Bidirectional OTDR testing means shooting the same fiber from both ends and averaging the two loss readings for every splice and connector. You do it because a single-direction OTDR reading of a splice is not the true splice loss. It is the true loss plus or minus an error caused by the two fibers scattering light differently. Averaging both directions cancels that error.
On single-mode links that contain splices or connectors, IEC 61280-4-2 calls for testing from both ends. If your job is going through formal acceptance, assume it’s required unless the project specification explicitly says otherwise.
Why one direction lies to you
An OTDR doesn’t measure loss directly. It measures backscattered light and infers loss from how much the backscatter level drops across an event.
That works perfectly when the fibers on both sides of a splice are identical. They rarely are. Fiber from different manufacturers, different production runs, or even different sections of the same spool can have slightly different backscatter coefficients and mode field diameters.
When light crosses from a low-backscatter fiber into a higher-backscatter fiber, the OTDR suddenly receives more returned light than it expects. The splice looks better than it is. If the difference is large enough, the trace shows a “gain” at the splice, which is physically impossible for a passive joint. That’s the gainer.
Shoot the same splice from the other end and the effect reverses. The same splice now looks worse than it really is. The FOA calls this the “loser” side, and it’s the one that causes the most trouble in the field: a splicer re-makes a perfectly good splice three times because the one-way trace keeps showing 0.25 dB, when the true loss is a fraction of that.
The actual splice loss is the same in both directions. Only the OTDR’s view of it changes.
How the averaging works
The math is simple. For each event:
True loss = (Loss from A→B + Loss from B→A) ÷ 2
IEC 61280-4-2 defines the individual attenuation exactly this way, as the half sum of the attenuation recorded from each end.
A worked example with illustrative numbers:
| Splice | A→B reading | B→A reading | Bidirectional average |
|---|---|---|---|
| Splice 1 | -0.06 dB (gainer) | 0.18 dB | 0.06 dB |
| Splice 2 | 0.22 dB | -0.02 dB (gainer) | 0.10 dB |
| Splice 3 | 0.04 dB | 0.06 dB | 0.05 dB |
Look at Splice 1. Tested only from end B, it reads 0.18 dB and might fail a 0.15 dB splice spec. Tested only from end A, it shows a gain and tells you nothing. The average, 0.06 dB, is the number that belongs in the report.
Splice 3 is what a splice between well-matched fibers looks like: both directions agree closely and the average barely moves. When both readings are close, the one-way reading was already trustworthy. When they’re far apart, the one-way reading was mostly measurement artifact.
Most current OTDRs and post-processing software will pair the two traces and compute this for you. Check that the software has matched the events correctly by distance before you trust the result. A missed or shifted event pairing produces an average of two different splices.
What the standards actually say
Standards language gets paraphrased loosely online, so here is what can be traced back to the source documents (as summarized by Fluke Networks’ review of the IEC texts):
IEC 61280-4-2 (installed single-mode cabling): OTDR testing is carried out from both ends when the cabling contains splices or connectors. Launch and tail cords stay in place during both measurements. Testing without a tail cord is described as a non-recommended practice, because the far-end connector isn’t measured; the standard accepts it only when qualifying a repair on cabling that was already tested before the damage.
IEC 61280-4-1 (installed multimode cabling): bidirectional averaging is the method for properly measuring the first and last connections. A one-direction test may be enough only if the launch and tail cords have identical backscatter characteristics and you only need total link attenuation, which in practice is rarely the case.
TIA documents: you’ll see conflicting claims online about whether TIA premises standards make bidirectional OTDR testing mandatory. Don’t rely on a blog summary for this. Read the version your contract cites, and check whether it lists OTDR testing as required or supplementary.
In practice, the controlling document is usually the project specification or the network owner’s acceptance procedure. Carriers and utilities commonly write bidirectional testing into their OSP acceptance requirements even when a referenced standard leaves room. Read the spec before you mobilize. Finding out at sign-off that the owner wanted bidirectional traces means sending a crew back to the far end.
When you can reasonably skip it
Not every OTDR shot needs to be bidirectional:
- Fault location and troubleshooting. You’re looking for where the break or the bend is, not certifying a loss value. One direction is fine.
- Repair verification on a link with a known-good baseline, where you’re comparing against an earlier trace. IEC 61280-4-2 explicitly allows this.
- Short links with no splices, where the OTDR is just confirming length and connector condition and loss is certified separately with a light source and power meter.
If the result will go into an acceptance report with pass/fail splice values, test both directions.
The field procedure
The rule that matters most, and the one most often broken: move only the OTDR. Leave the launch and tail cords connected to the link.
- Set up end A. Connect the launch cord to the near end of the fiber under test and the tail (receive) cord to the far end. See our launch cable guide for length selection.
- Shoot A→B. Run the acquisition with your chosen pulse width, wavelength, and averaging time. Save the trace with a clear fiber ID.
- Move the OTDR, not the cords. Disconnect the OTDR from the launch cord, carry it to end B, and connect it to the tail cord. The tail cord now acts as your launch cord.
- Shoot B→A with identical settings. Same wavelength, pulse width, averaging time, and index of refraction. Different settings make the averaging meaningless.
- Pair and average. Load both traces into the OTDR’s bidirectional function or your reporting software, confirm events line up by distance, and generate the averaged event table.
- Repeat per wavelength. If the job requires 1310 nm and 1550 nm, each wavelength gets its own bidirectional pair.
Why the cords must stay in place: the first and last connectors in the link are mated to your test cords. If you swap or reconnect a cord between the two shots, you’re measuring a different connection the second time, and averaging two different connections gives you a number that describes neither.
Three ways to get to the other end
The procedure above assumes one OTDR carried between ends. That’s the cheapest setup and the slowest. There are alternatives, each with a trade-off:
| Method | Equipment | Best for | Trade-off |
|---|---|---|---|
| Conventional | One OTDR, moved between ends | Small jobs, short campus links | Travel time; more chances for setup differences between shots |
| Automated (two units) | Master + remote OTDR communicating over the link or a network | High-count cables, carrier acceptance | Higher equipment cost; both units must be compatible |
| Loopback | One OTDR, a loopback jumper at the far end joining two fibers | Testing fiber pairs when far-end staffing is limited | Still needs one visit to the far end to place the loop; analysis is more involved |
For a contractor testing a 144-fiber cable between two splice cases, the conventional method can mean dozens of trips or a very long wait for one technician to finish all of end A before driving to end B. That’s where a two-unit automated setup usually pays for itself.
What bidirectional testing won’t fix
Averaging cancels backscatter differences. It does not cancel bad technique. If your connectors are dirty, your pulse width is wrong for the link length, or the trace is noisy because you averaged for five seconds, you’ll get two bad traces and an averaged bad result.
It also doubles the opportunities for setup mistakes. The most common ones we see in submitted reports:
- Different index of refraction entered at each end, which shifts event distances so the software pairs the wrong events
- Different pulse widths between the two shots
- A tail cord that was swapped for a different one between measurements
- Fiber IDs that don’t match between the A-end and B-end files, so fiber 12 from one end gets averaged with fiber 13 from the other
A quick sanity check: after averaging, total link length and the number of events should match in both directions. If they don’t, find out why before you sign off.
Preguntas frecuentes
No. A gainer usually indicates a backscatter or mode field diameter difference between the two fibers, not a problem with the splice itself. You can only judge the splice after averaging both directions. A large gainer in one direction will usually show up as a large loss in the other.
No. A light source and power meter give you total link loss, not individual splice loss. Bidirectional averaging only works between two OTDR traces of the same fiber taken with matching settings.
Gainers are less common on multimode, but IEC 61280-4-1 still points to bidirectional averaging for accurate first and last connector measurement. Follow the project specification.
That depends on the contract. Common specifications fall in the 0.1 to 0.3 dB range per fusion splice, judged on the bidirectional average. If a splice reads high in one direction but the average is within spec, it passes.
Related reading:
- Interpreting events on the trace: How to Read an OTDR Trace
- Choosing launch and tail cord length: OTDR Launch Cable Guide