SFP Module Not Detected? Check Mod_ABS, I2C, 3.3 V and Connector Contacts

When an SFP module is not detected on a board you designed, check four things in order: presence, power, the management bus and connector contact. First, Mod_ABS (pin 6) must go low when the module is fully seated: the module ties it to ground and the host pulls it up, so a line that stays high means the host sees an empty cage. Second, measure 3.3 V on VccT and VccR (pins 16 and 15) at the connector, not at the regulator. Third, SDA and SCL (pins 4 and 5) should idle high, and the module should acknowledge an identifier read. If a known-good module still fails in only one port, inspect that port's connector seating and solder joints. Pin numbers follow SFF-8431. A switch that reads the module but reports it as unsupported has a firmware or vendor-coding issue, not a board fault.

The checks below follow one path, from the module edge through the board-side connector and its solder joints to the host's presence input and I2C bus. VOOHU supplies those SFP connectors and cage assemblies.

#Why is my SFP module not being detected?

"Not detected" covers three different faults, and each one points to a different part of the board:

• The host never sees the module: Mod_ABS does not change state.
• The module is present but cannot be read: vendor, part or serial data is missing, so look at 3.3 V and the I2C bus.
• The module is read but the link stays down: look at Tx_Disable, Rx_LOS, port configuration, the high-speed channel and the fiber.

Do not debug optical power or port speed before the host detects presence. Record presence, module power, management access and link state for every attempt; the first layer that fails tells you where to measure next.

#SFP Mod_ABS: what should the module absent signal do?

Mod_ABS is pin 6 of the 20-pin SFP+ host connector. Section 2.4.4 of SFF-8431 Rev 4.1 (now archived by SNIA) states that the module connects Mod_ABS to its ground (VeeT or VeeR), that the host may pull the contact up to Vcc_Host with a resistor of 4.7 kΩ to 10 kΩ, and that Mod_ABS is high when no module is in the slot. The original SFP specification, INF-8074i, gives the same contact the same function and calls it MOD-DEF0.

Where the low-speed SFP+ signals are pulled up, per SFF-8431 Rev 4.1. Schematic for illustration; follow the specification and your module's datasheet.

An inserted module should therefore pull the line low. When the host still reports an empty slot:

• Measure Mod_ABS at a test point while inserting and removing a known-good module. If the voltage changes but software does not, check the host input's pin function, polarity and driver configuration.
• If the voltage never changes, check that the pull-up is fitted to the intended rail, then check continuity from the connector contact through its solder joint to the host input.
• Check full insertion. SFF-8431 Table 3 mates ground contacts first, power second and signal contacts, including Mod_ABS, third, so make sure the module is fully latched before you measure.
• Pressing or rocking the module is not a fix. If light pressure changes the presence state, inspect seating, contact and solder with power removed.

For every pin and pull-up value on the connector, see our SFP connector pinout.

#Is 3.3 V reaching the module?

An SFP+ module takes 3.3 V on two contacts: VccR (pin 15) for the receiver and VccT (pin 16) for the transmitter. SFF-8431 section 2.8 gives a maximum of 500 mA for each connector contact, and says every module powers up at Power Level I (up to 1.0 W) until a host that supports Power Level II (up to 1.5 W) enables it through the two-wire interface.

A correct voltage at the regulator does not prove a correct voltage at the module: load switches, ferrite beads, contacts and solder joints all add drop. Measure on the module side during insertion, cold boot and warm restart, and fix any power cycling before you debug I2C. On a multiport board, compare one module with a full cage to separate a shared supply limit from a local fault.

#SFP I2C not working: how do you check SDA and SCL?

SDA (pin 4) and SCL (pin 5) form the two-wire management interface; INF-8074i calls them MOD-DEF2 and MOD-DEF1. SFF-8431 section 2.4.5 places their pull-up resistors on the host, tied to the host's two-wire supply, and the specification leaves the management address and register map to SFF-8472. The right resistor value depends on bus capacitance and clock rate; our pinout article lists the SFF-8431 limits.

• Idle levels first. With no traffic, both lines should sit high. A line held low points to a solder bridge, swapped routing, a leaky protection device, contamination or a faulty module.
• Pull-up rail. SFF-8431 does not allow any module contact to see more than the module supply plus 0.5 V, so check which rail the pull-ups go to.
• Clock and acknowledge. Probe an identifier read and look for the module's acknowledge. Slow edges or heavy ringing point to bus capacitance, pull-up value or layout, not to the optical channel.
• Repeat reads. Read the base identity, then the advertised diagnostics, across cold boot, warm boot and re-insertion. One good read does not prove margin.
• Multiplexed buses. On boards that route several cages through an I2C multiplexer, check the multiplexer channel for the failing port before you suspect the connector.

#SFP detected but no link: what do the Tx_Disable pin, Tx_Fault and Rx_LOS tell you?

Tx_Disable (pin 3) is a module input with a 4.7 kΩ to 10 kΩ pull-up to VccT inside the module (SFF-8431 Table 3). When it is high or left open, the module transmitter is off (section 2.4.2). A host that leaves Tx_Disable floating, or drives it high by default, ends up with a module that is detected and readable but never transmits. Drive it low to transmit.

Tx_Fault (pin 2) and Rx_LOS (pin 8) are open-drain outputs pulled up on the host (sections 2.4.1 and 2.4.6). Tx_Fault high reports a transmitter fault. Rx_LOS high means the received optical signal is below the level the relevant standard requires, which typically points to a missing or broken fiber or to the far-end transmitter. These signals explain a recognized module with no traffic; they do not replace Mod_ABS and I2C for detection.

#Can the SFP connector or cage cause intermittent detection?

Yes. A latched module does not guarantee that every contact sits in a stable region. Cage-to-PCB offset, a misaligned bezel or connector-position error can put a module pad at the edge of its contact zone, and the fault may affect only Mod_ABS, SDA or SCL. The result is an absent indication, presence without management access, or a state that changes when the module is touched.

On solder-tail connectors, insufficient solder, opens, coplanarity error or rework damage can leave a low-speed contact conducting over only part of its mechanical or temperature range; debris and flux residue do the same. Use magnified inspection and continuity checks, and correlate any board-strain or temperature response with the failing signal. On press-fit cage assemblies, also inspect the press-fit pin array and the bezel alignment.

On a 1×N or 2×N assembly, a failure tied to one port points to that connector, its joints and its routing; a whole row failing together points to shared 3.3 V, I2C multiplexing, host configuration or assembly alignment. Record port, module, boot condition, insertion count and read result as a matrix, not as a note saying "detection is intermittent".

#Which VOOHU SFP connectors and cages fit these checks?

Signal names on any 20-pin SFP+ host connector follow SFF-8431, so pin 6 is Mod_ABS whichever connector you fit. The data below comes from each part's VOOHU drawing.

Fix the SFP+ or SFP28 type, port count and solder or press-fit process first, then validate footprint, cage, connector, bezel and transceiver as one assembly; do not swap connectors on appearance alone. Compare SFP connectors and cages or SFP28 connectors and cages, and see our SFP connector selection guide.

#How do you make an intermittent detection fault repeatable?

Test the suspect module in a known-good port, a known-good module in the suspect port, and both again in the original port, recording Mod_ABS, module-side 3.3 V, identity read, Rx_LOS and link state each time. A fault that follows the module points to the module or its compatibility; a fault that stays with one port points to the board path.

Some margin failures appear only at cold start, on hot insertion, after repeated plug cycles or at high temperature, so change one condition at a time. In production, a first article should do more than read the module once: repeat reads on every port, check latching, and run multiport assemblies fully populated to expose shared-power, multiplexer or press-fit alignment issues.

#FAQ

QHow do I check if an SFP module is working or not?

Try it in a known-good port on a known-good host, ideally the same model of equipment. If that host reads the module's identity, the module is basically working and the fault is in your board's presence, power, I2C or contact path. On a commercial switch, a message that the module is unsupported is a vendor-coding or firmware policy, not a board fault; check the switch vendor's compatibility list.

QWhat does the SFP module absent (Mod_ABS) signal mean?

Mod_ABS is pin 6 of the SFP+ host connector. The module ties it to ground and the host pulls it up, so high means the slot is empty and low means a module is present (SFF-8431 section 2.4.4). It reports presence only; it says nothing about the 3.3 V supply or the I2C bus.

QDoes missing DDM data mean the SFP connector is bad?

Not on its own. First confirm that the module implements digital diagnostics and that its base identifier reads reliably. Suspect the connector or board when the same module works in another port, or when reads change with touch or temperature.

QOnly one port of a multiport SFP cage fails detection. Where should I start?

Cross-test with the same known-good module. If the failure stays with one port, check that port's Mod_ABS, SDA, SCL, local power and connector joints. If several ports fail together, check the shared 3.3 V rail, I2C multiplexing and host configuration first.

Need a second check on your SFP port design?

Send us your SFP port schematic and PCB footprint through our sample and support form, and tell us which connector or cage you are using. Our FAE team can suggest suitable parts and help check the design, from the footprint and pin 1 position to the Mod_ABS and I2C pull-ups against the connector drawing. Samples are available for evaluation, and selected parts are free of charge. Our FAE team will be glad to confirm the options for your project. VOOHU supplies RJ45 and SFP connectors, LAN and PoE transformers and common mode chokes from Suzhou, China, to more than 1,500 customers.