 to 1 straight through, Type B reverses fully so 1 maps to 12, Type C swaps adjacent pairs. Full field-level definitions are in the MPO patch cord selection guide.
Direct link: transceiver to transceiver
One cord between two parallel transceivers needs a single crossover somewhere in the channel, and Type B delivers it in the cord itself. This is the default for new builds.
Structured link: trunk, cassette, jumper
With cassettes in the channel, the crossover moves. Order Type A trunks and let the cassette perform the polarity compensation, which is the arrangement most structured systems are designed around.
Type A demands attention at the duplex ends. Because a Type A trunk maps position 1 to position 1 with no flip, the channel needs an A-B duplex patch cord at one end and an A-A duplex patch cord at the other. Use A-B at both ends and fiber 1 transmits to fiber 1, putting Rx against Rx, and the link fails.
Fibconet supports MPO cassettes built for this compensation.
Why two Type B cords in series cancel out
Each Type B cord flips the mapping once. Two of them in series flip it twice, which returns the channel to straight-through.
That matters when a channel grows. A direct Type B link between two transceivers works. Add a second Type B jumper later to extend the reach through a panel, and the channel now has two flips: position 1 arrives at position 1, transmit meets transmit, and the link that worked yesterday stops working.
Count the flips across the whole channel, not per cord. The channel needs an odd number of crossovers between the two transceivers — one, or three, but not two.
For a new build, keeping every cord Type B and every count odd is the arrangement that survives future changes with the least thinking. Type C exists for legacy duplex systems, and we would not use it on a new parallel link.
Pin Configuration — Get This Wrong and It Won’t Plug In
Male MPO connectors carry two guide pins; female connectors have the matching holes. Every mated pair needs one of each. Two male connectors have nowhere for the pins to seat and will not mate at all.
Fibconet builds all three combinations, ordered per assembly:
| Configuration | Where it fits |
|---|---|
| Female to female | Both ends land on equipment ports or cassette ports, which are usually male |
| Female to male | One end to equipment, one end into a female adapter or panel |
| Male to male | Both ends land in female adapters or cassette ports |
Mapping gender across four connection points
Gender is a property of the mated pair, not of the cord. On a channel with a trunk and two jumpers there are four connection points, and each needs one pin and one hole.
Work backwards from the fixed ends. Transceiver ports and cassette ports have a gender you cannot change, so start there and alternate along the channel. The common arrangement puts female connectors on the equipment side and male on the fixed trunk side, though the cassette in front of you decides the actual answer.
Ordering an mpo male to female patch cord when the link needs female-to-female is not recoverable on site. The pins are set during assembly, so a mis-ordered gender means re-ordering the cord. Check the port at each end before you release the order rather than after the cords arrive.
What Happens If You Get It Wrong
Two different failures get called “a polarity problem”, and they have different owners.
Manufacturing errors happen when fiber sequence or polarity is mis-built during ribbon splicing.
Fibconet tests every core on every assembly for insertion loss, return loss, end face, polarity, fiber sequence, length and appearance, so this class is caught before shipment rather than on your site.
Specification errors happen when the ordered configuration does not match the channel. The factory builds exactly what the order says, every core passes test, and the cord still will not work in that particular link. No amount of factory testing catches this one, because nothing is wrong with the cord.
The table below covers the second class:
| Mis-specified | When it shows up | How hard to diagnose | Fixable on site |
|---|---|---|---|
| Gender | The moment you try to plug it in | Immediate and obvious | No. Pins are set at assembly; re-order |
| Fiber count | At the port, or at power-up | Low. Visible on the connector | No. Re-order |
| Polarity | At power-up. Physically mated, link down | Highest. Everything looks connected | No. Re-order the correct type |
| Loss grade | Link comes up, then runs short on margin | High. Presents as a transceiver or fiber problem | No, but a link budget recount confirms it |
None of these four is fixable in the field, which puts the whole cost of getting it wrong into the RFQ. Confirming four fields before the order costs one email.
Configuration Matrix by Use Case
Find the row that matches your channel:
| Use case | Form | Fiber count | Polarity | Gender | Grade |
|---|---|---|---|---|---|
| Transceiver to transceiver, same rack, 40G/100G SR4 | Jumper | 12F (or 8F) | Type B | Female to female | Standard, single mating point |
| Transceiver to transceiver, 400G/800G parallel | Jumper | 8F or 16F | Type B | Female to female | Low loss |
| Backbone between cabinets, feeding cassettes | Trunk | 12F or 24F | Type A | Male to male (into cassette) | Low loss if ≥2 mating points |
| Cassette to switch, structured channel | Jumper | 12F | Type B | Female to female | Low loss |
| High-density backbone, hyperscale row | Trunk | 24F to 144F | Type A | Per cassette | Low loss |
| Extending an existing Type B direct link | Jumper | Match existing | Count total flips, keep odd | Per ports | Match existing |
| Legacy duplex-based system | Either | 12F | Type C | Per ports | Standard |
Stocking guidance for distributors: new-build data centre demand concentrates on Type B female-to-female jumpers in 12F multimode, since that is what direct transceiver connections take.
Structured-cabling projects pull Type A trunks with the crossover handled in the cassette. Holding both patterns covers most incoming enquiries without stocking every permutation.
Send the matrix row plus length and quantity with your enquiry and we can quote against a fixed configuration.
How These Assemblies Are Built and Tested
Every assembly is tested core by core before shipment: insertion loss, return loss, end face, polarity, fiber sequence, length and appearance. Electronic test reports are available after order placement.
Core-by-core testing exists because of where MPO defects originate. Polarity reversal and sequence errors happen during ribbon splicing, and they produce an assembly that looks correct and measures correct on any single core. Sampling one core in ten does not find them. Fibconet splices with high-precision ribbon fusion equipment and verifies the full mapping on the finished assembly.
End-face geometry is measured by 3D interferometry against these limits:
| Parameter | Limit |
|---|---|
| Radius of curvature, X | ≥ 2000 mm |
| Radius of curvature, Y | ≥ 50 mm |
| Angle, APC | 7.85° to 8.15° |
| Fiber height | 1000 to 3500 nm |
| Adjacent fiber height difference | ±300 nm |
Those numbers decide whether twelve fibers make physical contact simultaneously.
