FEC Performance with ARQ and Adaptive Burst Profile Selection

The contribution compares different FEC schemes in terms of data throughput under different ARQ scenarios. System is assumed to implement adaptive modulation and coding, i.e. adaptive burst profile selection. Among several interesting conclusions, it is shown that the coding scheme specified in the current working document provides good support for systems without ARQ but is disadvantageous for systems that employ ARQ to obtain higher throughput. Purpose Assist 802.16 in finalizing 802.16ab working document. Notice This document has been prepared to assist IEEE 802.16. 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 grants a free, irrevocable license to the IEEE to incorporate material contained in this contribution, and any modifications thereof, in the creation of an IEEE Standards publication; to copyright in the IEEE’s name any IEEE Standards publication even though it may include portions of this contribution; and at the IEEE’s sole discretion to permit others to reproduce in whole or in part the resulting IEEE Standards publication. The contributor also acknowledges and accepts that this contribution may be made public by IEEE 802.16. Patent Policy and Procedures The contributor is familiar with the IEEE 802.16 Patent Policy and Procedures (Version 1.0) , including the statement “IEEE standards may include the known use of patent(s), including patent applications, if there is technical justification in the opinion of the standardsdeveloping committee and provided the IEEE receives assurance from the patent holder that it will license applicants under reasonable terms and conditions for the purpose of implementing the standard.” Early disclosure to the Working Group of patent information that might be relevant to the standard is essential to reduce the possibility for delays in the development process and increase the likelihood that the draft publication will be approved for publication. Please notify the Chair as early as possible, in written or electronic form, of any patents (granted or under application) that may cover technology that is under consideration by or has been approved by IEEE 802.16. The Chair will disclose this notification via the IEEE 802.16 web site . 2001-11-10 IEEE 802.16abc-01/60 1 FEC Performance with ARQ and Adaptive Burst Profile Selection Lingjie Li, Octavian Sarca Redline Communications Inc. Introduction This might look like “just another contribution on FEC performance”. The novelty of the present contribution is that, instead of analyzing FEC performance as a separate unit, FEC, modulation and ARQ are jointly analyzed as a system. The benchmark used in this analysis is the data throughput since maximizing the throughput is the final target of 802.16 FWA systems. The relationship between throughput and signal to noise ratio (SNR) is analyzed under three ARQ scenarios: • No ARQ, i.e. no retransmission. • Each data packet/fragment is allowed one retransmission (for minimal delay) • There is no limit on the number of retransmissions allowed for each data packet/fragment For all scenarios, we assume the system employs adaptive modulation and coding, i.e. it selects the burst profile (modulation and coding) such that it maximizes the throughput while maintaining the packet error rate (PER) under a preset threshold PER0. Simulations We considered the 6 burst profiles currently specified by 802.16ab working document for the 256-FFT OFDM mode, using concatenated RS/CC FEC scheme. For consistency, the parameters of the 6 burst profiles are shown in Table 1. Burst profile Modulation Uncoded Bytes per Symbol NUBPS Coded Bytes per Symbol NCBPS Overall Rate RS Code CC Rate 1 QPSK 24 48 1/2 (32,24,4) 2/3 2 QPSK 36 48 3/4 (40,36,2) 5/6 3 16QAM 48 96 1/2 (64,48,8) 2/3 4 16QAM 72 96 3/4 (80,72,4) 5/6 5 64QAM 96 144 2/3 (108,96,6) 3/4 6 64QAM 108 144 3/4 (120,108,6) 5/6 Table 1: Burst profiles using concatenated RS/CC scheme We considered also 6 additional burst profiles that use the same combinations of modulation and coding rate but with convolutional coding (CC) instead of concatenated coding. The parameters of these 6 burst profiles are shown in Table 2. Burst profile Modulation Uncoded Bytes per Symbol NUBPS Coded Bytes per Symbol NCBPS CC Rate 1 QPSK 24 48 1/2 2 QPSK 36 48 3/4 3 16QAM 48 96 1/2 2001-11-10 IEEE 802.16abc-01/60 2 4 16QAM 72 96 3/4 5 64QAM 96 144 2/3 6 64QAM 108 144 3/4 Table 2: Burst profiles using CC-only PER simulations All 12 burst profiles have been first simulated to obtain the PER(SNR) curves for a super-packet of 864 bytes (smallest common multiple of all block sizes). Then the PER was scaled to a packet size of 128 bytes, which is the size of the ARQ block according to the current 802.16ab working document. The results for all 12 burst profiles are shown in Figure 1. 0 5 10 15 20 25 30 10 -6 10 -5 10 -4 10 -3 10 -2 10 -1 10 0 PER vs. SNR scaled to 128 byte packets SNR [dB] P ac k et E rr o r R a te ( P E R ) QPSK RS/CC 1/2 QPSK RS/CC 3/4 16QAM RS/CC 1/2 16QAM RS/CC 3/4 64QAM RS/CC 2/3 64QAM RS/CC 3/4 QPSK CC 1/2 QPSK CC 3/4 16QAM CC 1/2 16QAM CC 3/4 64QAM CC 2/3 64QAM CC 3/4 Figure 1: PER vs. SNR for 128-byte packets We used several ideal parameters to obtain the maximum achievable performance for both coding schemes and to facilitate better statistical estimation. Here are the parameters used for this simulations: • More than 10 data bits used to estimate PER=10 for each burst profile • AWGN channel • SNR does not include losses due to guard interval and pilots • Ideal channel equalization • Perfect bit interleaving • Soft-decision Viterbi decoding for both coding schemes with floating-point precision • Tail-biting block equal to the size of an OFDM symbol • Length of the trace-back memory larger that the symbol size 2001-11-10 IEEE 802.16abc-01/60 3 • RS decoder with perfect erasures • Perfect erasures generated by comparing the Viterbi output with the transmitted bit stream (very ideal setup). Throughput without ARQ A system without ARQ has no chance to improve the PER provided by the physical layer. Such a system will maximize the throughput by choosing the highest burst profile that meets PER < PER0 for the given SNR. For each burst profile, the throughput for a given SNR can be computed from PER(SNR) as: NBPS(SNR) = NUBPS (1-PER(SNR)) if PER(SNR) < PER0 = 0 otherwise where NBPS is the average number of bytes transmitted successfully in one OFDM symbol after ARQ and NUBPS is the number of uncoded bytes per symbol provided by the corresponding burst profile. The overall system throughput for a given SNR is the maximum across throughputs for all burst profiles. Throughput with one retransmission A system with ARQ that allows only one retransmission can improve significantly the PER provided by the physical layer. We assume that retransmissions are not constrained to use the same burst profile as the first transmission. Let PER1(SNR) and NUBPS1 be the PER and number of uncoded bytes per symbol for the burst profile used for the first transmission and let PER2(SNR) and NUBPS2 be the same for the retransmission. Then, the overall packet error rate is PER1(SNR)⋅PER2(SNR) and the overall throughput is given by: NBPS(SNR) = (1 – PER1(SNR)⋅PER2(SNR))⋅NUBPS1/(1+PER1(SNR)⋅NUBPS1/NUBPS2)... if PER1(SNR)⋅PER2(SNR) < PER0 = 0 otherwise where NBPS is again the average number of bytes transmitted successfully in one OFDM symbol. The overall system throughput for a given SNR is the maximum across throughputs for all possible pairs of burst profiles. Throughput with arbitrary number of retransmissions A system with ARQ that allows an unlimited number of retransmissions can theoretically improve the PER provided by the physical layer to the point that the overall PER = 0. We let again retransmissions use a different burst profile than the first transmission but, for simplicity, we assume that all retransmissions use the same burst profile. Let PER1(SNR) and NUBPS1 be the PER and number of uncoded bytes per symbol for the burst profile used for the first transmission and let PER2(SNR) and NUBPS2 be the same for the retransmissions. Then, if k retransmissions are allowed, the overall packet error rate is PER1(SNR)⋅PER2 (SNR) and the overall throughput is given by: NBPS(SNR) = (1 – PER1(SNR)⋅PER2 k(SNR))⋅NUBPS1/(1+PER1(SNR)⋅(1PER2 (SNR))/(1-... ...PER2(SNR))⋅NUBPS1/NUBPS2) if PER1(SNR)⋅PER2 (SNR) < PER0 = 0 otherwise where NBPS is again the average number of bytes transmitted successfully in one OFDM symbol. If unlimited number of retransmissions are allowed (i.e. k → ∞), the overall packet error rate can be made arbitrarily small and the overall throughput is given by: NBPS(SNR) = NUBPS1/(1+PER1(SNR)/(1-PER2(SNR))⋅NUBPS1/NUBPS2) In both cases (k is finite and k is infinite), the overall system throughput for a given SNR is the maximum across throughputs for all possible pairs of burst profiles. 2001-11-10 IEEE 802.16abc-01/60 4 Results For both ARQ scenarios, we have simulated the following four combinations: • First transmission and retransmission(s) use RS/CC • First transmission uses RS/CC and retransmission(s) use CC-only • First transmissions and retransmission(s) use CC-only • First transmission uses CC-only and retransmission(s) use RS/CC The target PER for all simulations was PER0 = 10 . Figure 2 shows the computed throughput for all four combinations with the ARQ limited to a single retransmission. We note that, when CC-only is used for the first transmission, the system provides a higher throughput, regardless of the coding used for retransmissions. The advantage provided by the CC-only scheme can be as high as 1.2dB (SNR=8-10dB). We also note that there is some advantage (less than 0.3dB around SNR=5dB and less than 0.1dB around SNR=17dB) in using RS/CC scheme for retransmissions. Figure 3 shows the computed throughput for all four combinations without any limit on the number of retransmissions. We note again that that, when CC-only is used for the first transmission, the system provides a higher throughput, regardless of the coding used for retransmissions. The advantage provided by the CC-only scheme can be as high as 1.2dB (SNR=8-10dB). We also note that for SNR>4dB there is little difference between using CC or RS/CC as the coding for retransmissions. There is just a little advantage (less than 0.1dB) towards using CC-only in several places. However, for SNR<4dB, the CC-only scheme offers tremendous advantages (up to 3dB) when used for retransmissions. Another interesting analysis is to see how much throughput improvement the ARQ can bring to the system. Figure 4 and Figure 5 compare the throughput with and without ARQ for both coding schemes RS/CC and CConly (same scheme used for first transmission and for retransmissions). Figure 4 shows the results for a single retransmission and Figure 5 shows the results for multiple retransmissions. We note that ARQ improves the SNR with up to 1.5dB for the RS/CC scheme and up to 4dB for the CC-only scheme. Overall the best system without ARQ is up to 2dB worse than the best system using ARQ. Looking closely at Figure 2 and Figure 3 we see little difference between using one retransmission and multiple retransmissions. This is because we allow the retransmission use a lower burst profile than the one used for the first transmission. Regardless of how high is the PER on the first transmission, the system can choose the profile for the retransmission enough low such that the overall PER is smaller than PER0. This is true as long there is a lower burst profile to choose from, and thus we see a significant difference only for SNR < 4.2dB. More details about this subject can be found in the next section. Another interesting remark is that systems without ARQ suffer a stair-like degradation of the throughput with the SNR. Therefore, in order to accommodate channel variations, a larger link margin must be used. Systems using ARQ with concatenated RS/CC scheme have a similar stair-like degradation and thus they require also a large link margin. However, systems with ARQ and CC-only scheme have a very smooth throughput degradation with the SNR. Such systems may operate properly with a much smaller link margin. 2001-11-10 IEEE 802.16abc-01/60 5 0 2 4 6 8 10 12 14 16 18 20 0 20 40 60 80 100 120 Throughput vs. SNR, PER<=1e-6, single SNR [dB] A vg . by te s /s ym bo l 1st RS/CC + 2nd RS/CC 1st RS/CC + 2nd CC 1st CC + 2nd RS/CC 1st CC + 2nd CC Figure 2: Throughput with RS/CC and CC for single retransmission 0 2 4 6 8 10 12 14 16 18 20 0 20 40 60 80 100 120 Throughput vs. SNR, PER<=1e-6, multiple SNR [dB] A vg . by te s /s ym bo l 1st RS/CC + 2nd RS/CC 1st RS/CC + 2nd CC 1st CC + 2nd RS/CC 1st CC + 2nd CC Figure 3: Throughput with RS/CC and CC for multiple retransmissions 2001-11-10 IEEE 802.16abc-01/60 6 0 2 4 6 8 10 12 14 16 18 20 0 20 40 60 80 100 120 Throughput vs. SNR, PER<=1e-6 SNR [dB] A vg . by te s /s ym bo l no ARQ, 1st RS/CC no ARQ, 1st CC single, 1st RS/CC + 2nd RS/CC single, 1st CC + 2nd CC Figure 4: Throughput with and without ARQ for single retransmission 0 2 4 6 8 10 12 14 16 18 20 0 20 40 60 80 100 120 Throughput vs. SNR, PER<=1e-6 SNR [dB] A vg . by te s /s ym bo l no ARQ, 1st RS/CC no ARQ, 1st CC multiple, 1st RS/CC + 2nd RS/CC multiple, 1st CC + 2nd CC Figure 5: Throughput with and without ARQ for multiple retransmissions 2001-11-10 IEEE 802.16abc-01/60 7 Comments and explanations In order to understand why the CC-only performs better than RS/CC with ARQ and reversely without ARQ, let us imagine that the system starts operating with an excellent (very high) SNR and that the SNR slowly degrades. Systems without ARQ Having an excellent SNR, the adaptive modulation and coding will initially select the highest burst profile, in our case 6, to maximize the system throughput. If the SNR degrades to the point that the burst profile 6 does not meet anymore the PER(6,SNR)5%. 2001-11-10 IEEE 802.16abc-01/60 8 0 2 4 6 8 10 12 14 16 18 20 0 1 2 3 4 5 6 7 Burst profile vs. SNR, 1st CC + 2nd CC PER<=1e-6, single SNR [dB] B u rs t pr of ile First transmission Retransmission 0 2 4 6 8 10 12 14 16 18 20 0 1 2 3 4 5 6 7 Burst profile vs. SNR, 1st RS/CC + 2nd RS/CC PER<=1e-6, single SNR [dB] B u rs t pr of ile First transmission retransmission Figure 6: Burst profiles vs. SNR for PER0 = 10 , single retransmission 0 2 4 6 8 10 12 14 16 18 20 10 -2 10 -1 10 0 First transmission PER vs. SNR, 1st CC + 2nd CC PER<=1e-6, single SNR [dB] P a c ke t E rr o r R a te ( P E R ) 0 2 4 6 8 10 12 14 16 18 20 10 -2 10 -1 10 0 First transmission PER vs. SNR, 1st RS/CC + 2nd RS/CC PER<=1e-6, single SNR [dB] P a c ke t E rr o r R a te ( P E R ) Figure 7: First transmission PER vs. SNR for PER0 = 10 , single retransmission 2001-11-10 IEEE 802.16abc-01/60 9 SNR [dB] Burst profile 1st transmission Burst profile retransmission PER [%]