A Capacity-Based Approach for Designing Bit-Interleaved Coded GFSK with Noncoherent Detection Rohit Iyer Seshadri and Matthew C. Valenti Lane Dept. of Computer Science and Electrical Engineering West Virginia University iyerr, mvalenti @csee.wvu.edu Problem “ Which is the optimal combination of channel coding rate and continuous phase modulation (CPM) parameters for a given bandwidth efficiency and decoder complexity?” 7/12/2006 2/24 Continuous Phase Modulation CPM is a nonlinear modulation scheme with memory – Modulation induces controlled inter symbol interference (ISI) Phase continuity results in small spectral side lobes Well suited for bandwidth constrained systems Constant envelope makes it suitable for systems with nonlinear amplifiers CPM is characterized by the following modulation parameters – Modulation order M – Type and width of the pulse shape – Modulation index h Different combination of these parameters result in different spectral characteristics and signal bandwidths 7/12/2006 3/24 Challenges CPM includes an almost infinite variations on the modulated signal – Full response, partial response, GFSK, REC, RC etc.. CPM is nonlinear – Problem of finding realistic performance bounds for coded CPM systems is non-trivial When dealing with CPM systems with bandwidth constraints, lowering the code rate does not necessarily improve the error rate System complexity and hence the detector complexity must be kept feasible 7/12/2006 4/24 Uncoded CPM System Bit to Symbol u a Modulator x Channel r’ Filter ^ r a Detector Symbol to Bit ^ u u: data bits a: message stream comprised of data symbols from the set { ±1, ± 3,…, ±(M-1)} x: modulated CPM waveform r’: signal at the output of the channel. The filter removes out-of band noise ^ a: symbol estimates provided by the detector ^ u: bit estimates provided by the detector 7/12/2006 5/24 An Uncoded System with Gaussian Frequency Shift Keying u Bit to Symbol a GFSK x Channel r’ Filter ^ r a Detector Symbol to Bit ^ u Gaussian frequency shift keying (GFSK) is a widely used class of CPM e.g. Bluetooth, GSM Baseband GFSK signal during kT ≤ t ≤ (k+1)T GFSK phase 7/12/2006 6/24 GFSK Pulse Shape and Uncoded Power Spectrum The pulse shape g(t) is the response of a Gaussian filter to rectangular pulse of width T g (t ) [Q(cBt ) Q(cB(t T ))]/ T 0 BT =0.5 -5 BT =0.5, 2B T =1.04 BT is the normalized 3 dB bandwidth of the filer – – Width of the pulse shape depends on BT Wider the pulse, greater is the ISI Smaller values of BT result in a more compact power spectrum – – Here M =2 and h =0.5 2B99Tb quantifies the bandwidth efficiency Power Spectral Density (dB) 99 b -10 BT =0.25 -15 BT =0.25, 2B T =0.86 99 b BT =0.2 -20 BT =0.2, 2B T =0.79 99 b -25 -30 -35 -40 0 0.2 0.4 0.6 0.8 1 1.2 Frequency (normalized by T) 7/12/2006 7/24 Coded GFSK System u Encoder b GFSK a Channel x ^ r Filter Detector 2. 3. 4. PSD for GFSK using rate Rc code is now S ( f ) Rc S x ( Rc f ) c x cIt is not immediately clear if the performance loss must meet the required spectral efficiency xcaused be lowering h and/or BT will be overcome Decoder 0 Power Spectral Density (dB) 1. The value of BT needs to lowered, with h unchanged Find the OR power spectral density for uncoded GFSK S x ( f ) Both can be lowered u 10 Suppose weimproves need 2B99 Tb =1.04 while using Channel coding energy efficiency at a rate ½ code , of bandwidth efficiency the expense For The our system, belowered, done without value ofcoding h needsmust to be with BT bandwidth expansion, i.e. 2B T should 99 b unchanged remain unchanged OR ^ a -10 -20 M =2, BT =0.5, h =0.125, R =1/2 c -30 M =2, BT =0.5, h =0.5, uncoded -40 M =2, BT =0.075, h =0.5, R =1/2 c S (f) -50 by the coding gain This implies the GFSK parameters have to be modified for the coded signal 7/12/2006 0 2 4 6 8 10 Frequency (normalized by T) 8/24 Proposed Coded GFSK System u Encoder b’ Bit Intrlv. b x GFSK Channel r' ^ r Filter SO-SDDPD b ^ Bit Deintrlv. b’ ^ Decoder u Noncoherent detection used to reduce complexity Detector: Soft-Decision differential phase detector (SDDPD), [Fonseka, 2001]. Produces hard-estimates of the modulated symbols SO-SDDPD generates bit-wise log-likelihood ratios (LLRs) for the code bits Bit-wise interleaving between encoder and modulator and bit-wise soft-information passed from detector to decoder (BICM) Shannon Capacity under modulation and detector design constraints used to drive the search for the “optimum” combination of code rates and GFSK parameters at different spectral efficiencies The availability of capacity-approaching turbo and LDPC codes make the capacity under BICM a very practical indicator of system performance 7/12/2006 9/24 System Model Bit-interleaved codeword b is mapped to symbol sequence a, which is modulated to produce x The baseband GFSK signal x is sent through a frequency nonselective Rician channel Received signal at the output of the channel, before filtering r’(t, a) = c(t) x(t, a) + n’(t) c(t ) Ps Pd (t ) , Ps Pd 1, K Ps Pd Received signal after filtering r(t, a) = c(t) x(t, a) + n(t) Received signal phase (t, a) = (t, a) + (t ) 7/12/2006 10/24 SO-SDDPD Detector finds the phase difference between successive symbol intervals k (k (tk ) (tk T )) mod 2 We assume that GFSK pulse shape causes adjacent symbol interference k (ak0 ak 11 ak 11 ) mod 2 i h iT T g (t )dt iT The phase difference space from 0 to 2 is divided into R sub-regions Detector selects the sub-region Dk in which The sequence of phase regions (D0, DI, …) is sent to a branch metric calculator 7/12/2006 k lies 11/24 SO-SDDPD Let ( oi , 1i ,...) be the phase differences corresponding to any transmitted sequence ( aoi , a1i ,...) (P(Do | 0i ), P( D1 | 1i ),...) A branch metric calculator finds the conditional probabilities Branch metrics sent to a 4-state MAP decoder whose state transition is from Sk 1 ak 1 , ak to Sk ak , ak 1 The SO-SDDPD estimates the LLR for code bits 7/12/2006 12/24 Capacity Under Modulation, Channel And Receiver Design Constraints Channel capacity denotes maximum allowable data rate for reliable communication over noisy channels C max I ( X ; Y ) p( x) p ( x, y ) dxdy p( x) p( x) p( y ) In any practical system, the input distribution is constrained by the choice of modulation C max p ( x, y ) log 2 – Capacity is mutual information between the bit at modulator input and LLR at detector output C I ( X ;Y ) Constrained capacity in nats is; [Caire, 1998] C E[log(2) log p(bi | r )] 7/12/2006 13/24 Capacity Under Modulation, Channel And Receiver Design Constraints Constrained capacity for the proposed system is now C log 2 M log(2) E a ,c , n , s s ' i 1 In bits per channel use C log 2 M 7/12/2006 log 2 M i 1 1 Ea,c,n,s s ' [log{exp(0) exp( zi (1)bi )}] log(2) Constrained capacity hence influenced by – – – – [log{exp(0) exp( zi (1)bi )}] Modulation parameters (M, h and BT) Channel Detector design Computed using Monte-Carlo integration The constrained capacity is used to find the minimum Eb/No required for reliable signaling 14/24 Capacity Under Modulation, Channel And Receiver Design Constraints Scenario: BICM capacity under constraint of using the SOSDDPD 2 M = 4, h = 0.21, BT = 0.2 1.8 SDDPD specifications: R=26 uniform sub-regions for 4-GFSK 1.6 C (bit s/ channel use) 1.4 Channel specifications: Rayleigh 1.2 GFSK specifications : M =4, h =0.21, BT =0.2, 2B99Tb =0.6 with Rc =2/3 1 0.8 min{Es/No} if found at C=Rclog2M 0.6 min{Eb/No} = min{Es/No} /C 0.4 0.2 0 -10 0 10 20 E / N (dB) s 7/12/2006 30 40 50 o 15/24 Optimum Combination of Code Rates And GFSK Parameters In An Ergodic Channel The search space is – – – – At a particular Rc – – – M ={2, 4}- GFSK Rc ={6/7, 5/6, 3/4, 2/3, 1/2, 1/3, 1/4, 1/5} BT ={0.5, 0.25, 0.25} 2B99Tb ={0.4, 0.6, 0.8, 0.9, 1.0, 1.2} Find h for each value of BT and M that meets a desired 2B99Tb Find min{Eb/No} for all allowable combinations of M, h, BT at every 2B99Tb At each 2B99Tb, select GFSK parameters yielding the lowest min{Eb/No} Select the combination of Rc and GFSK parameters that have the lowest min{Eb/No} at the desired 99% bandwidth 7/12/2006 16/24 Optimum Combination of Code Rates And GFSK Parameters In An Ergodic Channel Scenario: Information theoretic minimum Eb/No at different 2B99Tb with Rc =5/6 24 Inform a t ion t heoret ic m inim um Eb/ No (dB) M = 2, BT = 0.5 M= 2, BT = 0.25 22 SDDPD specifications: R=40 uniform sub-regions for 2-GFSK R=26 uniform sub-regions for 4-GFSK M = 2, BT = 0.2 20 18 M =4, BT =0.5 Channel specifications: Rayleigh M= 4, BT = 0.25 0.14 M = 4, BT = 0.2 Search specifications: At each 2B99Tb, there are 6 combinations of M, h and BT 16 0.26 14 0.33 0.29 12 The numbers denote h values corresponding to GFSK parameters with the lowest min{Eb/No} at the particular bandwidth efficiency 0.48 0.7 10 0.4 0.5 0.6 0.7 0.8 2B T 0.9 1 1.1 1.2 1.3 At 2B99Tb =1.2, selecting M =2, h =0.7 and BT =0.25 yields the lowest min{Eb/No} 99 b 7/12/2006 17/24 Optimum Combination of Code Rates And GFSK Parameters In An Ergodic Channel Scenario: Best GFSK parameters for various code rates at 2B99Tb =0.9 24 M = 4, BT = 0.5, h = 0.35 M = 4, BT = 0.5, h = 0.33 Informat ion t heoret ic minimum Eb/ No (dB) 22 20 SDDPD specifications: R=40 uniform sub-regions for 2-GFSK R=26 uniform sub-regions for 4-GFSK M = 4, BT = 0.5, h = 0.285 Rayleigh M = 4, BT = 0.5, h = 0.24 M = 4, BT = 0.5, h = 0.14 18 M = 4, BT = 0.5, h = 0.07 Channel specifications: AWGN, Rayleigh M = 4, BT = 0.5, h = 0.046 16 M = 4, BT = 0.25, h = 0.05 14 12 Search specifications: The combination of code rates and GFSK parameters with lowest min{Eb/No} can be identified at the particular 2B99Tb At 2B99Tb =0.9: M =4, h =0.24, BT =0.5 with Rc =2/3 (Rayleigh) M =4, h =0.285, BT =0.5 with Rc =3/4 (AWGN) yield the best energy efficiency AWGN 10 8 6 4 0.2 7/12/2006 0.3 0.4 0.5 0.6 Code rat e 0.7 0.8 0.9 1 Notice the trade-off between code rate and energy efficiency 18/24 Combination of Code Rates And GFSK Parameters Rayleigh Fading 7/12/2006 2B99Tb Rate M BT h min{Eb/No} dB 0.4 3/4 4 0.2 0.195 18.15 0.6 2/3 4 0.2 0.21 18.08 0.8 3/4 4 0.5 0.25 12.38 0.9 2/3 4 0.5 0.24 11.99 1.0 2/3 4 0.5 0.3 11.44 1.2 5/6 2 0.25 0.7 11.34 19/24 Combination of Code Rates And GFSK Parameters Rician Fading (K =6 dB) 7/12/2006 2B99Tb Rate M BT h min{Eb/No} dB 0.4 3/4 4 0.2 0.195 15.38 0.6 5/6 4 0.5 0.18 11.67 0.8 5/6 4 0.5 0.29 9.09 0.9 3/4 4 0.5 0.285 8.87 1.0 2/3 4 0.5 0.3 8.83 1.2 6/7 2 0.25 0.76 8.39 20/24 Conclusions BICM with a soft-output SDDPD is used for noncoherent detection of GFSK signals The Shannon capacity of BICM under modulation, channel and detector constraints is evaluated using Monte-Carlo integration The constrained capacity is used to identify combination of code rates and GFSK parameters with the best energy efficiency and outage probability at a desired spectral efficiency 7/12/2006 21/24 Future Work Extend the search space to include – M >4 – Different pulse shapes and signal bandwidths – Alternative receivers A smarter method to comb the search space – Evolutionary algorithm 7/12/2006 22/24 Performance In Block Fading In block-fading a is broken into F blocks, which are transmitted over independent channels Channel coefficient c(t) =c, remains constant for the entire duration of a block Instantaneous SNR of the bth block is b | c |2 Es No When code combining is used at the receiver, the instantaneous capacity for the entire code word is 1 C (1 , 2 ,..., F ) F F C ( ) b 1 b The information outage probability po [ F ] P[C (1 , 2 ,..., F ) Rc log 2 M ] 7/12/2006 23/24 Optimum Combination of Code Rates And GFSK Parameters In A Block Fading Channel 0 10 0 Scenario: Information outage probability with code combining in block fading at F =1 and F = 100 for SO-SDDPD based BICM at 2B99Tb =0.9 10 -1 10-1 10 F =100 Inform Outaage gePProba roba bilit Informaattion ion Out bilit y y F =1 10 10 At F =1, M =4, h =0.285, BT =0.5 with Rc =3/4 has the lowest information outage probability -2 -2 At F =100, M =4, h =0.24, BT =0.5 with Rc =2/3 has the lowest information outage probability M = 4, BT = 0.5, h = 0.35, R = 6/ 7 -3 c -3 M = 4, BT = 0.5, h = 0.35, R = 6/ 7 10 M = 4, BT = 0.5, h = 0.33, R = 5/ 6 c 10 The capacity based search also helps in identifying the combination of code rates and GFSK parameters with the lowest outage probability in block fading c M = 4, BT = 0.5, h = 0.285, R = 3/ 4 M = 4, BT = 0.5, h = 0.33, R = 5/ 6c 10 c -4 4, BT = 0.5, h = 0.24, MM = 4,= BT = 0.5, h = 0.285, R =R3/ 4= 2/ 3 c c M = 4, BT = 0.5, h = 0.24, R = 2/ 3 M = 4, BT = 0.5,h = 0.14, c R = 1/ 2, c 10 -4 10 -5 M = 4, BT = 0.5, h = 0.14, R = 1/ 2 c M = 4, B T = 0.5, h = 0.07, R = 1/ 3 M = 4, BT =g0.5, h = 0.07, R = 1/ 3c c 4, B= 0.5, T = h0.5, h = 0.046, MM = 4,= BT = 0.046, R = 1/R4 = 1/ 4 g c c MM = 4,= BT h = 0.05, = 1/R5 = 1/ 5 4, B= 0.25, T = 0.25, h = R0.05, c g -6 -5 10 10 -5 -10 00 c 10 5 20 10 30 EE b// N No (dB) (dB) b 7/12/2006 15 40 20 50 25 60 o 24/24