A coherent optical transceiver has a transmitter (TX) laser and a local oscillator (LO) laser, which can be based on two separate lasers or a single laser. The laser specification requirements are different for different transmission distances. Here we discuss a few key laser requirements and where they will be used.
1. Low power consumption and small size
As can be seen in Fig.1, a tunable laser with low power consumption and small size is always required for pluggable coherent transceivers. Line-cards, on the other hand, can have a better tolerance toward these two requirements. Note that the new small form factor pluggables such as DD-QSFP and OSFP have a very tight space and naturally special care needs to be taken to ensure these two laser requirements are met.
2. High optical power
This is especially important for a coherent transceiver with (a) a higher modulator insertion loss (e.g., silicon photonics-based), and/or (b) a higher “modulation loss” due to a higher order of modulation. The latter is owing to the decreased average signal power when the modulation order goes higher. For example, the modulation loss due to 64QAM is higher than that of QPSK by 3~3.5dB. Certainly, the above description is based on the assumption that a booster EDFA cannot be built into the transceiver due to the space limitation; or even if an internal EDFA is available, higher optical power laser is needed to achieve a certain TX OSNR requirement.
A high optical power laser is also needed in the case when a single laser is used simultaneously as a TX and a LO laser, i.e., its power is shared between TX and LO. Typically, +16dBm or above is considered a high output power (this applies equally to tunable or fixed-wavelength lasers).
As shown in Fig.1, a high power laser is not needed only in a few cases: (i) In-line optical amplifiers are used in a DWDM system with a lower data rate and a short distance, or (ii) a grey link with very low data rates and a short distance, and a less stringent link budget requirement (due to small patch panel loss, for example)
3. Wide tuning range
This requirement can be divided into four areas: (a) fixed (or partially tunable) frequency; (b) tunable over the C-band; (c) tunable over the C- and L-bands using two separate tunable lasers; and (d) tunable over 1.5x frequency range of the C-band. As shown in Fig.1, within a data center, a widely tunable laser is generally not required. In an inter-data center (DCI) 10-80km link, laser frequency tunability is not a must, although tunable lasers would be preferred for easier operations. For metro systems and beyond, laser frequency tunability is always necessary due to the sporadic and random laser frequency deployment. For DCI, C-band and L-band tunable lasers can help boost the total capacity to 9.6THz (= 4.8THz x 2), but would require two separate models of tunable lasers and two models of erbium-doped fiber amplifiers (EDFAs). For telecom networks, the most recently developed tunable laser can cover 6THz (as opposed to the commonly used 4 or 4.8THz tunable laser), and only requires a single EDFA model to make the network more cost-effective and easier to operate than C+L. The original idea of 6THz total bandwidth comes from the thought of trying to avoid total fiber capacity loss due to the new 75GHz-spacing requirement when the signal baud rate goes beyond 64Gbaud. By having tunable lasers covering 6THz, the total number of channels can be maintained at 80 (= 6THz/75GHz), which is the same as in a regular 50GHz-spaced system. As a result, when the baud rate is doubled and the channel number is unchanged, the total capacity can indeed be doubled.
4. Low phase noise
A laser with a low phase noise also means that it has a low frequency noise. A low frequency noise laser implies that, in its frequency noise power spectral density, it has low flicker noise and interfering tones below 1~100MHz, and low white frequency noise between ~10MHz and ~1GHz in a state-of-the-art semiconductor tunable laser. Multiplying the one-sided white frequency noise power spectral density by a factor of π gives the laser linewidth.
Fig.2 shows the impact on a received 16QAM constellation diagrams in a coherent system without and with the presence of laser phase noise. We can clearly see the laser phase noise-induced phase rotations for the constellation points along the three constant radii in a 16QAM constellation in Fig.2(b). The negative effect is that some of the neighbor constellation points cannot be clearly distinguished and results in a higher system bit-error-rate.
Generally speaking, low laser phase noise is required for (i) a system with a high order modulation ≥ 64QAM (one can imagine that the laser phase noise can cause its dense constellation points to interfere with each other easily), and (ii) a system with a high baud rate and/or a long transmission distance (caused by a phenomenon called “equalizer-enhanced phase noise” which imposes stringent requirements on LO phase noise).
As shown in Fig.1, laser phase noise is of particular importance for a long-haul or subsea coherent system, or a metro distance with a high data rate. For a metro system with a low data rate, and for short distances such as DCI (10-80km) and intra-data centers, low phase noise is not as critical, as long as a high order modulation ≥ 64QAM is not used. Note that for 10-80km DCI running at ≥400Gb/s, a laser linewidth of less than 200~500 KHz is still required, regardless of whether a tunable or a fixed-wavelength DFB laser is used.
With the above description of the requirements on coherent lasers, we include in Fig.1 for each application the proper candidates of commercially available tunable and fixed-wavelength lasers.