HighSpeed Wireline Communication Systems Prof Brian L Evans

High-Speed Wireline Communication Systems Prof. Brian L. Evans Dept. of Electrical and Comp. Eng. The University of Texas at Austin http: //signal. ece. utexas. edu Current graduate students: Ming Ding, Zukang Shen Ex-graduate students: Güner Arslan (Silicon Laboratories), Biao Lu (Schlumberger), Milosevic (Schlumberger) Ex-undergraduate students: Wade Berglund, Jerel Canales, David Love, Ketan Mandke, Scott Margo, Esther Resendiz, Jeff Wu http: //www. ece. utexas. edu/~bevans/projects/adsl

Schlumberger Downhole Data Communications • Downhole drilling – – Cable of several miles in length Power and data delivered on cables (harmonics) Motors downhole turning on and off (harmonics) Downhole borehead faces high temperatures, vibrations, etc. • Need for speed – Uplink: digitized images/properties of ground (high data rate) – Downlink: command, control, and programs (low data rate) • Need for asymmetric data communications – High-the-better uplink data rates – Lower-the-better bit error rates on both links • Approaches – Single channel, single carrier (e. g. quadrature amplitude modulation) – Single channel, multiple carriers (e. g discrete multitone modulation) – Multiple channels 2

Quadrature Amplitude Modulation (QAM) Q Modulator I Bits 00110 Constellation encoder I Q cos(2 p fc t) Lowpass filter Transmit - Bandpass sin(2 p fc t) magnitude channel fc frequency – Single carrier – Single signal, occupying entire available bandwidth – Symbol rate is bandwidth of signal being centered on carrier frequency – Mike Kuei-che Cheng, Improving the Performance of a Wireline Telemetry Receiver, MS Thesis, UT Austin, 1997. 3

Multicarrier Modulation • Divide broadband channel into narrowband subchannels – No ISI in subchannels if constant gain in every subchannel and if ideal sampling – Each subchannel has a different carrier pulse DTFT-1 • Discrete multitone modulation w – Based on fast Fourier transform -wc – Standardized for ADSL and VDSL – Used in Schlumberger downhole modems magnitude sinc n wc channel carrier Subchannels are 4. 3 k. Hz wide in ADSL and VDSL subchannel (QAM signal) frequency 4

Interne t Digital Subscriber Line (DSL) Broadband Access DSLAM high data rate Central Office DSL modem Voice Switch DSL modem low data rate LPF Customer Premises Telephone Network DSLAM - Digital Subscriber Line Access Multiplexer LPF – Low Pass Filter (passes voiceband frequencies) 5

Simulation Results for 17 -Tap Equalizers Parameters Cyclic prefix length 32 FFT size (N) 512 Coding gain (d. B) 4. 2 Margin (d. B) 6 Input power (d. Bm) 23 Noise power (d. Bm/Hz) -140 Crosstalk noise 24 ISDN disturbers High rate direction Figure 1 in [Martin, Vanbleu, Ding, Ysebaert, Milosevic, Evans, Moonen & Johnson, submitted] 6

Multichannel Discrete Multitone Transmission Duplex Channel NEXT FEXT NEXT: Near End Crosstalk FEXT: Far End Crosstalk • Different levels of coordination – Multiuser detection (no coordination) – Joint spectra optimization (coordination of transmit spectra usage) – Vectored transmission (full signaling coordination at both ends) 7

Improving Data Rates • Per channel improvements – – Symbol synchronization (embed makers in transmitted data) Multicarrier modulation (number of channels, bit swapping) Equalization (training sequence and time) Error detection and correction (choice of coding methods) • Multichannel improvements – Coordination of transmit specta – Coordination of signaling at both ends (training sequence and time) – Interference cancellation 8

Backup Slides

Multiuser Detection Duplex Channel NEXT FEXT NEXT: Near End Crosstalk FEXT: Far End Crosstalk • No coordination between duplex channels • Different service providers bundled in same/adjacent cable • Must combat near-end and far-end crosstalk – Crosstalk identification: estimate crosstalk channel and power – Crosstalk cancellation 10

Joint Spectra Optimization Duplex Channel NEXT FEXT NEXT: Near End Crosstalk FEXT: Far End Crosstalk • Coordination in joint spectra design • Goal: find multiuser power allocation to maximize sum of data rates • Solution: For all users, regard others as additional noise and perform single user water-filling and iterate 11

Vectored Transmission Duplex Channel NEXT FEXT NEXT: Near End Crosstalk FEXT: Far End Crosstalk • Signal level coordination – Full knowledge of downstream transmitted signal and upstream received signal at central office – Block transmission at both ends fully synchronized • Channel characterization – Pertone basis – Multi-channel 12

Crosstalk Cancellation • NEXT is suppressed by frequency division duplexing • FEXT is cancelled per tone via QR decomposition of Ti – Downstream • Pertone MIMO precoding • No crosstalk after channel – Upstream • QR leads to a back-substitution structure • decode last user, decision feedback as crosstalk • Successive crosstalk cancellation 13

ADSL Equalization Simulation Results for 17 -Tap TEQs (con’t) Parameters Cyclic prefix length 32 FFT size (N) 512 Coding gain (d. B) 4. 2 Margin (d. B) 6 Input power (d. Bm) 23 Noise power (d. Bm/Hz) -140 Crosstalk noise 24 ISDN disturbers Downstream transmission Figure 3 in [Martin, Vanbleu, Ding, Ysebaert, Milosevic, Evans, Moonen & Johnson, submitted] 14

Data Transmission in an ADSL Transceiver N/2 subchannels N real samples Bits 00110 quadrature amplitude modulation (QAM) mapping S/P TRANSMITTER N/2 subchannels QAM decoder add cyclic prefix P/S D/A + transmit filter each block programmed in lab and covered in one full lecture in EE 345 S each block covered in one full lecture RECEIVER P/S mirror data and N-IFFT invert channel = frequency domain equalizer P/S parallel-to-serial channel N real samples N-FFT and remove mirrored data S/P serial-to-parallel remove S/P cyclic prefix time domain equalizer (FIR filter) receive filter + A/D FFT fast Fourier transform 15

Introduction Discrete Multitone (DMT) DSL Standards ADSL – Asymmetric DSL Maximum data rates supported in G. DMT standard (ideal case) Echo cancelled: 14. 94 Mbps downstream, 1. 56 Mbps upstream Frequency division multiplexing: 13. 38 Mbps downstream, 1. 56 Mbps up Widespread deployment in US, Canada, Western Europe, Hong Kong Central office providers only installing frequency-division multiplexed (FDM) ADSL: cable modem market 1: 2 in US & 5: 1 worldwide ADSL+ 8 Mbps downstream min. ADSL 2 doubles analog bandwidth VDSL – Very High Rate DSL Asymmetric Faster G. DMT FDM ADSL 2 m subcarriers m [8, 12] Symmetric: 13, 9, or 6 Mbps Optional 12 -17 MHz band 16

Introduction Spectral Compatibility of x. DSL Plain Old Telephone Service ISDN Any overlap with the AM radio band? 1. 1 MHz ADSL - USA ADSL - Europe HDSL/SHDSL Home. PNA Any overlap with the FM radio band? VDSL 10 k 100 k 1 M 10 M Frequency (Hz) Upstream Downstream 100 M 12 MHz Mixed 17

Modulation A Digital Communications System • • Encoder maps a group of message bits to data symbols Modulator maps these symbols to analog waveforms Demodulator maps received waveforms back to symbols Decoder maps the symbols back to binary message bits Message Source Encoder Modulator Transmitter Noise Channel Decoder Message Sink Demodulator Receiver 18

Modulation Amplitude Modulation by Cosine Function • Example: y(t) = f(t) cos(w 0 t) F(w) f(t) is an ideal lowpass signal Assume w 1 << w 0 Y(w) is real-valued if F(w) is real-valued 1 -w 1 ½F(w + w 0) Y(w) 0 w 1 w ½F(w - w 0) ½ -w 0 - w 1 -w 0 + w 1 0 w 0 - w 1 w 0 + w 1 w • Demodulation is modulation then lowpass filtering • Similar derivation for modulation with sin(w 0 t) 19

Modulation Amplitude Modulation by Sine Function • Example: y(t) = f(t) sin(w 0 t) F(w) f(t) is an ideal lowpass signal Assume w 1 << w 0 Y(w) is imaginary-valued if F(w) is real-valued j ½F(w + w 0) 1 -w 1 Y(w) j -w 0 ½ -w 0 + w 1 -j w -j ½F(w - w 0) w 0 - w 1 -w 0 - w 1 0 w 0 + w 1 w ½ • Demodulation is modulation then lowpass filtering 20

Modulation Multicarrier Modulation by Inverse FFT Q g(t) x x I g(t) x + g(t) : pulse shaping filter Discrete time x + x Xi : ith symbol from encoder 21

ADSL Transceivers Multicarrier Modulation in ADSL Q 00101 QAM I X 0 N/2 subchannels (carriers) X 1 X 2 XN/2 Mirror complex data (in red) and take conjugates: XN/2 -1* X 2 * N-point Inverse Fast Fourier Transform (IFFT) x 0 x 1 x 2 x. N-1 N realvalued time samples forms ADSL symbol X 1 * 22

ADSL Transceivers Multicarrier Modulation in ADSL Inverse FFT v samples CP N samples s y m b o l (i) CP s y m b o l ( i+1) copy CP: Cyclic Prefix copy D/A + transmit filter ADSL frame is an ADSL symbol plus cyclic prefix 23

ADSL Transceivers Multicarrier Demodulation in ADSL S/P N/2 subchannels (carriers) N-point Fast Fourier Transform (FFT) N time samples 24

ADSL Transceivers Bit Manipulations • Serial-to-parallel converter 110 00110 Bits • Parallel-to-serial converter 110 S/P 00 00 Words • Example of one input bit stream and two output words 00110 Bits • Example of two input words and one output bit stream 25

Combating ISI Inter-symbol Interference (ISI) 2. 1 • Ideal channel 1. 7 111 1 1 . 7. 4 . 7. 1 = * -1 1 Channel impulse response Received signal – Impulse response is an impulse – Frequency response is flat • Non-ideal channel causes ISI – Channel memory – Magnitude and phase variation • Received symbol is weighted sum of neighboring symbols Threshold at zero 11 1 Detected signal – Weights are determined by channel impulse response 26

Combating ISI Single Carrier Modulation • Ideal (non-distorting) channel over transmission band – Flat magnitude response – Linear phase response: delay is constant for all spectral components – No intersymbol interference • Impulse response for ideal channel over all frequencies nk – Continuous time: g d(t-T) Channel Equalizer yk xk rk ek – Discrete time: g d[k-D] + w + h + • Equalizer – Shortens channel impulse response (time domain) – Compensates for frequency distortion (frequency domain) Ideal Channel z- g Discretized Baseband System 27

Combating ISI Combat ISI with Equalization • Problem: Channel frequency response is not flat • Solution: Use equalizer to flatten channel frequency response • Zero-forcing equalizer – Inverts channel (impulse response forced to impulse) – Flattens frequency response – Amplifies noise • Minimum mean squared error (MMSE) equalizer – Optimizes trade-off between noise amplification and ISI Zero-forcing Equalizer frequency response MMSE Equalizer frequency response Channel frequency response • Decision-feedback equalizer – Increases complexity – Propagates error 28

Combating ISI Cyclic Prefix Helps in Fighting ISI subsymbols to be transmitted cyclic prefix mirrored subsymbols to be removed equal 29

Combating ISI Cyclic Prefix Helps in Fighting ISI • Provide guard time between successive symbols – No ISI if channel length is shorter than n +1 samples • Choose guard time samples to be a copy of the beginning of the symbol – cyclic prefix – Cyclic prefix converts linear convolution into circular convolution – Need circular convolution so that symbol channel FFT(symbol) x FFT(channel) – Then division by the FFT(channel) can undo channel distortion v samples CP N samples s y m b o l (i) copy CP s y m b o l ( i+1) copy 30

Combating ISI Channel Impulse Response frequency (k. Hz) 31

Combating ISI Channel Impulse Response frequency (k. Hz) 32

Combating ISI Combat ISI with Time-Domain Equalizer • Channel length is usually longer than cyclic prefix • Use finite impulse response (FIR) filter called a timedomain equalizer to shorten channel impulse response to be no longer than cyclic prefix length channel impulse response shortened channel impulse response 33

ADSL Equalization Eliminating ISI in Discrete Multitone Modulation • Time domain equalizer (TEQ) – Finite impulse response (FIR) filter – Effective channel impulse response: convolution of TEQ impulse response with channel impulse response • Frequency domain equalizer (FEQ) – Compensates magnitude/phase distortion of equalized channel by dividing each FFT coefficient by complex number – Generally updated during data transmission n+1 channel impulse response effective channel impulse response D: transmission delay n: cyclic prefix length • ADSL G. DMT equalizer training – Reverb: same symbol sent 1, 024 to 1, 536 times – Medley: aperiodic sequence of 16, 384 symbols – At 0. 25 s after medley, receiver returns number of bits on each subcarrier that can be supported 34

ADSL Equalization Time-Domain Equalizer Design • Minimizing mean squared error – Minimize mean squared error (MMSE) method [Chow & Cioffi, 1992] – Geometric SNR method [Al-Dhahir & Cioffi, 1996] • Minimizing energy outside of shortened channel response – Maximum Shortening SNR method [Melsa, Younce & Rohrs, 1996] – Minimum ISI method [Arslan, Evans & Kiaei, 2000] • Maximizing achievable bit rate – Maximum bit rate method [Arslan, Evans, Kiaei, 2000] – Maximum data rate method [Milosevic, Pessoa, Evans, Baldick, 2002] – Bit rate maximization [Vanblue, Ysebaert, Cuypers, Moonen & Van Acker, 2003] • Other equalizer architectures – Dual-path (DP) design uses two TEQs [Ming, Redfern & Evans, 2002] – TEQ filter bank design [Milosevic, Pessoa, Evans, Baldick, 2002] – Per tone equalization [Acker, Leus, Moonen, van der Wiel, Pollet, 2001] 35

ADSL Equalization Minimum Mean Squared Error TEQ Design Channel xk h nk + z- yk TEQ w b rk ek - + bk-D • Minimize E{ek 2} [Chow & Cioffi, 1992] – Chose length of b (e. g. n in ADSL) to shorten length of h * w – b is eigenvector of minimum eigenvalue of channel-dependent matrix – Minimum MSE achieved when • Disadvantages where Amenable to real-time fixedpoint DSP implementation – Does not consider bit rate – Deep notches in equalizer frequency response (zeros out low SNR bands) – Infinite length TEQ case: zeros of b on unit circle (kills n subchannels) 36

ADSL Equalization Maximum Shortening SNR Solution • Minimize energy leakage outside shortened channel length • For each possible position of a window of +1 samples, h w • Disadvantages – Does not consider channel capacity – Requires Cholesky decomposition and eigenvector calculation – Does not consider channel noise • Amenable to real-time fixed-point DSP realization 37

ADSL Equalization Maximum Shortening SNR Solution • Choose w to minimize energy outside window of desired length – Locate window to capture maximum channel impulse response energy nk xk yk h + rk w hwin, hwall : equalized channel within and outside the window • Objective function is shortening SNR (SSNR) 38

ADSL Equalization Matlab DMT TEQ Design Toolbox 3. 1 • Single-path, dual-path, per-tone & TEQ filter bank equalizers Available at http: //www. ece. utexas. edu/~bevans/projects/adsl/dmtteq/ default parameters from G. DMT ADSL standard 23 -140 various performance measures different graphical views 39

Multicarrier Modulation • Advantages – Efficient use of bandwidth without full channel equalization – Robust against impulsive noise and narrowband interference – Dynamic rate adaptation • Disadvantages – Transmitter: High signal peak-to-average power ratio – Receiver: Sensitive to frequency and phase offset in carriers • Open issues for point-to-point connections – – Pulse shapes of subchannels (orthogonal, efficient realization) Channel equalizer design (increase bit rate, reduce complexity) Synchronization (timing recovery, symbol synchronization) Bit loading (allocation of bits in each subchannel) • Open issues for coordinating multiple connections 40

Notes

Notes Applications of Broadband Access Residential Business 42

Notes DSL Broadband Access Standards Courtesy of Mr. Shawn Mc. Caslin 43

Notes ADSL and Cable Modems • Need for high-speed (broadband) data access – Voiceband data modems can yield 53 kbps (kilobits per second) – Telephone voice channel capacity ois 64 kbps (the Central Office samples voice signals at 8 k. Hz using 8 bits/sample) – Integrated Services Digital Network (ISDN) modems deliver 128 kbps – New modem standards are necessary to meet the demand for higher bandwidth access for telecommuting, videoconferencing, video-ondemand, Internet service providers, Internet access, etc. • Two standards tested in 1998 and now widely available – Cable modems – Asymmetric Digital Subscriber Line (ADSL) modems • Cable Modems – Always connected to the Internet – Your neighbors on the same local area network share the bit rate – Local area network provides either 27 or 36 Mbps downstream, and between 320 kbps and 10 Mbps upstream. 44

Notes ADSL Modems • ADSL modems – Always connected to the Internet – Call central office using a dedicated telephone line which also supports a conventional Plain Old Telephone Service (POTS) line for voice – Connection time is 5 -10 seconds – ADSL modems are capable of delivering 1 -10 Mbps from the central office to the customer (downstream) and 0. 5 -1 Mbps from the customer to the central office (upstream) – Although ADSL lines have been available from Southwestern Bell since the Fall of 1997, ADSL modems were not commercially available until Fall of 1999. 45

Notes Discrete Multitone (DMT) Modulation • DMT uses multiple harmonically related carriers – Implemented as inverse Fast Fourier Transform (FFT) in transmitter – Implemented using forward FFT in receiver • Transmission bandwidth – 1. 1 MHz downstream and 256 k. Hz upstream – Limit of 1. 1 MHz is due to power constraints imposed by the FCC – For 18 kft telephone lines, the attenuation at 1. 1 MHz is -120 d. Bm. • Frequency domain is divided into 256 4. 3 -k. Hz bins – Channel 0 is dedicated to voice – Channels 1 -5 are not used due to compatibility with ISDN services. 46

Notes Two Types of Transmission • Two versions of ADSL 1. Frequency Division Multiplexing: the upstream and downstream channels do not overlap: the upstream uses channels 6 -31 and the downstream uses channels 32 -255. 2. Echo Cancelled: the upstream and downstream channels overlap: the upstream uses channels 6 -31 and the downstream uses channels 6 -255. • According to available SNR in each bin, bin carries – QAM signal whose constellation varies from 2 -15 bits or – no signal if SNR is less than 12 d. B in that subchannel • • Constellations chosen so that overall bit error rate < 10 -7 Maximum transmission rate with symbol rate of 4 k. Hz – Downstream: 248 channels x 15 bits/channel x 4 k. Hz = 14. 88 Mbps – Upstream: 24 channels x 15 bits/channel x 4 k. Hz = 1. 440 Mbps 47

Notes Channel Attenuation • Reliable transmission of high-frequency information over a telephone line is wrought with several challenges. – Telephone lines are unshielded and bundled 50 wires to a trunk. The other lines in the bundle can cause severe crosstalk – Telephone lines attenuate signals. The attenuation increases with increasing frequency. At 1. 1 MHz, which is the highest transmitted frequency, the attenuation of a 24 gauge wire is 10 kft 12 kft 14 kft -70 d. Bm/Hz -90 d. Bm/Hz -100 d. Bm/Hz 16 kft 18 kft -110 d. Bm/Hz -120 d. Bm/Hz • Because of severe effects in the channel, the ADSL standard defines channel coding using cyclic prefixes and employs error correcting codes 48

Notes Bridge Taps • Bridge Taps are unterminated lines – During modem initialization, effect of bridge taps is included in channel estimate. Their effect would be to lower the possible channel capacity. – During data transmission, bridge taps may saturate the front-end at a least will be unpleasant for the echo canceller. The echo canceller should have an estimate of the echo channel including the bridge taps. Given that the reflected echo is almost instantaneous than the echo canceller channel estimate should capture them too. • In G. lite, echo cancellation is optional – Modems who use it can still use it – A bigger problem in G. lite is the phone due to the splitterless environment – Transmitters that do not have an echo canceller system can rely on their receive filters to reduce the echo. 49

Notes ADSL Modems • ADSL modem consists of a line driver plus 3 subsystems: 1. analog front end (15 V) 2. digital interface (3 V) 3. discrete multitone processor (3 V) • Analog front end provides the analog-to-digital and digitalto-analog interfaces to the telephone line. • Digital interace manages the input and output digital message streams. • Discrete multitone processor implements the digital communications and signal processing to support the ADSL standard. An ADSL modem requires much greater than 200 Digital Signal Processor MIPS. 50

Notes Motorola Copper. Gold ADSL Chip • • Announced March 1998 5 million transistors, 144 pins, clocked at 55 MHz 1. 5 W power consumption DMT processor contains – Motorola MC 56300 DSP core – Several application specific ICs • 512 -point FFT • 17 -tap FIR filter for time-domain channel equalization based on MMSE method (20 bits precision per tap) • DSP core and memory occupies about 1/3 of chip area • It gives up to 8 Mbps upstream and 1 Mbps downstream 51

Notes Motorola Copper Gold ADSL Transceiver • Contains all 3 ADSL modem subsystems on a single chip. – Has programmable bit to tell it whether it is at customer's or central office site – Analog front end operates at a sampling rate of 2. 208 MHz and gives 16 bits/sample of resolution. It uses sigma-delta modulation with an oversampling factor of 55 / 2. 208 = 25. • Discrete multitone processor consists of a Motorola MC 56300 DSP Onyx core and several application-specific digital VLSI circuits to implement – 256 -point FFT for downstream transmission or 512 -point FFT for downstream reception if it is at the central office or customer's site, respectively – 17 -tap adaptive FIR filter for channel equalization (20 bits of precision per tap) running at 2. 208 MHz – DSP core computes the 32 -point FFT for the downstream transmission or the 64 -point FFT for the downstream reception. 52

Notes Minimum Mean Squared Error TEQ nk xk h z- + yk w b rk - + ek zk Matrix O selects the proper part out of Rx|y corresponding to the delay 53

Notes Simulation Results for 17 -Tap TEQ Cyclic prefix length FFT size (N) Coding gain Margin 32 512 4. 2 d. B 6 d. B Input power Noise power Crosstalk noise POTS splitter 23 d. Bm -140 d. Bm/Hz 8 ADSL disturbers 5 th order Chebyshev 54

Notes Simulation Results for Three-Tap TEQ Cyclic prefix length FFT size (N) Coding gain Margin 32 512 4. 2 d. B 6 d. B Input power Noise power Crosstalk noise POTS splitter 23 d. Bm -140 d. Bm/Hz 8 ADSL disturbers 5 th order Chebyshev 55