Modified EESM Link Adaptation Method with Multiple Constellation for Future Wireless Networks

Transcription

1 Communication Technology, Vol, Issue 7, July-3 ISSN(Online) ISSN (Print) Modified EESM Link Adaptation Method with Multiple Constellation for Future Wireless Networks Sandanalakshmi R, Kadhiravan D Department of Electronics and Communication Engineering, Pondicherry Engineering College, Puducherry. sandanalakshmi@pec.edu Department of Electronics and Communication Engineering, Pondicherry Engineering College, Puducherry. redphoenixkaran@gmail.com Abstract Wireless communication systems require high spectral efficiency and throughput in order to be money-spinning. Link adaptation schemes are known to be the best solution to achieve this aspiration. In customary link adaptation scheme for OFDM, a look-up table is maintained at the system level and best Modulation and Coding Scheme (MCS) is chosen based on the link generalization information from the link level simulator. One of the well known link generalization method is Exponential Effective SNR Mapping (EESM), where the received SNR vectors at the receiver are being abstracted to a single metric, SNR eff that represents the link quality. One Channel State Information (CSI) representing the whole group of sub-carriers over a time varying fading channel may lead to be better or worse for few sub-carriers in the group and reduces the Bit Error Rate. Hence multiple feedbacks and multiple constellations may within one group of sub-carriers may improve the performance and utilization of the channel. Modified EESM approach is proposed in the paper with multiple feedbacks and constellations to improve the BER performance. Improvement in the throughput performance and reduction in feedback overhead has been shown as performance metrics. KEYWORDS : EESM, Link Adaptation, CSI.. INTRODUCTION Innovative progress in Wireless Communication Technologies during the last decade has had a major brunt on modern society and cost-cutting measure. While early Wireless Communication Systems focused on low data rate services like voice services, current Wireless Communication Systems are focused on high data rate service like multimedia services. Moreover, the applicable range of Wireless Data Communication Systems has lengthened from Single Cellular Systems to Multi-cellular systems. Data communication technologies which have been mainly deployed in Wireless Local Area Network (WLAN) are now expected to be used in multi-cellular environments. The key prerequisite to achieve the required high data capability is the increase of spectral efficiency over wireless channels that suffer from fading. Orthogonal Frequency Division Multiplexing (OFDM) a multicarrier data transmission technique is one acknowledged solution for high data capability in Wireless Communication Systems. OFDM has been applied in many Wideband Digital Communications such as Digital Audio Broadcasting (DAB), High Definition Television (HDTV) WLAN such as HIPERLAN/, IEEE 8.a, IEEE 8.g and IEEE 8.6 Broadband Wireless Access system. This multicarrier modulation scheme has been an immense success in WLAN systems and has been adopted as the chief transmission technology over various fourth generation wireless communication standards. Transmission of Channel State Information (CSI) between transceivers is one of key predicament in meeting the requirement of elevated spectral efficiency in OFDM systems in order to be moneyspinning. By using proper CSI, spectral efficiency in OFDM systems can be effectively enhanced by a Link Adaptation scheme. Moreover, a Link Adaptation scheme in concurrence with CSI can also be exploited to amplify multiuser diversity gain with multiuser scheduling. However, the prerequisite of the exact CSI for link adaptation schemes normally dictate a large amount of feedback that averts practical implementation because Wireless Communication is likely to take place in a restricted feedback environment. Furthermore, implementation is also restricted due to the additional complexity requirements, as optimally distribute resources, such as power, frequency, and modulation levels. In this perspective, recent investigation is focused on the realistic constraints for wireless OFDM systems, particularly feedback sum and implementation complexity. In addition, investigation has been particularly focused on performance improvement Page 366

2 Communication Technology, Vol, Issue 7, July-3 ISSN(Online) ISSN (Print) in the Low Signal-to-Noise Ratio (SNR) or Signalto-Interference-plus-Noise Ratio (SINR) regimes, which largely determine the performance of systems.. RELATED WORK Halil Yigit and Adnan Kavak [], discussed the essentials and the necessity for Link Adaptation in Wireless Networks. They proposed the key idea of performing a Link Adaptation using a Neural Network. They considered a simple x Multi Input Multi Output (MIMO) OFDM model in a frequency selective channel and employed the instantaneous SNR report to the transmitter using neural network. They also have discussed the significance of regression achievement in a Neural Network. They were able to accomplish a regression of.5 to.6. Peng F. et al., [3] have simulated the basic Link Adaptation scheme based on the instantaneous Packet Error Rate (PER) and Instantaneous Signal to Noise Ratio (SNR) report. They have compared the throughput performance for both Frequency Flat channel and Frequency fast selective channel. Their spectral efficiency plot shows the significance and necessity for feed back over head diminution in Channel State Information (CSI) report. Daniels R. C. et al.[] have developed a Neural Network using k-nearest neighbor algorithm. They demonstrated exceptional performance in frequency selective channels (uniform power delay profile). Perhaps the most valuable observation was that this algorithm was easily extendable to diverse channel models without retraining the system. Since their choice of the training set is universal and because their feature set does not utterly depend on the channel distribution, this approach is capable and well suit for real systems. 3GPP TSG RAN WG #37 [,6], presented an exponential ESM (Effective SIR Mapping) function, as a candidate ESM for the system-level evaluation e.g. OFDM performance. It has been shown that a comprehensive exponential ESM, including an MCS-dependent parameter, can provide good accuracy in case of QPSK modulation. For 6QAM modulation the exponential ESM is less truthful. However, by a suitable parameter selection, the exponential ESM could provide an, optimistic estimate of the OFDM performance, also in case of 6QAM modulation. Guowang Miao, Zhisheng Niu [5] focused on realistic feedback and adaptive signaling designs of OFDM systems. Based on the investigations of the characteristics of frequency selective channels, the Dynamic Neighboring Subcarrier Grouping Scheme (DNSGS) was proposed to cut down the feedback and adaptive signaling overhead, and to abridge the implementation of OFDM adaptive techniques. Then the capacity of DNSGS based OFDM systems was also given. The simulation results impressively show that DNSGS is flexible enough for diverse nature of communication environments, and can in effect cut down the feedback and signaling overhead with a certain extent limited performance penalty. Besides, DNSGS outperforms the on hand grouping schemes greatly. 3. EESM 3. EESM principle EESM is used to chart the instantaneous values of SINRs to the consequent Bit Error Rate (BER) value []. Although EESM was introduced to work with Signal to Interference Ratio (SIR), it also works with SNR as well. EESM is a simple mapping scheme used when all the subcarriers of a particular subscriber are modulated using the same Modulation and Coding Scheme (MCS) level. The essential initiative of EESM is to discover a compression function that maps the set of SINRs to a single value that is a first-rate predictor of the actual BER. Figure shows the focal intention behind using EESM function. Note that average SINR is not a first-rate predictor of actual BER. EESM is a channel reliant principle that maps power level as well as MCS level to SINR values in the AWGN channel domain. Such function allow its mapping along with AWGN hypothesis (such as consequence of augment in power, CINR/MCS threshold tables) to predict the consequence of MCS and make better amendments. The scheme has been publicized to acquiesce an truthful estimation of the AWGN equivalent SINR (from now on referred to as effective SINR.) for frequency selective channels. Figure. Principle of EESM In case of Multi-carrier transmission as in WiMAX, the set of subcarrier SINRs are mapped with the help of EESM formula, into a scalar instantaneous effective SINR value. A guesstimate of the BER value is then attained, using the effective SINR value, from fundamental AWGN Link-level performance. The mapping of the Page 367

3 Communication Technology, Vol, Issue 7, July-3 ISSN(Online) ISSN (Print) effective SINR value to the consequent BER value will use either a look-up table for the mapping function or use a fairly accurate investigative expression if available. The EESM method guesstimate the effective SINR using the following formula [] N EESM(, ) ln i e N / eff () Where, γ is a vector [γ, γ,., γ N ] of the persubcarrier SINR values, which are different in a frequency selective channel. β is the parameter to be determined for every Modulation Coding Scheme level, and this value is used to regulate EESM function to recompense the divergence between the actual BER and the predicted BER. 3. EESM unequal MCS using FSFA The customary EESM abstracts all the received SNR vectors into a single vector branded as effective SNR. This method of generalization may be truthful but have an effect on the throughput performance to a certain level. So, in order to augment the throughput performance supplementary, the perception of grouping of subcarriers shall be considered [5]. The initiative is to cluster the received SNR vectors into three diverse groups that is to say A, B, and C. Cluster A consists of the SNR vectors of superlative strength. Cluster B consist of SNR vectors of fine strength. Cluster C consist of the SNR vectors of feeble strength. Here it should be renowned that the clusters are made up to the fitting FFT sizes at the transmitter. For each cluster the EESM perception is applied and a separate effective SNR is being calibrated and then reported to the transmitter. The transmitter then come to a decision an fitting MCS for each group. Cluster C may be a no transmission mode if the sub carriers are found to be below feeble level. If stipulation arises the same cluster shall be assigned as the lowest order modulation. The following Figure shows the perception of EESM based clustering of subcarriers. 3.3 Signaling Information The conventional EESM scheme follows equal constellation for all the subcarriers. Hence there is no necessity for reporting the individual subcarrier potency information. In EESM unequal modulation comes the perception of clustering of subcarriers where the transmitter have to be intimated the subcarrier indexing information and SNR threshold of each group so that the transmitter shall carry out the unequal constellation assignment to each cluster. Hence a practical condensed feedback scheme has to be considered. Realistic Feedback design based OFDM Link Adaptive Communications over Frequency Selective Channels [5] has been incorporated here. The conventional OFDM systems report to 6 bits per subcarrier, where the proposed scheme reports only the indexing information in bits per subcarrier. 3. EESM Implementation in a Neural Network This section converse the technique used to develop a Neural Network and working out it to carry out EESM based Link Adaptation. The Network formation for EESM Link Adaptation consist of two stages in which the stage work out the Effective SNR value and the stage make a decision, the superlative MCS based on effective SNR. The subsequent Figure 3 gives an idea about the Exact Radial Basis Neural Network structure that was developed and trained for Link adaptation using EESM method. The learning rate was set to.5 and the momentum was set to.5. Within the Training Criteria module, pattern selection was set to random. Within the training module, the network was set to involuntarily save the weights for the superlative test set. The network was then trained within the Training module. The trained network was then applied to the 5 sample experiment patterns at the outset and later simulated with the full input set that was used for training. Figure. Perception of EESM based clustering Page 368

4 Communication Technology, Vol, Issue 7, July-3 ISSN(Online) ISSN (Print) Effective SNR Calibrstion Matlab Simulated Neural Network Simulated Effective SNR Figure 5. Effective SNR calibration by neural network over channel instant Figure 3. EESM Neural Network structure. PERFORMANCE EVALUATION A SISO OFDM system with 8 subcarriers in an AWGN and Rayleigh channel environment is considered. The simulation platform used is MATLAB 9b. - - QAM-AWGN QAM-Rayleigh 6QAM-AWGN 6QAM-Rayleigh 6QAM-AWGN 6QAM-Rayleigh Figure. illustrate OFDM bit error performance plot under AWGN channel and Rayleigh Fading channel circumstances for the modulation schemes QAM, 6 QAM, and 6 QAM. The AWGN performance curve acts as the reference for mapping the BER which is realized at real time channel. Figure 5 illustrates the effective SNR calibration MATLAB simulated vs. Neural Network simulated for 5 time instants. Figure 6. illustrates the EESM link adaptation simulation result realized by MATLAB and Neural Network for 5 time instants. The adaptation is conceded out between MCS schemes which are QAM (/ coded), QAM uncoded, 6 QAM, 6 QAM. At any time instant the MCS that gives the superlative throughput is selected first, provided the target Bit Error Rate is ( - ). A regression of.99983, surmise that a very truthful performance is achieved by the Neural Network. BER Actual Neural predicted SNR in db Figure. BER vs. SNR plot for Rayleigh channel Modulation order > QAM -> 6 QAM 3 -> 6 QAM Figure 6. EESM based link adaptation using neural network MCS selection over channel instant. Figure 7. illustrate the clustering of subcarriers for 8 subcarriers with three cluster levels such as cluster A, cluster B, cluster C. Page 369

5 Communication Technology, Vol, Issue 7, July-3 ISSN(Online) ISSN (Print) Cluster A is the superlative cluster with FFT size of 5 subcarriers and cluster B is the good quality cluster with FFT size of 5 subcarriers and cluster C is of no transmission. If stipulation arise cluster C shall also be made use of. A regression of.853 was accomplished > QAM(/) -> QAM 3 -> 6 QAM -> 6 QAM Group actual Neural Simulated Group actual Neural Simulated Group No Group No Group No Signaling information for grouping of subcarriers-matlab Simulated 6 8 Subcarriers Signaling information for grouping of subcarriers-neural Network simulated 6 8 Subcarriers Rounded Neural network result 6 8 Subcarriers MCS Selection Figure 8. EESM based link adaptation using neural network for unequal constellation Figure 7. Signaling information for EESM unequal modulation The Figure 8. illustrates the EESM based Link Adaptation using Neural network for unequal constellation. The plot gives an idea about the MCS selection comparison by Matlab simulation and Neural Network simulation. Figure 9. demonstrates the throughput performance of conventional OFDM systems vs. EESM Link Adapted system. For conventional system there is a perimeter in throughput realization even when the SNR value increases. But in case of EESM Link Adapted system, the MCS scheme is toggled from one to another based upon the SNR upgrading so that the throughput accomplished equivalent to the higher order modulation scheme is achieved. Here in this figure the colored lines shows the conventional system throughput where the black circular marked line shows the Link Adapted throughput enhancement. Figure 9. EESM throughput performance comparison over conventional OFDM system BER - - qam(/) qam 6qam 6qam Link Adapted s Figure. EESM error performance comparison over conventional OFDM system Page 37

6 Communication Technology, Vol, Issue 7, July-3 ISSN(Online) ISSN (Print) Throughput x Link Adapted qam(/) qam 6qam 6qam whichever time instant the subcarriers are clustered into three clusters A, B, and C. Cluster A and cluster B are prearranged separate MCS while group C is considered as no transmission mode. The figure also infers that a throughput average of.5 Mbps was accomplished through the proposed EESM unequal modulation Link Adaptation procedure, while the conventional EESM equal modulation gave a throughput of only Mbps and Mbps at infrequent time instants respectively. EESM unequal modulation scheme gives a superior throughput performance than matched up to that of the conventional EESM procedure s Figure. Throughput performance of EESM over channel instant Throughput in bps x 6 LA TP Fixed QAM Fixed 6 QAM Fixed 6 QAM Fixed QAM(/) Figure. Throughput performance comparison traditional vs. EESM unequal modulation Figure. illustrates the error performance of EESM over conventional OFDM system. Figure. illustrates the throughput performance of EESM Link Adapted system for time instants (channel instants). The system become accustomed to any one of the MCS schemes (/ convolutionally coded QAM, uncoded QAM, uncoded 6 QAM, and uncoded 6 QAM) for every time instant depending upon the channel state information conditional from the effective SNR value sent by the receiver. Likewise, for the next set of time instants the superlative MCS scheme that which gives the maximum throughput is chosen endow with it fulfills the target BER requirement. Figure. illustrates the throughput accomplish by EESM unequal modulation by clustering of subcarriers over the conventional fixed system (equal modulation) throughput. At Figure 3. illustrates the feedback overhead diminution in bits by clustering of subcarriers over the conventional instantaneous SNR feedback scheme for OFDM system with subcarriers 5, 8, 56, 5,, 8 respectively. In conventional scheme for every instantaneous subcarrier a feedback overhead of to 6 bits is sent, whereas in clustering of subcarriers an overhead of bits per subcarrier intimating that in which the subcarrier falls with group threshold is sent. This surmise that there is a diminution of in the region of 5% overhead over the conventional feedback scheme, which in turn also infers condensed control channel bandwidth consumption. The cluster division is done as three clusters which are feeble, good quality and superior. The group size is kept to be least feasible FFT size. For case in point consider an OFDM system with three clusters A, B, C. Group A has maximum FFT size of 8 and group B has minimum FFT size of 7. Group C is considered as no transmission mode where there is no data bits transmitted. Overhead bits Overhead comparison Conventional vs Grouping Conventional Feedback Feedback using grouping of subcarriers No of subcarrriers Figure 3. Feed back overhead reduction traditional vs. EESM unequal modulation 5. CONCLUSION The realm of this project work was to investigate various Link adaption algorithms for the IEEE Page 37

7 Communication Technology, Vol, Issue 7, July-3 ISSN(Online) ISSN (Print) n standard which selects with dynamism a suitable MCS for the in progress wireless channel realization. The Link Adaption algorithm decides on a suitable MCS for the in progress channel realization which renders high throughput while maintaining a certain target BER ( - ). The appropriate MCS is recommended to the transmitter via the Link Adaption protocol defined in the standard IEEE 8.n. In this perspective, EESM based Link Adaptation for equal and unequal constellation has been simulated using Matlab 9b platform. The Link Adaptation has been performed for four diverse MCS orders, QAM ½ convolutional coded, QAM, 6 QAM, & 6 QAM keeping a target Bit Error Rate of -. The throughput performance comparison has been shown for EESM equal and unequal constellation with conventional system. It is observed that EESM on Neural Network have attained upto.6 regression REFERENCES [] Halil Yigit, and Adnan Kavak, Adaptation using Neural Network in Frequency Selective MIMO-OFDM Systems, International Symposium on Wireless Pervasive Computing,. [] Daniels R. C., Caramanis C. M., and Heath R. W., A supervised learning approach to adaptation in practical MIMO-OFDM wireless systems, Proceedings of IEEE GLOBECOM, New Orleans, LA, USA, pp. 5. November 3 December, 8. [3] Peng F., Zhang J., and Ryan W.E., Adaptive modulation and coding for IEEE 8.n, Proceedings of IEEE WCNC, Kowloon, Hong Kong, pp , March. 5, 7. [] 3GPP TSG RAN WG #37, Modelling of Performance with Coloured Interference Using the EESM (Exponential Effective SIR Mapping), Montreal, Quebec, Canada, May -,. [5] Guowang Miao, and Zhisheng Niu, Practical Feedback Design based OFDM Link Adaptive Communications over Frequency Selective Channels, Proceedings of IEEE ICC, 6. [6] Krishna Sayana, Jeff Zhuang, Ken Stewart and Motorola Inc, Link Performance Abstraction based on Mean Mutual Information per Bit (MMIB) of the LLR Channel, Proceedings of IEEE 8.6 Broadband Wireless Access Working Group. Page 37