BiDi SFP modules: every single-fibre optic from 155M to 100G

A duplex optical module assumes two fibres: one to transmit on, one to receive on. That assumption is free in a data centre and expensive almost everywhere else. In access networks, municipal ducts, rail and utility routes, mast sites and anything fed by a historic lead-in, the number of strands you have is the number somebody installed decades ago, and getting another one means a civil engineering project.

A BiDi module solves that by putting one wavelength in each direction down a single strand, with a simplex LC on the front instead of a duplex pair. It doubles the capacity of fibre that is already in the ground, needs no multiplexer, and costs a fraction of trenching. The trade is that the two ends are no longer identical, and that single fact is behind most of the support calls this product class generates.

Our guide to BiDi transceivers covers the technology, the pairing rule and the reflection failure mode. This article is the companion to it: the complete range, 147 part numbers from 155 Mbit/s to 100 Gigabit, what each speed can actually reach, and the four things to settle before you order.

The rule that catches everybody

BiDi modules are sold in matched upstream and downstream halves, and two of the same half will never link.

A 1310/1550 nm pair means one end transmits at 1310 and listens at 1550 while the other does exactly the reverse. Order two upstream parts and you have two modules shouting on the same wavelength and deaf to each other. There is no configuration that fixes it and no diagnostic that says so clearly — both modules report a healthy transmitter and no received power, which looks like a broken fibre.

You can see the rule in the part numbering throughout the lists below. Every Carritech BiDi design exists as a U and a D: CT-SFP-BX-40U pairs with CT-SFP-BX-40D, and so on at every reach and every speed. OEM codings follow the same convention with their own letters — -D e -U, BXD e BXU, or Juniper’s descriptive 13R15 e 15R13, which spell out transmit-1310-receive-1550 and the reverse. That last scheme is the clearest in the industry and the only one you can read without a data sheet.

The practical advice is simple and we follow it on every order: buy pairs, label both ends, and record which wavelength plan a route uses somewhere other than one engineer’s memory. Tell us the reach and we will supply a matched pair rather than leaving you to work out which half is which.

What each speed can actually reach

This table is drawn from the stocked Carritech designs rather than from a standard, because for most of this range there is no standard to draw from — a point the next section deals with properly.

Two things in that table are worth more than the rest of it.

The slowest parts go the furthest, by a long way. CT-MSFP-BX-180U is a 155 Mbit/s BiDi module specified for 180 kilometres on a 29 dB link budget, drawing under a watt. That is not a typo and it is not exotic engineering: a slow signal needs far less received power per bit to decode cleanly, so the same laser buys enormously more distance. The identical logic gives the 1.25G range 180 km at CT-SFP-BX-180U, while 10G stops at 100 and 25G at 80. If you have a very long single-strand route and a modest bandwidth requirement, the cheap slow optic is the right answer and the fast one will not reach at all. Our guide to 155M and 1.25G SFP modules makes the same argument for duplex parts.

At 100G the wavelength pair is 1304 and 1309 nm — five nanometres apart. Everywhere else in this range the two directions are separated by twenty to two hundred and sixty nanometres, which makes the filtering inside the module straightforward. At 100 Gigabit both directions have to sit inside the narrow window where a single-wavelength PAM4 interface works at all, so CT-QSFP28-BX-40U separates them by five. That is a genuinely harder optical filter, and it is why 100G BiDi is a recent product rather than an old one.

Where the standards are, and where they are not

This matters commercially, because it decides what "compatible" can mean.

IEEE 802.3 defines single-strand bidirectional working only in specific places — principally the 100BASE-BX and 1000BASE-BX port types that came in through the Ethernet-in-the-first-mile work. There is no 10GBASE-BX, no 25GBASE-BX and no 100GBASE-BX. Everything above a gigabit in the lists below is an industry design: a manufacturer has taken a standard optical interface, split the directions across two wavelengths, and specified the result themselves.

Two consequences follow, and both are practical rather than academic.

• Interoperability between two vendors’ BiDi optics is a question of specification overlap, not standards compliance. Two modules both labelled "10G BiDi 40 km" may use different wavelength pairs, and if they do they will not link at all. Compare the wavelengths first and the reach second. This is the opposite of the duplex case, where an LR is an LR.
• The reach figures are the manufacturer’s, derived from a reference link. Work the optical link budget for the actual route, on ITU-T G.652 fibre, with the real splice and connector count in it.

The one exception is at the top of the range. CT-QSFP28-BX-10U is built as 100GBASE-LR1, the single-wavelength 100G interface standardised in IEEE Std 802.3cu-2021, approved on 9 February 2021, and the 40 km design additionally references ITU-T G.959.1 "Optical transport network physical layer interfaces", in-force edition January 2024, along with the industry 100G Lambda MSA. So at 100 Gigabit the underlying optical interface is standardised even though the bidirectional arrangement is not — which is a better position than 10G or 25G BiDi, where neither half of the statement holds.

Four things to settle before you order

1. The pairing is optical, so the two ends need not be the same form factor

This is the most useful thing in this article and it is widely missed. What has to match between the two halves of a BiDi link is the wavelength plan and the reach budget. The package does not. A 10G BiDi module in an XFP cage will link to a 10G BiDi module in an SFP+ cage on the same wavelength pair, because the fibre between them has no idea what shape the modules are.

That is exactly the arrangement you want when one end of a span is an older transport or routing line card and the other is a modern access switch — which, in the networks that buy single-fibre optics, is the normal case rather than the exception. It means you can modernise one end of a single-strand route without touching the other. Our guide to XFP transceivers covers the same argument for duplex optics.

2. A single strand means reflections come back to the receiver

On a duplex link, light reflected from a dirty or badly mated connector travels back towards the transmitter, where nothing much cares. On a single strand, the receiver is on the same fibre as the transmitter, so a reflection arrives at a detector that is straining to hear a signal from eighty kilometres away. The symptom is errors that do not correlate with anything, on a link whose received power reads perfectly normal.

The fix is unglamorous and cheap. Clean and inspect every connector on the route, use angled-physical-contact connectors where the budget is tight, and be more careful about mating quality than you would be on a duplex pair. Fibre optic connector cleaning covers the method, and it is worth doing before you reach for an attenuator or blame the module.

3. A sensitive receiver overloads as easily as it reaches

The long-reach parts in this range buy their distance with an avalanche photodiode — CT-SFP-BX-120U and the 10G, 25G and 100G long parts all use one — while the short-reach designs use a simpler PIN detector. An APD hears far more, and it also has a narrower usable input range.

So a 120 km module on a 15 km span can be saturated by its own far end, and the symptom looks exactly like a dirty connector that does not improve when you clean it. Specify the reach for the route you have, not the longest one in the range, and if you must use a long part on a short span, plan for an attenuator. This cuts both ways in a single-fibre design: because there is no second strand to move to, there is nothing to swap in order to diagnose it.

4. Do not trust the diagnostics to two decimal places

Every module in this range reports digital diagnostics against the SFF management specifications held in the SNIA SFF specifications library, which is genuinely useful: temperature, laser bias, transmit power and received power, live from the port. What those figures are not is calibrated instrumentation. They are an internal estimate with a tolerance, and on a single-fibre link a little of your own transmitter’s light can find its way into the reading.

Use them the way they are good for — watching a trend over weeks and catching a link that is degrading before it drops — rather than as a measurement to design against. Spotting a failing fibre link before it drops covers how to read them.

When BiDi is the wrong answer

Three alternatives are worth checking first, because single-fibre working is not always the cheapest way out of a strand shortage.

• Two services down one strand: use CSFP. A CSFP fits two independent single-fibre links into one SFP cage, doubling a cabinet’s port count with no new hardware. See 1,25G CSFP and the 1,25G CSFP range.
• Many services down one pair: use coloured optics. CWDM on the 20 nm grid in ITU-T G.694.2 gives eighteen channels; DWDM on the frequency grid in ITU-T G.694.1 gives dozens. CWDM vs DWDM covers the choice; 10G DWDM and CWDM SFP+ covers it at ten gigabits.
• Inside a building or a rack: use cable, or copper. Nobody should be buying BiDi optics for a three-metre link. DAC and AOC cables e copper SFP modules are both cheaper, and media converters cover the jobs a pluggable cannot do.

And if the strand count is genuinely not the constraint, duplex optics are cheaper, simpler, easier to stock and not sold in halves. The 1.25G duplex SFP, 10G duplex SFP+ e 100G QSFP28 ranges are the default for a reason.

Six checks before you order

1. Order a matched pair, and say which end is which. Two upstream halves will never link, and nothing in the diagnostics will tell you that is the problem.
2. Confirm the wavelength pair, not just the reach. Two parts both described as "40 km BiDi" on different wavelength plans are not compatible with each other.
3. Match the reach to the route, including the short ones. A 120 km APD receiver can be overloaded by a span that is too short.
4. Budget the real fibre. Splices, connectors and the actual route length, not the straight-line distance.
5. Inspect and clean both connectors before you commission. Reflections matter far more on one strand than on two.
6. Match the coding to the platform at each end independently — see transceiver coding explained. The two ends of a BiDi link are frequently different vendors’ equipment, which means two different identification strings for one link.

The BiDi range Carritech stocks

One hundred and forty-seven tested part numbers across five speeds and three packages, every one coded for the platform it is going into and measured against its specification before despatch. Grouped by speed below, with the Carritech base designs named at each one.

155M BiDi SFP — 20 parts, to 180 km

622-131T155R-I1, 622-155T131R-I1, CT-MSFP-BX-120D, CT-MSFP-BX-120U, CT-MSFP-BX-160D, CT-MSFP-BX-160U, CT-MSFP-BX-180D, CT-MSFP-BX-180U, CT-MSFP-BX-20D, CT-MSFP-BX-20U, CT-MSFP-BX-40D, CT-MSFP-BX-40U, CT-MSFP-BX-80D, CT-MSFP-BX-80U, EX-SFP-FE20KT13R15, EX-SFP-FE20KT15R13, MFB-FA20, MFB-FB20, SFP-100BX1310-20-D, SFP-100BX1550-20-D

Seven Carritech reaches from 20 to 180 km in matched halves, plus Nokia, Juniper, Zyxel and PLANET codings. These are SDH and Fast Ethernet access optics, and the segment is stable because the networks that use it are twenty-year assets. Browse the 155M BiDi SFP range and the 155M transceivers page, and see 100M BiDi SFP.

1.25G BiDi SFP — 31 parts, to 180 km

CT-SFP-BX-120D, CT-SFP-BX-120U, CT-SFP-BX-160D, CT-SFP-BX-160U, CT-SFP-BX-180D, CT-SFP-BX-180U, CT-SFP-BX-20D, CT-SFP-BX-20U, CT-SFP-BX-3D, CT-SFP-BX-3U, CT-SFP-BX-40D, CT-SFP-BX-40U, CT-SFP-BX-80D, CT-SFP-BX-80U, DEM-331R, DEM-331T, E1MG-BXD-TA, E1MG-BXU-TA, EX-SFP-GE40KT13R15, EX-SFP-GE40KT15R13, GLC-BX40-D-I, GLC-BX40-DA-I, GLC-BX40-U-I, GLC-BX80-D-I, GLC-BX80-U-I, MGBIC-BX120-D, MGBIC-BX120-U, MGBIC-BX40-D, MGBIC-BX40-U, SFPGE40KT13R15, SFPGE40KT15R13

The broadest range in the family: eight Carritech reaches from 3 to 180 km, plus Cisco, Juniper, Extreme, Brocade and D-Link codings. The 3 km design uses FP lasers and PIN receivers at 1310/1550 nm; the 180 km design uses DFB transmitters and APD receivers at 1490/1550. Browse the 1.25G BiDi SFP range and the 1.25G transceivers page, and see 1.25G BiDi SFP products.

10G BiDi SFP+ — 50 parts, to 100 km

02310QDT, 02311BJB, 02311JNF, 02314REC, 02314RED, 10G-SFPP-BXU-ER, 10G-SFPP-BXU-S, 10GB-BX10-U, 10GB-BX40-U, 34060544-002, 3FE48235AA, 3FE55743AA, 3HE05037AA, CT-SFP+-BX-100D, CT-SFP+-BX-10D, CT-SFP+-BX-20D, CT-SFP+-BX-3D, CT-SFP+-BX-40D, CT-SFP+-BX-60D, CT-SFP+-BX-80D, CT-SFP+BX-100U, CT-SFP+BX-10U, CT-SFP+BX-20U, CT-SFP+BX-3U, CT-SFP+BX-40U, CT-SFP+BX-60U, CT-SFP+BX-80U, DEM-436XT-BXD, DEM-436XT-BXU, JL737A, JL739A, JNP-SFP-10G-BX10D, JNP-SFP-10G-BX10U, JNP-SFP-10G-BX40D, JNP-SFP-10G-BX40U, R9X55A, S6H21A, S6H24A, SFP-10G-BX40D-I, SFP-10G-BX40U-I, SFP-10G-BX80U-I, SFP-10G-BXD-I, SFP-10G-BXU-I, SFP-10G-BXU1, SFP-10G-ERBD-D, SFP-10G-ERBD-U, SFPP-10GE-BX10D-IT, SFPP-10GE-BX10U-IT, SFPP-10GE-BX40D-IT, SFPP-10GE-BX40U-IT

The largest single group, and the one with the widest OEM spread — Cisco, Juniper, Huawei, Nokia, HPE, Brocade, Extreme, Arista and D-Link all code this optic, which tells you how common single-strand 10G has become in access and metro. Seven Carritech reaches from 3 to 100 km in matched halves. The underlying optical interfaces derive from IEEE Std 802.3ae-2002, approved at the June 2002 Standards Board meeting, though the bidirectional arrangement itself is an industry design. Browse the 10G BiDi SFP+ range and the 10G transceivers page, and see 10G BiDi SFP.

10G BiDi XFP — six upstream designs

CT-XFP-BX-100U, CT-XFP-BX-10U, CT-XFP-BX-20U, CT-XFP-BX-40U, CT-XFP-BX-60U, CT-XFP-BX-80U

Six Carritech XFP upstream designs from 10 to 100 km, for transport and routing line cards whose cages are XFP rather than SFP+. Per the pairing point above, these link to the SFP+ downstream halves listed in the previous group on the same wavelength plan — which is the whole reason to stock them. The following parts are held in the same category and are the SFP+ downstream halves of pairs, listed here because that is how they are bought: 10G-SFPP-BXD-ER, 10G-SFPP-BXD-S, 10GB-BX10-D, 10GB-BX40-D, 3FE48235AB, 3FE55744AA, 3HE05037AB, CT-SFP+-BX-100D, CT-SFP+-BX-10D, CT-SFP+-BX-20D, CT-SFP+-BX-40D, CT-SFP+-BX-60D, CT-SFP+-BX-80D, JL738A, JL740A, R9X54A, S6H20A, S6H22A. Browse the 10G BiDi XFP range, and see XFP transceivers e 10G BiDi XFP explained.

25G BiDi SFP28 — 12 parts, to 80 km

CT-SFP28-BX-20D, CT-SFP28-BX-20U, CT-SFP28-BX-40D, CT-SFP28-BX-40U, CT-SFP28-BX-60D, CT-SFP28-BX-60U, CT-SFP28-BX-80D, CT-SFP28-BX-80U, S0V70A, S0V71A, SFP-25G-BX40D-I, SFP-25G-BX40U-I

Four Carritech reaches from 20 to 80 km in matched halves, plus Cisco and HPE codings. The designs use DML transmitters with an internal clock and data recovery circuit on both the transmit and receive paths, and APD receivers on the longer parts. This range exists largely for mobile fronthaul, where 25G runs over whatever single strand reaches the cell site, and FEC configuration is the usual reason a 25G link stays down — a host setting rather than a module property. See 25G SFP28 transceivers e forward error correction explained, and browse the 25G SFP28 range.

100G BiDi QSFP28 — 16 parts, to 80 km

02315GHV, 02315GHW, 02315GHX, 02315GHY, CT-QSFP28-BX-10D, CT-QSFP28-BX-10U, CT-QSFP28-BX-20D, CT-QSFP28-BX-20U, CT-QSFP28-BX-30D, CT-QSFP28-BX-30U, CT-QSFP28-BX-40D, CT-QSFP28-BX-40U, CT-QSFP28-BX-80D, CT-QSFP28-BX-80U, QSFP-100G-B20D4-I, QSFP-100G-B20U4-I

The newest and most interesting group: 100 Gigabit down a single strand. Five Carritech reaches from 10 to 80 km, plus Cisco and Huawei codings. The 10 km design is a 100GBASE-LR1 PAM4 interface with an EML transmitter and PIN receiver under 4.5 W; the 40 km design uses the 1304/1309 nm pair described above; the 80 km design is a bidirectional ZR4 arrangement supporting 100GBASE-ZR4 and OTU4 transport. Note that "100G BiDi" means two different products in the market — the single-strand parts here, and the dual-wavelength multimode modules that run 40G or 100G over an existing duplex OM3 pair. Our 100G QSFP28 transceivers guide separates them. Browse the 100G BiDi QSFP28 range and the 100G transceivers page.

One note on how this range is filed

Worth saying plainly, because you will notice it browsing the categories. A number of SFP+ downstream halves are held in the 10G BiDi XFP category alongside the XFP upstream designs they pair with. That is a filing decision rather than a specification claim: the pairing is a wavelength property, the two halves are bought together, and keeping them adjacent is how orders actually get placed. But the modules themselves are SFP+ and will not fit an XFP cage. If you are ordering from a category listing rather than from a part number, check the form factor of each half separately. We are tidying it; in the meantime, tell us the cages at both ends and we will send the right two modules.

Keeping single-fibre routes supplied

Carritech Optics holds the BiDi range in UK and EU stock across all five speeds: 155M, 1.25G, 10G SFP+, 10G XFP, 25G SFP28 e 100G QSFP28, with 1,25G CSFP alongside for two services on one strand. Patch cords, copper SFPs and adapters e media converters are there too.

Every module is coded for the target platform and measured against its specification before despatch — see the 22 checks a module passes, which verify the wavelength accuracy and receiver sensitivity a long single-strand link lives or dies by — and carries a Garanzia a vita, with UK and EU stock and support behind it. BiDi is the range where coding at the point of deployment saves the most, because a single-fibre link routinely has a different vendor’s equipment at each end: one optical design, two identification strings, one Opticode coding box. The case is set out in building a smarter transceiver stock strategy, e your optics will go end-of-sale before your switch does is worth reading before you standardise a wavelength plan across an access network.

On the procurement questions, compatible optical transceivers vs OEM, OEM vs compatible: cost, risk and performance compared e do third-party optical transceivers void your warranty? answer them straight, and our accreditations, why choose us e support pages cover how we work. The knowledge centre holds the rest of the technical library.

Not sure which half you need, whether an XFP end will talk to an SFP+ end, or how far a single strand of unknown vintage will actually carry 10 Gigabit? Tell us the two platforms, the wavelength plan and the route and we will say. Our compatibility checker returns the tested Carritech equivalent of an OEM part number in seconds, upload a whole part-number list and we will price it, or request a quote.