TL;DR
This paper derives new upper and lower bounds for the capacity of underspread fading wireless channels, considering realistic constraints like limited peak power and non-coherent reception. These bounds are explicit in the channel's scattering function and provide insights into optimal bandwidth and system design, particularly for scenarios with peak constraints in time and frequency.
What Problem Does This Paper Address?
Understanding the maximum achievable data rate (channel capacity) is crucial for designing efficient wireless communication systems. However, real-world wireless channels are complex: they experience fading (signal strength fluctuations) that varies in both time and frequency, and transmitters have limitations on their peak power output. Furthermore, receivers often don't know the exact channel conditions in real-time (non-coherent reception). Existing capacity calculations often struggle with these combined complexities, especially for channels that are "underspread" (delay and Doppler spreads are small), a common characteristic of most wireless environments. This paper aims to provide more accurate and practical capacity bounds for such channels, considering both peak power limitations and non-coherent operation.
Methods
The authors tackle this problem by first discretizing the continuous-time wireless channel model into a discrete-time, discrete-frequency representation. This discretization is specifically tailored for "underspread" channels, minimizing approximation errors. They then derive upper and lower bounds on the channel capacity under two main scenarios:
- Peak constraint in time and frequency: This scenario models systems where both the signal's duration and its frequency spread are limited. The bounds are expressed in terms of the channel's "scattering function," a key characteristic that describes how the channel scatters signals in time and frequency. They also derive an asymptotic expression for capacity in the limit of very large bandwidth.
- Peak constraint in time only: This scenario allows signals to be "peaky" in frequency. Here, they focus on the infinite-bandwidth capacity and find cases where their derived bounds exactly characterize this capacity, relating it to Viterbi's earlier work.
The analysis relies heavily on advanced mathematical tools, including approximations of channel operators, properties of Toeplitz matrices, and information-theoretic concepts like mutual information and minimum mean-square error (MMSE).
Datasets
Not discussed in the paper.
Key Insights and Ideas
- Underspread channels are key: The paper leverages the "underspread" property of most wireless channels to develop a tractable mathematical framework. This property simplifies the channel model significantly.
- Scattering function is central: The derived capacity bounds are explicit functions of the channel's scattering function. This is important because the scattering function can be estimated from measurements, making the bounds practical for real-world channel characterization.
- Peak constraints drastically alter capacity: The paper highlights how different peak power constraints (time vs. time-and-frequency) lead to fundamentally different capacity behaviors, especially in the large-bandwidth regime. Allowing signals to be "peaky" in frequency can prevent capacity from vanishing, unlike strict time-and-frequency peak constraints.
- Bounds are tight in the wideband regime: For certain scenarios, the derived upper and lower bounds are shown to be tight (coincide) in the wideband limit, providing exact capacity characterizations.
- Approximation is powerful: The use of an approximating channel operator, based on Weyl-Heisenberg sets, allows for a structured analysis and leads to practical results, even though it introduces approximation errors.
Key Results
- New capacity bounds: The paper provides novel upper and lower bounds for the noncoherent capacity of underspread WSSUS channels, considering both average power and peak power constraints (in time and frequency, or just in time).
- Bandwidth optimization insights: The bounds allow for a coarse identification of the capacity-optimal bandwidth as a function of peak power and channel characteristics.
- Wideband behavior characterized: A closed-form expression for the first-order Taylor series expansion of capacity in the infinite-bandwidth limit is derived, showing tightness of the bounds in this regime.
- Exact infinite-bandwidth capacity found: For peak-constrained-in-time-only signals, the paper establishes exact characterizations of the infinite-bandwidth capacity in certain cases.
- Numerical example confirms theory: Figure 1 illustrates the derived bounds and approximations, showing how capacity behaves with bandwidth and how the bounds relate to each other.
Limitations and Future Work
- Approximation error: While the paper uses an approximation for the channel operator, a complete characterization of the approximation error between the original and approximated operators remains an open problem. This includes quantifying the differences in their null and range spaces.
- Overspread channels: The analysis focuses on "underspread" channels. Characterizing the capacity of "overspread" channels (where the spread is large) is identified as a challenging open problem due to the inability to diagonalize the channel kernel with deterministic eigenfunctions.
- Practical implementation: While the bounds are explicit in the scattering function, evaluating them might still require numerical methods, especially for complex scattering functions.
- MIMO channels: Extensions to spatially correlated MIMO channels are mentioned as provided in other work [68], suggesting a direction for future research.
Summary updated