Chapter Summary

Chapter Summary

Key Points

• 1.

  The normal approximation $R^*(n, \epsilon) = C - \sqrt{V/n}\,Q^{-1}(\epsilon) + O(\log n/n)$ is the fundamental formula for finite-blocklength coding. The channel dispersion $V = \text{Var}[\iota(X;Y)]$ governs the speed of convergence to capacity: two channels with the same capacity but different dispersions behave very differently at short blocklengths.

• 2.

  The RCU bound and meta-converse provide tight, computable, non-asymptotic bounds on $R^*(n, \epsilon)$ that replace the asymptotic achievability-converse pair (random coding + Fano). Together they sandwich the true maximum rate to within a fraction of a bit for most practical channels, even at $n \sim 100$.

• 3.

  The rate-reliability-blocklength tradeoff is the central design tool for URLLC. At $n = 200$ and $\epsilon = 10^{-5}$, the achievable rate can be 20-40% below the Shannon capacity, depending on the channel and SNR. Using the capacity formula for short-blocklength design leads to significant under-provisioning of resources.

• 4.

  The AWGN dispersion $V = \text{SNR}(\text{SNR}+2)/(2(1+\text{SNR})^2)$ in nats$^2$ approaches $1/2$ at high SNR. In fading channels, the dispersion includes an additional term from fading variance, which dominates at low diversity orders and makes multi-antenna systems essential for URLLC.

• 5.

  Multi-user finite-blocklength theory extends the normal approximation to MAC and BC. The MAC dispersion is a matrix governing the shrinkage of the capacity region. The sum-rate dispersion equals the point-to-point dispersion at the sum power. Superposition coding remains second-order optimal for the degraded BC.

• 6.

  Massive MTC and grant-free access operate in the many-access regime where $K_a$ grows with $n$. The per-user energy-per-bit requirement grows as $\Theta(\log K_a)$, fundamentally limiting short-packet random access scalability beyond the $\sqrt{V/n}$ penalty.