Chapter Summary

Chapter Summary

Key Points

• 1.

  The DD channel acts by 2D convolution with a sparse kernel. The central structural result: $y_{DD} = h \star\star x_{DD} + \mathbf{w}_{DD}$, where $h(\tau, \nu) = \sum_{i=1}^P h_i\,\delta(\tau - \tau_i)\,\delta(\nu - \nu_i)$. The 2D convolution is on the delay-Doppler torus and is doubly circular. The Zak covariance of Chapter 2 is what converts the time-domain channel's delay-and-Doppler-shift action into pure translations on the DD plane.

• 2.

  Discrete form has $P$ taps per DD cell. On the $(M, N)$ grid with integer-indexed paths: $Y_{DD}[\ell, k] = \sum_{i=1}^P h_i\,e^{j\alpha_i(\ell,k)}\,X_{DD}[(\ell-\ell_i)\bmod M, (k-k_i)\bmod N] + W_{DD}[\ell, k]$. The equation has exactly $P$ terms — not $MN$. This is the concrete sparsity on the discrete grid.

• 3.

  The DD channel matrix is a Kronecker sum of sparse shifts. $\mathbf{H}_{DD} = \sum_i h_i (\boldsymbol{\Pi}_N^{k_i}\boldsymbol{\Delta}_N^{\ell_i}) \otimes \boldsymbol{\Pi}_M^{\ell_i}$ has density $P/(MN) \ll 1$, is block circulant with circulant blocks, and is diagonalized by the 2D DFT. These three properties enable efficient detection algorithms with $O(PMN)$ or $O(MN\log(MN))$ complexity.

• 4.

  The DD channel is time-invariant over one OTFS frame. Under block-fading (constant path geometry over $T \leq 10$ ms), the DD kernel $h(\tau, \nu)$ does not depend on time. One channel estimate suffices for the entire frame; one uniform detector works at every DD cell. Contrast with OFDM, where $H(f, t)$ evolves visibly across a frame under mobility.

• 5.

  Sparsity converts estimation from $O(MN)$-dimensional to $O(P)$-dimensional. Pilot overhead drops from $\sim 10\%$ (per-subcarrier DMRS in OFDM) to $\sim 1\%$ (single embedded pilot + guard region in OTFS). Estimation MSE scales with $P$, not with $MN$, because there are only $P$ unknowns. Detection complexity scales linearly in $MN$ (up to log factors) rather than quadratically.

• 6.

  Integer Doppler is a convenient fiction. Fractional Doppler offsets $\epsilon = \nu_i - k_i/T$ create inter-Doppler interference that spreads energy across neighboring cells. The exact treatment is in Chapter 10. For the analysis of Chapters 7-9, we assume integer indices and treat fractional effects as perturbations.