Quick Answer: A fiber optic receiver converts incoming light into an electrical logic signal through a photodiode and a transimpedance amplifier. It only works when three things match: transmitter wavelength, fiber core size and optical budget. Sensitivity is quoted as a minimum received power, typically -17 dBm for 1000BASE-SX and -14.4 dBm for 10GBASE-LR. Subtract sensitivity from the transmitter’s minimum output, then subtract connector and splice loss. If the remaining margin is under 3 dB, the link is a field failure waiting for a dirty connector.
Receivers are specified in a narrow window, not at a single threshold. Too little light and the bit error rate climbs; too much and the photodiode saturates, producing the same symptom from the opposite direction. A 1310 nm receiver fed by an 850 nm transmitter sees almost nothing, because the diode’s responsivity falls away outside its band. Every datasheet number assumes the right wavelength, rate and fiber.
What Sits Inside a Receiver
The optical front end is a photodiode, usually InGaAs for 850 to 1625 nm work, feeding a transimpedance amplifier that turns photocurrent into voltage. A limiting amplifier restores logic levels, then a clock and data recovery block retimes the bit stream.
Receiver versus transceiver
A stand-alone receiver appears in video links and test gear, where one direction carries the signal. Data networks use transceivers, because Ethernet needs both directions for link negotiation. A receiver-only module in a switch port gives you a link that never comes up.
Digital diagnostics
Most modules expose received optical power through DDM registers over I2C, which separates a dead laser from a dirty end face in seconds.
Reading the Optical Budget
The budget is transmitter minimum output minus receiver sensitivity: the total loss the link may have and still work. Each mated connector pair costs roughly 0.3 to 0.5 dB, a fusion splice about 0.1 dB, and multimode fiber about 3.5 dB per kilometre at 850 nm against 0.5 dB per kilometre for single mode at 1310 nm. Those per-unit figures are planning rules of thumb, not standardised values.
| Standard | Wavelength | Tx power range | Rx sensitivity | Budget |
|---|
| 100BASE-FX | 1310 nm SMF | -19 to -14 dBm | -31 dBm | 12 dB |
| 1000BASE-SX | 850 nm MMF | -9.5 to -3 dBm | -17 dBm | 7.5 dB |
| 1000BASE-LX | 1310 nm SMF | -9.5 to -3 dBm | -19 dBm | 9.5 dB |
| 10GBASE-SR | 850 nm MMF | -7.3 to -1 dBm | -11.1 dBm | 3.8 dB |
| 10GBASE-LR | 1310 nm SMF | -8.2 to 0.5 dBm | -14.4 dBm | 6.2 dB |
10GBASE-SR is where budgets vanish. A 3.8 dB allowance covers two connector pairs at 0.5 dB each, leaving 2.8 dB for fiber, which at 850 nm is roughly 800 m of OM3 or 400 m of OM2 on a hot day. That is why a 300 m OM3 link works and a 400 m OM2 link does not, though both are sold as 10G capable.
Pairing Rules That Prevent Field Failures
Wavelength and fiber type
An 850 nm source belongs on multimode, a 1310 nm source on single mode. Mixing the two works over short distances, then fails unpredictably as launch conditions change the modal distribution. The single mode vs multimode fiber choice should be made once per campus, not per link.
Connector polish and end face
UPC and APC ferrules are both 2.5 mm, but an APC end face is angled 8 degrees and mated against UPC reflects enough light to break a 10G link. Check the colour code, then the scope. A 5 micron scratch in the core is worth 2 dB.
Rate and coding
A receiver built for a 1.25 Gbps line rate will not decode a 10.3 Gbps stream, whatever the budget says. Gigabit Ethernet uses 8B/10B coding, 10G uses 64B/66B, and the CDR has to lock to the right rate. An SFP fiber module programmed for the wrong speed shows a link light on some switches and nothing on others.
Saturation and Dirty Connectors
Overload is the failure nobody checks. A short link delivering -3 dBm to a receiver rated for a -1 dBm maximum sits near saturation, and errors appear only when the temperature rises. Attenuators of 5 to 10 dB fix short single-mode spans between buildings. Dust is the other silent killer: a contaminated fiber optic patch cords end face scatters light and can damage a transmitter mated while dirty.
Practical Scenarios
- Warehouse link, 250 m: 1000BASE-SX on OM3 is comfortable at 7.5 dB; do not substitute 10GBASE-SR without recalculating at 3.8 dB.
- Two buildings on separate earths: use single mode with 1310 nm optics, never copper, so a potential difference cannot travel the screen.
- Existing 62.5 micron campus fiber: 1000BASE-SX reaches about 220 m; longer spans need 1000BASE-LX with mode conditioning cords.
- Link that works cold and fails at noon: measure received power. A reading within 2 dB of sensitivity points to a connector, not the module.
FAQ
Can I use any receiver with any transmitter?
Only if wavelength, fiber type, line rate and connector match. The receiver does not negotiate; it simply fails to decode. Vendor coding is separate, since some switches refuse a module they did not brand even when the optics are correct.
What received power should I measure?
Aim for the middle of the window. For 1000BASE-SX that means -10 to -5 dBm, well above the -17 dBm floor and below overload. Record the reading at installation so later comparison shows connector degradation.
Why does a short link show errors?
Receiver overload. Two modules in the same rack can deliver -2 dBm into a diode rated for -1 dBm maximum. Add 5 to 10 dB of attenuation at the patch panel and the error counter stops climbing.
