1. April 13, 2005Topic 4 1 Telecommunications Engineering Topic 4: Spread Spectrum and CDMA James K Beard, Ph.D. jkbeard@temple.edu http://astro.temple.edu/~jkbeard/

2. Topic 42 April 13, 2005 Attendance

3. Topic 43 April 13, 2005 Essentials Text: Simon Haykin and Michael Moher, Modern Wireless Communications SystemView  Use the full version in E&A 603A for your term project Web Site  URL http://astro.temple.edu/~jkbeard/http://astro.temple.edu/~jkbeard/  Content includes slides for EE320 and EE521  SystemView page  A few links Office Hours  E&A 349  Hours Tuesday afternoons 3:00 PM to 4:30 PM  MWF 10:30 AM to 11:30 AM  Others by appointment; ask by email

4. Topic 44 April 13, 2005 Topics Today we explore the third tool  FDMA, uses separate channels for each user  TDMA, uses time multiplexing to time multiplex the channel between users  Now, CDMA with spread spectrum enables multiple simultaneous users of the channel Direct-sequence modulation Spreading codes Code synchronization

5. Topic 45 April 13, 2005 Direct Sequence Modulation We begin with BPSK or QPSK We replace the simple pulse shape  Each “pulse” is a more complex wide band pulse  The bandwidth of the resulting signal is that of the new wide band pulse Spectrum of new signal is given by the convolution theorem

6. Topic 46 April 13, 2005 Base Performance Equations

7. Topic 47 April 13, 2005 Performance in Noise Base equation (Hayken & Moher equations (5.12) page 263, (E.11) page 518) Adding spreading function -- E b and N 0 are invariant through matched filter

8. Topic 48 April 13, 2005 Performance in Interference Consider a tone as interference In base coded signal  Matched filter spreads tone over channel  Tone energy becomes part of noise floor In spread spectrum signal  Matched filter spreads tone over channel  Effective additional noise reduced by spreading factor

9. Topic 49 April 13, 2005 Spreading Codes and CDMA Common method is to use a code for each pulse in a signal This is the spreading code CDMA is achieved when the spreading code is one of an orthogonal set for each user of the channel

10. Topic 410 April 13, 2005 Spreading Codes and CDMA Use a coded pulse for each bit in the message The coded pulse is the symbol-shaping function Make the code one of an orthogonal set for each user of the same broadened channel Result  BER performance is unchanged for each user  Users of other spreading codes look like the noise floor

11. Topic 411 April 13, 2005 The Symbol-Shaping Function

12. Topic 412 April 13, 2005 Walsh-Hadamard Sequences A simple way to formulate orthogonal code sequences Based on recursive augmentation of Walsh-Hadamard matrices

13. Topic 413 April 13, 2005 Properties of Walsh-Hadamard Sequences Matrices are symmetrical Matrices are self-orthogonal Each matrix has rows or columns are a sequence of orthogonal sequences of length 2 k Cross-correlation properties  Excellent for zero lag  Poor for other lags

14. Topic 414 April 13, 2005 Maximal-Length Sequences Bit sequence is essentially random Pseudo-random noise (PRN) code Codes Construction  Shift registers with feedback  Recursive modulo-2 polynomial arithmetic PRN codes are then selected for good cross-correlation properties

15. Topic 415 April 13, 2005 Desirable PRN Code Properties Maximal length – 2 m codes before repeating Balance – equal number of (+1) and (-1) pulses Closed on circular shifts Contain shorter subsequences Good autocorrelation properties

16. Topic 416 April 13, 2005 Galois Field Vector Extensions of Order 2 m Polynomials modulo 2 of order m-1 Arithmetic is done modulo a generating polynomial of the form Proper selection of generating polynomial  Sequence of powers produces all 2 m elements  Set is closed on multiplication

17. Topic 417 April 13, 2005 An Important Isomorphism Shift registers with feedback  Bits in shift register are isomorphic with polynomial coefficients  Shift is isomorphic with multiplication by x  Modulo the generating polynomial is isomorphic to multiple-tap feedback Shift registers with feedback can produce a Galois field in sequence of powers of x These codes are also called m-sequences

18. Topic 418 April 13, 2005 Gold Codes R. Gold, optimal binary sequences for spread spectrum multiplexing, IEEE Trans. Inform. Theory, Vol. IT-14, pp. 154-156, 1968. Based on summing the output of two m- sequence generators

19. Topic 419 April 13, 2005 Code Synchronization Two phases  Recover timing  Recover phase  Timing must be recovered first To recover timing  Use code bits known to be 1’s  Matched filter for symbol-shaping function  Step timing in increments of T c until match is found

20. Topic 420 April 13, 2005 Assignment Read 5.2, 5.3, 5.5, 5.7, 5.11, 5.15 Do problem 5.7 p. 273 Next time  Power control  Frequency hopping  An example

21. Topic 421 April 13, 2005 Chinese Remainder Theorem Over numbers from 0 to 2. 3. 5=30 The method works when N has no repeated prime factors Arithmetic advantages?

22. Topic 422 April 13, 2005 A Finite Field Integers mod a prime A reciprocal of a positive integer always exists Addition, subtraction, multiplication, division, all defined and commutative

23. Topic 423 April 13, 2005 Power Control and CDMA The near-far problem  The spreading loss will vary up to 70 dB over the coverage area  Code rejection factors are usually less than this  Result is that interference can occur between closely- spaced handsets or near base stations Solution is power control  Reduce handset power to make received power constant

24. Topic 424 April 13, 2005 Frequency Hopping Definition: Changing from channel to channel at regular intervals Mitigates these problems  The near-far problem between handsets  Narrow band interference But, non-coherent detection is necessary Advantages also include  Full and best use of available spectrum for QoS  Can be combined with spread spectrum (FH-SS)

25. April 13, 2005Topic 4 25 EE320 March 28

26. Topic 426 April 13, 2005 Topics Term Project Problem 5.1 p. 262 Problem 5.17 p. 299 Theme Example: WCDMA

27. Topic 427 April 13, 2005 The Term Project Continue with the start that you turned in with the first quiz backup  Input Frequency sweep 1000 Hz to 3500 Hz Noise to obtain 20 dB SNR  Sampling to obtain good performance Do NOT pitch your beginning and pick up the ADC to bitstream modules as a template Sample and encode/decode as instructed Measure BER vs. Eb/N0 as instructed Compare hard decoding with soft decoding

28. Topic 428 April 13, 2005 Problem 5.1 p. 262 What is the equation for the spectrum of the spreading sequence given by Eq. (5.5) p. 261? The chips c(q) are +1 or -1 and the chip shape g c (t) is

29. Topic 429 April 13, 2005 Use the Convolution Theorem The spreading sequence is The Fourier transform of each term in the sum is

30. Topic 430 April 13, 2005 Problem 5.17 p. 299 Do you expect FEC codes to have a greater or lesser benefit in Rayleigh-fading channels? Discuss your answer Rayleigh fading channels have higher BER than otherwise similar Gaussian channels – more opportunity for improvement Interleavers are necessary to make sure that d free or fewer bits are exposed in a coherency interval

31. Topic 431 April 13, 2005 WCDMA (1 of 3) From Theme Example 4 pp.323-328 Cell phone technology generations  First: analog cell phones  Second: TDMA, IS-95, GSM  Third: Universal Mobile Terrestrial Telecommunications systems (UMTS) WCDMA is a UMTS

32. Topic 432 April 13, 2005 WCDMA (2 of 3) Functional differences  Simultaneous voice and data transmission  Other data such as real-time TV Performance improvements  Three times the bandwidth  Four times the maximum spreading factor  Optional turbo codes

33. Topic 433 April 13, 2005 WCDMA (3 of 3) Other differences  Multiple simultaneous CDMA downlink  Downlink power control  Asynchronous base stations Bottom line  Broadband or ISDN in a cell phone  Near-far problems mitigated  Higher density of base stations and users

34. Topic 434 April 13, 2005 Problem 5.19 page 305 (1 of 3) Define the cellular spectral efficiency nu, in bits/second/Hz/cell; this is the total number of bits/second/Hz transmitted by all users in a cell. For a QPSK base modulation, assume that the spectral efficiency of a single CDMA user is 1/Q bits/second/Hz, where Q is the length of the spreading code. Suppose the receiver requires a specified SINR. Using Eq. (5.85) page 304, develop an expression for nu that depends on the received I 0 /N 0, SINR, and f. Whay does the result not depend explicitly on Q? How does it depend implicitly on Q?

35. Topic 435 April 13, 2005 Problem 5.19 page 309 (2 of 3) The spectral efficiency for  K users in the cell  Each transmitting 2/Q bits/second/Hz From Eq. (5.85) page 304

36. Topic 436 April 13, 2005 Problem 5.19 page 309 (3 of 3) Rolling up these two equations gives nu as The spreading factor Q influences  The interference factor f  The interference to noise ratio I 0 /N 0

37. Topic 437 April 13, 2005 Theme Example 1: IS-95 Section 5.12 Page 311 Wireless cellular generations  Analog systems  Initial digital systems – GSM, IS-54, IS-95  Integrated voice and data systems Cell bands  Uplink 869-894 MHz, downlink 24 MHz lower  Uplink 1930-1990 MHz, downlink 80 MHz lower

38. Topic 438 April 13, 2005 IS-95 Specifications and Usage Most CDMA cell phones use the IS-95 standard Data rate is 9.6 kbps  Mainly voice  Some data, trend is increasing amounts Direct sequence spread to 1.2288 megachips per second Channel bandwidth is 1.25 MHz Emerging standard based on IS-95 is CDMA2000

39. Topic 439 April 13, 2005 Channel Protocol of IS-95 Making an IS-95 call – the Mobile Terminal  Searches for Pilot channel and synchronizes with it  Locks to the Sync channel that is synchronized with the Pilot channel, and gets system information (spreading code) of the access and paging channels  Sends a request to set up a call to the Access channel  Listens to Paging channel for traffic channel assignment  Transmits up assigned uplink channel, receives on assigned downlink channel

40. Topic 440 April 13, 2005 Channel Protocol of IS-95 Receiving an IS-95 call – the Base Station  Transmits a short message on the paging channel  Accepts Mobile Terminal request for call Differences  Request for call has the phone number to initiate a call  Paging channel has Mobile Terminal phone number in the paging message

41. Topic 441 April 13, 2005 What The Pilot Channel Is Shared by all users of the base station Transmitted at higher power than the data channels – about 20% of total power Unmodulated signal – no CDMA here Provides fast synch and reliable channel tracking to support coherent demodulation and robust CDMA Mobile terminal  Tracks the pilot channel of the current cell  Searches for other pilot channels  Switches cells when another pilot signal is stronger  Transparent to the user

42. Topic 442 April 13, 2005 The Four Downlink Channels Separated by use of Walsh-Hadamard codes of length 64  Pilot used Walsh #0  Sync uses Walsh #32  Paging using Walsh #1  Traffic uses one of the other codes See Figure 5.29 page 314

43. Topic 443 April 13, 2005 The Traffic Channel Multiplexed with control bits for power control Rate ½ FEC encoded and interleaved Scrambling with long code sequence follows interleaving (42 bits) Block diagram in Figure 5.30 page 315

44. Topic 444 April 13, 2005 Problem 5.2 Page 263 Filtering with an integrate-and- dump filter is equivalent to convolving with a rectangular pulse of length T. Show, by using Parseval’s theorem, that the noise bandwidth of an integrate-and- dump is 1/T.

45. Topic 445 April 13, 2005 Parseval’s Theorem For Fourier transform pair see Table A.2 p. 482 For Parseval’s theorem see Eq. (A.36) p. 491

46. Topic 446 April 13, 2005 Noise Bandwidth Definition: ratio of  The variance of the output of a transfer function to a white noise with two-sided power spectral density N 0 /2  The power spectral density N 0 Equation

47. Topic 447 April 13, 2005 Power Control: The Near-Far Problem Haykin & Moher Section 5.7 pp. 294-297 Received signal from K CDMA transmitters is, from Eq. (5.38) p. 279

48. Topic 448 April 13, 2005 SINR of First User More detail in 5.4.1 pages 279-283

49. Topic 449 April 13, 2005 Degradation in Multi-User Performance

50. Topic 450 April 13, 2005 FEC Coding and CDMA Haykin & Moher Section 5.8 pp. 297-299 Direct Sequence Spread Spectrum (DS- SS) spreads spectrum without added redundancy Use of FEC spreads spectrum and adds redundancy

51. Topic 451 April 13, 2005 Spreading Rate and Degradation The maximum spreading rate is Degradation in multi-user performance is

52. Topic 452 April 13, 2005 Example 5.5 Pages 298-299 Suppose a system has an information rate R b =4800 bps and Q=R c /R b =32. The system is error protected by a rate- 1/2 convolutional code. Compare the degradation D b with and without FEC coding at a BER of 10 -5 when there are seven interfering users.

53. Topic 453 April 13, 2005 Base Parameters Without FEC Encoding

54. Topic 454 April 13, 2005 Base Parameters With FEC Encoding With rate-1/2 constratin-length 7 convolutional FEC encoding BER improved to 10 -5 with E b /N 0 decreased to 4.5 dB Result is D g increased to 0.62 Improvement is about 2.7 dB

55. Topic 455 April 13, 2005

56. Topic 4 56 EE320 Telecommunications Engineering James K Beard, Ph.D. jkbeard@temple.edu E&A 349

57. Topic 457 April 13, 2005 Quiz 2 Not a difficult quiz Some problems were a slight variation of the text material such as substitution of one code for another I allowed 2 ½ hours for a 50-minute quiz The curve from this quiz should be definitive

58. Topic 458 April 13, 2005 Crunch Time We have about four weeks left  Last day of class is Monday May 2  Final exam is Monday May 11, 11:00 AM - 1:00 PM Some of you are in trouble  Some quiz grades are low  Not everyone will pass  Department has been notified

59. Topic 459 April 13, 2005 Watch for a Warning If you are heading toward a grade lower than C  You will receive a warning with your Quiz 2 grade  The cover page of your quiz  Your minimum Final Exam grade will be given Final Exam  By the book  Two hours, no talking

60. Topic 460 April 13, 2005 The Problem EE320 Telecommunications Engineering is…  A tough course  A required course  Material packed with new concepts and technology But the Perception of some is…  An easy course  A required course that everyone will pass  Watch the slides, read the text before each quiz, and everything will be OK

61. Topic 461 April 13, 2005 The Solution Take notes  Regular class notebook  On the slides  The act of taking notes helps retention Study a little  Even the best student needs to do two or three homework problems per chapter  The study guide can help you pick them Do well on the Final Examination  A good grade there can bring up your final grade  Don’t wait until Study Day to catch up on four courses

62. Topic 462 April 13, 2005 Problem 5.3 page 265 Fill in the missing details of Eq. (5.19) A non-spread link The jammer is on for T seconds Spectrum  After multiplying by de-spreading sequence  Development as in Eq. (5.17)  Eq. (5.17) with T c ->T, Q->1 because of no spreading

63. Topic 463 April 13, 2005 Problem 5.3 page 265 Jammer spectral density at baseband is The noise bandwidth of an integrate-and-dump is 1/T (see problem 5.2) The noise variance is

64. Topic 464 April 13, 2005 Properties of m-Sequences Length property: Each m-sequence is of length 2 m -1 Balance property: Each m-sequence has 2 m-1 ones and 2 m-1 -1 zeros Shift property: The modulo-2 sum of an m-sequence and any circularly-shifted version of itself produces another circularly-shifted version of itself Subsequence property: Each m-sequence contains a subsequence of 1, 2, 3,…,m-1 zeros and ones Autocorrelation property: See Equations (3.30) and (3.31) pages 272, 272

65. Topic 465 April 13, 2005 Problem 5.7 page 272 Prove the autocorrelation property of Eq. (5.31) for m-sequences (Hint: use preceding Properties 1 through 5 as needed.) Eq. (5.30) and (5.31) pages 271 and 272

66. Topic 466 April 13, 2005 Proof of Autocorrelation Property of m-Sequences By the Shift property, the circular autocorrelation, a modulo-2 sum of an m- sequence and a circularly-shifted version of itself, is another circularly-shifted version of iteslf From the Balance property an m- sequence has one more 1 than zeros. QED

67. Topic 467 April 13, 2005 Course Material Overview The problem  Make a cell phone system work  Deal with mobile terminals  Deal with urban fading The solution  Construct the network layer infrastructure  Dispense the data link layer mobile terminals  Exploit the physical layer successfully

68. Topic 468 April 13, 2005 Designing the Data Link Layer The problem  Addressing multiple users Data push – making calls Data pull – accepting calls SINR  Dealing with urban fading The solution  IS-95 and other 2 nd generation standards  Third generation standards

69. Topic 469 April 13, 2005 How IS-95 Meets the Challenges 64 Walsh-Hadamard codes Uplinks 45 MHz below downlinks Synchronization  The pilot channel (Walsh code 0) allows coherent detection  The synchronization channel (Walsh code 32) provides spreading codes of access and paging channels Paging channel (Walsh code 1) assigns access channel Access channels (other Walsh codes)

70. Topic 470 April 13, 2005 Meeting the Challenge of Fading Forward Error Correction Codes (FECs)  Allows robust operation with high bit error rates (BER) Spread spectrum  Allows higher BER in frequency-selective fading Interleaving  Helps bridge dropouts from fading and interference over intervals of a few milliseconds

71. Topic 471 April 13, 2005 Meeting the Challenge of Higher Traffic and New Uses Use of CDMA to allow channel sharing Use of power control to limit SINR at the base station In next-generation standards such as Universal Mobile Terminal Terrestrial Telecommunication Systems (UMTSs)  Base station power control to limit SINR at the mobile station  Higher bandwidths and data rates  More sophisticated coding to approach Shannon channel limit  More versatile data formats for text, video, etc.

72. Topic 472 April 13, 2005 Problem 5.8 Page 274 To show that scramblers based on m- sequences are not very good encryption devices, determine the minimum number of consecutive bits that would need to be known to reconstruct the initial state. The generating polynomial is known. Use Figure 5.10, f(x)=x 7 +x 3 +1 as an example

73. Topic 473 April 13, 2005 Solution to Problem 5.8 If the generating polynomial is known then the entire m-sequence is known The problem is reduced to determining how many successive bits in the m- sequence are necessary to uniquely determine the position in the sequence From Figure 5.10 m, the number of lags – seven for this example – bits in a row determines the state of the shift register

74. Topic 474 April 13, 2005 RAKE Receiver RAKE  Not an acronym  Based on signal flow diagram that looks like a garden rake Receiver architecture used for CDMA systems Concept addresses multipath environments Consists of  An array of up to Q parallel receiver  Timing between these receivers varies in steps of T c

75. Topic 475 April 13, 2005 RAKE Signal Flow Each channel  Multiplied by spreading code g(t)  Integrate and dump filter of length T=Q. T c  Weight by expected corresponding multipath channel amplitude All are then summed into a single-channel processor Result is “matched filter” to multipath channel

76. Topic 476 April 13, 2005 Example 5.5 Pages 298-299 Given  BPSK  Information rate R b =4800 bps  Spreading factor Q=R c /R b =32  Rate ½ convolutional code  BER is 10 -5  Number of interfering users is 7 (K=8)  Compare degradation D g with and without FEC Use Eqs. Page 272 and (5.72) just preceding

77. Topic 477 April 13, 2005 Example 5.5 (Continued) The BER of 10 -5 indicates single-user The degradation factor is

78. Topic 478 April 13, 2005 Example 5.5 (Continued) With rate ½ constraint length 7FEC encoding  Spreading factor Q s =16, total Q=32  Single-user Degradation factor is

79. Topic 479 April 13, 2005 Example 5.5 (Concluded) Degradation improves 2.7 dB with FEC Degradation vs. loading for rate ½ codes shown in Figure 5.23 page 298  No improvement for single user  Gains of about 2 dB for K near Q  Improvement doesn’t vary much with BER Conclusion: BER is important power- bandwidth tradeoff with multiple users

80. Topic 480 April 13, 2005 Problem 5.49 Page 336 Describe how the use of a rate ¼ FEC doe would affect the implementation and performance of a RAKE Receiver Effects of change in FEC code  Delay-line parallelism is not affected  Channel tracking (see 5.6 pages 292-294) is affected because algorithm operates before FEC and E b /N 0 is lower with better codes Measures for use with better codes include  Use a known pilot signal, as with IS-95  Use training sequences (standard messages) for channel tracking, as with WCDMA

81. April 13, 2005Topic 4 81 EE320 Telecommunication Engineering Wireless Architectures

82. Topic 482 April 13, 2005 Open System Interconnection (OSI) Model Seven-layer model  Physical layer (modem)  Data link layer  Network layer  Transport layer (packetizing, ACK/NAK)  Session layer (Service selection and access)  Presentation layer (encryption, compression)  Application layer (HMI) Layers designed together as a system

83. Topic 483 April 13, 2005 Power Control Architectures Open Loop  Mobile terminals measure strength of pilot channel  Transmit power decreased for strong pilot channels  Fast and simple, but must be approximate Closed Loop  Base station measures mobile terminal signal strength  Mobile station receives signal strength by downlink  Accurate but delay and averaging must be smaller than channel coherence time Outer Loop Control  Base station uses expected signal strength in control algorithm  Complexity can result in a slow loop

84. Topic 484 April 13, 2005 Power Control: Summary Power control minimizes SINR in busy cells Handset power control minimizes SINR in the base station but not at the mobile terminal Methods still evolving Next generation standards will implement  Newer techniques such as outer-loop control  Base station power control for SINR control at the mobile station

85. Topic 485 April 13, 2005 Next Time Assignment: Read parts of Chapter 7  7.3, OSI  7.6, Power Control  7.7, Handover  7.8, Network Layer