Summary

Chapter 19 Summary: Multiple Access Techniques

Key Points

• 1.

  Orthogonal multiple access (FDMA, TDMA, OFDMA) eliminates multi-user interference by partitioning the time-frequency resource among users. However, the sum rate of any orthogonal scheme is strictly less than the MAC capacity for $K \geq 2$ users: by Jensen's inequality, $\sum_k \alpha_k \log(1 + \text{SNR}_{k}/\alpha_k) \leq \log(1 + \sum_k \text{SNR}_{k})$. OFDMA partially recovers the loss through frequency-domain scheduling and multi-user diversity.

• 2.

  Non-orthogonal multiple access (NOMA) allows all users to share the same resource simultaneously, resolving them via successive interference cancellation (SIC). Superposition coding with SIC achieves the entire boundary of the Gaussian MAC capacity region, including the sum-rate point $C_{\text{sum}} = \log_2(1 + \sum_k \text{SNR}_{k})$. The gain over orthogonal access is most significant when user channel gains are highly asymmetric.

• 3.

  CDMA uses spreading codes to allow simultaneous transmission over the full bandwidth. The processing gain $N$ provides interference suppression of approximately $N/(K-1)$ with a matched-filter receiver. The near-far problem is the critical challenge: without tight power control, nearby users overwhelm distant ones. Rake receivers exploit multipath diversity by coherently combining resolvable path components.

• 4.

  Random access protocols (ALOHA, CSMA/CA) enable uncoordinated transmission, trading throughput for simplicity. Slotted ALOHA achieves a maximum throughput of $1/e \approx 36.8\%$ at offered load $G = 1$, exactly double the $1/(2e)$ of pure ALOHA. Both protocols are susceptible to congestion collapse at high load, requiring stabilisation mechanisms such as binary exponential backoff.

• 5.

  Duplexing determines how uplink and downlink share the spectrum. FDD uses paired frequency bands (independent channels requiring explicit feedback), while TDD uses a single band with time-domain switching (enabling channel reciprocity). The CSI overhead of FDD scales as $O(M)$ with the number of BS antennas, while TDD overhead scales as $O(K)$, making TDD overwhelmingly preferred for massive MIMO. This overhead asymmetry is the primary reason 5G NR massive MIMO deployments use TDD.