This type works, but as the EEPROM data does not allow to distinguish
the -U and -D sides, the WRS cannot correctly set the delay parameters.
Therefore they are more difficult to use in White Rabbit installations.
Despite the price of these SFPs is rather low please be aware that VN of
both mentioned SFPs is the same. It is (almost) impossible to
distinguish these two SFPs (which has TX 1490nm) and apply correct
calibration values. Even worse, there are kinds of SFPs from this
manufacturer that are impossible to distinguish at all. For example:
For both TX wavelength is the same (1310nm), but RX differs (1490nm vs
1550nm). The content of EEPROM is almost the same. The same VN, VS and
even TX wavelength (in DOM area). SFP's EEPROM contains VN, VS, SN and
TX wavelength (in DOM area), but not RX wavelength. Right now we check
only VN, VS and SN during the SFP matching.
identical 'calib.sfp.part_num' (as of Sept 2017), modules indistinguishable for WR switch
In our testbench, there are 8 WR nodes and 1 WR switch. They are working
well for about two weeks with full UDP bandwidth and WR protocol. We
also tested the temperature coefficient of these SFPs, which is almost
the same with the SFPs produced by AXCEN. From mail on wr-dev mailing
1550/1310nm - single strand
The following SFP transceivers use 1550/1310nm, which is not
compliant to the 1000BASE-BX10 standard. Therefore they should not be
used in White Rabbit installations.
wavelengths 1550/1310nm instead of 1490/1310nm (Achim)
1270/1330nm - single strand
Out of curiosity we made a preliminary test using 10 Gbit/s SFP+ transcievers from fs.com specified at 1330nm and 1270nm (we got their 40km models). Measured in wavelength, this TX-RX spacing is 60nm, or 3-fold closer than the ‘standard’ BX10 SFPs at 1490/1310. However the zero-dispersion wavelength is somewhere around 1300nm, which has an effect on asymmetry.
Our preliminary result using the fs.com SFP+ BiDirs on 1330/1270nm on a 40km test-spool-fiber in the lab is an asymmetry-coefficient (alpha) of ~14 PPM. This is almost 18-fold lower than with BX10 (different fibers seem to cluster around 250 PPM with BX10).
I’d be interested in hearing if other people have tried SFP+ BiDirs and what economical and easily available choices there are for low-asymmetry links.
40km reach 10GBASE-LR/LW full-duplex on SMF single strand. Wavelengths 1270/1330 nm instead of 1490/1310nm (Erik)
1271/1331nm - single strand
The following SFP transceivers use 1271/1331nm DFB laser, which is
not compliant to the 1000BASE-BX10 standard, but with 10GBASE-LR/LW.
Therefore they should not be used in White Rabbit installations.
However, for specific applications where it is important to have a much
closer spacing of the wavelengths, it may be interesting to use these
10km reach 10GBASE-LR/LW full-duplex on SMF single strand. Wavelengths 1271/1331 nm instead of 1490/1310nm (Erik)
1470nm to 1610nm in 20nm steps - dual strand
The following SFP transceivers use DFB lasers with wavelengths from
1470nm to 1610nm in 20nm steps* (selected on purchase). This is not
compliant to the 1000BASE-BX10 standard, but with 1000BASE-ZX or other
speeds. Therefore they should not be used in White Rabbit
installations. However, for specific applications where it is important
to have a specific wavelength to be sent over CWDM networks (coarse
wavelength division multiplexing), it may be interesting to use these
Some types are even specified for a transmission over 120 km.
VSL in Delft, The Netherlands (E.Dierikx)
have used in August 2014 the following SFPs for a link running 120km.
Note that they use special light amplifiers halfway the link. For the
optical path between the WRS and WR Slave 2, these SFPs are used that
can cover longer distances:
The SFPs have two LC connectors; one for Tx and one for Rx. This means
that the first part of the fibre path (both at the master and at the
slave) needs to be a fibre pair. This fibre pair goes to a 2:1 CWDM
multiplexer or demultiplexer. These MUX and DEMUX devices are from
Stable Technology. These devices are ordered for the wavelengths 1470
and 1490 nm.
In the lab experiment, the signal from the MUX on the master side goes
into a 5 km fibre spool and is attenuated by another 20 dB and then
goes into the DEMUX on the slave side (later this fibre spool and
attenuator was replaced with the fibre link that goes to Amsterdam).
Note that it is important to have sufficient attenuation in the test
link. The laser power from these 120 km SFPs is quite strong and may
damage the photo diode on the receiver side!
Achim Vollhardt, Erik van der Bij - 7 September 2020