Digital Wireless Communication: Physical Layer Exploitation Robert W. Heath Jr. Ph.D, P.E. KE5NCG Wireless Networking and Communications Group Department of Electrical and Computer Engineering The University of Texas at Austin, Austin TX USA http://www.profheath.org [email protected] Thursday, August 30, 12

Wireless is Everywhere

cellular networks

local area networks

personal area networks

emerging applications

Thursday, August 30, 12

Where is Wireless Taught? Undergraduate

Signals and Systems Digital Signal Processing

time in EE

Analog Communication Graduate

Digital Communication Intro to Wireless Advanced Wireless

a typical curriculum (many exceptions of course) 3 Thursday, August 30, 12

Why at the Graduate Level? It involves many different areas of expertise Digital communication Propagation Antennas Signal Processing Probability Etc.

Hot area of research Requires some depth in areas of expertise

Curriculum is not widely available (vs e.g. signals and systems) 4 Thursday, August 30, 12

Where Could it be Taught? Undergraduate

Signals and Systems Digital Signal Processing

Graduate

time in EE

Wireless Digital Comm. Digital Communication Advanced Wireless

Lab-based approach to teach wireless to undergraduates 5 Thursday, August 30, 12

Wireless Communications Lab @ UT Premises of the course

EE 471C / EE 381V

Wireless communication can be taught to undergraduates Wireless communication can be taught without a communication background Students can implement what they learn while they learn it

Key ideas Teach digital communication from a digital signal processing perspective Incorporate modulation, channel estimation, equalization, synchronization Use algorithmic design examples, not comprehensive theory Leverage flexible software defined radio prototyping Exploit LabVIEW & USRP

Developed and tested over 7 years 6 Thursday, August 30, 12

DSP Approach to Wireless Inputs

System

0110110

h[n]

Outputs

0110110

h(t)

time

time

time

time

Use systems approach for communication 7 Thursday, August 30, 12

How this Fits with the Lab transmitter

Source

Channel Coding

Modulation

D/A

RF Upconversion channel

receiver

Sink

Channel Decoding

Demodulation

Laptop with LabVIEW (all digital signal processing)

A/D

RF Downconversion

NI USRP 2921

Real world

8 Thursday, August 30, 12

How it Works at UT The course is cross-listed for undergrads and grads Pre-requisites: a course in digital signal processing and a course on probability Undergraduates take in 3rd or 4th year as a 4 credit course Graduate students take their 1st or 3rd semester

Structure of the course 3 hours of lecture per week, covers the theory of the course 3 hours in the lab per week, demonstrate what has been learned Homework assignments test the theory Prelabs, labs, lab reports test what has been learned in the lab Yes there are exams too......(why do the students complain of high workload??)

9 Thursday, August 30, 12

Content of the Course Digital comm overview Signals, stochastic processes Mathematical preliminaries Transforms, sampling theorm Frequency response, power spectrum, bandwidth Upconversion, downconversion, complex baseband Quadrature pulse amplitude modulation Basic digital comm Optimal pulse shapes Maximum likelihood detection in AWGN Sample timing offset, sample timing algorithms Channel impairments Frequency selective channels, least squares channel estimation Frequency offset estimation and correction, frequency domain equalization Single carrier frequency domain equalization, OFDM, the cyclic prefix Standards IEEE 802.11a, GSM standard Introduction to propagation, large-scale fading, link budgets, path-loss Fading Small-scale fading, coherence time, coherence bandwidth Probability of error in fading channels Sources of diversity, Alalmouti space-time code, maximum ratio combining MIMO Introduction to MIMO communication, spatial multiplexing Introduction to MIMO-OFDM, highlights of the IEEE 802.11n standard

10 Thursday, August 30, 12

Content of the Course Digital comm overview Signals, stochastic processes Transforms, sampling theorm Frequency response, power spectrum, bandwidth Upconversion, downconversion, complex baseband Quadrature pulse amplitude modulation Optimal pulse shapes Maximum likelihood detection in AWGN Sample timing offset, sample timing algorithms Frequency selective channels, least squares channel estimation Frequency offset estimation and correction, frequency domain equalization Single carrier frequency domain equalization, OFDM, the cyclic prefix IEEE 802.11a, GSM standard Introduction to propagation, large-scale fading, link budgets, path-loss Small-scale fading, coherence time, coherence bandwidth Probability of error in fading channels Sources of diversity, Alalmouti space-time code, maximum ratio combining Introduction to MIMO communication, spatial multiplexing Introduction to MIMO-OFDM, highlights of the IEEE 802.11n standard

Done in the Lab

11 Thursday, August 30, 12

Materials Developed for Lecture Textbook that describes the theory Introduction to Wireless Digital Communication: A Signal Processing Perspective by Robert W. Heath Jr. Book is 200 pages, unpublished but available for use for free (has been used as is for several years) (0) gTX [ n]

s[n]

L





Target completion date is end of 2012

D/C

g

( L 1) TX

[ n]

L

Ex

x(t )

Tx

1

z z 1

Figure 20: An implementation of transmit pulse shaping using upsampling. 2 Implementing pulse-shaping at the receiver

Lecture notes In LaTeX form approximately 84 pages

• Objective: Implement discrete-time pulse shaping at the receiver using oversampling combined with downsampling. • Objective: Apply downsampling identities to implement and simplify receive pulse shaping • Thus far we have considered the following QAM receiver

Slides are forthcoming • This structure does not map to a flexible implementation Requires analog matched filtering Take advantage of flexible digital processing Does not lend itself to various synchronization tasks • How can we implement pulse-shaping in discrete-time? • Let Tz = T /M for some integer M . If 1/Tz > Nyquist rate then this is called oversampling • Suppose that we implement the analog filter using discrete-time processing • We have already solved this problem using discrete-time processing of a bandlimited continuoustime signal z (t )

Thursday, August 30, 12

C/D Tz

g RX [n]

D/C

C/D

Tz

T  MTz

y[n]

12

Materials Developed for Lab Laboratory manual DIGITAL COMMUNICATIONS: Physical Layer Exploration Using the NI USRP front.pdf

1

9/12/11

4:46 PM

141 pages 8 Laboratory experiments

Lab experiments Background information Include prelab to be completed prior to lab Laboratory experiments Postlab

Complete software framework TA guide with solutions

DIGITAL COMMUNICATIONS PHYSICAL LAYER EXPLORATION LAB USING THE NI USRP™ PLATFORM

Dr. Robert W. Heath, University of Texas at Austin

Included with the Digital Communications Teaching Bundle http://sine.ni.com/nips/cds/view/p/lang/en/nid/210385 Thursday, August 30, 12

13

Contents

Outline of Lab Manual Preface

vii

About the Author

xi

Lab 1: Part 1 Introduction to NI LabVIEW

1

Lab 1: Part 2 Introduction to NI RF Hardware

10

Lab 2: Part 1 Modulation and Detection

22

Lab 2: Part 2 Pulse Shaping and Matched Filtering

35

Lab 3: Synchronization

51

Lab 4: Channel Estimation & Equalization

63

Lab 5: Frame Detection & Frequency Offset Correction

82

Lab 6: OFDM Modulation & Frequency Domain Equalization

99

Lab 7: Synchronization in OFDM Systems

115

Lab 8: Channel Coding in OFDM Systems

130

Appendix A: Reference for Common LabVIEW VIs

139

Bibliography

141 14

Thursday, August 30, 12

!

Sample Pages !

!

“book” — 2011/9/29 — 15:18 — page 43 — #55

Lab 2: Part 2 Pulse Shaping and Matched Filtering

!

!

!

“book” — 2011/9/29 — 15:18 — page 51 — #63

!

!

Lab 3: Synchronization: Symbol Timing Recovery in Narrowband Channels

43

Summary In this lab you will consider the problem of symbol timing recovery also known as symbol synchronization. Timing recovery is one of several synchronization tasks; others will be considered in future labs. The wireless communication channel is not well modeled by simple additive white Gaussian noise. A more realistic channel model also includes attenuation, phase shifts, and propagation delays. Perhaps the simplest channel model is known as the frequency flat channel. The frequency flat channel creates the received signal z(t)

Figure 3: Hierarchy of code framework for new simulator.

=

αej φ x(t − τd ) + v(t),

(1)

where α is an attenuation, φ is a phase shift, and τd is the delay. The objective of this lab is to correct for the delay caused by τd in discretetime. The approach will be to determine an amount of delay kˆ and then to delay the filtered received signal by kˆ prior to downsampling. This will modified the receiver processing as illustrated in Figure 1. Two algorithms will be implemented for symbol synchronization in this lab: the maximum energy method and the Early Late gate algorithm. The maximum energy method attempts to find the sample point that maximizes the average received energy. The early–late gate algorithm implements a discrete-time version of a continuous-time optimization to maximize a certain

top rx.vi and provides each with the appropriate inputs. The parts of the simulator you will be modifying are located in transmitter.vi and receiver.vi shown in Figures 4 and 5 respectively. You will be putting your VIs into transmitter.vi and receiver.vi, replacing the locked versions that are already there. After doing this, you will then open up simulator.vi, that you will use to confirm your VIs operate correctly before implementing them on the NI-USRP. Notice that pulse shaping.vi and matched filtering.vi do not take any parameters as inputs. All of the pulse shaping and oversampling parameters you need to use for these VIs can be accessed from the modulation parameters in cluster. After building these VIs, replace the existing code in the simulator with your code. Replace a subVI in the transmitter or receiver with your code

z (t )

C/D Tz =

ˆ

zk

g RX [n]

M kˆ

T M

Symbol Sync

Figure 1: Receiver with symbol synchronization after the digital matched filtering. Figure 4: Block diagram of transmitter.vi.

!

!

Thursday, August 30, 12

! !

!

51

!

!

!

15

Methodology for the Lab All labs have already been implemented, functions are locked Both over-the-air and simulation capable functions

Students work in groups of 2 or 3 Each week students implement a new block For example modulation, or demodulation, or some synchronization

During the pre-lab Student implements software, verifies it is correct, answers prelab questions

During the lab Student demonstrates correction function to TA, answers inlab questions

After the lab Students write a short lab report documenting what they learned

16 Thursday, August 30, 12

Lab Size and Equipment Course has 30-40 students enrolled every year Half undergraduates, half graduates

Lab has ten workstations for transmit / receive Students work in teams of 2 or 3 Accommodate 20 students in the lab for 3 hours One TA services two lab sessions, grades homework, holds office hours

17 Thursday, August 30, 12

Operational Challenges Covering material required before the lab Only an issue the first few weeks of the class

Keeping PCs up-to-date Solved by requiring students to use their own laptops

Finding high quality TAs with enough experience Not an issue after a couple of years

Avoiding copying of code, plagiarism Need better tools for detecting plagiarism in graphical code

Helping students that fall behind The labs build on each other every week Falling behind can be a big problem

18 Thursday, August 30, 12

Possible Evolution of the Lab Lab equipment can be checked out, experiments performed How to avoid all students doing the same thing? Do they really do the experiment themselves?

Lab equipment can be networked and accessed remotely How to manage access to the equipment? How to configure equipment for experiments? What is the difference between this and simulation?

Lab equipment in the cloud Perhaps not owned by the university, equipment pooled together Some examples of this already for research applications Is it still “real”??

19 Thursday, August 30, 12

Conclusion Wireless communication can be taught to undergraduates Laboratory approach makes wireless more concrete Avoids simply drowning in mathematics Useful for graduate students as well

Students build a practical foundation for further study Good preparation for industry Practical insights make for more relevant research

20 Thursday, August 30, 12

Questions?

Thursday, August 30, 12