News
Unidirectional transmission with optical transceivers is not a niche trick — it is a deliberate design choice that solves real problems. Whether you are building a physically isolated security link, maximizing bandwidth on a single fiber strand, or setting up a one-way data diode for a critical network segment, the rules are different from standard bidirectional fiber links. Get the pairing wrong, mix up your wavelengths, or ignore the polarity, and the link simply will not come up. Or worse, it comes up silently and drops packets for weeks before you notice.
This guide covers the exact specifications and procedures you need to follow when deploying unidirectional optical transceiver links.
Before you plug anything in, you need to know what unidirectional actually means in practice — because it is not just "one fiber instead of two."
In a standard bidirectional link, two fibers carry traffic — one for transmit, one for receive. In a unidirectional single-fiber setup, both directions share the same strand using different wavelengths. One end transmits at 1310nm and receives at 1550nm. The other end does the opposite — transmits at 1550nm and receives at 1310nm. This is wavelength division multiplexing, and it is what makes single-fiber unidirectional links possible.
The catch is that both ends must be paired correctly. An A-end module (1310nm TX / 1550nm RX) will only talk to a B-end module (1550nm TX / 1310nm RX). Connect A to A or B to B and you get nothing — no link, no light, no errors. Just silence.
Bidirectional links split the fiber bandwidth between two directions. A unidirectional link gives the full bandwidth to one direction. If you are pushing massive data sets — think data center replication, video surveillance aggregation, or backup streams — unidirectional single-fiber transmission lets you use 100% of the fiber capacity for that one flow. No sharing, no contention, no head-of-line blocking.
It also creates a physical one-way barrier. Data can flow from point A to point B but not back. This is why unidirectional transceivers are the backbone of data diodes and air-gapped security architectures.
This is where most installation failures happen. The wavelengths must match, and the ends must be opposite.
Single-fiber unidirectional transceivers always operate on two wavelengths: 1310nm for one direction and 1550nm for the other. The module at each end of the link is pre-configured for one role.
The A end transmits on 1310nm and receives on 1550nm. The B end transmits on 1550nm and receives on 1310nm. You cannot swap these roles by changing a setting — the laser and the receiver are tuned to specific wavelengths at the factory. If you plug an A-end module into a port expecting a B-end, the lasers will be talking past each other and the receivers will see nothing.
Manufacturers label the ends clearly, usually with "A" and "B" printed on the module body or on the pull tab. Some use color coding — a beige boot for the 1310nm side and a black boot for the 1550nm side. If your modules do not have visible labels, check the specification sheet or the port documentation on your switch. Never guess. Guessing means you will pull the module out, reseat it, pull it out again, and waste an hour of your day.
A single-fiber unidirectional module has one optical bore. A dual-fiber bidirectional module has two bores — one for TX, one for RX. They look similar but they are not interchangeable. If you insert a dual-fiber module into a port wired for single-fiber, you will only get one direction of traffic. The link may show as up but it will be half-duplex at best, or completely dead at worst.
The physical installation follows the same ESD and cleanliness rules as any transceiver, but there are a few unidirectional-specific checks you need to add.
Before you push the module into the cage, look at the label and confirm the wavelength matches the port requirement. A 1310nm-only module will not work on a link that expects 1550nm TX. This sounds obvious, but in a hurry it is easy to grab the wrong module from a bin of identical-looking parts.
Also confirm the fiber type matches. Single-mode unidirectional transceivers must use single-mode fiber. Multi-mode fiber will not carry the 1310nm or 1550nm signals over any meaningful distance — the core size is wrong and the modal dispersion will kill your signal.
Even though this is a unidirectional link, the fiber crossing rule still applies. The transmit side of the A end must connect to the receive side of the B end. On a single-fiber link, this is handled by the wavelength separation — but on the patch panel side, you still need to make sure the fiber from the A-end TX port routes to the B-end RX port.
Use a continuity tester or a visual fault locator to verify the path before you power up the link. A crossed patch at the panel will look fine on the LED but will give you zero throughput.
Push the transceiver straight into the cage until the latch clicks. For SFP modules, press the ejector tab down to release the latch before pulling. For QSFP modules, press both side tabs simultaneously. A module that is not fully seated will show an intermittent link — the LED blinks, the port flaps, and your monitoring system sends you alerts at three in the morning.
Unidirectional transmission shines in security-critical deployments where you need to guarantee that data can only flow one way.
A unidirectional optical link creates a true one-way data path. Malware, commands, or exfiltration attempts cannot travel backward through the fiber because there is no receive path on the sending side. This is fundamentally different from a firewall rule — it is a physical constraint. Even if someone compromises the receiving device, they cannot send anything back through the fiber.
This is why government, military, and financial networks use unidirectional transceivers for data diode architectures. The transmission is physically impossible in the reverse direction, not just logically blocked.
Some advanced unidirectional devices use free-space optical coupling instead of a physical fiber connection. The transmit end shoots a collimated laser beam across an air gap to the receive end. The beam can be physically interrupted by placing an object in the path — and the link drops instantly. Remove the object and the link restores. This gives you a physical kill switch that no software can override.
If you are deploying this type of system, make sure the alignment is precise. Even a few millimeters of offset can cause enough signal loss to drop the link. Use the alignment tools that come with the device and verify the optical power with a meter before you go live.
Do not assume the link is good just because the LED is on.
Use an optical power meter to verify both the 1310nm and 1550nm channels. The A end should show strong output on 1310nm and healthy receive power on 1550nm. The B end should be the mirror image. If one wavelength is weak or absent, you have a pairing mismatch or a dirty connector.
On a unidirectional link, the LED behavior is different from bidirectional. A steady green means the wavelength pair is locked and traffic is flowing. An amber or blinking LED usually means the module is detecting light but the wavelength pairing is wrong — the module is receiving on the wrong channel. If you see no light at all, the module is not seated or the fiber is disconnected.
Light up the link and run a sustained traffic test for at least five minutes. Unidirectional links can show a green LED but still drop frames if the fiber is stressed, the bend radius is too tight, or the connector is slightly misaligned. A five-minute traffic test catches these issues before they become production problems.
This is the single most common error. Two A-end modules cannot talk to each other. Two B-end modules cannot talk to each other. You need one A and one B, period. If you have a spool of single-fiber patch cable and both ends are the same type, the cable itself is fine — but the modules on each end must be opposite.
A 1310nm module and a 1550nm module are not the same thing, even if they fit the same cage. The laser wavelength, the receiver sensitivity, and the fiber type all have to match the link design. Mixing wavelengths gives you a link that looks alive but passes zero data.
If you are replacing a bidirectional link with a unidirectional one, your applications need to handle the fact that there is no reverse path. TCP sessions that expect acknowledgments will hang. DNS lookups from the receiving side will fail. Plan your application architecture before you cut the return fiber.
