[Paper Review] Spreading Signals in the Wideband Limit

Elchanan Zwecher, Dana Porrat|ArXiv.org|Apr 8, 2008

Ultra-Wideband Communications Technology17 references3 citations

TL;DR

This paper establishes that spreading signals in wideband systems suffer from diminishing achievable data rates due to channel uncertainty when the number of multipath components grows sub-linearly with bandwidth. Using the I-MMSE relationship, it proves that communication becomes impossible in the wideband limit if SNR scales slower than log(W/L)/(W/L), where W is bandwidth and L is the number of paths.

ABSTRACT

Wideband communications are impossible with signals that are spread over a very large band and are transmitted over multipath channels unknown ahead of time. This work exploits the I-mmse connection to bound the achievable data-rate of spreading signals in wideband settings, and to conclude that the achievable data-rate diminishes as the bandwidth increases due to channel uncertainty. The result applies to all spreading modulations, i.e. signals that are evenly spread over the bandwidth available to the communication system, with SNR smaller than log(W/L)/(W/L) and holds for communications over channels where the number of paths L is unbounded by sub-linear in the bandwidth W.

Motivation & Objective

• To analyze the fundamental limits of wideband communication systems using spreading signals over multipath channels with unknown path characteristics.
• To identify the SNR regime in which reliable communication becomes impossible in the wideband limit due to channel uncertainty.
• To extend prior results on PPM and impulse radio to a broader class of spreading signals, including DS-SS and OFDM, under a unified framework.
• To quantify the impact of channel eigen-mode uncertainty on achievable data rates using the I-MMSE relationship.
• To demonstrate that modulation schemes using only a few eigen-modes (e.g., FSK) are less affected by uncertainty than those using all eigen-modes.

Proposed method

• Models the wideband communication system using a discrete, block-coherent channel model with bandwidth W and coherence time Tc, sampling at rate W to form a vector of length Kc = WTc.
• Applies the I-MMSE relationship to bound the achievable data rate by analyzing the minimum mean square error (MMSE) estimate of the unknown channel impulse response H.
• Derives upper bounds on key terms in the likelihood ratio test, including the signal-to-noise ratio terms a_i, b_i, and the dominant noise term c_k^*, using order statistics and concentration inequalities.
• Uses asymptotic extreme value theory to show that the maximum noise term c_k^* scales as √(K_c SNR / L) × √(2 log(K_c / L)) in the large-bandwidth limit.
• Establishes that the likelihood ratio J(H) decays to zero when SNR = o(log(K_c / L) / (K_c / L)), implying failure to detect signals in the wideband limit.
• Demonstrates that the dominant error source is the uncertainty in the channel's eigen-values (complex gains), especially when signals excite all eigen-modes simultaneously.

Experimental results

Research questions

• RQ1 What is the fundamental SNR threshold below which spreading signals cannot achieve reliable communication in the wideband limit due to channel uncertainty?

Key findings

• The achievable data rate of spreading signals diminishes to zero in the wideband limit when the SNR scales slower than log(W/L)/(W/L), where W is bandwidth and L is the number of multipath components.
• The critical factor limiting performance is the exposure to uncertainty in a large number of channel eigen-values, which occurs when signals spread power across all available frequency modes.
• The likelihood ratio J(H) decays to zero as bandwidth increases when SNR = o(log(K_c / L) / (K_c / L)), proving that reliable detection becomes impossible under such SNR scaling.
• The result applies to all spreading modulations, including PPM, impulse radio, DS-SS, and OFDM, provided the entire bandwidth is used.
• Modulation schemes like FSK, which use only a small number of eigen-modes per symbol, are less affected by channel uncertainty and remain viable in the wideband regime.
• The analysis generalizes prior results on PPM and impulse radio by showing that the fundamental limitation is not signal structure per se, but the number of unknown channel parameters excited by the signal.