Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs)

Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [CCA algorithm proposal for MB-OFDM] Date Submitted: [9 Sept, 2004] Source: [Charles Razzell, Yifeng Zhang] Company [Philips] Address [1109, McKay Drive, San Jose, CA 95131, USA] Voice:[+1 408 474 7243], FAX: [+1 408 474 5343], E-Mail:[charles.razzell@philips.com, yifeng.zhang@philips.com] Re: [ ] Abstract: [Two algorithm proposals based on the same basic principle are evaluated for performance and complexity. Fast CCA is achieved requiring neither preamble correlation nor pre-existing, tight time and frequency synchronization. Initial estimations of algorithmic complexity are given.] Purpose: [] Notice: This document has been prepared to assist the IEEE P802.15. It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Release: The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P802.15. C. Razzell, Y.Zhang, Philips

CCA algorithm proposal for MB-OFDM C. Razzell, Y.Zhang, Philips

Background • Previous proposals for MB-OFDM CCA have focused on time domain correlation of preamble • Energy detect may also be used as coarse detection for strong input signal • Ambiguity of media status appears when frame preamble is missing and SNR is low (close to sensitivity level) • Ambiguity of media status impacts MAC decisions, and may degrade the throughput C. Razzell, Y.Zhang, Philips

Requirements • CCA should be able to sense the media at any time • The channel clear state should be sensed within a few micro seconds (within 3.75ms is the target) • CCA should be reliable at SNR as low as possible, and be robust in multipath channels • CCA should be reliable even in SOP environment • CCA implementation should be as simple as possible C. Razzell, Y.Zhang, Philips

Scenarios analysis for CCA without preamble • Two scenarios have been identified • Receiver is not time-synchronized to the channel Time-Frequency Interleaving (TFI) code • This is rarely a relevant case for media access, since timing must be known to access the media. It may, however, act as an aid to initial acquisition. • Receiver has at least coarse time synchronization for the relevant TFI code to be used • This is often the case after initial acquisition: a small amount of timing drift occurs depending on the time elapsed after the most recent successful burst reception. C. Razzell, Y.Zhang, Philips

Signal analysis for Case 1 • If the TFI channel timing information is not available to the receiver, the receiver has to stay at one channel to receive at least 1/3 of the symbols; otherwise, cases like hopping chasing could happen, and nothing will be received. C. Razzell, Y.Zhang, Philips

Signal analysis for Case 1 (continued) • If receiver stays on one frequency channel, the signal will be gated by the channel selectivity. C. Razzell, Y.Zhang, Philips

Proposed algorithm for case 1 • A 128 moving average (MA) algorithm is proposed for this case. The sample power is calculated and MA over 128 samples. For a properly down-converted UWB symbol, the power MA over 128 samples is equivalent to frequency domain MRC, which maximizes the SNR at the MA output. C. Razzell, Y.Zhang, Philips

Mathematical explanation of MA algorithm Received UWB Symbol Parseval’s theorem for Periodic signal RHS is recognizable as Frequency domain maximal ratio Combining (MRC) C. Razzell, Y.Zhang, Philips

MA based CCA detector C. Razzell, Y.Zhang, Philips

MA based CCA detector description • A peak detector will be used to find the peak of the MA output • A pre-determined threshold (Pth) is used to compare with the peak • A peak window is used to detect the periodicity of the peak (Pwindow) • A peak counter is used to count the number of peak occurred (Npeak) • Properly chosen Pth, Pwindow and Npeak can achieve optimal false alarm rate and missed frame rate C. Razzell, Y.Zhang, Philips

MA based CCA detector performance C. Razzell, Y.Zhang, Philips

Complexity of MA based CCA • Since MA is calculated sample by sample, it needs only 5 multipliers, 5 adders and size 128 buffer, and results in modest gate count. C. Razzell, Y.Zhang, Philips

Signal analysis for Case 2 • If TFI channel timing information is coarsely available to the receiver, the receiver can follow the assigned frequency/time sequence and almost all the symbols can be down-converted to base band. C. Razzell, Y.Zhang, Philips

Signal analysis for Case 2 (continued) • If TFI timing information is coarsely available to the receiver, due to ZP, the signal can be looked as the product of a continuous random signal with a 3.2MHz square wave clock C. Razzell, Y.Zhang, Philips

Proposed algorithm for case 2 • A symbol based cross-correlation (CC) algorithm is proposed. By using symbol by symbol cross correlation of the input signal and a 3.2MHz sinusoid, the existence of UWB signal can be detected. C. Razzell, Y.Zhang, Philips

Mathematical explanation of CC algorithm Instantaneous power of received UWB signal, remove impact from frequency offset Remove DC to enhance 3.2MHz frequency component, N is number of samples per period Cross correlation with 3.2MHz sinusoid C. Razzell, Y.Zhang, Philips

CC based CCA detector C. Razzell, Y.Zhang, Philips

CC based CCA detector description • Several symbols (Nsym) are used to provide averaged CC output • A pre-determined threshold (Pth) is used to compare with the CC output • Properly chosen Pth, and Nsym can achieve good false alarm rate and missed frame rate • A counter can be used to count the number of contiguous CC outputs which are larger than Pth. This will enhance both the false alarm rate and missed frame rate performance. C. Razzell, Y.Zhang, Philips

CC based CCA detector performance C. Razzell, Y.Zhang, Philips

Complexity of CC based CCA • Since CC is also calculated sample by sample, it needs only 5 multipliers, 5 adders and size 64 buffer (due to the symmetric property of sinusoid), which has even less gate count compared with MA. C. Razzell, Y.Zhang, Philips

Comparison parameters MA algorithm CC algorithm Channel dependency Robust Robust False alarm rate Relatively low Relatively lower Detection duration 3.75ms 3.75ms Sensitivity to TFI sync Robust Robust piconet identify capability Low High SOP tolerance High Mid Complexity Very low Very low, and lower than MA algorithm Comparison of two schemes C. Razzell, Y.Zhang, Philips

Conclusions • Two CCA schemes for systems missing preamble detection is proposed • Both schemes provide reliable detection (detect rate > 90%) within 3.75us, regardless of channel condition, at SNR > -2 dB. • With SOP, the MA shows robust performance while CC shows acceptable performance • CC method also shows robust performance to the TFI synchronization condition. • Both schemes can be realized with moderate complexity. C. Razzell, Y.Zhang, Philips

Supporting materials Simulation setup and conditions C. Razzell, Y.Zhang, Philips

Simulation setup C. Razzell, Y.Zhang, Philips

Simulation condition for MA scheme in different channels • 12 UWB symbols(3.75ms) are used for CCA detection • Pth = 1.3 • Window size is 90 samples • Number of frames simulated for each SNR is 20000 • 100 channel realization are used, therefore, 200 frames per channel realization • Peak counter > 1 for a valid detection • The complex channels are normalized such that the shadowing effect is excluded. C. Razzell, Y.Zhang, Philips

Simulation condition for MA scheme in SOP environment • 12 UWB symbols (3.75ms) are used for CCA detection • Pth = 1.3 • Window size is 90 samples • CM4 is used as multipath channel, number of frames simulated for each SNR is 20000, 100 channel realization are used, therefore, 200 frames per channel realization • Peak counter > 1 for a valid detection • The complex channels are normalized such that the shadowing effect is excluded. • There are two piconets in the system, piconet 1 is the designated signal with TFI pattern 1 2 3; piconet 2 is the interference with hopping pattern 1 3 2 C. Razzell, Y.Zhang, Philips

Simulation condition for MA scheme in SOP environment (continued) • SNR = 0 dB • piconet 1 and 2 has different relative phase shift that reflects the unsynchronized effect • piconet 2 and 1 has different power level at receiver input, defined as ISR. ISR = -1dB means the power of interference piconet 2 is 1 dB lower than that of piconet 1 at receiver input. C. Razzell, Y.Zhang, Philips

Simulation condition for SOP environment (continued) Phase misalignment is represented by number of samples in the simulation plot, sampling rate is 1/2112MHz C. Razzell, Y.Zhang, Philips

Simulation condition for CC scheme in different channels • 12 UWB symbols (3.75ms) are used for CCA detection • Pth = -12.5dB • Number of frames simulated for each SNR is 20000 • 100 channel realization are used, therefore, 200 frames per channel realization • The complex channels are normalized such that the shadowing effect is excluded. • Ideal frequency hopping synchronization C. Razzell, Y.Zhang, Philips

Simulation condition for CC scheme in SOP environment • 12 UWB symbols (3.75ms) are used for CCA detection • Pth = -11.9dB • CM4 is used as multipath channel, number of frames simulated for each SNR is 20000, 100 channel realization are used, therefore, 200 frames per channel realization • The complex channels are normalized such that the shadowing effect is excluded. • There are two piconets in the system, piconet 1 is the designated signal with hopping pattern 1 2 3; piconet 2 is the interference with hopping pattern 1 3 2 • SNR = 0 dB • piconet 1 and 2 has different relative phase shift that reflects the unsynchronized effect • piconet 2 and 1 has different power level at receiver input, defined as ISR. ISR = -1dB means the power of interference piconet 2 is 1 dB lower than that of piconet 1 at receiver input. C. Razzell, Y.Zhang, Philips

Simulation condition for CC scheme with imperfect hopping synchronization • 12 UWB symbols (3.75ms) are used for CCA detection • Pth = -11.9 • CM4 is used as multipath channel, number of frames simulated for each SNR is 20000, 100 channel realization are used, therefore, 200 frames per channel realization • The complex channels are normalized such that the shadowing effect is excluded. • SNR = 0 dB • Timing illustration is given on next slide… C. Razzell, Y.Zhang, Philips

Simulation condition for CC scheme with imperfect hopping synchronization Phase misalignment is represented by number of samples in the simulation plot, sampling rate is 1/2112MHz C. Razzell, Y.Zhang, Philips

Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs)