The name is a specification
Every module you will be asked to quote follows the same grammar: [speed]BASE-[type]. The speed is the Ethernet rate. BASE means baseband signalling, as opposed to the broadband modulation of a cable system. The type is the optical part – wavelength, fibre, reach – and it is the only segment where vendors are allowed to invent their own words.
| Segment | What it fixes | Examples | Where mismatches happen |
|---|---|---|---|
1000, 10G, 25G, 40G, 100G | The rate the port negotiates | 1000 = 1 Gbps, 10G = 10 Gbps | A 1G port cannot run a 10G module; a 10G port can usually run a 1G module |
BASE | Baseband signalling | Constant in every module name | Nothing – it is the one part that never changes |
| Type suffix | Wavelength, fibre type, reach | SX, LX, SR, LR, ER, ZR, BX | Most mismatches. Read the table below, not the name |
| Trailing number | Lane or wavelength count | SR4 = four lanes, LR4 = four wavelengths | Ends in the wrong connector: SR4 is MPO, LR4 is duplex LC |
Figure 1 — The name decoded, and the 1G reach ladder from shortest to longest. Grey is copper, green is multimode, blue is single-mode fibre, and amber marks the suffixes that are not IEEE standards.
The 1G family: wavelength and distance
This is the table that resolves most 1G disputes. Two columns matter more than the others: which standards body defines it, and the wavelength.
| Type | Defined by | Fibre | Wavelength | Reach | Connector | Notes |
|---|---|---|---|---|---|---|
| 1000BASE-T | IEEE 802.3ab | Cat5e or better copper | – | 100 m | RJ45 | An electrical port in an SFP-shaped case, not an optic |
| 1000BASE-SX | IEEE 802.3z | Multimode | 850 nm | 220 m on OM1, 550 m on OM2 | Duplex LC | The default for in-room and in-building links |
| 1000BASE-LX | IEEE 802.3z | Single-mode, or multimode | 1310 nm | 10 km single-mode, 550 m multimode | Duplex LC | Over legacy multimode, needs a mode conditioning patch cord |
| 1000BASE-LX10 | IEEE 802.3ah | Single-mode or multimode | 1310 nm | 10 km | Duplex LC | The common 10 km single-mode workhorse |
| 1000BASE-EX | Vendor extension | Single-mode | Quoted as 1310 nm by some, 1550 nm by others | Commonly 30 to 40 km | Duplex LC | Fills the gap between LX and ZX – and proves the naming point |
| 1000BASE-LH | Vendor extension | Single-mode | Usually 1310 nm | Commonly 20 to 40 km | Duplex LC | Some vendors treat LH and EX as the same thing, others do not |
| 1000BASE-ZX | Vendor extension | Single-mode | 1550 nm | 70 to 80 km typical, up to 120 km in EZX variants | Duplex LC | Needs an attenuator on short spans; check the receiver overload point |
| 1000BASE-BX10-D / -U | IEEE 802.3ah | One strand of single-mode | D: 1490 nm out / 1310 nm in. U: the reverse | 10 km | Simplex LC | Must be bought as a complementary pair. Two identical modules link nowhere |
Fibre colour is a useful sanity check before you patch: yellow is single-mode, orange is OM1 or OM2, aqua is OM3 or OM4, lime green is OM5. If the patch cord colour does not match the module’s fibre type, stop and check the part number.
10G and above: the same grammar, different words
Above 1G the suffixes change from SX/LX to SR/LR/ER/ZR, and the trailing number starts carrying information. The standards position repeats itself: the mainstream reaches are IEEE, the extremes usually are not.
| Type | Defined by | Fibre and wavelength | Reach | Connector |
|---|---|---|---|---|
| 10GBASE-SR | IEEE 802.3ae | Multimode 850 nm | 300 m OM3, 400 m OM4 | Duplex LC |
| 10GBASE-LR | IEEE 802.3ae | Single-mode 1310 nm | 10 km | Duplex LC |
| 10GBASE-ER | IEEE 802.3ae | Single-mode 1550 nm | 40 km | Duplex LC |
| 10GBASE-ZR | Vendor extension | Single-mode 1550 nm | 80 km | Duplex LC |
| 25GBASE-SR / LR | IEEE 802.3by | Multimode 850 nm / single-mode 1310 nm | 70 to 100 m on OM4 / 10 km | Duplex LC |
| 40GBASE-SR4 | IEEE 802.3ba | Multimode 850 nm, four lanes | 100 m OM3, 150 m OM4 | MPO |
| 40GBASE-LR4 / ER4 | IEEE 802.3ba | Single-mode, four CWDM wavelengths near 1310 nm | 10 km / 40 km | Duplex LC |
| 100GBASE-SR4 | IEEE 802.3ba | Multimode 850 nm, four lanes of 25G | 100 m OM4 | MPO |
| 100GBASE-LR4 / ER4 | IEEE 802.3ba / 802.3bm | Single-mode, four LAN-WDM wavelengths near 1310 nm | 10 km / 40 km | Duplex LC |
| 100GBASE-CWDM4 | MSA, not IEEE | Single-mode, four CWDM wavelengths, 1271 to 1331 nm | 2 km, with extended variants quoted at 10 km | Duplex LC |
| 100GBASE-PSM4 | MSA, not IEEE | Single-mode 1310 nm, four parallel lanes | 500 m | MPO |
| 100GBASE-DR / FR | 100G Lambda MSA | Single-mode 1310 nm, single wavelength | 500 m / 2 km | Duplex LC |
The distance on the label is a budget, not a promise
Ten kilometres on a datasheet means the module has enough optical budget to tolerate roughly the loss a well-built 10 km link presents. It does not mean the link will reach 10 km, because it says nothing about how much loss your route actually has, how many splices are in it, how many patch panels it crosses, or whether the light at the far end is high enough to sit inside the receiver’s usable window.
Two independent checks decide whether a link works. The first is whether enough light arrives. The second is whether too much does.
Check one: is there enough light?
Three terms, and the arithmetic is simple enough to do on the back of an enquiry.
| Term | Formula | Whose property it is |
|---|---|---|
| Power budget | Tx minimum power − Rx sensitivity | The optics. Use the guaranteed minimum launch, never the typical value |
| Loss budget | Fibre loss + connector loss + splice loss + patch panel and passive loss | The installed plant. Use measured insertion loss where you have it |
| Margin | Power budget − loss budget | The design decision. At least 3 dB indoors, more on aerial, buried or ageing routes |
Planning values that are accurate enough for a first pass: single-mode attenuation around 0.35 dB/km at 1310 nm and 0.25 dB/km at 1550 nm, multimode higher and strongly dependent on grade; 0.2 to 0.5 dB per mated connector pair; 0.05 to 0.1 dB per fusion splice; and a patch panel or adapter contribution at each cross-connection. Then compare the result against the allowance the standard gives your optic – 6.2 dB of channel insertion loss for 10GBASE-LR, for example.
Figure 2 — The two failure directions. Left: enough light, with a worked 10GBASE-LR budget that fits on paper and still leaves only 0.5 dB of margin. Right: too much light, which is what a long-reach optic does on a short route.
Check two: is the light too strong?
This is the check that gets skipped, and it produces the more confusing failure. A long-reach transmitter puts out enough power to survive its rated distance. Put it on a two-metre patch cord and the receiver sees far more than its overload threshold, so the link comes up and immediately starts producing CRC errors, drops and flapping – on a route that should have been trivial.
The symptoms look like a bad module; the cause is the wrong reach class. Fix it by ordering the correct class for the distance rather than by adding attenuation to compensate for an over-specified optic. Where a long-reach model genuinely has to serve a short span, add a fixed inline attenuator of the value the module’s own datasheet specifies – and note that these values are not portable between vendors. One vendor’s ZR family needs roughly 15 dB below 5 km, 10 dB between 5 and 25 km, and 5 dB between 25 and 45 km; a different vendor’s ZR will publish different numbers.
Compatible versus OEM: the price, the coding, and the warranty truth
This is the part of the conversation where most suppliers either overclaim or hedge. Here is the position, separated into what is true about the hardware, what is true about the money, and what you actually give up.
The hardware is defined by the same document
The physical and electrical interface of every SFP, SFP+ and SFP28 is set by the Multi-Source Agreement, and its diagnostic interface by SFF-8472: the same pin-out, the same voltages, the same signal levels, the same 256-byte memory map. A compatible module built to those documents is not an approximation of an OEM part at the electrical level; it is the same interface, implemented by a different manufacturer. What the MSA does not standardise is the firmware conversation between the module and the host, and that is the entire source of the compatibility problem.
Where the price difference actually comes from
The gap is real and it is large – market coverage commonly reports compatible modules at a fraction of OEM list, with the OEM premium quoted anywhere between three and five times the price of equivalent hardware. But the percentage is not the interesting part, because the premium is not paying for the laser. It is paying for four things the OEM funds and a compatible supplier does not: the qualification programme against every platform the module is offered for, a single-source availability commitment, the support and warranty policy described below, and the amortisation of an R&D budget across a product line rather than one part.
Two practical consequences follow. The gap is widest on high-volume commodity types – 1G and 10G, multisource, thousands of units – and narrowest on specialist parts such as CWDM and DWDM channels and the newest speeds, where volume is low. And the decision should not be “compatible or OEM” for the whole network. It should be per link, by how much a failure on that link costs.
The warranty truth, including the part that favours compatible modules
| Question | What actually happens |
|---|---|
| Does inserting a compatible module void the switch warranty? | Not automatically. In the United States, the anti-tying provision of the Magnuson-Moss Warranty Act means a manufacturer cannot void the whole warranty merely because a third-party part was used; it can refuse a specific claim it can demonstrate that part caused. Other jurisdictions differ, so treat this as a US position and check locally. |
| What do you give up with the OEM? | Active support on that port. This is the real cost. Published OEM policy is that support under warranty or a support contract may be withheld where a fault can be traced to a third-party transceiver, and in practice the first request in a support case is to install a genuine module and reproduce the problem. You are self-insuring the diagnostic time, not the module. |
| What about monitoring? | A switch vendor validates and stands behind the telemetry of its own optics. A compatible module still reports temperature, voltage, bias current and Tx/Rx power per SFF-8472, but the switch vendor does not vouch for those values, so proactive alarm thresholds become advisory rather than authoritative. If you depend on DOM alarming for an SLA, that matters. |
| How long is the warranty on the module itself? | Usually longer on the compatible part. OEM optics typically carry a limited warranty measured in months, commonly 90 days to a year. Compatible suppliers typically publish three to five years. This is the part of the comparison buyers usually have backwards. |
| Who pays if a switch firmware upgrade breaks the coding? | Whoever agreed to. A firmware release can change the validation behaviour and invalidate previously working coding. Ask before you buy whether re-coding is free and how fast it turns around – this is a contract term, not a technical one. |
| When is genuine the right answer? | Production uplinks and spine paths where one failure degrades a redundant pair; links under a support contract with a committed response time that you will actually invoke; anywhere you rely on DOM alarming for a customer SLA; and any project with a country-of-origin or trade-agreement requirement, where provenance is the requirement and price is irrelevant. Most networks should run a mixed policy, and saying so is more useful than pretending one answer fits every link. |
Coding is what you are really buying
A host switch reads the module’s EEPROM before it enables the port. The identity block at address A0h holds the vendor name, the organisationally unique identifier, the part number and the compliance codes; the diagnostic block at A2h holds the alarm thresholds and the live telemetry. Some platforms accept any MSA-conformant module, some log a warning, some restrict diagnostic monitoring, and some keep the port down. Because the check is a firmware behaviour, it changes between model generations and between firmware trains – which is why “compatible with Cisco” is not a specification and “coded for Catalyst 9200 running this IOS-XE train” is.
Some platforms expose a command that forces the port to accept an unrecognised module. Two things are worth knowing about it: it does not improve the optic, it only stops the port being disabled, so it will happily mask a marginal laser; and entering it prints an explicit warning that support may be withheld for problems traced to that module. Use it as a diagnostic tool, not as a deployment method.
Four questions to ask a supplier before you order
Compatibility problems are almost always settled before the purchase order, by a buyer who asked the right four questions and read the answers. Each one below includes what a good answer contains and what should make you slow down.
| Question | A good answer contains | Slow down if |
|---|---|---|
| 1. Which exact platforms and firmware versions will you code for, and can you show me the identity string you will write? | The make and model, the firmware train, and the vendor name and part number to be written into the A0h block – offered before you ask for it | The answer is a brand name, or “compatible with all major brands”, or a general assurance that “our modules work everywhere” |
| 2. Will you send a coded sample to test in my own equipment, and what does the received optical power look like at my link distance? | A sample coded for your platform, plus the module’s Tx window, Rx sensitivity and Rx overload, so you can check both ends of the power window yourself | A minimum order quantity is required before any sample, or the only optical figures offered are typically values with no minimum and no overload point |
| 3. What is the module warranty, who pays return freight, and is re-coding free if a firmware upgrade breaks it? | Warranty length in writing, the return process and who pays carriage, and a clear position on re-coding after a platform upgrade | The warranty is described as lifetime but nothing about the process is written down, or re-coding is treated as a new sale |
| 4. What evidence ships with the batch, and at which temperature grade was it measured? | CE, FCC and RoHS certificates per batch, test records, and confirmation that the assembly was tested at the commercial, extended or industrial grade you specified | Certificates are described as available on request in a way that means they do not exist per batch, or the temperature grade is not stated at all |
The third and fourth questions are the ones buyers leave out, and they are the ones that decide the cost of the first failure. A cheap module with an unwritten warranty and no batch record is not cheap at the point where a hundred of them are in the field.
How to order so the risk is gone before the volume order
The sequence that works is short, and it costs one line on the purchase order.
1.Freeze the four parameters. Rate, form factor, reach class and fibre type – from the tables above, not from the previous project’s part number.
2.Calculate both ends of the power window for the worst link in the order, and write the numbers into the enquiry. A supplier who sees your budget arithmetic will quote the right class instead of the nearest stock item.
3.Send the platform and firmware train, and ask for the identity string in writing.
4.Order a coded sample set and run them in your own equipment: read the transceiver inventory from the switch, confirm the module is accepted, confirm the DOM values are visible, and compare them against the thresholds on the datasheet.
4.Then place the volume order, against the same coding and the same temperature grade as the sample. A batch that differs in either is a new part until proven otherwise.
If you are choosing the host at the same time, the two industrial converters that take a pluggable SFP module are wsCom1005-2GST-4G/DC-INT and wsCom3010-2GST-8G/DC-INT, the 10G pair is WS1005-2GESFP-4GE/DC-INT, and on the switch side WS3010-8GSFP-2G/DC-INT carries two SFP ports. Modules are sold separately on every one of them, which is exactly why the coding question belongs in the same enquiry as the host.
Send us the list in that form – rate, form factor, reach, fibre, temperature grade, quantity, and the switch make, model and firmware version – and an engineer will come back with the coding position and a written quote within 24 hours. Our range covers 100M, 1G, 2.5G, 10G, 25G, 40G and 100G in SFP, SFP+, SFP28, QSFP+ and QSFP28 form factors, dual-fibre and single-fibre BiDi, CWDM and DWDM on request, in commercial, extended and industrial temperature grades, every unit tested before it ships. The range and the coding statement are on the transceiver pages, and the cabinets the modules go into are on the industrial media converter and switch pages. If the link needs terminating at both ends, the patch cords, pigtails and adapters are in the cable and accessory ranges. Or send the bill of materials through the contact form; everything is built in our own factory in Shenzhen – Winsure since 2016, the Hansun group since 2004, ISO 9001, ISO 14001 and ISO 45001.
Frequently asked questions
1.What is the difference between 1000BASE-SX and 1000BASE-LX?
Fibre type and wavelength. 1000BASE-SX uses 850 nm over multimode fibre, reaching 220 m on OM1 and 550 m on OM2, and it is the right choice for links inside a room, a rack row or a building. 1000BASE-LX uses 1310 nm and reaches 10 km over single-mode, with 550 m over multimode if you add a mode conditioning patch cord. They are not interchangeable: SX will not work over single-mode, and LX over legacy multimode needs the conditioning cord or modal dispersion will make even a short link unstable.
2.Why does a vendor call a module 1000BASE-EX when EX is not an IEEE standard?
Because the gap between 1000BASE-LX at 10 km and 1000BASE-ZX at 70 to 80 km needed filling, and the standard never filled it. EX, LH, ZX and EZX are industry conventions adopted by module manufacturers, so each vendor defines the wavelength and reach for itself. A module called EX may be quoted at 1310 nm or at 1550 nm, and at 30 km or 40 km, depending on who made it. Treat the suffix as a hint and the datasheet as the specification.
3.Will using compatible SFP modules void my switch warranty?
The switch hardware warranty is not automatically voided. In the United States the anti-tying provision of the Magnuson-Moss Warranty Act prevents a manufacturer from voiding the whole warranty simply because a third-party part was installed; it can only refuse a claim it can prove that part caused. What you do lose is the manufacturer’s active support on that port: published policy allows support to be withheld where a fault traces to a third-party transceiver, and support engineers will typically ask you to fit a genuine module before going further. Check the position for your own jurisdiction, and decide per link rather than per network.
4.Can I use a 10 km module on a fibre run that is only 200 metres long?
Sometimes yes, sometimes it will produce errors you will spend a week chasing. The problem is receiver overload: a long-reach transmitter can deliver more power than the receiver is allowed to accept, so the link comes up and immediately errors. Calculate the received power as transmit power minus the actual path loss, and compare it against the receiver’s maximum input from the datasheet. If it exceeds it, either drop to a shorter-reach class or fit a fixed inline attenuator of the value the module’s own datasheet specifies. Attenuator values are not portable between vendors.
5.How do I know whether a compatible module will be accepted by my switch?
Physically and optically, any MSA-conformant module will fit and pass light. Acceptance is a firmware decision based on the identity fields in the module’s EEPROM, and it varies between platforms and firmware trains. The reliable test is a coded sample in your own equipment: read the transceiver inventory from the switch, confirm the module is recognised without warning, and confirm that DOM values appear and sit inside the datasheet thresholds. Code for the exact model and firmware version, not for the brand.
6.What should I put in an enquiry for optical modules?
Seven things: rate, form factor, reach class, fibre type (and which side of a BiDi pair, if single-fibre), temperature grade, quantity, and the host switch make, model and firmware version. Add the calculated loss budget for your worst link if you have it, because that is what decides whether the reach class you asked for is the one you need. Send it as a list or a spreadsheet and an engineer will review it, flag anything ambiguous, and return a written quote within 24 hours.