400GBASE Standardization Trend
The 400G Ethernet standard was approved as IEEE802.3bs in December 2017. In 2019, there will be cases of actually introducing the corresponding equipment. I am wondering if the optical fiber to be laid can support 400G. I tried to make a table including the ones proposed by MSA other than IEEE.
|400GBASE-SR16||MMF 32-core||bs||100m||25G x 16C|
|SMF 8-core||bs||500m||50G x 2 (PAM4) X 4C||yellow yellow|
|400GBASE-FR8||SMF 2-core||bs||2km||25G x 2 (PAM4) x 8 (LWDM)||green|
|400GBASE-LR8||SMF 2-core||bs||10km||25G x 2 (PAM4) x 8 (LWDM)||blue|
|400GBASE-eLR8||SMF 2-core||30km||25G x 2 (PAM4) x 8 (LWDM)|
|400GBASE-CWDM8||SMF 2-core||2km||50G x 8 (CWDM)|
|400GBASE-CWDM8-10||SMF 2-core||10km||50G x 8 (CWDM)|
|400GBASE-FR4||SMF 2-core||cu||2km||50G x 2 (PAM4) x 4 (CWDM)||green|
|400GBASE-LR4-6||SMF 2-core||cu||6km||50G x 2 (PAM4) x 4 (CWDM)|
|400GBASE-eFR4 / LR4||SMF 2-core||10km||50G x 2 (PAM4) x 4 (CWDM)||blue|
|400GBASE-LR4-15||SMF 2-core||cu||15km||50G x 2 (PAM4) x 4 (LWDM)|
|400GBASE-SR8||AOC (MMF)||cm||100m||25G x 2 (PAM4) x 8C||beige|
|400GBASE-PSM8||SMF 16-core||2km||25G x 2 (PAM4) x 8C|
|MMF 8-core||cm||100m||25G x 2 (PAM4) x 8C (BiDi)||beige|
|400GBASE-ER4 Lite||SMF 2-core||30km||50G x 2 (PAM4) x 4 (LWDM)|
|400GBASE-ER8||SMF 2-core||cn||40km||25G x 2 (PAM4) x 8 (LWDM)||red|
|400GBASE-SR4||MMF 8-core||bd||30m||50G x 2 (PAM4) x 4C|
|400GBASE-VR4||MMF 8-core||bd||50m||50G x 2 (PAM4) x 4C|
SR16 uses 16x 2-core of multimode fiber. The connector will define a new 16-core 2-row instead of the 12-core 2-row 24 fibers for 100G SR10. However, there are many 32 cores. Even with optical fiber, it seems to be reasonably thick, and the distance of 100m is difficult to use. I think it's better to give up using MMF. Since there are no vendors planning to release SR16 with QSFP-DD specifications at the end of 2018, it is a specification that has virtually disappeared. SR8 is defined instead. AOC also adopts SR8. Introducing 400G SR 4.2 that can be used with 8-core MMF MPO for 40G / 100G SR4.
SR4 is also being considered, which is up to 100m and makes it 30m with OM3.
The SMF's FR8 / LR8 is no longer a specification packed with the features of a large DWDM transmission device, so the cost of the transceiver will be quite expensive. An ER8 that can be used up to 40km may be worth the cost.
On the other hand, in the CWDM grid, the 4-wave FR4 has a great advantage and is attracting attention. In MSA, the CWDM grid is still 10km LR4, but IEEE is likely to be 6km.
Whether it will be a supplement to that is the LR4-15, which uses LAN WDM as the wavelength grid to improve the transmission distance. Since IEEE has a direction to narrow down the number of specifications, I think that LR4-15 will be forgotten as IEEE 802.3cu.
DR4 has an advantage in terms of transceiver cost . It is an 8-core single mode fiber that uses the same MPO-12 connector as 100G PSM4. Breakout connections to 100G DR are also likely to be used.
The PSM4 has a relatively good distance of 2km at 40G and 500m at 100G, but it was difficult to use because of the resistance to consuming multiple cores of SMF. However, it should be noted when considering the reuse at 400G.
Considering 400G, the design of 100G also using PSM4 and wiring with MPO-12 SMF fiber is worth considering.
400GBASE-SR4 / VSR4 (OM4 50m)
IEEE 802.3cm (2020/01 close)
Aim to achieve 400G with fewer cores than 16 pairs of MMFs. Two methods: 400G-SR8 and 400G-SR4.2 of BiDi method.
SR8 exhibited the OSFP product at interop 2018 as a reference, the method of connecting 25G baud PAM5 with 8 lane / 8 pair MMF makes the technical level more difficult than SR16 by the amount of PAM4, but the number of lasers is half and fiber is also half, I think that AOC with integrated cable has an advantage.
SR4.2 BiDi realizes 8 lanes of bidirectional 25G baud PAM4 by wavelength division multiplexing used in cisco's 100G SRBD with 4 pairs of MMF fiber with MPO adopted in 100G SR4 and obtains 400G. It makes a lot of sense to use the existing MPO-12 type MMF. Although it is technically and costly disadvantageous, the 100G SRBD has already been commercialized, and it seems to be highly feasible because it is configured to bundle four of them.
- 400GBASE-SR8 option A: 2x12 MPO, option B: 16 MPO.
- 400GBASE-SR4.2 844/863nm BiDi x4
IEEE802.3cu (2021/02/09 Approved)
|tx||-3.3 dBm||-2.8 dBm|
|rx||-7.3 dBm||-9.1 dBm|
|channel insertion loss||4.0||6.0|
- 400GBASE-LR4-6: 6km CWDM4
- 400GBASE-LR4-15: 15km LWDM4
|tx||-2.0 dBm||-2.4 dBm||-1.1 dBm|
|rx||-5.9 dBm||-6.4 dBm||-7.7 dBm|
|channel insertion loss||4.0||6.3|
- 400GBASE-ER8-30: 30km LWDM8
- 400GBASE-ZR 75GHz 64ch
|channel insertion loss||15.0||18.0|
400G CWDM8 MSA
CWDM8 was announced on September 17, 2017. Since it is a 50G NRZ x 8 CWDM grid, it is used from 1271 to 1411nm in 20nm steps. It is said that a 10km specification will be released soon, but is it possible to go with this wavelength? It seems that the dispersion is large and there is no room.
The 10km version was also announced on December 20, 2017. 50G modulation is also possible with CWDM8, which greatly expands the range of dispersion support and uses wavelengths far from zero dispersion.
It disappeared completely at OFC 2019 in March 2019. It is believed that it has been replaced by the 400G FR4.
100G Lambda MSA
100G x 4 CWDM 400G FR4. A balance between the cost increase of the PAM4 circuit and the cost reduction of reducing the number of lasers to four. Costs are expected to go down in the long run.
|baud||25G + PAM4||25G + PAM4||53.125G||53.125G||53.125G + PAM4|
|Positive dispersion (max)||19.3||96.4||6.7|
|Negative dispersion (min)||-11.9||-59.3||-11.9|
400Gbps 20km with CWDM Grid
SRm.nm = #fiber pairs n = #wavelengths
Personally, I don't think I'll ever use multimode fiber at 400G, but the IEEE study group is considering various combinations.
|Baud||Modulation method||Number of wavelengths||Number of fibers||Number of lasers|
|400G SR4.2 BD||28G||PAM4||BiDi||8||8|
It seems to be demand for a specification (SR4.x) that can use the same MMF MPO-12 8-core as the 100G-SR4. Since SR4.4 only needs to bundle four 100G SWDM4s, it seems that the commonality of parts is high. It seems that PAM4 by VCSEL at 850nm wavelength has also gathered data, so SR4.2 is also possible. Saving the number of lasers will be a trade-off between the cost of the PAM4 circuit.
finisar is presenting research with OM5 105m with SR1.4 configuration. The practical distance including the loss due to the connector is unknown.
400G BiDi MSA
Just expanded 100G SRBD to 4 pairs. Technically a simple extension of 100m on OM4
Difference due to multiplex method
|parallel fiber 850nm 4 lane||SR4|
|parallel fiber 850/950 BiDi 4 lane||SR4.2|
|parallel fiber O band 4 lane||DR4||DR4 +|
|CWDM O band 4 wave||FR4||LR4|
|LAN WDM O band 8 wave||LR8||ER8|
|DWDM (C band single)||ZR|
Color Code (from OSFP Specification)
|Product Type||Example PMD||Color||
Color Pantone Code
|OSFP copper cables||400G-CR8||Black||N / A|
|OSFP AOC Cables||400G-AOC||Gray||422U|
|OSFP 850nm solutions||400G-SR8, SR4||Beige||475U|
|OSFP 1310nm solutions for up to 500m||400G DR4||Yellow||107U|
|OSFP 1310nm solutions for up to 2km||400G FR4, FR8||Green||354C|
|OSFP 1310nm solutions for up to 10km||400G LR8||Blue||300U|
|OSFP 1310nm solutions for up to 40km||400G ER8||Red||1797U|
|OSFP 1550nm solutions for up to 80km||400G ZR8||White||N / A|