83 CHAPTER 4 IMPROVED MULTIUSER DETECTIO SCHEMES FOR ITERFERECE MAAGEMET I TH PPM UWB SYSTEM WITH m-zcz SEQUECES 4.1 ITRODUCTIO Accommodating many users in a small area is a major issue in the communication system. Frequency spectrum sharing and time sharing are used in the conventional systems. Hence, they have limitation on the capacity of the system. With the advent of spread spectrum, many users can be accommodated within the fixed bandwidth by making use of certain coding properties over the bandwidth. However, this system suffers from MAI caused by simultaneous multiple access. The key to this problem is MUD technique. These detection schemes help to detect the users data in the presence of MAI, ISI and noise (Verdu 1989). guyen Van Yen (27) proposed certain MUD schemes namely SIC and PIC and compared these schemes in UWB system. Thomas Richardson et al (21) stated that the current UWB systems utilize LDPC codes as it surpasses turbo codes in terms of BER and coding gain and turbo codes are the most suitable for the lower code rates. Michal Pietrzyk (21) employed LDPC codes in WiMAX and WiFi standards due to their outstanding communication performance.
84 Yang Jie et al (27) presented and analyzed the image transmission in TH PPM UWB using AWG channel. Wen Jun Lin et al (28) had investigated the performance of MAI cancellations for the asynchronous M-ary TH PPM UWB systems. Joanne Gomes et al (21) had proposed performance of binary double error correcting code for UWB communication for a specific application of image transmission. In this chapter, an improved interference management scheme in TH PPM UWB system with m-zcz sequences using MUD schemes is proposed and BER performance of various MUD schemes such as MMSE detector, DD, SIC and PIC under UWB channels are compared. Also, the proposed system is applied with LDPC codes and MMSE detection to improve the BER performance of the system in multiuser environments. 4.2 TYPES OF MUD The basic principle of the interference cancellation scheme is to first estimate the MAI for each user and then cancels out the MAI contributed by each user. Spread spectrum techniques are widely accepted. It also promises to play a key role in the future of wireless communication applications because of their efficient use of the channel. Hence, techniques that can improve the capacity of UWB systems are needed. The mobile communication systems based on UWB are inherently subject to MAI. Hence, it is impossible to maintain orthogonality between different users. MAI limits the capacity of conventional detectors. MUD techniques exploit the characteristics of the MAI by removal of the MUI from each user s received signal before making data decision. Hence, it offers significant gains in capacity over the conventional receiver.
85 UWB receivers are divided into single user and multiuser detectors. A single user receiver detects the data of one user at a time whereas a multiuser receiver jointly detects several users information. Single user and multi user receivers are also called as decentralized and centralized receivers respectively. The receiver restores the signal that is corrupted by the channel to its original form. Multiuser detectors can be broadly classified as linear and nonlinear detectors. The most popular linear multiuser detectors are DD and MMSE detectors. SIC and PIC are popular non-linear multiuser detectors. Recently, there has been a considerable interest in linear MMSE MUD scheme. In comparison to the other detection schemes, MMSE detector, has the advantage of not requiring the explicit knowledge of interference parameters. Although it does not achieve minimum BER, MMSE detector has been proved to achieve the optimal near far resistance. In the MMSE receiver, a linear transformation on the matched filter output minimizes the MSE whereas the DD utilizes the cross correlation between the signature sequences (Verdu 1998). 4.2.1 Conventional Matched Filter (CMF) Figure 4.1 shows CMF detector. In this method, the signal received is demodulated using a bank of matched filters.
86 Figure 4.1 Conventional matched filter detector The received signal is obtained for the first time interval m and for th j user is given by, r t A b S j j j t k u 1,k A b k j k t S k t n t (4.1) signal MAI noise where A is the amplitudes of the j th j users, b j is the data vector of the th j user and n t is a zero mean Gaussian random vector with noise power B. The received signal of th j has three terms. The first term refers to the recovered data, the second term refers to MAI due to all the users other than j, the third term refers to the thermal noise. As MAI is different from the noise it has to be treated differently. It is an in band interference unlike noise. It cannot be rejected through a band pass filter. If CMF detector is designed for the case of orthogonal spreading waveforms, it
87 does not consider MAI but for the non orthogonal case, MAI is treated as noise term. Hence, correlating the received signal with the user s th j chip sequence yields, y j T b r t S j t dt (4.2) A b j j k u 1, k A b k j k kj n j (4.3) dˆ sign y (4.4) j j where th kj is the cross correlation between k user and th j user and n represents noise term. j The CMF has the advantage of easy to implement and does not require knowledge of the channel or user amplitudes. 4.2.2 Decorrelating Detector (DD) The DD suppresses the interference by a linear transformation of cross correlation matrix inversion on the soft output of the matched filter bank. The only source of interference is the background noise resulting in noise enhancement. In DD, the received signal is fed into a bank of matched filters, each matched to the signature sequence of a different user. This detector has the advantage of MAI suppression and hence is near far resistant. Moreover, it does not require estimates of channel parameters.
88 However, the computation of the inverse of K K cross correlation matrix becomes complex when the number of users increase. The decorrelating receiver applies the inverse of the correlation matrix to the output of the matched filter to decouple the data. The output of the matched filter can be written in matrix form as, Y match RAb n (4.5) where R is the normalized cross correlation matrix and A is the diagonal matrix with A 1,A 2...A l as the amplitudes of different users and b represents the data vectors of k users and n is the zero mean Gaussian random vector with noise power. The output of DD with hard decision corresponding to the first user as the desired user is given by, b sign R (4.6) 1 Ymatch 1 where the interference caused by the other users is eliminated completely, but the noise component in the received signal are being scaled by the inverse of the correlation matrix. Hence, it s performance suffers from noise enhancement. 4.2.3 Minimum Mean Square Error (MMSE) Detector In MMSE, the MSE between the actual data and the soft output of the MAI is minimized. The advantage of this detector is that it requires no information about the interferers parameters. However, it has the problem of slow tracking.
89 The purpose of MMSE detector is to minimize the MSE between the transmitted signal and the detected signal transformed by matrix L linearly, i.e L Ymatch 2 E b (4.7) This linear transformation can also maximize the SIR ratio (Kao 21, Verdu 1998). Thus, detection scheme can be written as, b sign (4.8) L Ymatch 4.2.4 Successive Interference Cancellation (SIC) This is a non-linear type of MUD scheme in which the users are successively decoded. This scheme successively cancels the strongest users by re-encoding the decoded bits. After making an estimate of the channel, the interfering signal is recreated at the receiver and subtracted from the received waveform. SIC improves the BER performance of all the users. It is a serial approach, user by user. This detector has the advantage of easy implementation but suffers from the drawback of error propagation. 4.2.5 Parallel Interference Cancellation (PIC) The data estimates obtained from the matched filter denoted by d,... d corresponding to all the k users are multiplied by the 1 k amplitude estimates, spread using the corresponding spreading codes. Thus, the received signal is regenerated. All the regenerated signals except the desired user are partially summed. This output is the MAI and is subtracted from the received signal. Thus, the MAI cancelled output is passed on to a second bank of MF to produce the second set of data
9 estimates. This process is repeated for multiple stages till the desired BER performance is achieved. The merit of this scheme over SIC is less delay. However, it requires fast processors to perform the cancellation in parallel. 4.3 LDPC CODES LDPC code is a linear block code which has the parity check matrix H with low density ones. From (n-k) n matrix H, k n generator matrix G is derived. It encodes k information bits into n codeword bits. The received code word is decoded by (n-k) check nodes. An LDPC code is regular if the number of ones in each row and that of each column of H are both uniform, otherwise irregular. The weight of a row or column is the number of ones in it (Richardson et al. 21). Code rate is equal to k n, that is (n-k) redundant bits have been added to the information bits to correct errors. LDPC codes are represented by a bi-partite graph called a Tanner graph (Tanner 1981). A bi-partite graph is a graph in which the nodes are separated into two classes and the edges are connected to the two nodes that are not residing in the same class. The two classes of the nodes in a Tanner graph are bit nodes and check nodes. The Tanner graph is drawn according to the following rule, check node j is connected to bit node i whenever element h in H is a 1. Hence, there are m=n-k check ji nodes, one for each check equation and n bit nodes, one for each code bit. Figure 4.2 shows a Tanner graph for a regular parity check matrix H.
91 1 1 1 1 1 1 1 1 H (4.9) 1 1 1 1 1 1 1 1 Figure 4.2 Tanner graph for a regular parity check matrix H 4.3.1 LDPC Encoding To achieve the desired BER, longer LDPC codes with higher code rate are preferred. A systematic LDPC code, generator matrix for a code with parity-check matrix H is found by applying Gauss-Jordan elimination on H in order to get it in the form, H A, (4.1) I n k
92 n where A is a (n k identity matrix. k) k binary matrix and I n k is the size of The generator matrix is then, T G Ik, A (4.11) In LDPC encoding, fixed bit positions are identified. In systematic LDPC codes, the value of transmission code word must be the same as the values of H matrix s message word. parity check matrix of LDPC code, code word and fixed bit set is H is the h i, j c c1,c2,c3...c the encoder s s ci1,ci 2,......c i... c j ik ij n m n, order c or c 1 the encoded output is then c c i j ij,c... c,c......c ij,cij,......c ik.... c that is the 1 2 i1 i 1 1 message of bit nodes and check nodes are not encoded but only the fixed number of s or 1 s are encoded. The decoder uses the value of fixed bits which improves the accuracy of decoding (Tadashi Wadayama et al 21). Decoding is done by using hard message passing algorithm of LDPC codes known as bit flipping algorithm. 4.3.2 Bit Flipping Decoding It is a hard decision message passing algorithm. A hard decision about each received bit is found by the detector and this is passed to the decoder. In bit flipping algorithm, the messages that are passed along the Tanner graph edges are binary. Bit node sends a message declaring that it is one or zero. In check node, a message is connected to the bit node, declaring the value of the bit based on the information available to the check node.
93 The check node determines whether the parity check equation is satisfied if the modulo-2-sum of incoming bit values are zero. If many of the messages received by a bit node are different from its received value, the bit node flips its current value. This process is repeated until all the parity equations are satisfied or until the maximum number of decoder iterations has passed and the decoder gives up. The bit flipping decoder can immediately terminate when a valid codeword has been found by checking if all of the parity check equations are satisfied. It has the advantage of avoiding additional iterations once a solution has been found. 4.4 PROPOSED TH PPM UWB SYSTEM WITH m-zcz AD IMPROVED MUD SCHEMES USIG LDPC CODES Figure 4.3 Proposed TH PPM UWB system with m-zcz and improved MUD schemes using LDPC codes
94 Figure 4.3 shows the proposed TH PPM UWB system with m-zcz and improved MUD schemes using LDPC codes. Bernoulli sequence generator is used to generate binary data for the different users. The random binary sequence of s and 1 s are fed into TH PPM modulator to get the time shifting of sequence and it is applied to pulse shaper to get UWB Gaussian pulse. Each Gaussian pulse is transmitted through the LDPC encoder. Each users encoded output is multiplied by the m-zcz sequence of the corresponding user. The spread sequences are multiplied by the amplitude of each user. The amplitude values decide the power of each user. The transmitted sequences of different users are summed to represent the transmitted signal from different multiple mobile users. This signal is transmitted through UWB channel models CM1, CM2, CM3 and CM4 under different fading conditions. The received signal is then fed into LDPC decoder and the decoded outputs are fed into MMSE detector and BER is obtained. 4.5 RESULTS AD DISCUSSIO In this section, BER performance of the proposed TH PPM UWB with different MUD schemes have been compared. Linear MMSE combined with LDPC codes in TH PPM UWB for the frame size of 1. umber of bits per frame is 1. The proposed system uses Bit Flip LDPC decoder to get back the original information in the receiver side.
95 Table 4.1 Input parameters for MUD combined TH PPM UWB with LDPC Power level of users Constant power level P 1 Random power level P 2 Type of spreading sequences m-sequences/m-zcz sequences Frame length 1 Type of the channel AWG/ Rayleigh/CM1/CM2/CM3/CM4 Eb 2-12 db for AWG / UWB channel models o Eb 2-2 db for Rayleigh fading channel o Channel coding rate Multiuser Detector 1 2 MMSE Output Parameter BER is found out for various number of energy to noise density ratio and for different number of users. The simulation parameters are the same as discussed in Chapter 3. To study near far effect, the simulation is done for different power levels P 1 and P 2 and PSR is calculated using Equation (3.23) as defined in Chapter 3 in Section 3.5. 4.5.1 Comparison of MUD Schemes Without LDPC Codes In this section, the BER comparison of the various MUD schemes with m-zcz for TH PPM UWB system with varying Eb and a o constant power level 1 P are given for all UWB channels from Figures 4.4 to 4.7.
96 1 1-1 1-2 1-3 1-4 CMF:mZCZseq:CM1 MMSE:mZCZseq:CM1 DD:mZCZseq:CM1 PIC:mZCZseq:CM1 SIC:mZCZseq:CM1 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.4 BER performance of the various MUD schemes for the proposed TH PPM UWB system with m-zcz under CM1 with constant power level 1 P for 1 users Figure 4.4 shows the BER performance of various MUD schemes for the proposed TH PPM UWB system with m-zcz under CM1 with constant power level 1 P for 1 users. For E b 8 db, a BER of.9 is obtained in DD MUD scheme. However, the BER value is.22 in CMF. In MMSE and SIC the BER values are.68 and.15 respectively. A BER of.4 is obtained in PIC.
97 1 1-1 1-2 1-3 1-4 CMF:mZCZseq:CM2 MMSE:mZCZseq:CM2 DD:mZCZseq:CM2 PIC:mZCZseq:CM2 SIC:mZCZseq:CM2 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.5 BER performance of the various MUD schemes for the proposed TH PPM UWB system with m-zcz under CM2 with constant power level 1 P for 1 users Figure 4.5 shows the BER performance of various MUD schemes for the proposed TH PPM UWB system with m-zcz under CM2 with constant power level 1P for 1 users. For E b 8 db, a BER of.36 is obtained in DD MUD scheme. However, the BER value is.232 in CMF. In MMSE and PIC the BER values are.72 and.15 respectively. A BER of.7 is obtained in SIC.
98 1 1-1 1-2 1-3 1-4 CMF:mZCZseq:CM3 MMSE:mZCZseq:CM3 DD:mZCZseq:CM3 PIC:mZCZseq:CM3 SIC:mZCZseq:CM3 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.6 BER performance of the various MUD schemes for the proposed TH PPM UWB system with m-zcz under CM3 with constant power level 1 P for 1 users Figure 4.6 shows the BER performance of various MUD schemes for the proposed TH PPM UWB system with m-zcz under CM3 with constant power level 1P for 1 users. For E b 8 db, a BER of.62 is obtained in DD MUD scheme. However, the BER value is.235 in CMF. In MMSE and SIC the BER values are.78 and.1 respectively. A BER of.15 is obtained in PIC.
99 1 1-1 1-2 1-3 1-4 CMF:mZCZseq:CM4 MMSE:mZCZseq:CM4 DD:mZCZseq:CM4 PIC:mZCZseq:CM4 SIC:mZCZseq:CM4 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.7 BER Performance of the various MUD schemes for the proposed TH PPM UWB system with m-zcz under CM4 with constant power level 1 P for 1 users Figure 4.7 shows the BER performance of various MUD schemes for the proposed TH PPM UWB system with m-zcz under CM4 with constant power level 1 P for 1 users. For E b 8 db, a BER of.112 is obtained in DD MUD scheme. However, the BER value is.258 in CMF. In MMSE and SIC the BER values are.95 and.2 respectively. A BER of.5 is obtained in PIC. From Figures 4.4 to 4.7, it is observed that CMF shows the highest BER as it does not distinguish between the channel noise and MAI. DD has the best BER performance because it eliminates MAI. ext to DD,
1 SIC has better performance. When the number of users in the proposed system is less, the performance difference between the PIC and SIC are small and this difference become larger when the number of users is increased or E b is higher because of the increase in MAI (guyen Van Yen 27). MMSE has better performance than CMF as it reduces noise amplification to a certain extent, but results in higher residual MAI. Even though DD has better BER, it has high computational complexity. SIC is overall less computationally intensive and the delay per stage of cancellation is more. Hence, the proposed system performs well with MMSE. Table 4.2 the shows capacity of TH PPM UWB system using various MUD schemes with m-zcz sequences under CM1 channel for E b 4 db with varying number of users for constant power level 1 P. From Table 4.2, it can also be seen that a BER of.186 is obtained for 4 db at 1 users in CMF. At E b 4 db, for 1 users, BER values for DD, MMSE, PIC and SIC are.25,.49,.45 and.35 respectively. When the number of users increases to 1, the BER values of CMF, DD, MMSE, SIC and PIC are.255,.43,.127,.55 and.67 respectively. Thus, the BER value increases as the number of users increase.
11 Table 4.2 Capacity of the TH PPM UWB system using various MUD schemes with m-zcz sequences under CM1 channel for E b 4 db with varying number of users umber of Users BER for E b 4 db CMF DD MMSE PIC SIC 1.186.25.49.45.35 2.195.26.51.5.42 3.21.28.55.59.44 4.21.27.59.6.47 5.224.31.62.62.49 6.237.33.64.64.51 7.244.39.69.67.52 8.247.4.72.68.535 9.25.42.924.69.542 1.255.43.127.67.55 Table 4.3 shows the capacity of TH PPM UWB system using various MUD schemes with m-zcz sequences under CM2 channel for E b 4 db with varying number of users for constant power level P. 1 From Table 4.3, it can be seen that a BER of.1924 is obtained for 4 db at E 1 users in CMF. At b 4 db, for 1 users, BER values for DD, MMSE, SIC and PIC are.28,.79,.37 and.47 respectively. When the number of users increases to 1, the BER values of CMF, DD, MMSE,
12 SIC and PIC are.263,.455,.116,.5 and.72 respectively. Thus, the BER value increases as the number of users increase. Table 4.3 Capacity of TH PPM UWB system using various MUD schemes with m-zcz sequences under CM2 channel for E b 4 db with varying number of users umber of Users BER for E b 4 db CMF DD MMSE PIC SIC 1.1924.28.79.47.37 2.194.32.81.51.41 3.195.39.85.55.45 4.1969.41.89.57.455 5.212.42.92.59.46 6.229.43.94.62.47 7.232.438.99.64.48 8.245.44.11.66.485 9.253.45.114.67.49 1.263.455.116.72.5
13 Table 4.4 Capacity of TH PPM UWB system using various MUD schemes with m-zcz sequences under CM3 channel for E b 4 db with varying number of users umber of Users BER for E b 4 db CMF DD MMSE PIC SIC 1.181.326.9.38.32 2.192.344.112.44.36 3.22.38.115.45.38 4.226.4.116.53.44 5.231.42.117.56.48 6.234.43.118.57.52 7.243.44.117.59.53 8.254.43.122.59.54 9.259.45.123.57.55 1.265.469.123.69.56 Table 4.4 shows the capacity of TH PPM UWB system using various MUD schemes with m-zcz sequences under CM3 channel for E b 4 db with varying number of users for constant power level P. 1 From Table 4.4, it can be seen that a BER of.181 is obtained for 4 db at E 1 users in CMF. At b 4 db, for 1 users, BER values for DD, MMSE, PIC and SIC are.326,.9,.38 and.32 respectively. When the number of users increases to 1, the BER values of CMF, DD, MMSE,
14 PIC and SIC are.265,.469,.123,.69 and.56 respectively. Thus, the BER value increases as the number of users increase. Table 4.5 Capacity of TH PPM UWB system using various MUD schemes with m-zcz sequences under CM4 channel for E b 4 db with varying number of users umber of Users BER for E b 4 db CMF DD MMSE PIC SIC 1.185.49.99.5.53 2.195.5.112.54.56 3.213.55.116.57.58 4.224.56.12.6.6 5.235.59.12.62.65 6.241.6.121.69.71 7.244.61.121.75.74 8.258.62.123.82.73 9.264.625.133.95.78 1.271.637.138.12.82 Table 4.5 shows the capacity of TH PPM UWB system using various MUD schemes with m-zcz sequences under CM4 channel for E b 4 db with varying number of users for constant power level P. 1
15 From Table 4.5, it can be seen that a BER of.185 is obtained for 4 db at E 1 users in CMF. At b 4 db, for 1 users, BER values for DD, MMSE, PIC and SIC are.49,.99,.5 and.53 respectively. When the number of users increases to 1, the BER values of CMF, DD, MMSE, PIC and SIC are.271,.637,.138,.12 and.82 respectively. Thus, the BER value increases as the number of users increase. 4.5.2 BER Performance of MUD Schemes Under Random Power Level P 2 in UWB Channel Models (ear Far Effect) In this section, comparison of the BER performance of various MUD schemes with varying E b and a random power level for each user under UWB channels CM1, CM2, CM3 and CM4 are given in the Figures 4.8 to 4.11. Figure 4.8 shows the BER performance of the various MUD schemes under CM1 channel with random power level P 2 for 1 users. From Figure 4.8, it is clear that the different types of detectors behave differently under the near-far effect. The BER performance of the CMF shows a constant BER than all the other detectors. The CMF is robust against the near-far effect. This is because CMF treats all MAI as noise.
16 1 1-1 1-2 1-3 1-4 CMF:mZCZseq:CM1 DD:mZCZseq:CM1 MMSE:mZCZseq:CM1 PIC:mZCZseq:CM1 SIC:mZCZseq:CM1 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.8 BER performances of the various MUD schemes under CM1 channel with random power level P 2 for 1 users From the Figure 4.8, it can be seen that for E b 8 db the BER of MMSE, PIC, SIC and DD are.6,.8,.6 and.35 respectively. BER performance of the SIC detector is better than that of the MMSE. This is due to the reason that at low E b values, the background noise power is more due to random power levels. As E b value increases, the MAI is more than the background noise and the SIC detector performs well to suppress the MAI. Hence, the performance of the proposed SIC
17 detector is closer to MMSE at low E b values and better than the MMSE at high E b values. 1 1-1 1-2 1-3 1-4 CMF:mZCZseq:CM2 DD:mZCZseq:CM2 MMSE:mZCZseq:CM2 PIC:mZCZseq:CM2 SIC:mZCZseq:CM2 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.9 BER performances of the various MUD schemes under CM2 channel with random power level P 2 for 1 users Figure 4.9 shows the BER performance of the various MUD schemes under CM2 channel with random power level P 2 for 1 users. The BER performance of the CMF shows a poor BER performance than all the other detectors. This is because CMF treats all MAI as noise. From the Figure 4.9, it can be seen that for E b 8 db, the BER of MMSE, PIC, SIC and DD are.65,.24,.12 and.42 respectively.
18 BER performance of the SIC detector is better than that of the MMSE. This is due to the reason that at low E b values, the background noise power is more due to random power levels. As E b value increases, the MAI is more than the background noise and the SIC detector performs well to suppress the MAI. Hence, the performance of the proposed SIC detector is closer to MMSE at low E b values and better than the MMSE at high E b values. Comparing the linear and non-linear detectors, it is seen that there is a significant improvement in the performance of DD and their corresponding BER is.42 at E b 8 db. 1 1-1 1-2 1-3 1-4 CMF:mZCZseq:CM3 DD:mZCZseq:CM3 MMSE:mZCZseq:CM3 PIC:mZCZseq:CM3 SIC:mZCZseq:CM3 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.1 BER performance of the various MUD schemes under CM3 with random power level P 2 for 1 users
19 Figure 4.1 shows the BER performance of the various MUD schemes under CM3 channel with random power level P 2 for 1 users. The CMF performs very poorly. Its bit error probability reaches an irreducible error floor of approximately.25 at E b 8 db. This is because CMF treats the MAI as noise. From Figure 4.1, it is observed that for E b 8 db the BER of MMSE, PIC, SIC and DD are.98,.6,.48 and.114 respectively. The performance of both the SIC and PIC degrades as E b increases. The SIC shows significant improvement as power disparity allows more accurate bit decisions to be made for the stronger users, based on which interference is cancelled. Hence, the performance of the SIC detector is closer to MMSE at low E b values and better than the MMSE at high E b values. Figure 4.11 shows the BER performance of the various MUD schemes under CM4 channel with random power level P 2 for 1 users. The BER performance of the CMF shows a poor BER than all the other channels. This is because CMF treats all MAI as noise.
11 1 1-1 1-2 1-3 1-4 CMF:mZCZseq:CM4 DD:mZCZseq:CM4 MMSE:mZCZseq:CM4 PIC:mZCZseq:CM4 SIC:mZCZseq:CM4 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.11 BER performance of the various MUD schemes under CM4 with random power level P 2 for 1 users From the Figure 4.11, it can be seen that for E b 8 db the BER of MMSE, PIC, SIC and DD are.114,.72,.516 and.216 respectively. BER performance of the SIC detector is better than that of the MMSE. This is due to the reason that at low E b values, the background noise power is more due to random power levels. As E b value increases, the MAI is more than the background noise and the SIC detector performs well to suppress the MAI. Hence, the performance of the proposed SIC
111 detector is closer to MMSE at low E b values and better than the MMSE at high E b values. DD performs better than the other detectors. It is seen that from Figures 4.8 to 4.11, that BER performance of the DD is better than that of the SIC. This is because the DD operation performs a linear transformation by which the MAI is completely cancelled. In the SIC scheme, the MAI is serially cancelled. Due to imperfect cancellation of the interfering users, the MAI is not completely cancelled. Hence, it exhibits poor performance compared to that of the DD at higher E b values where the MAI more than the background noise. 4.5.3 BER Performance of MMSE Based TH PPM UWB With LDPC Codes for Constant Power Level P 1 The random binary data are sent to the LDPC encoder. The coded sequence is the information. This coded data is sent under UWB channel CM1. Adding noise involves channel impulse realization due to cluster and ray arrival rates. The noisy signal is then fed into the LDPC bit flipping decoder. Five iterations are performed in the bit flipping decoder to get very low BER.
112 1 PPM-MMSE-LDPC:AWG[Exist-m] PPM-MMSE-LDPC:AWG[Propd-mZCZ] 1-1 1-2 1-3 1-4 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.12 BER performance of TH PPM UWB with MMSE using LDPC under AWG for 1 users Figure 4.12 shows the BER performance of TH PPM UWB with and without channel coding over AWG for 1 users. From Figure 4.12, it is observed that BER of.55 is achieved at E b 4 db for the proposed system using m-zcz sequences and a BER of.175 is obtained for the system using m-sequences under AWG channel. A BER improvement of 68.5% is achieved in the proposed system over that of the system with m-sequences under AWG channel.
113 1 PPM-MMSE-LDPC:CM1[Propd-mZCZ] PPM-MMSE-LDPC:CM1[Exist-m] 1-1 1-2 1-3 1-4 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.13 BER performance of TH PPM UWB with MMSE using LDPC under CM1 for 1 users Figure 4.13 shows the BER performance of TH PPM UWB with and without channel coding over UWB channel model CM1 for 1 users. From Figure 4.13, it is observed that BER of.62 is achieved at E b 4 db for the proposed system using m-zcz sequences and BER of.169 is obtained for the system using m-sequences under channel model CM1. A BER improvement of 63.3% is achieved in the proposed system than that of the system with m-sequences under CM1 channel.
114 1 PPM-MMSE-LDPC:CM2[Propd-mZCZ] PPM-MMSE-LDPC:CM2[Exist-m] 1-1 1-2 1-3 1-4 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.14 BER performance of TH PPM UWB with MMSE using LDPC under CM2 for 1 users Figure 4.14 shows a BER performance of TH PPM UWB with and without channel coding over AWG and UWB channel model CM2 for 1 users. From Figure 4.14, it is observed that BER of.79 is achieved at E b 4 db for the proposed system using m-zcz sequences and BER of.189 is obtained for the system using m-sequences under channel model CM2. A BER improvement of 58.2% over the system with m-sequences has been achieved under CM2 channel.
115 1 PPM-MMSE-LDPC:CM3[Propd-mZCZ] PPM-MMSE-LDPC:CM3[Exist-m] 1-1 1-2 1-3 1-4 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.15 BER performance of TH PPM UWB with MMSE using LDPC under CM3 for 1 users Figure 4.15 shows the BER performance of TH PPM UWB with and without channel coding over AWG and UWB channel model CM3 for 1 users. From Figure 4.15, it is observed that BER of.925 is achieved at E b 4 db for the proposed system using m-zcz sequences and BER of.19 is obtained for the system using m-sequences under channel model CM3. A BER improvement of 51.3% over that of the system with m-sequences has been achieved under CM3 channel.
116 1 PPM-MMSE-LDPC:CM4[Propd-mZCZ] PPM-MMSE-LDPC:CM4[Exist-m] 1-1 1-2 1-3 1-4 2 3 4 5 6 7 8 9 1 11 12 Eb/o(dB) Figure 4.16 BER performance of TH PPMUWB MMSE using LDPC under CM4 for 1 users Figure 4.16 shows the BER performance of TH PPM UWB with and without channel coding over AWG and UWB channel model CM4 for 1 users. From Figure 4.16, it is observed that BER of.99 is achieved at E b 4 db in the proposed system using m-zcz sequences and BER of.196 is obtained for the system using m-sequences under channel model CM4. A BER improvement of 49.48% over that of the system with m-sequences has been achieved under CM4 channel. 4.5.4 System Capacity of MMSE Based TH PPM UWB Using LDPC Codes In this section, the system capacity of MMSE based TH PPM UWB using LDPC codes under all the UWB channels are given in the Tables 4.6 to 4.9.
Table 4.6 Capacity of TH PPMUWB with MMSE using LDPC under CM1 umber of Users THPPM AWG Without MUD (m-zcz sequences) THPPM BER for AWG MMSE (m-sequences) E b 4 db THPPM CM1:MMSE (m-sequences) THPPM AWG:MMSE-LDPC (m-zcz sequences) TH PPM CM1:MMSE-LDPC (m-zcz sequences) 1.19.158.143.32.34 2.195.153.15.34.36 3.224.154.149.495.475 4.28.155.155.55.5 5.291.158.153.52.51 6.3.163.158.525.53 7.34.165.162.535.545 8.316.169.167.535.6 9.318.17.168.5.595 1.32.175.169.55.62
118 Table 4.6 shows the capacity of TH PPM UWB with and without channel coding over AWG and UWB channel model CM1 for 4 db. From Table 4.6, it is observed that when the number of users is 3, the corresponding BER is equal to.475 for the proposed TH PPM UWB system with MMSE using LDPC under UWB channel model CM1 at E b 4 db. When the number of users is 6, the BER increases to.53 for the same system under UWB channel model CM1 for E b E b 4 db. When the number of users is 3, the corresponding BER is equal to.149 for the proposed TH PPM UWB system with MMSE without LDPC codes using m- sequences under UWB channel model CM1. Hence, the system capacity of the proposed TH PPM UWB system with MMSE using LDPC is 68% better than the system without LDPC codes under UWB channel CM1. Table 4.7 shows the capacity of TH PPM UWB with and without channel coding over AWG and UWB channel model CM2 for E b 4 db. It is observed from the Table 4.7 that when the number of users is 3, the corresponding BER is equal to.55 for the proposed TH PPM UWB system with MMSE using LDPC under UWB channel model CM2 at E b 4 db. When the numbers of users is 1, the BER increases to.79 for the same system under UWB channel model CM2 for E b 4 When the number of users is 3, the corresponding BER is equal to.155 for the proposed TH PPM UWB system with MMSE without LDPC codes using db.
119 m-sequences under UWB channel model CM2. Hence, the system capacity of the proposed TH PPM UWB system with MMSE using LDPC is 67.4% better than that of the system without LDPC codes under UWB channel CM2. Table 4.8 shows the capacity of TH PPM UWB with and without channel coding over AWG and UWB channel model CM3 for E b 4 db. It is observed from the Table 4.8 that when the number of users is 3, the corresponding BER is equal to.645 for the proposed TH PPM UWB system with MMSE using LDPC under UWB channel model CM3 at E b 4 db. When the number of users is 1, the BER increases to.925 for the same system under UWB channel model CM3 for E b 4 When the number of users is 3, the corresponding BER is equal to.149 for the proposed TH PPM UWB system with MMSE without LDPC codes using m-sequences under UWB channel model CM3. Hence, the system capacity of the proposed TH PPM UWB system with MMSE using LDPC is 57% better than that of the system without LDPC codes under UWB channel CM3. db.
Table 4.7 Capacity of TH PPM UWB with MMSE using LDPC under CM2 umber of Users THPPM AWG MMSE (m-sequences) BER for THPPM CM2:MMSE (m-sequences) E b 4 db THPPM AWG:MMSE-LDPC (m-zcz sequences) TH PPM CM2:MMSE-LDPC (m-zcz sequences) 1.158.134.32.32 2.153.151.34.38 3.154.155.495.55 4.155.158.55.54 5.158.16.52.535 6.163.161.525.555 7.165.167.535.6 8.169.171.535.65 9.17.186.5.655 1.175.189.55.79
Table 4.8 Capacity of TH PPM UWB with MMSE using LDPC un umber of Users THPPM AWG MMSE (m-sequences) BER for THPPM CM3:MMSE (m-sequences) E b 4 db THPPM AWG:MMSE-LDPC (m-zcz sequences) 1.158.125.32 2.153.138.34 3.154.149.495 4.155.156.55 5.158.168.52 6.163.173.525 7.165.175.535 8.169.18.535 9.17.188.5 1.175.19.55
Table 4.9 Capacity of TH PPMUWB with MMSE using LDPC und umber of Users THPPM AWG MMSE (m-sequences) BER for THPPM CM4:MMSE (m-sequences) E b 4 db THPPM AWG:MMSE-LDPC (m-zcz sequences) 1.158.138.32 2.153.156.34 3.154.154.495 4.155.161.55 5.158.166.52 6.163.181.525 7.165.18.535 8.169.184.535 9.17.19.5 1.175.196.55
123 Table 4.9 shows the capacity of TH PPM UWB with and without channel coding over AWG and UWB channel model CM4 for E b 4 db. It is observed from the Table 4.9 that when there are 3 users, the corresponding BER is equal to.725 for the proposed TH PPM UWB system with MMSE using LDPC under UWB channel model CM4 at E b 4 db. When there are 1 users, the BER increases to.99 for the same system under UWB channel model CM4 for E b 4 db. When the number of users is 3, the corresponding BER is equal to.154 for the proposed TH PPM UWB system with MMSE without LDPC codes using m-sequences under UWB channel model CM4. Hence, the system capacity of the proposed TH PPMUWB system with MMSE using LDPC is 53% better than that of the system without LDPC codes under UWB channel CM4. 4.6 APPLICATIO OF THE PROPOSED SYSTEM WITH BIOMEDICAL IMAGES AS IPUT In this section, the performance of the proposed TH PPM UWB system with MMSE using LDPC codes with m-zcz sequences and m-sequences by applying biomedical sample images as input is obtained. The images are divided into frames and are sent to the LDPC encoder. The coded sequence is the information. This coded data is added with UWB channel model CM1. Five iterations have been performed in the Bit Flip decoder inspite of the extra latency since the image quality is better after 5 iterations and the decoded images are analyzed in terms of the PSR. The original image, the image corrupted by channel noise and the
124 decoded images for the chest x-ray image are shown in Figure 4.17. The same parameters are applied to the abdomen image and the transmission is simulated. The original image, the image corrupted by channel noise and decoded images after using bit flipping algorithm are shown in Figure 4.17. The PSR values have been noted for different iterations and are shown in Figure 4.17. In each iteration, the PSR value increases. Hence, the proposed TH PPM UWB with MMSE using LDPC system gives better PSR value after 5 iterations since the image quality is high. For chest x-ray, the PSR of 3.2 is obtained and for the abdomen image it is equal to 3.4.
TH PPMUWB with MMSE using LDPC TH PPMUWB with MMSE using LDPC TH PPMUWB with MMSE without LDPC TH PPMUWB with MMSE without LDPC a.(i) Input image b.(i) Input image c.(i) Input image d.(i) Input image a.(ii)after 1 st iterationpsr=2.27 b.(ii) After 1 st iteration,psr=21.78 c.(ii)decoded image without LDPC PSR=23.2 d.(ii)decoded image without LDPC PSR=24.5653 - - a.(iii)after 3rd iterationpsr=25.4 b.(iii) After 3rd teration,psr=25.65 - - (iv)decoded image (After 5 th iteration) using Bit flip LDPC decoder under CM1PSR=3.2 b.(iv) Decoded image(after 5 th iteration) using Bit flip LDPC decoder under CM1PSR=3.41 Figure 4.17 Transmission and reception of bio-medical images in the proposed TH PPMUWB with MMSE using LDPC system under CM1
126 4.7 COCLUSIO In this chapter, the proposed TH PPM UWB system using m-zcz with various MUD schemes are compared and their performances analyzed. Also, MUD combined TH PPM UWB system using LDPC codes with m-zcz sequences is proposed. Even though DD outperforms all the other detectors, it has the drawback of computational complexity. Since MMSE MUD scheme gives better BER at low SR, MMSE detector is incorporated with LDPC channel coder. In the proposed TH PPM UWB system with MMSE using LDPC has been optimally applied for transmission of random binary data. An average BER improvement of 61.4% is achieved in the proposed system than that of the system with MMSE under the four UWB channel models. A system model has been implemented where in biomedical images are subjected to UWB channel noise and additive white Gaussian noise. It has been observed that the PSR values of the images after decoding in the proposed TH PPM UWB system with MMSE using LDPC is 24% higher than with TH PPM UWB system using MMSE without LDPC under CM1. However, the proposed system has the limitation on frame length. When the data frame length increases, the computational complexity increases. Moreover for the detection of biomedical images, more number of iterations need to be performed at the decoder side. Hence, the computational complexity will be high. Therefore to obtain moderate computational complexity, an evolutionary technique called PSO is applied for further improvement of the system performance.