A Sample-Decimation Based Fast Preamble Detection Algorithm

Transcription

1 A Sampe-Decimation Based Fast Preambe Detection Agorithm Haining Zhang A Thesis in The Department of Eectrica and Computer Engineering Presented in Partia Fufiment of the Requirements for the Degree of Master of Appied Science in Eectrica Engineering Concordia University Montrea, Quebec, Canada March, 2008 Haining Zhang, 2008

2 * Library and Archives Canada Pubished Heritage Branch 395 Weington Street Ottawa ON KA0N4 Canada Bibiotheque et Archives Canada Direction du Patrimoine de I'edition 395, rue Weington Ottawa ON KA0N4 Canada Your fie Votre reference ISBN: Our fie Notre reference ISBN: NOTICE: The author has granted a nonexcusive icense aowing Library and Archives Canada to reproduce, pubish, archive, preserve, conserve, communicate to the pubic by teecommunication or on the Internet, oan, distribute and se theses wordwide, for commercia or noncommercia purposes, in microform, paper, eectronic and/or any other formats. The author retains copyright ownership and mora rights in this thesis. Neither the thesis nor substantia extracts from it may be printed or otherwise reproduced without the author's permission. AVIS: L'auteur a accorde une icence non excusive permettant a a Bibiotheque et Archives Canada de reproduire, pubier, archiver, sauvegarder, conserver, transmettre au pubic par teecommunication ou par Pntemet, prefer, distribuer et vendre des theses partout dans e monde, a des fins commerciaes ou autres, sur support microforme, papier, eectronique et/ou autres formats. L'auteur conserve a propriete du droit d'auteur et des droits moraux qui protege cette these. Ni a these ni des extraits substanties de cee-ci ne doivent etre imprimes ou autrement reproduits sans son autorisation. In compiance with the Canadian Privacy Act some supporting forms may have been removed from this thesis. Whie these forms may be incuded in the document page count, their remova does not represent any oss of content from the thesis. Conformement a a oi canadienne sur a protection de a vie privee, queques formuaires secondaires ont ete eneves de cette these. Bien que ces formuaires aient incus dans a pagination, i n'y aura aucun contenu manquant. Canada

3 ABSTRACT A Sampe-Decimation Based Fast Preambe Detection Agorithm Haining Zhang Random access is a commony used mutipe access scheme that aows mutipe users to share the same resource in a distributed fashion. In a Universa Mobie Teecommunication System (UMTS), the preambe of a random access channe (RACH) message is used by a mobie user to signa the base station for requesting network access or short data packets transportation. The base station is responsibe in a timey fashion for detecting the preambes and informing the user whether the request has been granted or denied through the acquisition indication channe (AICH). Preambe detection is one of the most computationay intensive functiona units of a base station. It has attracted many research attentions and investments in the past a few decades. The drawback of the existing preambe detection (PD) agorithms for UMTS base-station is that either their computationa compexity is high or the detection accuracy is ow. The conventiona fu search PD agorithm gives the best resut in terms of the detection probabiity, but its compexity is high. On the hand, the parae-seria code phase detector PD agorithm provides a reduced computationa compexity, but the detection accuracy becomes ow. In this thesis, a sampe-decimation based preambe detection technique is proposed in order to substantiay reduce the computationa compexity and at the same time retain a high detection accuracy. The proposed agorithm comprises two stages. Deay hypotheses or deay offsets which are unikey to have a strong correation power between the antenna sampes and the ocay generated preambe repica are iii

4 identified and discarded in the first stage. The second stage operates on the remaining offsets and empoys a the antenna sampes within the preambe signa. Extensive computer simuations are conducted under different eves of additive white Gaussian noise interferences. The resuts show that the proposed agorithm has a detection performance very cose to that of the conventiona fu search PD agorithm, whie at the same time it reduces the computationa compexity by more than sixty percent. IV

5 ACKNOWLEDGEMENTS I wish to express my deep gratitude to my thesis supervisors, Dr. M. Omair Ahmad and Dr. M.N.S. Swamy, for their guidance throughout the years of my graduate studies at Concordia University. I woud ike to thank them for their advice, insight, and care. Pursuing a Mater's degree in Eectrica and Computer Engineering was not ony a ong cherished dream of mine but aso a great chaenge as we. It woud have been impossibe for me to compete my graduate studies without the inspiration and support of my supervisors. I am indebted to my unce, Dr. Jiajun Zhang and his wife Dr. Li Zhang, who sefessy provided me with a their hep and support, technica or otherwise, during the entire course of my graduate studies and my ife in North America. I woud ike to thank my cousin, Ms. Min Yang, my friends, Mr. Feng Wan and Mr. Marc-Andre Laverdiere for their sincere advice, hep and company during these years of study at Concordia University. Lasty, and most importanty, I wish to thank my beoved parents, Guiing Xu and Jiafu Zhang. It is their unconditiona ove, sacrifices and understanding that have made a the things possibe in my ife. To them I dedicated this thesis. v

6 To my unce and aunt and my dearest parents VI

7 Tabe of Contents List of Figures List of Tabes List of Symbos List of Acronyms ix xi xii xiii Chapter Introduction. Genera.2 Literature Survey 5.3 Motivation, Scope and Organization of the Thesis 6 Chapter 2 The Fundamentas of a Preambe Detector 9 2. Upink Radio Frame Structure Downink Access Sots and AICH Channe 2.3 Time Reationship between AICH and PRACH RACH Sub-channe Preambe Signatures Access Service Cass Random Access Procedure in UMTS Preambe Signas Long scrambing sequence Preambe scrambing code Preambe signas Preambe Detection at Node-B Correation Search window Power deay profie Threshoding Probabiity of Detection and Probabiity of Fase Aarm 28 vn

8 2.2 Summary 29 Chapter 3 The Proposed Sampe-Decimation Based Agorithm Introduction Proposed Agorithm Correation in the sampe-decimated domain Correation in the origina sampe domain Computationa Compexity of the Proposed Agorithm Computationa oad of the fu search agorithm Computationa oad of the proposed agorithm Computationa savings Summary 43 Chapter 4 Simuation Study Introduction An Iustrative Exampe Simuation Resuts Test vector suite Simuation statistics Summary 67 Chapter 5 Concusion Concuding Remarks Scope for Future Work References 72 V

9 List of Figures Figure. Iustration of a base station and a user equipment 3 Figure 2. Frame structure in UMTS 0 Figure 2.2 The structures of an access sot and an access frame Figure 2.3 Access sot timing of downink physica channes 2 Figure 2.4 Timing reationship between AICH and PRACH 3 Figure 2.5 Reationship between access sots and sub-channes 5 Figure 2.6 Access service cass 7 Figure 2.7 Power ramping scheme in RACH procedure 2 Figure 2.8 An iustration of over-samping of an antenna stream 24 Figure 3. Operations performed in the first stage 36 Figure 4. Antenna sampes, (a) Rea part, (b) Imaginary part 46 Figure 4.2 The first 28 antenna sampes, (a) Rea part, (b) Imaginary part 47 Figure 4.3 Locay generated code sequence, (a) Rea part, (b) Imaginary part 48 Figure 4.4 The first 28 chips of the ocay generated code sequence, (a) Rea part. (b) Imaginary part 49 Figure 4.5 power deay profie in the origina sampe domain 50 Figure 4.6 Low-pass fitered antenna sampes, (a) Rea part, (b) Imaginary part 5 Figure 4.7 The first 28 sampes of the ow-pass fitered antenna sampes, (a) Rea part. (b) Imaginary part 52 Figure 4.8 Low-pass fitered ocay-generated code sampes, (a) Rea part. (b) Imaginary part 53 Figure 4.9 The first 28 sampes of the sequence in Figure 4.8. (a) Rea part. (b) Imaginary part 54 Figure 4.0 Decimated antenna sampes, (a) Rea part, (b) Imaginary part 55 Figure 4. The first 28 sampes of the decimated antenna signa, (a) Rea part. (b) Imaginary part 56 Figure 4.2 Locay generated decimated code sequence, (a) Rea part. (b) Imaginary part 57 IX

10 Figure 4.3 The first 28 sampes of the decimated code, (a) Rea part. (b) Imaginary part 58 Figure 4.4 Power deay profie in the decimated domain 59 Figure 4.5 Partia PDP of the iustrative exampe 60 Figure 4.6 Test vector with a deay offset 6 Figure 4.7 Detection performance with U Figure 4.8 Detection performance with U = x

11 List of Tabes Tabe 2. Avaiabe access sots for the sub-channes 4 Tabe 2.2 Signatures 6 Tabe 2.3 Ce sizes and search window sizes 27 Tabe 3. Search mode and correation mode 32 Tabe 4. Test vector suite parameters 6 Tabe 4.2 Reative computationa compexities of three agorithms with U=3 63 Tabe 4.3 Probabiity of detection for three agorithms with U=3 under various SNR vaues 64 Tabe 4.4 Reative computationa compexities of three agorithms with U=32 66 Tabe 4.5 Probabiity of detection for three agorithms with U=32 under various SNR vaues 66 XI

12 List of Symbos P s C S i g>s S r - P re,n S P re,n,s i j X(i) Xi(i) X2(i) Y n,s 0) Y n, s,i 0) Y n, s,2 0) W p signature s PRACH signature code for signature s nth PRACH preambe scrambing code PRACH preambe code for nth preambe scrambing code and signature s sampe index chip index received antenna signa at sampe i ow-pass fitered antenna signa at sampe i sampe-decimated antenna signa at sampe i Locay generated code sequence at chip/ ow-pass fitered code sequence at chip j sampe-decimated code sequence at chip/ search window size in sampes average of a powers in the search window M non-coherent accumuation ength M.2 non-coherent accumuation ength in the sampe-decimated domain N Coherent accumuation period ength N2 Coherent accumuation period ength in the sampe-decimated domain o deay offset in sampes with respect to the search window origin o org Odec P ns (o) U v a J3 offset in the origina domain offset in the sampe-decimated domain Signa power at the deay offset o for scrambing code n and signature s number of the finaists in the first stage set of deay offsets to be evauated in the second stage scae factor coefficient of detection threshod 9 Detection threshod xii

13 List of Acronyms 2G 3G 3GPP ASC ACK AI AS AWGN AICH BCH CDMA EDGE FD FSFC FSPC GSM GPRS ITU MRC MPS OSF PN PD PDP P-CCPCH P-SCH RACH PRACH Second Generation Third Generation Third Generation Partnership Project Access Service Cass Acknowedgement Acquisition Indication Access Sot Additive White Gaussian Noise Acquisition Indication Channe Broadcast Channe Code Division Mutipe Access Enhanced data rates for GSM evoution Finger De-spreader Fu Search with Fu Contribution Fu Search with Partia Contribution Goba System for Mobie Genera Packet Radio Service Internationa Teecommunication Union Maxima Ratio Combining Muti-path Search Over Samping Factor Pseudo-noise Preambe Detection Power Deay Profie Primary Common Contro Physica Channe Primary Synchronization Channe Random Access Channe Physica Random Access Channe X

14 PSFC PSPC RRC SR SFN SNR UMTS UE WCDMA Partia Search with Fu Contribution Partia Search with Partia Contribution Radio Resource Controer Symbo Rate System Frame Number Signa-to-Noise Ratio Universa Mobie Teecommunication System User Equipment Wideband Code Division Mutipe Access xiv

15 Chapter Introduction. Genera In the past two decades, mobie wireess communications have experienced a tremendous growth. From the first generation anaog mobie wireess networks and phones introduced in the 980s to the second generation (2G) digita mobie wireess communication systems, network capacity and user bandwidth have a grown substantiay. The goba system for mobie (GSM) communications [] and the code division mutipe access (CDMA) based IS-95 [2] are the most commony used technoogies in the 2G systems. To meet the increasing market demand for high bandwidth data services such as web browsing and internet TV, the third generation (3G) wireess teephone technoogies have been deveoped [3]. Compared to the 2G systems, the 3G systems can support a arge number of voice and data users at a higher data rate and a ower incrementa cost. Moreover, the efficient spectrum utiization of the 3G wireess technoogies makes it possibe for the system to support a wide range of wireess mutimedia services and to aow a goba roaming among the different 3G wireess systems [4].

16 One of the most popuar radio air interfaces is the wideband CDMA (WCDMA) [5]-[8]. The requirements for 3G mobie networks with IMT-2000 standard are defined by the Internationa Teecommunication Union (ITU). The Third Generation Partnership Project (3 GPP) has defined a mobie system that fufis the IMT-2000 standard. This system is caed the Universa Mobie Teecommunication System (UMTS) [9], [0]. The UMTS uses WCDMA as the underying air interface. Despite of the promising features of the 3G technoogies, the initia depoyment cost of the 3G systems has deayed a massive market production and depoyment of the UMTS networks, and has chaenged the academic community and the mobie industry to come up with more efficient and effective impementation techniques. In the mean time, the 2G systems have evoved into 2.5G by adapting themseves into enhanced data rates for GSM evoution (EDGE) and genera packet radio service (GPRS) [] techniques. Therefore, 3G mobie wireess communication remains a research attraction for researchers in the fieds of teecommunications, semiconductor and digita signa processing. In a UMTS system, the base station is known as node-b. The mobie user is referred to as user equipment (UE). The downink or forward ink is defined as the direction from node-b to UE, whereas the upink direction or reverse ink is defined as the direction from the UE to the node-b. A radio ink between node-b and UE comprises both a forward ink and a reverse ink. These concepts are iustrated in Figure.. 2

17 Base Station Figure. Iustration of a base station and a user equipment. To estabish a radio ink between a node-b and a UE, the UE needs to foow the procedures defined in the 3 GPP standard [3]. It sends a random access channe (RACH) message to the node-b to request a ca. The node-b has a preambe detector (PD) that continuousy monitors the RACH requests in the entire ce by detecting a preambe signa, which is sent immediatey before the RACH message. Upon a positive detection, the node-b grants the request by sending an acknowedgement (ACK) message through the downink acquisition indication (AI) channe. Upon receiving the ACK message, the UE proceeds with the RACH message transmission and radio ink setup. The PD provides an estimate of the ocation of the UE in the ce. The muti-path searcher takes the estimate of the ocation of the UE as the initia position for the search window, and continuousy finds and updates the ocations of the fingers during the ifespan of the radio 3

18 ink connection. The finger de-spreader (FD), on the other hand, is responsibe for descrambing and de-spreading the fingers, and then combining them to provide the resutant symbos for further symbo rate (SR) processing. The widey-used technique for finger combining is caed maxima ratio combining (MRC) [4]. The symbo rate processing is responsibe for decoding the channe coding and producing the fina binary vaues of the information bits. In order to inform the users within a given ce the system information, the base station sends out the access service cass (ASC) through the broadcast channe (BCH), which is a downink common channe. The ASC specifies which signatures and access sots in the ce are avaiabe for the UE to use. There coud be as many as 6 signatures in an ASC. From the node-b perspective, the preambe detector needs to check a the signature in the ASC at every access sot permitted by the ASC. In addition, since the UE coud be anywhere in the ce, the PD needs to search the entire ce to synchronize with the UE scrambing sequence, eading to a very arge search window. Further, the PD needs to compensate for the phase rotation due to the Dopper effect and carrier frequency shift. Therefore, it is we-known that the preambe detection is the most computationay intensive unit in the base station. Deveoping a fast preambe detection technique that requires a sma amount of correation computation and at the same time retains the detection accuracy of the fu search agorithm is a chaenging probem, which is of significant importance to the teecommunication industry. 4

19 .2 Literature Survey Ever since the introduction of spread spectrum concepts for communications [2]-[6] over thirty years ago, Pseudo-noise (PN) code synchronization has been the focus of attention for many researchers [7]-[33]. The seria search is the most widey used approach for PN code synchronization [28]-[33]. The simpest seria search is the fixed dwe time method [7]. With this method, each of the possibe PN code aignments, aso known as deay offsets or deay hypotheses, is examined for a fixed period of time in a seria fashion. This fixed period of time is caed correation ength. This method is caed singe dwe seria search, aso referred to as fu search with fu contribution in this thesis. This method is easy to impement in the sense that it empoys a fixed correation ength and exhaustivey computes/examines a possibe deay offsets. However, this singe dwe method is computationay intensive. In an attempt to reduce the computationa compexity, muti-dwe seria search has been proposed [3], [34]. In essence, both the preambe detection (PD) and muti-path search (MPS) used in UMTS are PN code synchronization techniques. The probem with the preambe detection, however, is that the detection has to be competed within a fixed time period and has to be based on a burst of sampes. That is why preambe detection is aso caed burst synchronization. Whie muti-dwe search is appicabe to appications such as sateite communication synchronizations and muti-path search in UMTS, it is not very suitabe for the preambe detection used in UMTS. To the best of the investigator's knowedge, no appication of the muti-dwe search for preambe detection has been reported in the iterature. 5

20 Recenty, Sheen et a. proposed a parae-seria code phase detector [35] suitabe for UMTS WCDMA base station. This method is simiar to the singe dwe search in that the period of the time used for cacuating the correation is aso fixed. However, it differs from the singe dwe seria search in that not a the antenna sampes are used for the detection. Specificay, the parae-seria code phase detector can be configured to use part of the 4096 chips of preambe signa for the correation. If we ca the singe dwe seria search using a the 4096 chips of the preambe signa as fu-search with fu contribution (FSFC), then we can ca Sheen's agorithm as fu-search with partia contribution (FSPC). By controing the correation ength of the FSPC agorithm, one can make trade-offs between compexity and detection performance. Thus, the detection accuracy of this method coud be expected to become cose to that of the singe dwe method (FSFC), whie at the same time the computationa compexity very cose to that of the singe dwe method. So far, no fast acquisition method has been reported in the iterature that has detection accuracy cose to that of the singe dwe method yet with substantiay ess computationa compexity..3 Motivation, Scope and Organization of the Thesis Efficient random access is one of the key designs in WCDMA ceuar systems. Detection of the preambe of the RACH message is the soe functionaity of the base station. Fast burst synchronization is essentia in random access in order to avoid excessive access deay and frequent retransmissions that may reduce the overa system capacity. Compared to the muti-path searcher and finger de-spreading, the preambe detection is 6

21 computationay the most intensive unit in a node-b. Thus, the effectiveness of impementing the PD directy affects the cost and performance of a node-b system. Deveoping new and effective PD agorithms remains a chaenging research probem and any possibe innovations regarding PD is probaby a we-guarded industry secret [36], [37]. Therefore, deveoping an efficient PD agorithm for the base station is essentia for a cost effective 3G UMTS network. Unti now, the most popuar PD search is the singe dwe seria search. This can provide the best detection accuracy, since a possibe aignments are exhaustivey searched with the contributions from a the received antenna sampes. With the high speed correation acceerators from Texas Instruments [38], [39], compex vaued sampe by sampe correation can be performed with a speed of 2048 times the chip rate. In other words, such a device has an equivaent of 2048 correators running in parae at the chip rate speed. However, even such a powerfu chip can ony support fewer than 64 mobie users in practice. The demand for an effective PD search agorithm sti remains very high. The mutipe dwe search method has undoubtedy reduced the computationa compexity of the singe dwe search method. But it is not suitabe for PD appications. The parae-seria phase detector based agorithm can hep to reduce the computationa compexity, but its detection accuracy becomes very ow whie the savings in the computationa compexity remain moderate to high. The work of this thesis is an attempt to deveop a technique for preambe detection that woud reduce the computationa compexity and at the same time provide a high detection accuracy. It is known from the 3 GPP standard that the antenna streams are pre- 7

22 processed by a puse shaping fiter before transmission [40]. Such a fitering, aong with channe fading and other interferences, actuay increases the correation among adjacent antenna sampes. That is, the information contained in the adjacent antenna sampes are more inter-reated. The proposed sampe-decimation based PD agorithm expoits the feature of the sampe correations and provides fast detection speed by performing the initia search ony on the sampe-decimated antenna stream and the corresponding ocay generated preambe signa repica. Generay speaking, a preambe detection agorithm consists of cacuating the power deay profie and detecting the presence of a true preambe signa based on the deay profie. In this research, however, we focus on cacuating the power deay profie that constitutes the major part of the preambe detection. It woud be shown that the proposed agorithm yieds amost the same power deay profie as that obtained by using the singe dwe method whie the overa computationa oad is substantiay reduced. This thesis is organized as foows: The basic concepts of preambe signa, correation, coherent accumuation, non-coherent averaging, search window, and power deay profie are reviewed in Chapter 2. In Chapter 3, a sampe-decimation based agorithm, with an objective of achieving a high detection accuracy and ow computationa compexity, is deveoped. In Chapter 4, a simuation study of the proposed agorithm in comparison with the singe dwe method and the parae-seria phase detector method is carried out. Chapter 5 concudes the thesis by highighting the contributions of this study and by providing some suggestions for future investigation. 8

23 Chapter 2 The Fundamentas of a Preambe Detector There are many good references avaiabe concerning the fundamentas of WCDMA and UMTS systems [5], [6], [0]. In this chapter, we focus our attention on the background knowedge of the working principes of random access process empoyed in the UMTS mobie wireess network [4]. 2. Upink Radio Frame Structure At the base station, the radio frequency signa is first demoduated from the carrier frequency signa into a baseband signa. Then this signa is samped and digitized. This digita sequence wi be oosey referred to as antenna signa in subsequent discussions. The signa received at the base station in essence is a deayed version of the transmitted signa by the mobie. To faciitate the subsequent discussions, et's first describe the structure of the antenna signa. An antenna signa (sequence) consists of many radio frames. A radio frame is 9

24 0 ms ong, which is divided into 5 sots. Each sot is further divided into 2560 chips. Thus, the duration of one chip period is 0.26 us. If the over samping factor is OSF=2, then each antenna sampe is 0.3 us in duration. The frame structure is iustrated in Figure ms Figure 2. Frame structure in UMTS. An access sot (AS) is equa to two system sots, thus having 520 chips. In other words, two radio frames compose one access frame which comprises 5 access sots. Since each frame has 5 sots, which is an odd number, access sot 8 straddes over the frame boundary as expected. The concepts of access sot and access frame are iustrated in Figure

25 0 ms Frame 0 Frame Frame 2 Frame 3 L. Access Frame 0 Access Frame Figure 2.2 The structures of an access sot and an access frame. 2.2 Downink Access Sots and AICH Channe The primary common contro physica channe (P-CCPCH) serves as the time reference for a the upink and downink physica channes and signas. Some channes are aigned with P-CCPCH such as the acquisition indication channe (AICH), and the primary synchronization channe (P-SCH). Figure 2.3 shows the timing reationships among the various physica channes and signas according to [42]. It can be seen that for every two radio frames with system frame number (SFN) moduo 2 = 0 and, there are 5 AICH access sots. Access sot #0 corresponds to the frame boundary with SFN moduo 2 = 0.

26 Primary SCH Secondary SCH Any CPICH P-CCPCH Radio framewith(sfn moduo 2) = 0 Radio framewith (SFN moduo 2) = k:th S-CCPCH ts-ccpch.k PICH for k:th S-CCPCH tpich + AICH access sots #0 i # i #2 i #3 i #4 i #5 i #6 i #7 i #8 i #9 i #0, #i #2, #3, #4 n:th DPCH tdpch.n p:th F-DPCH ^ Tf-DPCH.p ^ HS-SCCH subframes Subframe #0 Subframe # Subframe #2 Subframe #3 Subframe #4 0 ms 0 ms Figure 2.3 Access sot timing of downink physica channes. 2.3 Time Reationship between AICH and PRACH The physica random access channe (PRACH) is an upink shared physica channe. It carries the preambe signa and the RACH message. The timing reationship between AICH and PRACH is shown in Figure 2.4. As can be seen, the preambe signas are sent 2

27 in the PRACH access sots which have fixed time offset with respect to the corresponding AICH access sots. For a given AICH access sot, the corresponding PRACH access sot is ahead in time by a period x p. a. In other words, after sending a preambe signa at a given PRACH access sot, the UE woud expect a response from the node-b x p - a time units ater. When node-b sets the configuration parameter AICH_Transmission_Timing equa to 0, x p. a = 7680 chips. When node-b sets the configuration parameter AICHJTransmissionTiming equa to, x p. a = 2800 chips. The RACH message itsef is typicay 0 ms or 20 ms ong depending on the upper ayer configuration contro messages. Here upper ayer refers to MAC ayer that is above the physica ayer where the preambe detection takes pace. AICH access sots RX at UE One access sot Acq. Ind. ^D-a PRACH access sots TX at UE Preambe Preambe Message part Figure 2.4 Timing reationship between AICH and PRACH. 3

28 2.4 RACH Sub-channe A RACH sub-channe is a set of upink access sots, i.e., the PRACH access sots. There are 2 RACH sub-channes. Any one of the upink access sots must beong to one of the 2 sub-channes. Every two system (radio) frames comprise 5 access sots. Thus, eight system frames contain 60 access sots, giving each sub-channe 5 access sots within every 8 system frames. In other words, eight system frames is the period for resource partitioning of the upink access sots. It can be seen from Tabe 2. that sub-channe 0 contains access sots 0, 2, 9, 6, and 3. Tabe 2. Avaiabe access sots for the sub-channes P-CCPCH SFN mod Sub-channe number

29 The reationship between the access sots and the sub-channes is aso iustrated in Figure 2.5. The access sots are numbered from 0 through 4 within every two system frames. The sub-channes are numbered from 0 through. As seen from this figure, eight system frames form a compete period. The access sots beonging to sub-channe 0 are marked as gray in the figure. This corresponds to the second coumn of Tabe 2.. Frame 0 Frame Frame 2 Frame 3 Frame 4 Frame 5 Frame 6 Frame 7 Sub-channe Access sots H BHI Figure 2.5 Reationship between access sots and sub-channes. 2.5 Preambe Signatures A signature is a 6-bit code. There are a tota of 6 signatures in UMTS. The preambe signature corresponding to a signature s consists of 256 repetitions of a signature of ength 6, P s (n), n= This is defined as foows: C S ig,s(i) = P s (i moduo 6), i = 0,,..., 4095 (2.) 5

30 where C S j g;s (i) is the vaue of the preambe signature at chip i corresponding to the signature P s (.). The signature P s (.) is from the set of 6 Hadamard codes of ength 6. These are isted in Tabe 2.2. It can be seen that signature 0 has a 's in it. Tabe 2.2 Signatures Preambe signature Po(n) Vaue of n Pi(n) P 2 (n) - P 3 (n) - P 4 (n) P 5 (n) - P 6 (n) PKn) P 8 (n) - P 9 (n) Pio(n) Pn(n) - Pi2(n) Pi3(n) - Pi4(n) Pis(n) Access Service Cass In a UMTS network, mobie users are managed through access service cass (ASC). Each 6

31 user is associated with an ASC which specifies the set of signatures and the set of subchannes avaiabe for this cass. Figure 2.6 shows an exampe of ASC in which ASCO has signature in it and has a the sub-channes. On the hand, the ASCI has signatures 4 and 5 and sub-channes 2 and 3 in it Signatures CD c CD sz V n D CO ASC 9 0 Figure 2.6 Access service cass. 2.7 Random Access Procedure in UMTS Reference [4] describes in detai the compete random access procedure. For the sake of competeness; however, a simpified version of the procedure is summarized beow. 7

32 Before the physica random-access procedure is initiated, the physica ayer, aso known as Layer, receives the foowing information from the higher ayer which is the radio resource controer (RRC): The preambe scrambing code. The message ength in time, either 0 ms or 20 ms. The AICH_Transmission_Timing parameter, 0 or. The sets of avaiabe signatures and RACH sub-channes for each ASC. The power-ramping factor, Power Ramp Step. The parameter Preambe RetransMax. The initia preambe power, Preambe_Initia_Power. The physica random-access procedure is performed as foows: Derive the avaiabe upink access sots in the next fu access sot set for the set of avaiabe RACH sub-channes within the given ASC. Randomy seect one access sot among the ones previousy determined. If there is no access sot avaiabe in the seected set, randomy seect one upink access sot corresponding to the set of avaiabe RACH sub-channes within the given ASC from the next access sot set. 8

33 2 Randomy seect a signature from the set of avaiabe signatures within the given ASC. 3 Set the Preambe Retransmission Counter to Preambe Retrans_Max. 4 Transmit a preambe using the seected upink access sot, signature and preambe transmission power. 5 If no positive or negative acquisition indicator (AI * +, -) corresponding to the seected signature is detected in the downink access sot corresponding to the seected upink access sot: (a) Seect the next avaiabe access sot in the set of avaiabe RACH subchannes within the given ASC. (b) Randomy seect a new signature from the set of avaiabe signatures within the given ASC. (c) Increase the Commanded Preambe Power by APo = Power Ramp Step in db. If the Commanded Preambe Power exceeds the maximum aowed power by 6 db, the UE may pass LI status ("No ack on AICH") to the higher ayers (MAC) and exit the physica random access procedure. (d) Decrease the Preambe Retransmission Counter by one. 9

34 (e) If the Preambe Retransmission Counter > 0 then repeat from step 5. Otherwise pass LI status ("No ack on AICH") to the higher ayers (MAC) and exit the physica random access procedure. 6 If a negative acquisition indicator corresponding to the seected signature is detected in the downink access sot corresponding to the seected upink access sot, pass LI status ("Nack on AICH received") to the higher ayers (MAC) and exit the physica random access procedure. 7 If positive acquisition indicator is detected, transmit the random access message three or four upink access sots depending on the AICH transmission timing parameter after the upink access sot of the ast transmitted preambe. Pass LI status "RACH message transmitted" to the higher ayers and exit the physica random access procedure. Figure 2.7 shows an exampe of the RACH procedure described above. The UE first sends a preambe at access sot 0 with the minima transmission power and a randomy seected signature from the given ASC, and expects a response from the node-b at AICH access sot 0. The UE sends another preambe with increased power at PRACH access sot 3 with another randomy seected signature from the ASC, since no response (ACK or NACK) is detected at AICH access sot 0. This retransmission is repeated at PRACH access sots 6 and 9, since no response from node-b is detected at AICH access sots 3 and 6. Finay, at AICH access sot 9, an ACK (AI=) is detected by the UE. Thus, UE starts to transmit RACH message at PRACH access sot 2. 20

35 preambes message PRACH r~h i D^ : 0 \ 2/ 3 4\ 5/ 6 X 8/ 9 NO / 2 AICH ACK Figure 2.7 Power ramping scheme in RACH procedure. 2.8 Preambe Signas The preambe signas sent by a UE for a given scrambing code and signature are generated using the formuas given in the foowing sub-sections [43] Long scrambing sequence The ong scrambing sequences ci on g,i,n is constructed from the position wise moduo 2 sum of chip segments of two binary m-sequences generated by means of two generator poynomias of degree 25. Let x and y be the two m-sequences respectivey. The x sequence is constructed using the primitive poynomia X 25 +X 3 +. The y sequence is constructed using the poynomia X 25 +X 3 +X 2 +X+. The resuting sequences thus constitute segments of a set of God sequences [43]. Let n23... no be the 24-bit binary representation of the scrambing sequence number n with no being the east significant bit. The x sequence depends on the chosen scrambing 2

36 sequence number n and is denoted x n, in the seque. Furthermore, et x n (i) and y(i) denote the i th symbo of the sequence x n and y, respectivey. The m-sequences x n and y are constructed as foows: Initia conditions: x (0) =n 0,x n () = ni,... =x (22) = n 22,x (23) = n 23, x n (24)= (2.2) y(0)=y()=...=y(23hy(24)= (2.3) The recursive reations for the symbos are given by x n (i+25) =x n (i+3) + x n (i) moduo 2, i=0,..., 2-27 (2.4) y(i+25) = y(i+3)+y(i+2) +y(i+) +y(i) moduo 2, i=0,..., (2.5) Define the binary God sequence z n by: z n (i) = x n (i) +y(i) moduo 2, i = 0,, 2,..., 2-2 (2.6) The rea vaued God sequence Z n is defined by [+ ifzm) = 0 J n J Z(i) = \ " W ' ifz (i) = for i = 0,,.--, (27) { n Finay, the rea-vaued ong scrambing sequences ci ong,i,n is defined as foows: ciongjji) = Z n (i), i = 0,, 2,..., (2.8) 22

37 2.8.2 Preambe scrambing code The scrambing code for the PRACH preambe part is constructed from the ong scrambing sequences. In tota, there are 892 PRACH preambe scrambing codes. The nth preambe scrambing code, n = 0,,...,89, is defined as S r -pre,n(i) = C ong J, n (i), = 0,,..., 4095 (2.9) It shoud be noted here that the preambe scrambing code is ony 4096-chip ong even though the ong scrambing sequence ci 0 ng,i,n(-) is (2 25 )-chip ong Preambe signas The random access preambe signa or code, C pre)n, is a compex vaued sequence. It is buit from a preambe scrambing code S r _ pr e, n and a preambe signature C s i g;s as foows: C pre,n,s(k) = S r. pre, n (k) X C si&s (k) X J^^, k = 0,, 2, 3,..., 4095 (2.0) where k=0 corresponds to the chip transmitted first in time and S r. pre,n and C S i g>s are defined in (2.9) and (2.), respectivey. 2.9 Preambe Detection at Node-B The objective of a preambe detector is to find the presence of a vaid preambe signa in the antenna stream. The detector is responsibe for checking a the access sots that are 23

38 broadcasted to the users in their ASC. For each permissibe access sot, a the signatures permitted shoud be checked Correation The principe of the preambe detection is buit on the assumption that the antenna stream of the node-b contains a time deayed version of the transmitted signa at the UE with added transmission channe impairments incuding fading and additive white Gaussian noise. If the receiver at the node-b coud ocay generate a repica of the transmitted signa by the UE, then the correation between the received antenna stream and the compex conjugate of the repica wi give a strong autocorreation power, signifying the presence of an expected preambe signa. If the expected signature is not in the preambe signa or the aignment between the repica and the received signa is not correct, the correation wi give a weak power. Mathematicay, the correation operation can be expressed as foows. The received antenna signa is denoted as X(i) - Ix(i) + Qx(i), where the index i is the sampe position. The over-samping factor, OSF =2, gives a tota of 892 sampes for a compete preambe signa code of 4096 chips as shown in Figure 2.8. ' _X(0) Chi p 0 Chip ; ; X() X(2) X(3) Chip 4095 X(890)X(89) Figure 2.8 An iustration of over-samping of an antenna stream. 24

39 The ocay generated code denoted as Y n!s (j) = IY(J) + QY(J)> J = 0,, 2, 3,..., 4095 is the compex conjugate of C pr e, n,sg) which denotes chip j of the preambe signa with preambe scrambing code n and signature s. Y n>s (j) = C* pre, n M J = 0,, 2, 3,..., 4095 (2.) The correation between X and Y can be expressed as foows: 4095 accihao) = x(2y + o)y ntg (J) (2.2) where o is the offset in sampes between the input sampe stream and the ocay generated code. Since both X and Y are compex vaues, acciq n s (o) is aso compex-vaue. To find the power of the correation, we take the square of the ampitude of acciq n s (o) as foows: PnA ) =acc QnA ) 2 (2-3) Due to carrier frequency shift and the Dopper effect, there is a phase rotation between the received signa and the oca repica of transmitted signa. Moreover, the phase rotation changes over time. This has a negative effect on the correation towards the overa power accumuation when the correation ength (period) is too ong. Therefore, in practice, the compete accumuation period is divided into severa smaer 25

40 accumuation intervas. At the end of each of the intervas, the power of the intermediate resut is cacuated, and the tota accumuation is obtained as foows: M-\ P nm=y, m=0 JV- Y,X(2mN + 2j + o)y n<s (mn + j) 7=0 (2.4) where MN = The inner summation is caed coherent accumuation, and the outer summation as noncoherent accumuation. In this case, the coherent accumuation ength is N, and that of the non-coherent M Search window As stated earier, the received signa is a time deayed version of the origina transmitted signa. The amount of deay is determined by the distance between the user equipment and the base station. Quantitativey, deay by chip in time corresponds to 0.26 (j,s x m/s = 78 m in distance of propagation, since the radio wave traves at the speed of ight. The uncertainty region of the deay in time is caed the search window for the preambe detection. The node-b has to search the entire ce to determine if there is a preambe existing in the ce, since it has no a priori knowedge of the ocation of the UE. Therefore, the physica ce size determines the search window size of the PD. Tabe 2.3 gives the reationship between the physica size of a ce and the search window size of the PD. 26

41 Tabe 2.3 Ce sizes and search window sizes Ce size (km) Search window size (in chips) Search window size (in sampes) Power deay profie Within the search window, the node-b needs to find out the correation power at every possibe deay ocation in time. Each of the deay ocation is caed a deay offset with respect to the origin of the search window. The resuting reationship between the deay offsets and the correation powers is caed the power deay profie (PDP) for a given signature. If the expected signature does exist in the received antenna signa, then the PDP woud contain a peak power at some deay offset. 27

42 2.0 Threshoding There are many compicated methods to decide how to seect and appy the threshod on the PDP. No attempt is made to study a the existing threshoding methods since the focus of this thesis is to deveop an agorithm for a fast cacuation of the PDP. However, a commony used method that is empoyed in this thesis is briefy introduced here. The threshod 0 is defined as the average of powers at a the offsets within the search window and it is given by a w ~ x 0 = ^Y. P na ) (2.5) where W is the search window size in sampes and a is the scae factor. 2. Probabiity of Detection and Probabiity of Fase Aarm This subsection defines the concepts of the probabiity of detection and probabiity of fase aarm. A good detector shoud have a high probabiity of detection and a ow probabiity of fase aarm at the same time. (a) Probabiity of detection The probabiity of detection is defined as the ratio between the number of correct detections and the tota number of tests conducted. 28

43 (b) Probabiity of fase aarm The probabiity of fase aarm is defined as the ratio between the number of times a preambe is detected and the tota number of tests conducted when the signature is not present in the preambe. 2.2 Summary RACH preambe detection is an important function of the node-b in a UMTS network. The basic system eve knowedge that is reated to RACH preambe detection has been introduced in this chapter. Fundamenta concepts such as frame structure, access sot, signature, access service cass have been defined. The ce size of a node-b decides the search window size of the preambe detection. The correation ength affects the computationa oad and detection accuracy. To assess the performance of a RACH preambe detector, both the probabiity of detection and probabiity of fase aarm have to be taken into consideration. 29

44 Chapter 3 The Proposed Sampe-Decimation Based Agorithm 3. Introduction As described in Chapter 2, the computationa compexity of a preambe detector is affected by the ce size and the correation ength. The ce size is decided by the network depoyment needs, and thus, cannot be controed by a preambe detector agorithm. The correation ength, on the other hand, coud be reduced from the maxima ength of 4096 chips. However, this woud resut in affecting the detection accuracy. For a macro base station, where the ce size is typicay 0 km or arger and the number of UEs in a ce is 200 or more, the chance to have many users to send the preambe in the same access sot is very sim. This means, most of the deay offsets wi correspond to very ow correation power. In other words, most of the deay hypotheses are wrong. It is, therefore, desirabe to quicky discard the majority of the deay hypotheses (offsets) that are unikey to correspond to a strong correation power. For those offsets that 30

45 have strong correation power, a the avaiabe antenna sampes of 4096 chips shoud be used to perform the correation, thus giving the best detection accuracy. In order to expoit the fact that most of the deay offsets are not going to ead to strong correation powers, one can perform the cacuation of the PDP in two stages. The first stage performs a coarse search in the entire ce, and gives a ist of offsets that are ikey to have a strong correation power. The second stage then performs a fine search ony on those offsets that are seected by the first stage. The second stage utiizes the fu correation ength when the PDP is cacuated, thus preserving the accuracy of the PDP cacuation on those seected offsets. To faciitate the description of the proposed agorithm, the foowing terms and their abbreviations need to be introduced. As shown in Tabe 3. fu search with fu contribution (FSFC) refers to the conventiona singe dwe seria search method, where correation powers of a the deay offsets are cacuated with a correation ength equa to 4096 chips. Fu search with partia contribution (FSPC) refers to the search, where correation powers of a the deay offsets are cacuated with a correation ength smaer than 4096 chips. Partia search with fu contribution (PSFC) is refers to the search, where correation powers of some of the deay offsets are cacuated with a correation ength equa to 4096 chips. Finay, partia search with partia contribution (PSPC) refers to the search, where correation powers of some of the deay offsets are cacuated with a correation ength smaer than 4096 chips. 3

46 Tabe 3. Search mode and correation mode Search mode Correation (contribution) mode Fu Partia Fu FSFC FSPC Partia PSFC PSPC 3.2 Proposed Agorithm The proposed agorithm consists of two stages. In the first stage, a sampe-decimated domain is created by downsamping the input antenna sampes and the ocay generated repica of the transmitted signa, aso known as ocay generated code. Then, an FSFC is performed in the sampe-decimated domain. The PDP in the sampe-decimated domain is then used to indicate as to which offsets are ikey to have strong correation powers. This heps to narrow down the potentia ocations of the preambe with sma amount of computation. Then, in the second stage, the correation between the input sampe stream and the ocay generated code is performed in the origina sampe domain, but, the correation is restricted in the vicinities of those ocations that are ikey to have strong correation powers. Thus, second stage performs a PSFC. Detais of the agorithm are described in the foowing sub-sections. 32

47 3.2. Correation in the sampe-decimated domain The first step is to generate sampe-decimated antenna sampes and ocay-generated code. But before downsamping the antenna sampes and ocay generated code, a ow pass fitering is performed on the origina antenna sampes and ocay generated code. This heps to increase the association or correation among adjacent sampes, and thus gives better correation resuts in the sampe-decimated domain. (a) Preprocessing The origina input antenna sampes are represented by {X(i)}. The over samping factor (OSF) is equa to 2, thus there are two sampes in each chip period, and 892 input antenna sampes woud correspond to 4096 chips in time. The input sampe stream is first fitered with a ow pass fiter as foows: X x (i) = - (X(i - 2) + 2X(i -) + 3X(i) + 2X(i +) + X(i + 2)) (3.) 9 where i = 2,3,..., 892+W-3, and for other vaues of i, X (i) = 0, W bring the search window ength in sampes. It shoud be noted that in a rea system, the input antenna stream contains a constant fow of sampes at the chip rate speed with 3.84 x 0 6 chips per second. In describing the proposed agorithm, however, ony the first 892+W sampes are of concern. Next, the fitered sampe stream is down samped with a 4: ratio as foows: 33

48 X 2 (k) = X x (4k +), k = 0,,..., W/4- (3.2) Simiary, the ocay generated code which is 4096 in ength is preprocessed as foows: Y. J AJ)=fa,(j) + 2Y Jj) + Y, J U + )) (3.3) where j-.2,..., 4094, and W > =, W<>95) = 0 The fitered code is then down samped at a 2: ratio as foows: Y n, s,2 (*) = r,,,i (2k +), k = 0,,..., 2047 (3.4) It shoud be noted that the origina {X(i)} sequence is in the haf-chip resoution, whereas the origina code sequence {Y(j)} is in the chip resoution. That is why, 4: down samping is performed on the fitered antenna sampe sequence and 2: down samping is performed on the fitered code sequence. After the processing, {X2(k)} and {Y n, S) 2(k)} are in the 2-chip per sampe resoution. (b) Correation We refer to X 2 (k) and Y n^2 (k) as being in the sampe-decimated domain. The correation between X 2 {k) and Y ns2 (k) is then performed as foows: M, ^»=I m=0 N 2 -\ x 2 (m7v 2 + j)y ns2 (o + mn 2 + j) (3.5) 34

49 where M2N2 = 2048, o = 0,,.., W/4-. There are different possibe choices for the vaue of M2 and N2. One possibe choice is M2 = 4 and N2 = 52. (c) Sorting The correation operation mentioned above gives the PDP in the sampe-decimated domain. These offset-power pairs are then sorted according to the magnitude of the power. Then the U pairs corresponding to the U argest powers are seected as the finaists of the first stage. The vaue of U is programmabe. These candidates can be expressed by the set {(oi,pi), (o 2,P 2 ),., (ok,pk),-, (Pu,Pu)}, where (o k,pk) represents the power at offset Ok as Pk=P n,s(ok). A the steps invoved in the first-stage processing are iustrated in Figure 3.. Both the input antenna signa and the ocay generated code sequence are first ow-pass fitered. It shoud be noted that the antenna signa X is over-samped with OSF=2 as the input. Thus, every two sampes correspond to chip in time. On the other hand, the ocay generated signa Y not over-samped, i.e., OSF=. After the ow-pass fitering, the signas are down-samped to generate the sampedecimated signas. In the sampe-decimated domain, every signa sampe corresponds to two chips in time, or 0.52 us. This hods true for both the decimated antenna sampes and the decimated code. The origina search window size is W corresponding to W/2 chips in time. Now, the search window size in the decimated-sampe domain is W/4, which sti corresponds to W/2 chips in time. 35

50 Mi PDP correation ^T^ Downsamping A A ( «\ j \M f.,:. *: - i* *--.-. ;.. -, S '.IV Low-pass fitering Figure 3. Operations performed in the first stage. Performing correations between two decimated signas within the corresponding search window produces the PDP in the decimated signa domain. Finay, if U is set to 3 in the iustration of Figure 3., then the finaists woud consist of {(8,P 8 ),(22,P 2 2),(26,P 26 )}. (d) Average During the process of sorting the PDP, the average of a the powers in the search 36