Comments
Transcript
Design and Performance of Rate Compatible
Design and Performance of Rate Compatible-SCCC Alexandre Graell i Amat†‡, Guido Montorsi‡, Francesca Vatta* † Universitat Pompeu Fabra. Barcelona, Spain ‡ Politecnico di Torino. Torino, Italy * Università di Trieste. Trieste, Italy NEWCOM, Department 1-SPW1 meeting ENSEA, April 28th, 2005 Motivations ■ Standard SCCC for high-rates: Outer Encoder P Inner Encoder Politecnico di Torino – Universitat Pompeu Fabra 2 Motivations ■ Standard SCCC for high-rates: High-rate Encoder ■ ■ P Inner Encoder If the interleaver size is fixed different information block sizes for different rates For very high rates, the increasing value of the outer code rate causes an interleaver gain penalty error floor Politecnico di Torino – Universitat Pompeu Fabra 3 Motivations ■ Standard Rate-compatible SCCC: Outer Encoder P Inner Encoder Pi ■ Rate-compatibility restricts puncturing to the inner encoder ■ In general, the rate of the inner encoder is restricted to be Ri 1 the overall code rate is at most Ro Politecnico di Torino – Universitat Pompeu Fabra 4 A new class of SCCC RC-SCCC u Outer Encoder Po P Inner Encoder Psi Ppi M U X ■ The inner code may be punctured beyond the unitary rate RSCCC may be greater than the outer code rate ■ Puncturing is split between systematic and parity bits: rs : systematic permeability rp : parity permeability Politecnico di Torino – Universitat Pompeu Fabra 5 A new class of SCCC ■ Performance depend on puncturing patterns Po,Psi,Ppi rs and rp should be properly selected ■ We propose design criteria of this new class of SCCC by deriving the upper bounds to the error probability Outer Encoder C’’o C’o Po P Psi C’i Inner Encoder M U X Ppi Politecnico di Torino – Universitat Pompeu Fabra 6 Upper bounds to the error probability ■ We obtain: ■ The dominant contribution to the error probability for (asymptotic with N) is the largest exponent of N, aM. Politecnico di Torino – Universitat Pompeu Fabra 7 Upper bounds to the error probability ■ For recursive inner encoder: and ■ h(aM): weight associated to the highest exponent of N Politecnico di Torino – Universitat Pompeu Fabra 8 Upper bounds to the error probability ■ We obtain: ■ ■ ■ ■ do’f: free distance of C’o do’’(do’f): minimum weight of C’’o code sequences corresponding to a C’o code sequence of weight do’f di’f,eff: effective free distance of C’i h(3)m: minimum weight of C’i sequences generated by weight 3 input sequences Politecnico di Torino – Universitat Pompeu Fabra 9 Upper bounds to the error probability Outer Encoder C’o Po C’’o P Psi C’i Inner Encoder ■ ■ ■ ■ M U X Ppi do’f: free distance of C’o do’’(do’f): minimum weight of C’’o code sequences corresponding to a C’o code sequence of weight do’f di’f,eff: effective free distance of C’i h(3)m: minimum weight of C’i sequences generated by weight 3 input sequences Politecnico di Torino – Universitat Pompeu Fabra 10 Upper bounds to the error probability ■ We obtain: ■ ■ ■ ■ do’f: free distance of C’o do’’(do’f): minimum weight of C’’o code sequences corresponding to a C’o code sequence of weight do’f di’f,eff: effective free distance of C’i h(3)m: minimum weight of C’i sequences generated by weight 3 input sequences Politecnico di Torino – Universitat Pompeu Fabra 11 Upper Bound to the error probability ■ Then, Pb(e) (asymptotic with respect to N): do’f even do’f odd ■ For large Eb/N0 BER performance is given by: Politecnico di Torino – Universitat Pompeu Fabra 12 Upper Bound to the error probability ■ Design considerations: Po should be chosen to optimize C’o distance spectrum ■ Psi and Ppi should be chosen so that h(am ) and hm are maximized ■ Ppi must be optimized to yield the best C’i IOWEF ■ Psi must be selected to optimize do’’(do’f ) Psi turns out to be interleaver dependent ■ Politecnico di Torino – Universitat Pompeu Fabra 13 Rate-compatible SCCC ■ We designed well-performing rate-compatible SCCC following the aforementioned considerations ■ ■ ■ Psi to optimize do’’(do’f ) Ppi to optimize Ci’ IOWEF We used a searching algorithm that works incrementally, fulfilling the rate-compatible restriction, so that the punctured positions for a given outer rate are also punctured for all higher rates. Politecnico di Torino – Universitat Pompeu Fabra 14 The SCCC Scheme u Rate-1/2 4 state Fix punct. P Rate-1/2 4 state Psi Ppi M U X do’f=3 do’f=4 constituent codes outer code puncturing Politecnico di Torino – Universitat Pompeu Fabra 15 Performance Bounds Bounds of Rate-2/3 SCCC for several rp N=200. Po,1 rp =2/30 rp =4/30 rp =6/30 rp =8/30 rp =10/30 1.E+00 1.E-01 1.E-02 FER 1.E-03 1.E-04 1.E-05 1.E-06 1.E-07 1.E-08 1.E-09 0 1 2 3 4 Eb/N0 5 6 7 8 9 Politecnico di Torino – Universitat Pompeu Fabra 16 Performance Bounds Bounds of Rate-2/3 SCCC for several rp N=200. Po,2 rp =2/30 rp =4/30 rp =6/30 rp =8/30 rp =10/30 1.E+00 1.E-01 1.E-02 FER 1.E-03 1.E-04 1.E-05 1.E-06 1.E-07 1.E-08 1.E-09 0 1 2 3 4 Eb/N0 5 6 7 8 9 Politecnico di Torino – Universitat Pompeu Fabra 17 Simulation Results Performance of Rate-2/3 SCCC for several rp N=200. Po,1 1.E+00 1.E-01 1.E-02 FER 1.E-03 1.E-04 rp=2/30. Simulation rp =2/30. Bound rp =4/30. Simulation rp =4/30. Bound rp =8/30. Simulation rp =8/30. Bound rp =10/30. Simulation rp =10/30. Bound 1.E-05 1.E-06 1.E-07 1.E-08 1.E-09 0 1 2 3 4 5 6 7 8 9 Eb/N0 Politecnico di Torino – Universitat Pompeu Fabra 18 Simulation Results Performance of Rate-2/3 SCCC for several rp N=2000. Po,1 1.E+00 rp=2/30. Simulation rp =2/30. Bound rp =4/30. Simulation rp =4/30. Bound rp =8/30. Simulation rp =8/30. Bound UMTS PCCC SCCC (VTC’01) 1.E-01 1.E-02 1.E-03 FER 1.E-04 1.E-05 1.E-06 1.E-07 1.E-08 1.E-09 1.E-10 1.E-11 0 1 2 3 4 5 6 7 8 9 10 Eb/N0 Politecnico di Torino – Universitat Pompeu Fabra 19 Simulation Results Performance of Rate-9/10 SCCC for several rp N=2000. Po,1 1.E+00 rp =4/222. Simulation rp =4/222. Bound rp =10/222. Simulation rp =10/222. Bound rp =16/222. Simulation rp =16/222. Bound UMTS PCCC 1.E-01 1.E-02 FER 1.E-03 1.E-04 1.E-05 1.E-06 1.E-07 1.E-08 2 3 4 5 6 7 8 9 10 Eb/N0 Politecnico di Torino – Universitat Pompeu Fabra 20 Simulation Results Performance versus rp for several Eb/N0 . R=9/10. N=2000. Po,1 1.E+00 Eb/N0=4dB 4.2dB 1.E-01 4.4dB 4.6dB 1.E-02 4.8dB FER 5dB 1.E-03 5.2dB 5.4dB 1.E-04 5.6dB 5.8dB 6.2dB 1.E-05 6dB 6.4dB 6.6dB Eb/N0=6.8dB 1.E-06 22/222 200 20/222 202 18/222 204 16/222 206 14/222 208 12/222 210 rp Router 10/222 212 8/222 214 6/222 216 4/222 218 Politecnico di Torino – Universitat Pompeu Fabra 2/222 220 21 Simulation Results FER Performance comparison. N=428 1.E+00 SCCC (10 it.) PCCC (8 it.) LDPC (50 it.) 1.E-01 R=9/10 FER 1.E-02 1.E-03 1.E-04 R=5/6 R=1/3 1.E-05 1.E-06 0 1 2 3 4 5 6 7 8 Eb/N0 Politecnico di Torino – Universitat Pompeu Fabra 22 Conclusions ■ Derived lower bound to the error probability of a new class of SCCC ■ Derived suitable design guidelines ■ Derived optimal Rate-compatible SCCC families ■ The proposed scheme offers good performance for low to moderate block lengths in a large range of rates ■ The interleaver gain for low rates is kept also in the case of heavy puncturing ■ This code structure has been proposed as a candidate coding scheme for ESA MHOMS Politecnico di Torino – Universitat Pompeu Fabra 23 Open Problems ■ Convergence analysis EXIT charts and Density Evolution Techniques are difficult to apply ■ We are open to cooperations with other groups!!! Politecnico di Torino – Universitat Pompeu Fabra 24