NI Academic Day June 30, 2005 Beirut, Lebanon Modem Design, Implementation, and Testing Using NI’s LabVIEW Prof. Brian L. Evans Dept. of Electrical and Computer Engineering The University of Texas at Austin, Austin, Texas USA Visiting Associate Professor American University of Beirut, Beirut, Lebanon Contributions by Vishal Monga, Zukang Shen, Ahmet Toker, and Ian Wong, UT Austin
2 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Outline Real-Time Digital Signal Processing (DSP) Laboratory Course Single Carrier Transceiver Sinusoidal Generation Digital Filters Data Scramblers Pulse Amplitude Modulation Quadrature Amplitude Modulation Conclusion
3 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Real-Time DSP Course: Overview Objectives of undergraduate class Build intuition for signal processing concepts Translate signal processing concepts into real-time digital communications software Lecture: breadth (three hours/week) Digital signal processing algorithms Digital communication systems Digital signal processor architectures Laboratory: depth (three hours/week) Deliver voiceband modem “Design is the science of tradeoffs” (Prof. Yale Patt, UT) Test/validate implementation Over 600 served since 1997 Web site: Download site:
4 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Real-Time DSP Course: Overview Embedded system demand: volume, volume, … 400 Million units/year: automobiles, PCs, cell phones 30 Million units/year: ADSL modems and printers Consumer electronics products How much should an embedded processor cost? Source: CEA Market Reseach. Data for 2004 calendar year.
5 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Real-Time DSP Course: Overview Digital signal processor market 40% annual growth #1 in growth within semiconductor market Worldwide revenue (US dollars) $6.1B ‘00, $4.5B ‘01, $4.9B ‘02, $6.1B ‘03, $8.0B ‘04 Estimated annual growth of 23% for Market share (based on 2002 revenue) 43% TI, 14% Freescale, 14% Agere, 9% Analog Dev. Fixed-point vs. floating-point DSPs >90% of digital signal processors sold are fixed-point Floating–point DSPs used for initial real-time prototype How many digital signal processors are in a PC? Revenue figures from Forward Concepts (
6 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Real-Time DSP Course: Which DSP? Students are next-to-final year (junior) and final- year (senior) undergraduate students Fixed-point DSPs for high-volume products Battery-powered: cell phones, digital still cameras … Wall-powered: ADSL modems, cellular basestations … Fixed-point issues Using non-standard C extensions for fractional data Converting floating-point programs to fixed-point Manual tracking of binary point prone to error Floating-point DSPs Feasibility for fixed-point DSP realization Shorter prototyping time Program TI TMS320C67x DSP in C TI Code Composer Studio 2.2
7 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Real-Time DSP Course: Textbooks C. R. Johnson, Jr., and W. A. Sethares, Telecommunication Breakdown, Prentice Hall, Intro to digital communications and transceiver design Matlab examples S. A. Tretter, Comm. System Design using DSP Algorithms with Lab Experiments for the TMS320C6701 & TMS320C6711, Assumes DSP theory and algorithms Assumes access to C6000 reference manuals Errata/code: Bill Sethares (Wisconsin) Rick Johnson (Cornell) Steven Tretter (Maryland)
8 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Lab 1. QAM Transmitter Diagram Lab 4 Rate Control Lab 6 QAM Encoder Lab 3 Tx Filters Lab 2 Passband Signal LabVIEW demo by Zukang Shen (UT Austin)
9 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Lab 1. QAM Transmitter Diagram LabVIEW Control Panel QAM Passband Signal Eye Diagram LabVIEW demo by Zukang Shen (UT Austin)
10 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon square root raise cosine, roll-off = 0.75, SNR = raise cosine, roll-off = 1, SNR = 30 dB passband signal for 1200 bps mode passband signal for 2400 bps mode Lab 1. QAM Transmitter Diagram
11 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Lab 2. Sine Wave Generation Aim: Evaluate three ways to generate sine waves in signal quality vs. complexity Function call Lookup table Difference equation Three output methods Polling data transmit register Software interrupts Direct memory access (DMA) transfers Expected outcomes are to understand Signal quality vs. implementation complexity tradeoff C6701 EVM board’s stereo codec operation Interrupt mechanisms and DMA transfers
12 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Lab 2. Sine Wave Generation Evaluation procedure Validate sine wave frequency on scope, and test for various sampling rates (14 sampling rates on board) Method 1 with interrupt priorities Method 1 with different DMA initialization(s) LabVIEW DSP Test Integration Toolkit 2.0 Code Composer Studio 2.2 C6701 Fall 2003 HP 60 MHz Digital Storage Oscilloscope Spring 2004
13 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Lab 3. Digital Filters Aim: Evaluate four ways to implement discrete-time linear time-invariant filters FIR filter: convolution in C and assembly IIR Filter: direct form and cascade of biquads, both in C IIR filter design gotchas: oscillation & instability In classical designs, poles sensitive to perturbation Quality factor measures sensitivity of pole pair: Q [ ½, ) where Q = ½ dampens and Q = oscillates Elliptic analog lowpass IIR filter p = 0.21 at p = 20 rad/s and s = 0.31 at s = 30 rad/s [Evans 1999] Qpoleszeros ±j ±j ±j ±j classical Qpoleszeros ±j ±j ±j ±j optimized
14 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Lab 3. Digital Filters IIR filter design for implementation Butterworth/Chebyshev filters special cases of elliptic filters Minimum order not always most efficient Filter design gotcha: polynomial inflation Polynomial deflation (rooting) reliable in floating-point Polynomial inflation (expansion) may degrade roots Keep native form computed by filter design algorithm Expected outcomes are to understand Speedups from convolution assembly routine vs. C Quantization effects on filter stability (IIR) FIR vs. IIR: how to decide which one to use
15 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Lab 3. Digital Filters Test Equipment Agilent Function Generator HP 60 MHz Digital Storage Oscilloscope Spectrum Analyzer Evaluation Procedure Sweep filters with sinusoids to construct magnitude and phase responses Manually using test equipment, or Automatically by LabVIEW DSP Test Integration Toolkit Check filter output for cut-off frequency, roll-off factor… FIR: Compare execution times (in Code Composer) of C without compiler optimizations C with compiler optimizations C callable assembly language routine IIR: Compute execution times (in Code Composer)
16 Prof. Brian L. Evans NI Academic Day June 30, 2005 Beirut, Lebanon Conclusion Objectives Build intuition for signal processing concepts Translate signal processing concepts into real-time digital communications software Deliverables and takeaways Deliver voiceband transceiver Tradeoffs in signal quality vs. implementation complexity Test/validate implementation Extend hands-on experience to broadband modems Role of technology TI DSPs and Code Composer Studio NI LabVIEW and DSP Test Integration Toolkit