On the Capacity of Time-Varying Channels With Periodic Feedback

Mehdi Ansari Sadrabadi, Mohammad Ali Maddah-Ali, Amir K. Khandani

TL;DR

This paper investigates the theoretical limits (capacity) of communication over time-varying wireless channels when the receiver has perfect knowledge of the channel state and can periodically send this information back to the transmitter. They derive the capacity for two types of channels and show that for a common type of fading channel, optimal adaptation can be achieved with a single Gaussian codebook by adaptively adjusting power based on the feedback.

What Problem Does This Paper Address?

Wireless communication channels are rarely static; their characteristics (like signal strength and interference) change over time, often in unpredictable ways. This variability, known as "time-varying channels," poses a significant challenge for reliable data transmission. While it's ideal for the transmitter to know the channel's state to optimize its transmission strategy, obtaining this information perfectly is difficult. This paper focuses on a scenario where the receiver does know the channel state perfectly and can send this information back to the transmitter. The key challenge is to determine the maximum achievable data rate (capacity) under these conditions, specifically when the feedback is not continuous but occurs at regular intervals (periodic feedback). This is crucial for designing more efficient wireless systems that can adapt to changing channel conditions.

Methods

The authors tackle this problem by first defining two models for time-varying channels:

  1. Finite State Markov Channel (FSMC): This model assumes the channel can only be in a limited number of discrete states, and transitions between these states follow a Markov process.
  2. Time-Correlated Rayleigh Fading Channel: This is a more realistic model for wireless channels, where the signal strength fluctuates according to a Rayleigh distribution and these fluctuations are correlated over time.

For the FSMC, they leverage existing results and show that the capacity can be achieved by using multiple codebooks, each tailored to a specific channel state, and multiplexing them over time.

For the more complex time-correlated fading channel, they derive the capacity by considering the channel as a series of parallel subchannels, each corresponding to a feedback interval. They then use optimization techniques to find the optimal power allocation strategy. A key aspect of their approach is the use of information-theoretic tools like mutual information and Fano's inequality to establish upper and lower bounds on the channel capacity.

Datasets

Not discussed in the paper. The paper focuses on theoretical derivations and performance evaluation through simulations presented in a figure, rather than empirical datasets.

Key Insights and Ideas

  • Periodic feedback is surprisingly effective: Even with feedback only at discrete intervals, significant gains in capacity can be achieved compared to having no feedback at all.
  • Adaptive power is key for fading channels: For time-correlated fading channels, the optimal strategy isn't necessarily about using many different codebooks, but rather about adaptively adjusting the transmission power based on the received channel state information.
  • Single Gaussian codebook sufficiency: For the time-correlated fading channel, they demonstrate that a single Gaussian codebook, combined with adaptive power allocation, is sufficient to achieve the derived capacity. This simplifies practical implementation.
  • Impact of correlation and feedback period: The performance evaluation section (Figure 1) visually demonstrates how channel correlation and the feedback period (T) influence the achievable capacity. Higher correlation and shorter feedback periods generally lead to higher capacities.

Key Results

  • FSMC Capacity: The paper provides a formula for the capacity of an FSMC with periodic feedback, achievable through multiplexing T codebooks.
  • Gaussian Channel Capacity: For a time-correlated fading Gaussian channel, they derive the capacity as an expectation over the channel states and feedback intervals, subject to an average power constraint.
  • Optimal Adaptation Strategy: They show that for the Gaussian fading channel, the optimal adaptation can be achieved by adaptively scaling the power of a single Gaussian codebook based on the side information at the transmitter.
  • Performance Illustration: Figure 1 shows that the capacity of a time-correlated Rayleigh fading channel significantly increases with shorter feedback periods (smaller T) and higher channel correlation coefficients (closer to 1). The capacity with feedback is substantially higher than without any CSI at the transmitter.

Limitations and Future Work

  • Perfect CSI Assumption: The paper assumes perfect CSI at the receiver and a noiseless feedback link. In reality, CSI estimation is imperfect, and feedback links can be noisy and have delays, which would reduce the achievable rates.
  • Finite Alphabet Feedback: The paper considers a noiseless feedback link, but in practice, feedback information must be quantized, leading to a finite alphabet. Investigating the impact of finite-alphabet feedback is a natural extension.
  • Asymmetric Information: The paper primarily deals with scenarios where the receiver has perfect CSI. Extending the analysis to cases with asymmetric information (where neither transmitter nor receiver has perfect CSI, or they have different levels of information) would be valuable.
  • More Complex Channel Models: While the Rayleigh fading channel is common, other fading models or more complex channel dynamics could be explored.

The authors suggest that future work could explore the impact of feedback delay and estimation errors on the derived capacity.

Summary updated

BibTeX
@article{sadrabadi2007capacity,
  author = {Mehdi Ansari Sadrabadi and Mohammad Ali Maddah-Ali and Amir K. Khandani},
  title = {{On the Capacity of Time-Varying Channels With Periodic Feedback}},
  journal = {IEEE Trans. Information Theory},
  year = {2007},
  doi = {10.1109/tit.2007.901250},
}

Loading related content...

Figure