40G QSFP+ Transceivers Explained: SR4, LR4, ER4 and BiDi

Forty gigabit Ethernet was meant to be a stepping stone. It arrived in 2010, it was overtaken by 100G within five years, and every roadmap published since has treated it as a transitional speed on the way somewhere else. Sixteen years later there are 40G QSFP+ ports carrying production traffic in campus cores, firewall clusters, storage front-ends, spine-leaf fabrics and transport platforms all over Europe, and most of them are not going anywhere.

That is because a 40G port is usually four 10G lanes bonded together, and four 10G lanes is still a perfectly sensible amount of bandwidth for an uplink. The chassis works. The optics work. The fibre is in the ground. What has changed is the supply: the original vendors have moved their attention to 400G and 800G, and 40G optics are getting harder and more expensive to buy from them every year.

This guide covers what the 40G QSFP+ variants actually are, which cabling each one needs, the MPO trap that catches people out, how breakout to 4 × 10G works, why FEC does not come into it, and which modules fit which platform.

What QSFP+ is

QSFP+ is the "quad small form-factor pluggable plus" package: four lanes of 10G in a module roughly 72 mm long and 18 mm wide, managed over I²C, drawing between about 1.5 W and 3.5 W depending on the reach. It is defined mechanically and electrically in the SFF-8436 specification, now maintained through the SFF specifications at SNIA, and the Ethernet PMDs that run over it come from IEEE 802.3ba and its successors.

Two properties matter more than any datasheet line. First, QSFP+ is mechanically and electrically compatible with QSFP28 cages — a 40G module will generally work in a 100G port, which is why 40G optics keep appearing in equipment bought years after 40G stopped being specified. Second, the four lanes can be broken out, so one QSFP+ port can become four independent 10G interfaces. Together those two facts explain most of 40G’s longevity.

The 40G QSFP+ variants, compared

Eight optical variants cover essentially all 40G deployment. The names follow the Ethernet convention, with two widely-used additions that are not IEEE standards at all.

Variant	Comprimento de onda	Fibre and connector	Alcance	Typical power
SR4	850 nm × 4 lanes	OM3/OM4 multimode, MPO-12	100 m (OM3) / 150 m (OM4)	≤ 1.5 W
BiDi (SRBD)	850 / 900 nm	OM3/OM4 multimode, duplex LC	100 m (OM3) / 150 m (OM4)	~2.5 W
PSM4 / PLR4	1310 nm × 4 lanes	Single-mode, MPO ribbon	up to 10 km	≤ 3.5 W
LR4	1310 nm CWDM × 4	Single-mode, duplex LC	10 km	≤ 3.5 W
ER4	1310 nm CWDM × 4	Single-mode, duplex LC	30–40 km	≤ 3.5 W
ZR4	1310 nm LAN-WDM × 4	Single-mode, duplex LC	80 km	2.5 W typ.
ZR4+ / EZR4	1550 nm	Single-mode, duplex LC	100 km	≤ 3.5 W

The two non-standard entries are the interesting ones. BiDi is a vendor-originated design, not an IEEE PMD, and it exists for one reason covered in the next section. PSM4 and PLR4 are parallel single-mode designs that put each 10G lane on its own fibre pair rather than multiplexing four wavelengths onto one pair — cheaper optics, more fibre. Everything else on that list is a standard PMD: SR4 and LR4 came with 802.3ba in 2010, and ER4 was added later, in the 2015 amendment that also tidied up the 100G single-mode interfaces.

The MPO trap, and why BiDi exists

A 40GBASE-SR4 module does not use a duplex LC patch lead. It uses a 12-fibre MPO ribbon, of which it uses eight: four fibres transmitting, four receiving, with the middle four unused. That is a different cabling type, a different polarity convention, a different cleaning regime and a different set of failure modes from everything else in a typical 10G estate.

Which is why a large number of 40G links that should have been SR4 are BiDi instead. A 40G BiDi module carries two 20G channels in each direction over a single pair of multimode fibres using two wavelengths — 850 nm and 900 nm — so it plugs into the duplex LC multimode patch leads already installed for 10G. Same reach, same fibre, no re-cabling, no ribbon, no polarity exercise.

Three things to know before you choose BiDi over SR4:

Both ends must be BiDi, and they must agree. Some implementations are symmetrical and some are explicitly paired as upstream and downstream parts — QSFP-40G-BD-RX e QSFP-40G-SRBD-R are examples of named receive-side variants. Check the part pair, not just the part number.
It is not a standard, so it is not guaranteed to interoperate across vendors. In practice the common 850/900 nm designs do, and a correctly-coded compatible module behaves exactly like the OEM part it replaces, but it is a compatibility question rather than a standards question.
The fibre still has to be good. Two wavelengths over one multimode pair is less tolerant of a marginal link than four fibres each carrying one. Clean end faces matter more here, not less.

If you are going the other way and the MPO is already in, MPO polarity explained covers the three cabling methods and how to tell which one you have before you spend an afternoon on it. And if the fibre type itself is in doubt, single-mode vs multimode fibre is the place to start.

Breakout: one 40G port, four 10G interfaces

Because the four lanes are independent, most platforms can configure a QSFP+ port as 4 × 10G. With an SR4 module and an MPO-to-8×LC fan-out harness, one port becomes four 10GBASE-SR interfaces. With a PSM4 module and a single-mode harness, four 10GBASE-LR interfaces up to the module’s reach.

This is often the single most valuable thing about an elderly 40G line card: it is also a high-density 10G card. If you are short of 10G ports and have 40G ports going spare, breakout is usually cheaper than new hardware. The mechanics, the host-side configuration caveats and the cases where it goes wrong are the same at every speed — our guide to breakout explained works through them, and the 10G modules you will land on are in the 10G SFP+ range.

Two practical notes. The host has to support the split — it is a platform capability, not a module one, and some ports are fixed at 40G. And breakout changes your optical link budget: a fan-out harness adds connector pairs, and on a 100 m OM3 budget those are not free.

Inside the rack, use cable instead

For a short 40G link between adjacent racks, a module at each end plus a patch lead is the expensive answer. A direct attach cable is cheaper, draws almost nothing and adds no optical budget to manage; an active optical cable gets you to tens of metres with the same simplicity. The DAC QSFP+ 40G e 40G QSFP+ AOC ranges cover both, including breakout variants that fan one QSFP+ out to four SFP+ in a single assembly. DAC vs AOC sets out where the line falls.

No FEC, and why that matters for troubleshooting

Here is a genuine advantage of 40G that nobody mentions: 40GBASE-R has no forward error correction over fibre. The 10G-per-lane NRZ signalling that 802.3ba specified is clean enough not to need it, so the error counters on a 40G port mean exactly what an engineer trained on 10G expects them to mean. Errors are errors.

That is not true one generation up. From 25G per lane onwards FEC is in the path, corrected-error counters run continuously on healthy links, and the whole interpretation changes — which is the subject of our guide to forward error correction. If you are migrating a 40G estate to 100G or beyond, budget for the retraining as well as the optics.

What you do get on 40G is the full digital diagnostics set: temperature, supply voltage, laser bias and per-lane transmit and receive optical power. On a parallel module that per-lane breakdown is the most useful troubleshooting data available anywhere — one cold lane is one fibre, one connector or one MPO position, not a dead module. Our guide to spotting a failing fibre link before it drops covers how to read and trend it.

Reach, and the numbers behind ER4

The jump from LR4 to ER4 is a bigger engineering step than the distance suggests, and the public IEEE record shows why. The consensus proposal that became the 40GBASE-ER4 PMD allocated 17 to 18 dB of channel insertion loss at 40 km, against 6.7 dB for 40GBASE-LR4 at 10 km, with a receiver sensitivity of −18.5 dBm OMA. Roughly eleven more decibels of loss to absorb, which is what the avalanche photodiode receiver and the cooled laser array in an ER4 module are paying for.

Two consequences worth planning around. An ER4 module has real optical gain to spend, so it is intolerant of a short link — put one on a 2 km span without an attenuator and you can saturate the receiver. And the published reach is a specification point, not a promise about your fibre: 40 km of aged cable with six splices and four patch panels may not fit inside 18 dB. Work the link budget from the actual route.

Carritech’s own modules are specified with headroom against those figures: CT-QSFP+LR4 carries a 9.5 dB optical link budget where the standard allows 6.7 dB of channel loss at 10 km, and CT-QSFP+ER4 is specified at 18.5 dB for 40 km. Beyond that, CT-QSFP+ZR4 reaches 80 km on LAN-WDM optics with APD receivers, and CT-QSFP+EZR4 reaches 100 km at 1550 nm. For long single-mode spans the comparison in 100G ZR4 vs LR4 vs ER4 applies the same reasoning a speed up.

Power: the quiet argument for keeping 40G

A 40G QSFP+ SR4 module draws about 1.5 W. A 100G QSFP28 SR4 draws about 3.5 W for 2.5 times the bandwidth, so per bit 100G is clearly better — but per port, in a chassis with a fixed thermal budget and a front panel full of cages, 40G is the lighter load. On an elderly line card with marginal airflow that is not a trivial point, and it is the reason some operators deliberately leave 40G in place rather than re-populating with faster optics the card cannot cool. The numbers across the whole range are in optical transceiver power consumption.

40G or 100G? Four honest answers

1. Keep the 40G ports and source compatible modules

Usually right. The link does its job, the card is paid for, and a tested compatible QSFP+ costs a fraction of the OEM part — increasingly it is also available when the OEM part is not. The question is not whether 40G is current but whether the port needs to do more than 40G for the remaining life of the chassis. Compatible optical transceivers vs OEM covers the usual objections, and do third-party optical transceivers void your warranty? answers the one that always comes up.

2. Break the port out to 4 × 10G

If the bottleneck is port count rather than bandwidth, this is free capacity you already own.

3. Put QSFP28 optics in and run the port at 100G

Only if the port supports it — a QSFP+ module fits a QSFP28 cage, but not the reverse. Where the host does support 100G, the 100G QSFP28 range is a straight swap optically for LR4, since 100GBASE-LR4 and 40GBASE-LR4 both run four LAN or CWDM wavelengths over a duplex single-mode pair to 10 km on the same fibre.

4. Replace the hardware

Sometimes correct, rarely cheap. If you are costing it, do it with the optics included and read choosing 400G optics e the 800G and 1.6T migration path first, because the sensible jump from 40G is usually not to 100G.

Whichever route you take, decide it rather than discover it. 40G is now firmly in the territory described in your optics will go end-of-sale before your switch does: hardware with years of service left and a module supply quietly drying up. The Ethernet Alliance roadmap is a useful sanity check on where the industry’s attention is, and the IEEE 802.3 working group pages show what is still being actively specified. Neither mentions 40G.

The 40G QSFP+ modules Carritech stocks

A hundred and twenty tested part numbers across the 40G range, every one coded for the platform it is going into and measured against its specification before despatch. Grouped by OEM coding below.

The Carritech base parts

Eight optical designs sit behind all one hundred and twenty part numbers above. CT-QSFP+SR4 is the 850 nm MPO module to 100 m on OM3. CT-QSFP+SRBD is the BiDi equivalent over duplex LC to 100 m on OM3 and 150 m on OM4. CT-QSFP+PSM4 is the parallel single-mode design to 10 km over MPO ribbon. CT-QSFP+LR4 is the CWDM four-lane module to 10 km over duplex LC. CT-QSFP+ER4 reaches 40 km and CT-QSFP+ER4-L 30 km. CT-QSFP+ZR4 reaches 80 km on LAN-WDM optics, and CT-QSFP+EZR4 100 km at 1550 nm.

Eight designs, one hundred and twenty part numbers. The difference between them is almost entirely the vendor identification string in the EEPROM — which is the stock problem set out in building a smarter transceiver stock strategy and solved with the Carritech Opticode coding box, which lets you hold one design per reach and code at the point of deployment.

Six checks before you order

Confirm the connector, not just the reach. SR4 needs a 12-fibre MPO; BiDi, LR4, ER4 and ZR4 need a duplex LC pair; PSM4 needs single-mode MPO ribbon. Getting this wrong is the single most common 40G ordering error.
Confirm the fibre type and grade. OM3 and OM4 give different reaches for the same module, and a single-mode module on multimode fibre will not work at all.
Check whether the far end is BiDi or SR4. They are both "40G short reach over multimode" and they will not talk to each other.
Work the link budget for long spans. ER4 has 18 dB to spend and ZR4 more; both can be saturated by a link that is too short as easily as defeated by one that is too long.
Match the coding to the platform. Identification is policed per-platform — see transceiver coding explained.
Decide whether breakout gets you more than a module swap. Four 10G interfaces are often worth more than one 40G.

Keeping 40G ports supplied

Carritech Optics holds the 40G QSFP+ and 100G QSFP28 range in UK and EU stock. Browse 40G QSFP+ modules directly, the wider 40G transceiver range including cable assemblies, or the 100G range if that is where the port is heading. Patch cords e adapters are alongside.

Every module is coded for the target platform and measured against its specification before despatch — see the 22 checks a module passes and our transceiver testing process — and carries a garantia vitalícia, with UK and EU stock and support behind it. Our accreditations e why choose us pages cover the rest.

Not sure whether a port will take BiDi, or whether breakout beats a module swap on a given line card? Tell us the platform and the cabling and we will say. Our compatibility checker returns the tested Carritech equivalent of an OEM part number in seconds, upload a whole part-number list and we will price it, or request a quote.

Transceivers by speed

Cables & connections

Acessórios

40G QSFP+ transceivers: the generation nobody is upgrading from