Knowledge Center
Technology Insights · Industry Trends · Product Knowledge · Application Notes · News & Updates
A 100G link is not one product. LR4, ER4 and ZR4 modules all run 100G over duplex single-mode fiber, all fit the same QSFP28 cage, and all show the same headline rate in a BOM. Then the rack goes in: the 10 km LR4 pair comes up in minutes, the 30 km span quoted with ER4 needs an attenuator or refuses to train, and the site team asks why "the same 100G optic" behaves differently on every run.
The reach code is the answer. It fixes the channel plan, the laser and receiver class, the dispersion tolerance and the forward error correction the host has to run. Choose it from the fiber route and a 100G rollout stays predictable. Choose it from the price list and the wrong module becomes weeks of replacement logistics.
What Changes Between LR4, ER4 and ZR4
All three are 4x25G NRZ modules in QSFP28 form, delivering 100G over duplex LC single-mode fiber with DDM/DOM diagnostics and QSFP28 MSA compliance. The differences are in the optical layer:
- Channel plan. LR4 and ER4 use the LAN-WDM grid from 1295 to 1309 nm, four lanes spaced 4.5 nm apart. The 80 km ZR4 moves to a 1310 nm design with a semiconductor optical amplifier in the transmit path to raise launch power.
- Dispersion and OSNR margin. Chromatic dispersion and polarization mode dispersion grow with distance, not with rate. The reach rating states how much of that penalty the receiver can absorb, which is why an ER4 and an LR4 with identical lane count are not interchangeable.
- Receiver class. Longer reach means a more sensitive receiver and tighter control of extinction ratio, so the module costs more and draws more power.
- FEC expectation. Short and medium reach links often run with no host FEC or lighter coding. Beyond 40 km the host port must be configured for the FEC mode the module expects, or the link trains while error counters keep climbing.
100G QSFP28 Reach Codes at a Glance
| Reach code | Rated reach | Optical design | Connector | Where it fits |
|---|---|---|---|---|
| LR4 | 10 km | LAN-WDM, 1295 to 1309 nm | LC duplex | Metro access, campus backbone, spine to leaf in one building |
| ER4 | 40 km | LAN-WDM, 1295 to 1309 nm | LC duplex | Metro rings, regional aggregation, utility and rail backbones |
| ZR4 | 80 km | 1310 nm, EML with SOA | LC duplex | Metro to long-haul spans without an amplifier site |
| ZR4+ | 100 km | LAN-WDM, 1295 to 1309 nm | LC duplex | Long regional spans, unamplified DCI routes up to 100 km |
Under 2 km inside a data hall or campus node, the shorter members of the same QSFP28 family are the cheaper tool: SR4 over multimode for 100 m runs, DR1 for 500 m single-mode, and the 2 km CWDM option. Running an 80 km ZR4 across 300 m usually works, but only with fixed attenuation, because the receiver saturates and the link behaves worse than a short-reach part. Beyond 100 km, or when several 100G services share one fiber pair, DWDM line systems and coherent modules on a 50 GHz or 100 GHz grid take over.
Choosing a Reach Code from the Fiber Route
Work from the physical plant. The questions that decide the part number are almost always the same:
- What is the measured route distance, not the straight line between buildings, and how many splices and patch panels sit on it?
- Which fiber type is installed? Standard single-mode and legacy dispersion-shifted fiber carry the same reach code very differently.
- How many connectors and adaptors are in the path? Each one consumes budget that would otherwise be dispersion margin.
- Does the host line card support the FEC mode the longer reach codes need, and is that mode enabled by default?
- What receive power do you expect at the far end when measured, against the sensitivity of the module class?
- Is there an existing fleet to match? Using LR4 at one end and ER4 at the other is a common and expensive mistake.
- How much spares coverage do you want, and can one part serve several routes?
Where the Cost Difference Comes From
The price steps between reach codes follow the bill of materials: a stronger transmitter, a more sensitive receiver, tighter wavelength control and, at 80 km and beyond, transmit-side amplification. Power consumption rises with them, which matters when dozens of ports share one line card and the rack thermal budget was planned around short-reach optics.
Over-specifying carries an operational cost too. An ER4 part will link across 2 km, but with the receive signal far above its sensitivity, and the standard fix is a fixed attenuator at each end, one more component to order, label and maintain per span. If the route is short, order the short-reach part.
What APEX Supplies in the 100G QSFP28 Range
APEX covers the full 100G reach ladder in QSFP28, from 100 m parallel and CWDM options up to 100 km ZR4+. The four parts that answer most reach-planning questions are:
- APX-QSFP28-100G-10km-LR4 for 10 km metro and campus links. See the 100G QSFP28 LR4 10km transceiver.
- APX-QSFP28-100G-40km-ER4 for 40 km metro rings and regional aggregation. See the 100G QSFP28 ER4 40km transceiver.
- APX-QSFP28-100G-80km-ZR4 for 80 km spans on 1310 nm EML and SOA optics. See the 100G QSFP28 ZR4 80km transceiver.
- APX-QSFP28-100G-100km-ZR4+ for unamplified routes up to 100 km. See the 100G QSFP28 ZR4+ 100km transceiver.
Every unit reports DDM/DOM, so case temperature, transmit and receive power, bias current and supply voltage can be read from the host. Orders are coded to the host platform, not only to the MSA standard, and mixed projects combining 10 km, 40 km and 80 km routes ship together, so the spare at each site matches the module deployed there. The full range, including the 2 km and 100 m options, is listed on the 100G QSFP28 transceiver page.
Frequently Asked Questions
Can I use an ER4 module on a 10 km link?
It will usually link, but the receive level sits well above the module's sensitivity, so a fixed attenuator is normally required at both ends. For a permanent 10 km route, LR4 is the correct and cheaper choice.
Do LR4 and ER4 work together on the same span?
No. Both ends must use the same reach code. They share the LAN-WDM grid, but the reach class defines the receiver and dispersion design that link training assumes.
Is host FEC needed for the 80 km and 100 km parts?
The host port must be set to the FEC mode the module expects. With FEC off on a long span, the link may come up while the error counters continue to rise.
How much design margin should I plan for?
Allow a few decibels above the calculated power budget, and account separately for splice loss, connector loss and any aging allowance the site standard requires. Measure one span before ordering the rest of the route.
Next Step
Send the BOM with the route distance, fiber type, host platform and port configuration for each line. We will confirm the reach code, coding and FEC setting each span needs, quote the mix together, and ship a coded sample ahead of volume so the first production rack is not also the first test.
