Çift Yönlü Röleli Kanallarda Uyarlamalı Eşleme

Abstract

Son yıllarda telsiz iletişim sistemleri için kaliteli ses iletimi yanında yüksek hızlı veri iletimi gereksinimi de artmaktadır. Artan bu gereksinimi karşılamak için iletişim sistemlerinin başarım oranları ve sığaları artırılmalıdır. Bu amaçla bu çalışmada çift-yönlü röle ağları ele alınmaktadır. Çalışmada ele alınan sistem iki uç düğüm ve aralarında röleden oluşmaktadır. İki uç birim birbirlerine röle üzerinden bilgi iletmektedir. Klasik durumda bu iş dört evrede, geleneksel ağ kodlama yönteminde üç evrede, bu tezde ele alınan fiziksel katman ağ kodlamada (PNC) ise iki evrede gerçekleşmektedir. PNC’de rölenin yarı dupleks çalışmasından kaynaklanan zaman kaybı giderilmektedir. Bu yöntemde, telsiz iletişim sistemlerinde oluşan bozucu etki girişimden yararlanarak sistem başarımı iyileştirilmektedir. Fiziksel katman ağ kodlama yönteminde kanal sönümleme durumları dikkate alınmadığı için derin sönümleme noktaları oluşur. Yeni önerilen uyarlamalı ağ kodlama yönteminde, uç düğümlerde kullanılan modülasyon tipine göre rölede oluşan derin sönümleme noktaları dikkate alınmaktadır. Kanal durumlarına göre, rölede kullanılan ağ kodlama matrisinin seçilmesi hata başarımını artırmaktadır. Özellikle Rician sönümlemeli kanallarda kanal parametresi K’nın yüksek değerleri için kodlama kazancı artmaktadır. Kanal durumlarına göre, rölede kullanılan ağ kodlama matrislerinin oluşturulması için Latin kareler yöntemi basit ve etkin bir yöntemdir. Literatürde uyarlamalı eşleme ile ilgili pek çok çalışma vardır. Ancak asimetrik hızlarda çalışan uç düğümlerden oluşan röle ağı için uyarlamalı eşleme ile ilgili çalışma bulunmamaktadır. Bu tez çalışmasının ilk kısmında, iki uç düğümün röle yardımıyla bilgi alışverişi yaptığı çift yönlü röleli kanallarda fiziksel katman ağ kodlama ve uyarlamalı eşleme yöntemi incelenmektedir. Sistemde QPSK modülasyonu ve çöz-aktar iletim tekniği kullanılmaktadır. Her düğümün ve rölenin yarı dupleks çalıştığı ve bir anteni olduğu varsayılmaktadır. Farklı Rician sönümleme parametreleri için her iki sistemin bilgisayar benzetim sonuçları karşılaştırılmaktadır. Çalışmanın ikinci kısmında uyarlamalı eşlemenin asimetrik hızlarda çalışan çift yönlü röleli kanallara uygulanması gerçekleştirilmektedir. İlk olarak iki uç düğümde BPSK ve QPSK modülasyonu ile uyarlamalı eşleme yönteminin kullanıldığı sistem için derin sönümleme noktaları ve kanal bölgeleri elde edilmektedir. Bu sistem için sönümleme noktaları ve çevrelerindeki bölgelerde hata başarımını iyileştiren ağ kodlama matrisleri Latin dikdörtgenler yardımıyla oluşturulmaktadır. Bilgisayar benzetimleri yardımıyla elde edilen hata olasılığı eğrileri incelendiğinde, Rician kanal parametresi K=10 iken 〖10〗^(-4) lük bir bit hata başarımını sağlamak için E_b⁄N_0 oranında uyarlamalı eşleme yöntemi PNC yöntemine göre 4.5 dB kazanç sağlamaktadır. Daha sonra uç düğümlerde BPSK ve 8-PSK modülasyonu ile uyarlamalı eşleme yönteminin kullanıldığı sistem için derin sönümleme noktaları ve kanal bölgeleri elde edilmektedir. Bu sistemler için Latin dikdörtgenler yardımıyla ağ kodlama matrisleri bulunarak hata başarımları bilgisayar benzetimleri yardımıyla değerlendirilmektedir. Benzetim sonuçları incelendiğinde, Rician kanal parametresi K=10 iken 〖10〗^(-3) lük bir bit hata başarımı sağlamak için uyarlamalı eşleme yöntemi PNC yöntemine göre E_b⁄N_0 oranında 1.5 dB kazanç sağlamaktadır.

In recent years, besides the need for high quality voice transmission for wireless communication systems, the need for high-speed data transmission is also increasing. In order to respond to these needs, the reliability and capacity of the communication systems should be increased. In the current 3G systems, increase in data rates has been provided by larger bandwidths and multiple antennas techniques. However, in a typical cellular system, the signal to noise ratio required to use multiple antennas techniques is observed on a very small fraction of the cell. The motivation for relay-based architecture comes from achieving larger gains by using this techniques for all cell. For this reason, two-way relay channel has been studied in this thesis. In a wireless two-way relay channel, intermediate node called relay assists communication between two end nodes. The relay receives the signal from the source, processes and transmits the signal to the destination. The relays are classified into two classes as analog relays or digital relays. Analog relays that also known as nonregenerative relays just amplify and transmit the received signal to the destination. Digital relays also known as regenerative relays process the received signal. The received signal first demodulated and then modulated before transmission to the destination. In literature, analog relays are called amplify and forward whereas digital relays are called decode and forward. In relay-based wireless systems frequency division duplexing or time division duplexing techniques are used frequently. In a TDD technique, the transmissions need to be orthogonal in time in order to avoid from interference since the same frequency is used for all tranmissions. There are efficient coding strategies at the relay that provide utilization of the channel in literature. One of the well studied coding strategy is called physical-layer network coding. In PNC, the relay receives the signal from both sources, process and forwards an appropriate signal satisfying the exclusive law to the end nodes. In PNC, the bi-directional transmission is completed in two phases whereas in classical and conventional network-coding systems it is completed in four phases and three phases, respectively. Thus, the PNC protocol avoids the loss in spectral efficiency that occurs due to the half-duplex nature of the communication devices. PNC and its counterpart adaptive network coding (ANC) are the latest protocols for two-way communication systems. Conventional network coding is implemented in the network layer, where relay node combines bits by XOR operation and forwards to the end nodes. In PNC strategy, two end nodes send information to the relay simultaneously. Naturally, the signals from end nodes overlap in free space as electromagnetic waves. Then, the relay decodes the superposition of mixed signals. Although broadcast nature of wireless communication system makes interference a destructive nuisance in common case, PNC turns the broadcast property into a capacity-increasing advantage. Nevertheless, PNC scheme does not give the best performance for all channel fade coefficients. To reduce the impact of multiple access interference which occurs at the relay during the MA phase a new scheme that is called adaptive network coding is introduced. In ANC, network coding mapping employed at the relay during the broadcast phase is adaptively changed according to the channel fade coefficients. Thus, the impact of the multiple access interference is decreased. The set of all possible channel realizations is quantized into a finite number of regions. In each region a specific network coding map is used to reach the best error performance. These network coding maps should satisfy the exclusive law. In certain regions any network coding map which satisfies the exclusive law gives the best performance and changing network coding map doesn’t impact error performance. These regions are called non-removable singular fade states. The non-removable singular fade states ocur due to channel outage. In certain regions the choice of the network coding map affects the performance. These ones are called removable singular fade states. An upper bound on the average end-to-end Symbol Error Rate is given with and without adaptive network coding at the relay, for a Rician fading scenario. It is clear that without ANC, at high Signal to Noise Ratio, three error events contribute to the end-to-end SER. The first error event is associated with the removable singular fade states. The second one is associated with the non-removable singular fade states. The third one comes from the error event during the BC phase. All these error events decrease as SNR-1 without ANC. In the other hand, when adaptive network coding is used at the relay, the error events associated with the removable singular fade states decrease as SNR-2. As a result adaptive network coding strategy provides a coding gain over the physical layer network coding. In TWRC, the contribution to the average end-to-end SER during the MA phase is dominant for a Rician fading channel. Since adaptive network coding improves the error performance during the MA phase, ANC provides more coding gain in a Rician fading environment than in a Rayleigh fading environment. Network coding map that satisfies the exclusive law can be obtained by Latin Square. It is shown that the requirement of satisfying the exclusive law is same as the clustering being constructed by a Latin squares. An adaptive network coding has been well studied in the literature for systems where symmetric data rates are used at the end nodes, such as M-PSK modulations. However, there is not too much study for asymmetric systems for which transmission rates of the two end nodes differ. On the other hand, end-nodes can be located on different distances to the relay. The quality of the signal that reaches the relay from one of the end nodes can be weak while that of the other end node signal being strong. Thus, different modulation schemes can be used at the end nodes. In the first part of the thesis, adaptive physical layer network coding in two-way relay systems is investigated for the symmetric scheme where both end nodes employ the same signal constellation. It is assumed that the same QPSK modulation is employed at the end-nodes and decode-and-forward relaying strategy is applied in the considered system. In addition, each nodes and relay are assumed having each one antenna and cannot receive and transmit simultaneously. The BER and SER performances of the PNC and ANC schemes are evaluated via computer simulations are compared for both Rayleigh and Rician channels. In the second part, adaptive network coding is implemented for the scenario that the data rates of the end nodes are asymmetric. First, quantization of the complex plane is obtained for the scenario that the end nodes use BPSK and QPSK modulation. In this scenario network coding maps that give the best performance for a given region are determined by Latin Rectangles. It is shown that for the Rician fading scenarios considered for the Rician fading parameter K=10, at a BER of 〖10〗^(-4), the ANC provides an SNR gain of 4.5 dB over the classical PNC system. Second, quantization of the complex plane is obtained for the scenario that the end nodes employ BPSK and 8-PSK modulation. For all regions, specific network coding maps that satisfy the exclusive law are acquired by Latin rectangles. It is shown that to achieve a bit error rate of 〖10〗^(-3) ANC scheme provides a coding gain of 1.5 dB compared to the reference PNC scheme for a Rician fading parameter value of K=10. Finally, theoretical error probability upper bounds valid at high SNR values are obtained for the considered asymmetrical two-way relaying systems. As well as these upper bounds are loose, their relative behaviour compares well with the simulation curves.