Where is the coherent transmission going?
It is well known that the greatest demand for bandwidth and at the same time the greatest restrictions related to energy consumption come from data centers. Power consumption, chassis, port density on the front panel are just some of the aspects to consider when implementing coherent optics. Where is the coherent transmission going?
Each new fiber transmission technology goes through a very similar path to achieving standardization. First, when the technology is completely new, it is usually implemented on “line cards”. At this stage, each manufacturer can develop their own unique solutions and we are in vain looking here for any compatibility or interoperability with cards from other manufacturers. Then the same functions are built in the form of a transponder chip, usually 5 × 7 inches according to the MSA. These systems have clearly defined sizes and inputs/outputs, while the internal operations are specific to each manufacturer.
This means that on the line cards we can install modules from different manufacturers, while for correct communication on both sides of the optical track we will have to install line cards with chips from one manufacturer. At the next stage, the fiber optic transmission system is placed in a standardized housing of optical modules, thanks to which we will be able to install it in a working system, without interrupting its operation. This is the stage that 400G coherent transmission is currently entering.
The power budget of QSFP28 LR4 inserts is typically a few decibels. As we combine this with the fact that the transmission takes place in the second optical window, where the insertion losses are 0.35 dB per 1 km of fiber, obtaining greater distances of the inserts begins to be problematic. An additional difficulty is the fact that the standard clearly specifies the value of the maximum power consumption of the insert, which is 4.5 W. So how to cope with the transmission of a single “hundred” over a long distance? Can we somehow amplify this signal?
It is well known that the greatest demand for bandwidth and at the same time the greatest restrictions related to energy consumption come from data centers. This is where the key is how many optical modules we will be able to install on a standard 1RU faceplate. Another requirement for data centers is that the ports in network devices can be as universal as possible, i.e. that optical modules can be installed on the same port to connect to a switch in an adjacent rack, as well as to one that is hundreds of kilometers away. For 10G ports, this was achieved with the advent of SFP+/XFP DWDM inserts, while for 100G ports this role is played by the CFP-DCO and CFP2-DCO coherent modules. Currently, the situation is similar for 400G transmissions: optical modules appear in CFP2 and QSFP-DD format.
CFP2 — is the format in which 400G coherent transmission is certainly the easiest to implement. It is large enough and delivers enough power to allow the current generation of 7nm DSP chips to work. However, this will require special ports in devices supporting 400G, as the original CFP2 standard was designed to support up to 200G.
QSFP-DD — allows you to install many more slots in 1RU, making it the expected form of optical module by data centers. In contrast, current 7nm DSP technology allows 400G coherent optics to run at 15W power, which can be a challenge for some of today's QSFP-DD ports, as until recently this standard assumed power consumption of up to 12W.
Power consumption, chassis, port density on the front panel are just some of the aspects to consider when implementing coherent optics. However, we must not forget about what happens to the signal on the optical side. What we knew from 100G and 200G transmissions, that is, QPSK or 16QAM modulation with a transmission rate of 32Gbaud, would allow us to obtain 400G transmissions over very short distances, since leaving the 32Gbaud speed, we would have to use 64QAM modulation instead of 16QAM. The second option we have is to double the speed to 64Gbaud, which will allow obtaining 400G with 16QAM modulation. Unfortunately, this topic is much more complicated, well, if we can fit 32Gbaud in the 50GHz channel, then 64Gbaud no longer. Moving to the 100GHz network would mean a significant reduction in fiber bandwidth, from 96 channels on the 50GHz network to 48 channels on the 100GHz network. A signal with a speed of 64Gbaud can be transmitted in channels spaced every 75GHz. This means that the use of special filters will allow the transmission in a single fiber of 64 independent 400G channels. But the question arises: is this a good direction? Already today, work is underway on 800G optics and speeds exceeding 90Gbaud, which in turn will require a 100GHz network.
At Salumanus, we focus on the selection of solutions to meet the requirements of the rapidly developing world of telecommunications. Follow with us the directions of development and new technologies of fiber optic transmission.