How to verify a transceiver is genuine and compatible
A vendor logo on the shell proves nothing on its own. Four checks, run in the order that fails fastest, produce a written record of what a module is and whether your platform will accept it: the compatibility code in the EEPROM, the shell print against the vendor datasheet, the DDM readings from a live port, and a supervised link bring-up.
Platform selection is covered in the site’s guides hub.
What does “genuine and compatible” mean for an optical transceiver?
The two words test different things, and a module can pass one while failing the other. Genuine means the module was built by, or under licence from, the vendor whose name it carries: its EEPROM identity fields (vendor name, part number, serial number, date code, the SFF-8472 identification map) match that vendor’s records. Compatible means the module conforms to the relevant MSA footprint and carries the compatibility code the host platform expects, so the switch recognizes it at insertion.
Practically, “compatible” is a property of the pair, not the module: the same physical module is accepted by one platform and rejected by another. Cisco IOS-class switches compare the EEPROM vendor code against their allow-list at insertion and can refuse the port; Huawei VRP reads the same EEPROM fields and reports the vendor name through display transceiver (Huawei command reference).
Both claims are testable, which separates verification from trust: the genuine claim by comparing EEPROM identity fields against the vendor datasheet, the compatible claim by comparing the coded part number against the platform compatibility matrix. Refurbished genuine modules keep their original coding, so the same checks apply, with condition disclosed separately. The catalogue states coding and condition per listing, for example the 10G SFP transceiver product family, so the code can be cross-checked before ordering.
Which compatibility code must the module carry for your switch or router?
The module must carry the part code the host platform’s compatibility matrix names for that port. “A 10G LR module” describes a speed and a reach, and every platform’s matrix lists a specific code for that slot. The code sits in the module’s EEPROM identity fields (SFF-8472), and the platform compares it against its supported list at insertion.
Three ways to find the right code, in order of reliability:
- Read it off the module you are replacing. The part number on the shell and in the EEPROM is the code that already works in that port; disagreement between them is itself a finding.
- Check the host platform’s compatibility matrix. Cisco, Huawei and Juniper publish per-platform transceiver lists; find the port type and speed, then the approved part code.
- Ask the supplier to state the coding in writing. A supplier holding real stock can name the code per unit, including constraints such as a switch that refuses uncoded modules.
The distinction matters most at the wavelength and reach layer. QSFP28 CWDM4 (CWDM4 MSA, 2014) is specified to 2 km over G.652 single-mode fibre using four wavelengths in the 1310 nm region on the coarse grid described by ITU-T G.694.2, while 100GBASE-LR4 (IEEE 802.3bm) reaches 10 km on a much tighter LAN-WDM spacing. Only one matches a given link budget; the reference documents are the CWDM4 MSA technical specification and IEEE 802.3bm respectively.
Typical catalogue examples, each verified against its datasheet:
| Catalogue part | Host family | Optical standard | Nominal reach | Spec source |
|---|---|---|---|---|
| SFP-10G-LR | 10G SFP+ ports | IEEE 802.3ae 10GBASE-LR, 1310 nm | 10 km G.652 SMF | Cisco SFP-10G-LR datasheet |
| SFP-10G-SR | 10G SFP+ ports | IEEE 802.3ae 10GBASE-SR, 850 nm | 300 m OM3 / 400 m OM4 | Cisco SFP-10G-SR datasheet |
| GLC-LH-SMD | 1G SFP ports | IEEE 802.3z 1000BASE-LX/LH, 1310 nm | 10 km SMF | Cisco GLC-LH-SMD datasheet |
| QSFP-100G-CWDM4 | 100G QSFP28 ports | CWDM4 MSA (2014) | 2 km G.652 SMF | CWDM4 MSA technical specification |
| JNP-QSFP-100G-LR4 | Juniper 100G QSFP28 ports | IEEE 802.3bm 100GBASE-LR4 | 10 km G.652 SMF | IEEE 802.3bm |
The same optics family is offered under different platform codes: the JNP-QSFP-100G-LR4 carries Juniper’s coding while its optics follow the same IEEE 802.3bm standard as any 100GBASE-LR4 module. Which variant fits which link is covered in the site’s 100G QSFP28 guide; the compatibility code and the optical standard are separate questions, and a module can answer one correctly and the other wrongly.
Silkscreen and label checks on the module shell
Every printed field is cheap to fake, so the print counts only where two independent records agree with it: the vendor datasheet for that part code, and the EEPROM identity fields read from the module.
The comparison runs field by field, in this order:
- Read the EEPROM first. Insert the module into any platform that reports identity, Cisco’s
show interface transceiver detailor Huawei VRP’sdisplay transceiver interface verbose, and record vendor name, part number, serial number and date code. SFF-8472 stores the date code in YYMMDD format, so “260315” means 15 March 2026. - Compare EEPROM against the shell. Where the shell names one vendor and the EEPROM reports another, treat the unit as recoded or relabeled, and check the condition claim separately from the coding claim.
- Compare both against the datasheet. The part number must exist in the vendor’s datasheets with matching wavelength, reach and DOM support; a part number absent from vendor documentation cannot be verified at all.
- Check batch consistency. Within a lot, serials should be distinct and date codes agree with the declared build, the GLC-LH-SMD catalogue entry shows the pattern: 1310 nm, 10 km on single-mode fibre per the Cisco GLC-LH-SMD datasheet, condition and coding disclosed.
| Label field | Cross-checked against | What a mismatch indicates |
|---|---|---|
| Vendor name | EEPROM vendor-name field (SFF-8472 A0h) | Recoded or relabeled module; coding does not match shell |
| Part number | Host compatibility matrix + vendor datasheet | Wrong code for the platform; unverifiable part |
| Serial number | Uniqueness across the shipment | Duplicated serials indicate a relabeled batch |
| Date code (YYMMDD in EEPROM) | Box print and datasheet revision history | EEPROM date newer than shell print suggests rework |
| Wavelength marking | Datasheet wavelength and fibre type | An “LR 1310 nm” shell over SR 850 nm hardware |
The wavelength row survives every other check: a relabeled SR module looks, codes and plugs like the LR it pretends to be until the wavelength is inspected or it meets single-mode fibre, where it refuses to link or links with heavy errors. Silkscreen verification ends as a three-way agreement test between shell, EEPROM and datasheet.
How do you read DDM/DOM diagnostics to confirm the module is real and healthy?
DDM (digital diagnostic monitoring, also called DOM) works from inside the live port. SFF-8472 defines the diagnostic interface that exposes five real-time parameters, module temperature, supply voltage, transmitter bias current, TX output power and RX input power, together with alarm and warning flags whose thresholds are stored in the module’s EEPROM. A genuine, healthy module reports values inside its datasheet ranges that move plausibly when conditions change.
The reading procedure:
- Insert into the intended platform or a DDM-capable tester and let it thermally stabilise.
- Record the five parameters: Cisco’s
show interface transceiver detail, or Huawei VRP’sdisplay transceiver diagnosis interface, whose output the Huawei documentation lists as TX power, RX power, bias current, voltage and temperature. - Compare each parameter against the datasheet range for that exact part code; a sibling part’s range does not apply.
- Repeat after ten minutes of loaded traffic; values should hold steady without drifting alarm to alarm.
| DDM parameter (SFF-8472) | Healthy pattern | What an abnormal reading suggests |
|---|---|---|
| TX bias current | Stable inside the datasheet nominal range | Rising bias with steady TX power is the laser-ageing signature; near-zero bias with no light output indicates a dead laser or diagnostics not measuring reality |
| TX output power | Inside the datasheet TX range | An “LR” module reporting below its datasheet floor — the Cisco SFP-10G-LR datasheet gives −8.2 to +0.5 dBm — is degraded, mislabeled, or reporting fiction |
| RX input power | Tracks the light actually arriving | A receiver that reports the same power with and without fibre attached is not measuring anything |
| Module temperature | Near host ambient, inside the datasheet operating range (0–70 °C commercial is the SFP-10G-LR rating in the Cisco datasheet) | Reading hot in a cool rack, or pinned at an alarm threshold, points to a failing module or false EEPROM calibration |
| Supply voltage | Stable around the nominal 3.3 V rail | Voltage alarms are per-module EEPROM thresholds (SFF-8472); flapping alarms implicate the host supply or a marginal module |
Two patterns matter most. The first is static diagnostics: values that never change whatever the link does. Real receivers vary with the light they receive, so a fixed, plausible RX power is a number stored in the EEPROM. The second is threshold mismatch: alarms tripping at readings inside the datasheet range, meaning the EEPROM thresholds were written carelessly. Such a module may pass traffic, but its diagnostics cannot be trusted to warn anyone.
DDM also serves the compatibility check: refusing platforms expose the rejection through the same diagnostic path, so the DDM read doubles as evidence for the coding claims above. The same five-parameter pattern applies across form factors, the catalogue’s 25G SFP28 transceiver family and 100G QSFP28 family included, read through each platform’s CLI.
In-service compatibility testing, step by step
The first three checks can pass and the link still misbehaves, so the final step is a supervised bring-up in the intended host. “Trains a link for a minute” and “carries traffic cleanly for hours at rack temperatures” are different results, and only the second counts as acceptance.
The sequence, ordered so the fastest-failing check comes first:
- Insertion and recognition. Insert into the target port on the target platform. On enforcing Cisco platforms a refused module produces the unsupported-transceiver syslog and can err-disable the port; on Huawei VRP,
display transceivershould report the expected vendor and part fields. Any warning counts as a finding. - Link bring-up. Confirm the port trains at the expected speed over the planned fibre, and check the reported wavelength and fibre type: LR belongs on single-mode, SR on the OM3/OM4 grade the datasheet names, the Cisco SFP-10G-SR datasheet specifies up to 300 m on OM3.
- DDM baseline. Capture the five diagnostic parameters before loading traffic, so later degradation has a reference point.
- Loaded soak. Push real traffic for 30–60 minutes (an iperf run, or the platform’s loopback test) and watch error counters, CRC at lower speeds, FEC corrected/uncorrected on 25G and 100G. Incrementing counters point at dirty connectors, wrong reach class or a failing laser before the link drops.
- Optical budget cross-check. Compare measured RX power against estimated link loss: a 10 km-rated pair reading low on a 9 km span usually means unaccounted loss (patch points, aging fibre), and the module itself may be fine (the 10G SFP+ buying guide makes the same point about budget versus nominal reach).
- Temperature soak. Re-read DDM at the end of the loaded period. A module that starts near its temperature alarm in a normal rack will not survive a warm one.
Two cautions apply. Coding enforcement differs per platform and release, so a module accepted by a bench switch can be rejected by another of the same family. Sample several modules from the shipment; the first one unpacked says nothing about the rest. For 100G links, the CWDM4-versus-LR4 and breakout questions are covered in the site’s 100G QSFP28 guide.
Which documents and packaging details support a genuine claim?
Documents let a buyer repeat the verification independently, months later and without the seller present, so request them before payment. The set that matters:
- Spec sheet for the exact part code, the vendor or MSA datasheet naming wavelength, reach, TX/RX ranges and DOM support. Every other check compares against it.
- Per-unit DDM printout, the five parameters at room temperature, serials matching the shipped units. A supplier that tests each unit can produce this from stock records; one that cannot, will not.
- Coding statement, which platform family the module is coded for, in writing, converting the compatibility-code section from an assumption into a deliverable.
- Condition disclosure, new, refurbished or used, and what “tested” covered: DDM capture, error counters, link bring-up, or something less.
Packaging is secondary but free to check: modules should arrive in ESD-shielded bags whose labels match the shell part number and serial. Duplicated serials across bags are the classic sign of a relabeled batch, and the EEPROM date code (YYMMDD per SFF-8472) should not be newer than the box print.
No single document proves genuineness. What a claim rests on is the set: matching serials, a datasheet that lists the part, and packaging labels that agree with the shell. Where a stock list already states coding and condition per unit, as the 100G QSFP28 product page does, the buyer’s job is to confirm the delivered units match the listing.
Why do counterfeit transceivers behave differently from genuine ones?
Because the costs a counterfeit avoids sit in exactly the components whose health the diagnostic interface reports. The laser is the largest single cost, so a counterfeit typically substitutes a lower-grade or salvaged one; the substitution shows weeks later as rising TX bias current or falling TX output power. The receiver diode and calibration constants are the second saving: poor-sensitivity receivers pass a short bench link but fail at the datasheet reach, and careless calibration makes DDM alarms unreliable.
Counterfeits come in three broad types, each betrayed by a different check from this guide:
- Relabeled used-genuine modules. The EEPROM is original and internally consistent, so silkscreen and EEPROM agree; the betrayal is batch-level serial duplication and date codes inconsistent with the declared build. What is misdescribed here is the condition claim, while the hardware itself is genuine.
- Recoded third-party modules. Sound hardware carrying a borrowed vendor code: shell vendor and EEPROM vendor disagree, or the code passes on some platforms and not others.
- Fully fabricated modules. Invented part numbers, static DDM values, checksum errors. On Cisco platforms an invalid EEPROM checksum produces
GBIC_SECURITY_CRYPT-4-VN_DATA_CRC_ERROR-class syslog messages; elsewhere the module reports numbers that never move, which the static-diagnostics check catches.
The practical conclusion: counterfeits rarely fail visibly at hour zero. They fail in DDM trends, error counters, batch serials and platform security logs. The four checks run from documentary to physical for that reason, and a shipment cleared on one spot check leaves those signals unread.
Frequently asked questions
Can a third-party (compatible) transceiver damage my switch?
An MSA-compliant module follows the same mechanical and electrical footprint as the vendor original, so on enforcing platforms the documented risk is rejection instead of damage, the behaviour Cisco’s service unsupported-transceiver command exists to relax. What causes trouble is a non-compliant module with bad calibration or out-of-spec optics, which is why the DDM checks here matter more than the brand debate.
What if the EEPROM date code is newer than the shell print?
SFF-8472 stores the date code in the identification fields of address A0h as YYMMDD. A few days of build-to-ship lag are ordinary; a gap of months points to rework, a replaced shell or a relabeled module. Ask for the rework record before power-on, and check that the supplier’s test report carries a date covering the gap, because a report dated earlier than the EEPROM code cannot have been run on that unit.
A module reads clean on the bench but alarms after an hour in the rack. What does that point to?
Heat, most often. Bias current climbing while TX power holds is the laser-ageing pattern, and it appears only once the module reaches rack temperature. Read the five DDM parameters at insertion and again at the end of a loaded hour, then compare the two captures. If RX power falls at both ends at the same rate, check the patch cords and connector ferrules first; if it falls at one end only, the receiving module or its calibration is the better suspect.
I have CWDM4 modules in stock and the new span is 9 km. Can they carry it?
No. CWDM4 is an MSA (2014) specified to 2 km over G.652 single-mode fibre, and the wavelength plan does not change reach: the 20 nm grid of ITU-T G.694.2 does not buy distance. A 9 km span belongs to 100GBASE-LR4 (IEEE 802.3bm, 10 km), so keep the CWDM4 stock for intra-building and campus runs, and check the link budget before ordering the LR4 pair.
The supplier’s test report shows only PASS with no measured values. Is that enough?
Not for acceptance. A report that can arbitrate a dispute later carries the unit serial, the test items and the measured values: per-lane optical power, alarm-log status, throughput numbers. A bare PASS gives you nothing to compare against when the same unit fails in service. Ask for the measured sheet before payment; a supplier who tested the unit can export it from the bench records.
Does YT CIG supply transceivers coded for Huawei, Cisco and Juniper platforms?
The catalogue carries all three coding families: the SFP-10G-LR listing under Huawei-compatible coding, the GLC-LH-SMD listing (a Cisco part code per the Cisco GLC-LH-SMD datasheet), and the JNP-QSFP-100G-LR4 under Juniper coding, each linked in the compatibility-code table above. Condition is stated per listing, and the same four-document request set applies to any supplier.
Need a specific part?
Send the part number and quantity — we confirm condition, availability and price before you commit.