QAM (Quadrature Amplitude Modulation)

QAM (Quadrature Amplitude Modulation) is a modulation format that encodes data in both the amplitude and phase of the optical field, or equivalently in the in-phase (I) and quadrature (Q) components of the complex field amplitude. The number after the QAM label indicates the constellation size: 16-QAM has 16 distinct symbols, 64-QAM has 64, etc.

Constellation representation. QAM signals are visualized as constellation diagrams in the I-Q plane, where each symbol is a complex point $(I,Q)$ representing the field amplitude:

For optical communications, the standard format is dual-polarization QAM (DP-QAM): two independent QAM streams on the two orthogonal polarizations of the single-mode fiber, doubling the bit rate without doubling bandwidth.

Spectral efficiency. For DP-QAM:

Higher QAM order yields more bits per symbol but at the cost of needing higher SNR. The spectral-efficiency-vs-reach trade-off is the fundamental engineering choice in long-haul system design.

OSNR requirements. The SNR needed for a given BER rises with the number of bits per symbol. For a dual-polarization signal at symbol rate $R_s$, the OSNR in the usual 0.1 nm reference bandwidth ($B_{\text{ref}}$ = 12.5 GHz) is $\text{OSNR}=\text{SNR}\cdot R_s\mathrm{/}{B}_{\text{ref}}$. Theoretical values at 32 GBaud with additive white Gaussian noise, for a pre-FEC BER of $2\times {10}^{-2}$, a typical input threshold of soft-decision FEC:

Each step between these square formats, two more bits per symbol in each polarization, costs about 6 dB. The OSNR requirement also scales with the symbol rate, because the noise bandwidth grows with it: at 64 GBaud every value rises by 3 dB. Practical transceivers need more than these theoretical figures, by an implementation penalty. Once the pre-FEC BER is below the FEC threshold, the decoder output is error-free for practical purposes, so no additional OSNR is needed to reach a low post-FEC error rate; links are instead given a margin for aging and impairments.

Coherent detection required. QAM signals are normally received with coherent detection. A single direct-detection photodiode measures only intensity, so it cannot recover symbol phase without a co-transmitted carrier or a self-coherent receiver. A coherent receiver:

1. Mixes the incoming signal with a continuous-wave local oscillator (LO) of similar wavelength
2. Beat between signal and LO produces I and Q electrical signals proportional to the signal's complex field
3. Digital signal processing (DSP) demodulates the I/Q samples into symbol decisions

The local oscillator typically uses an ECDL (external-cavity diode laser) or similar narrow-linewidth source. Source linewidth must be $\ll 1\mathrm{\%}$ of the symbol rate to limit phase noise penalty.

Symbol rate × constellation rates. The total bit rate is:

$$R=\frac{B\cdot k_{\text{mod}}\cdot 2}{1+\text{FEC overhead}},$$

where $B$ is the symbol rate (baud), $k_{\text{mod}}$ is bits per QAM symbol, the factor of 2 is for dual-polarization, and FEC overhead is typically 20 – 25%.

For example, 200G with DP-16QAM at 32 GBaud and 25% FEC overhead:

$$\begin{array}{rl}{\displaystyle R}& {\displaystyle =\frac{32\text{ GBaud}\times 4\text{ bits/sym}\times 2\text{ pol}}{1.25\text{ (FEC)}}}\\ {\displaystyle }& {\displaystyle \approx 205\text{ Gb/s net},\sim 256\text{ Gb/s gross}\mathrm{.}}\end{array}$$

To exceed 200G net per channel, you need either higher baud or higher constellation.

Reach vs constellation choice.

Probabilistic constellation shaping (PCS). Modern coherent systems use PCS to fine-tune the trade-off between rate and reach. Instead of using all constellation points equally, inner points are used more often than outer points (Maxwell-Boltzmann distribution). PCS provides:

• Up to about 1 dB of SNR gain at the same modulation order and rate (0.72 dB for 64-QAM at 5 bits per symbol, against a theoretical ceiling of 1.53 dB)
• Continuous rate adjustment (without changing constellation)
• Better adaptation to varying channel conditions

For a single hardware design, PCS allows the same transceiver to operate at 200 G or 400 G or 500 G or 600 G depending on the SNR available.

Standards and product offerings.

The bit rate per coherent transceiver has approximately doubled every 3 – 4 years, driven by improvements in DSP, ADC bandwidth, and DAC linearity.

Why QAM enables capacity scaling. Capacity is the product of symbol rate and bits per symbol. Symbol rate is bandwidth-limited (today's ADCs/DACs cap at ~ 130 – 200 GBaud); bits per symbol scales with SNR. Modern transmission systems push both axes:

• Higher baud rate: 130 – 200 GBaud in 2024 – 2026 systems
• Higher constellation: 64-QAM, 256-QAM with PCS

The combination doubles capacity each generation.

DSP requirements at the receiver.

For DP-64QAM at 100 GBaud, the receiver's four ADCs each sample at more than 100 GSa/s, the DSP IC performs tens of trillions of operations per second on their output, and it consumes 15 – 25 W.

References: Agrawal, Fiber-Optic Communication Systems (4th ed., 2010), for the modulation-format treatment; K. Roberts, Q. Zhuge, I. Monga, S. Gareau and C. Laperle, "Beyond 100 Gb/s: capacity, flexibility, and network optimization," Journal of Optical Communications and Networking 9, C12 (2017), for modern QAM coherent transmission; ITU-T G.709 / OTN for standardized formats.