Confused by the name of 100G transceiver optics and cables?
The names I usually use for communication are vendor-specific dialects. It's a common story that different interpretations are commonplace next door.

The development trend of optical modules is higher speed, smaller packages and lower power consumption. As 100G, 200G, 400G and even 800G high-speed optical transceivers are wided used in data center interconnects and other fiber communication applications. Could 100G optical modules with smaller package sizes be developed to increase panel bandwidth density? It has become a common topic for system manufacturers and optical module manufacturers to discuss. 100G optical modules are mainly four channel 100G QSFP28 optical modules or single lambda 100G QSFP28 optical modules. Similar to the 100G QSFP28 single lambda optical transceivers, the SFP-DD also uses the single channel optical signal, PAM4 technology, to achieve 100G single channel optical signal transmission. The large-scale application of the SFP-DD can achieve interoperability with the current 100G QSFP28 single channel optical module and switch in the future. This forward compatible type is very conducive to network upgrading.

• QSFP28 = Quad Small Form Pluggable
• DSFP= Dual Small Form Factor Pluggable
• SFP-DD = SFP Double Densty (Small Form Factor Double Densty)
• SFP112 = Small Form Factor 112

QSFP28

The QSFP28 is a Quad (4-channel) small form factor hot pluggable fibre optical transceiver used for 100 Gigabit Ethernet (100GbE) data communications applications. The QSFP28 integrates 4 transmit and 4 receiver channels. The number 28 means each lane carries up to 28G data rate. QSFP28 supports 4x25G breakout connection, 2x50G breakout, or 1x100G depending on the type of QSFP28 transceiver used. The 100G QSFP28 Single Lambda optical transceivers convert four 25G electrical signals into a single-wave 100G PAM4 optical signal for data transmission by using a DSP. The optical transmitting and receiving units are reduced from four to one, effectively reducing the internal optical complexity of the optical module.

DSFP

Based on the trend of high throughput and large bandwidth in data center, server NICs and access switches are evolving to next-generation 100G interfaces. The biggest difference between the next-generation 100G interface and the QSFP28 interface at this stage is in the size. There are three interface standards for the 100G interface: DSFP, SFP-DD and SFP112. These standards have in common is the small mechanical size. Having the same size with SFP series transceiver, they can support 48*100G under the 1U switch panel to connect servers. Compared with SFP28 interface, the port density is increased by 4 times, greatly reducing the cost of equipment, operation and maintenance.
DSFP is similar to SFP-DD in that it has two electrical channels, each running up to 56 Gbps using NRZ and up to 112 Gbps using PAM4. Their main difference is in the number of pins: SFP-DD doubles the original SFP pins and has two rows of connecting fingers (20 pins to 40 pins); DSFP mainly follows the SFP pins but adds two pins on the basis of SFP.

SFP-DD VS SFP112 VS SFP-DD 112

• SFP-DD supports up to 100 Gb/s in aggregate over a 2 x 50Gb/s electrical interface.
• SFP112 supports 100 Gb/s over single electrical lane.
• SFP-DD112 supports up to 200 Gb/s in aggregate over a 2 x 100 Gb/s electrical interface.

The cage and connector design provides backwards compatibility to SFP+/SFP28 modules which can be inserted into an SFP-DD/SFP-DD112 cage and connector using 1 of the electrical channels.
SFP112 cage and connectors are compatible with SFP+/SFP28 modules. Furthermore, SFP112 modules maybe inserted into SFP-DD112 cage and connector using 1 of the electrical lanes.

SFP-DD

SFP-DD probably is the Next Generation SFP Module to Support 100G. By using this tiny SFP slot, it could realize the highest number of 100G front-panel pluggable ports in a Top-of-Rack switch. Eventually, two channels of 100G PAM4 will enable 200G per SFP-DD device. SFP28 are a speed enhanced variations of the predecessor SFP+ form factor, which was developed in 2014 under SFF Committee document SFF-8402 as well as earlier SFF-8472 and SFF-8432 specifications, define SFP28. SFP28 is targeted at enabling 25Gb/s to 28Gb/s applications, such as 25Gb/s Ethernet as well as 32GFC Fibre Channel. For the emerging generation of 50Gb/s to 56Gb/s PAM4 generation of SFPs, newer transceiver, connector and cage specifications might be required under a new name, called "SFP56". What's why SFP-DD also called "SFP56-DD".
SFP-DD electrical interfaces will employ 2 lanes that operate up to 25 Gbps NRZ modulation or 56 Gbps PAM4 modulation, providing solutions up to 50 Gbps or 112 Gbps PAM4 aggregate.

SFP112

SFP112 probably is the world's smallest form factor of the 100G pluggable optics, which is similar to the path of 10G with SFP+. It will be future proof, forward compatible, and target high manufacturability for high volume due to simplicity with less components, so ideally supports high-density applications of router switches and also provide an upgrade solution at 100G rate for the next generation wireless fronthaul networks.
The SFP112 will support 112 Gbps and the SFP-DD112 delivers an aggregate rate of 224 Gbps. This new revision also enables compatibility from SFP28/SFP56 to SFP112 and SFP-DD to SFP-DD112.

SFP112 has been renamed to SFP2

SFP-DD112 improved module paddle card 7.3 added option for double or triple split in the pre-wipe signal pads. SFP112 chapters are removed as they are now documented as SFP2 in the SFF-TA-1031.(SFP-DD MSA. October 27, 2023)

This is a specification for a 100Gbps transceiver package that is compatible with SFP+/SFP28, which was previously called SFP112. What is the origin of SFP112? Who decided on the name "SFP112"? The SFP-DD MSA Specifications are clearly defined. This is Nov 05,2011. After that, SFP112 appeared in Cisco's explanatory text and I think it gained recognition.

The case size is the same as SFP+/SFP28. Also applies to SFP56. For management, both SFF-8472 and CMIS can be applied, but CMIS is recommended for SFP56 and SFP112 (SFP2).
Regarding power consumption, it depends on the heat dissipation design, but it is expected to be up to 5.0W. There are also descriptions regarding more than that in the specifications. It is assumed that 48 ports can be arranged on the front panel of a 1U high device with a 19-inch rack width for heat dissipation.
As of July 2023, several SFP112 specification transceivers have been announced, but no Ethernet switches or NICs have been found that use them.