Digital Signal Processing Techniques

Size: px

Start display at page:

Download "Digital Signal Processing Techniques"

Jonas Randall Edwards
6 years ago
Views:

1 Digital Signal Processing Techniques Dmitry Teytelman Dimtel, Inc., San Jose, CA, 95124, USA June 17, 2009

2 Outline 1 Introduction 2 Signal synthesis Arbitrary Waveform Generation CORDIC Direct Digital Synthesis 3 Up- and Down-conversion I&Q Processing Frequency Conversion 4 Signal Blocks Phase Shifting

3 Linear and Non-linear DSP Yesterday we talked mostly about linear signal processing filtering. Of course sampling and quantization is non-linear; Sampling rate change (decimation). Many more exciting non-linear things to do.

4 Linear and Non-linear DSP Yesterday we talked mostly about linear signal processing filtering. Of course sampling and quantization is non-linear; Sampling rate change (decimation). Many more exciting non-linear things to do.

5 Linear and Non-linear DSP Yesterday we talked mostly about linear signal processing filtering. Of course sampling and quantization is non-linear; Sampling rate change (decimation). Many more exciting non-linear things to do.

6 Problem Partitioning When you do real-time signal processing, problem partitioning is critical. DSP complexity scale. Push floating point elaborate algorithms to a CPU, not in real time. Project the problem into a set of coefficient setpoints, controlling real-time hardware.

7 Problem Partitioning When you do real-time signal processing, problem partitioning is critical. DSP complexity scale. Push floating point elaborate algorithms to a CPU, not in real time. Project the problem into a set of coefficient setpoints, controlling real-time hardware.

8 Problem Partitioning When you do real-time signal processing, problem partitioning is critical. DSP complexity scale. Push floating point elaborate algorithms to a CPU, not in real time. Project the problem into a set of coefficient setpoints, controlling real-time hardware.

9 Problem Partitioning When you do real-time signal processing, problem partitioning is critical. DSP complexity scale. Push floating point elaborate algorithms to a CPU, not in real time. Project the problem into a set of coefficient setpoints, controlling real-time hardware.

10 Non-linear Methods Signal synthesis. Downconversion. Upconversion. Sampling rate changes: decimation and upsampling. Quadratic functions. Trigonometric functions.

11 Non-linear Methods Signal synthesis. Downconversion. Upconversion. Sampling rate changes: decimation and upsampling. Quadratic functions. Trigonometric functions.

12 Non-linear Methods Signal synthesis. Downconversion. Upconversion. Sampling rate changes: decimation and upsampling. Quadratic functions. Trigonometric functions.

13 Non-linear Methods Signal synthesis. Downconversion. Upconversion. Sampling rate changes: decimation and upsampling. Quadratic functions. Trigonometric functions.

14 Non-linear Methods Signal synthesis. Downconversion. Upconversion. Sampling rate changes: decimation and upsampling. Quadratic functions. Trigonometric functions.

15 Non-linear Methods Signal synthesis. Downconversion. Upconversion. Sampling rate changes: decimation and upsampling. Quadratic functions. Trigonometric functions.

16 Outline 1 Introduction 2 Signal synthesis Arbitrary Waveform Generation CORDIC Direct Digital Synthesis 3 Up- and Down-conversion I&Q Processing Frequency Conversion 4 Signal Blocks Phase Shifting

17 Arbitrary Waveforms Basic Approach A memory block is used to store and then continuously output a sequence of samples. Frequency quantization depends on sequence length. Any waveform can be produced. Limitations on square, sawtooth, and other signals. "White" noise. Shaped excitations.

18 Arbitrary Waveforms Basic Approach A memory block is used to store and then continuously output a sequence of samples. Frequency quantization depends on sequence length. Any waveform can be produced. Limitations on square, sawtooth, and other signals. "White" noise. Shaped excitations.

19 Arbitrary Waveforms Basic Approach A memory block is used to store and then continuously output a sequence of samples. Frequency quantization depends on sequence length. Any waveform can be produced. Limitations on square, sawtooth, and other signals. "White" noise. Shaped excitations.

20 Arbitrary Waveforms Basic Approach A memory block is used to store and then continuously output a sequence of samples. Frequency quantization depends on sequence length. Any waveform can be produced. Limitations on square, sawtooth, and other signals. "White" noise. Shaped excitations.

21 Arbitrary Waveforms Basic Approach A memory block is used to store and then continuously output a sequence of samples. Frequency quantization depends on sequence length. Any waveform can be produced. Limitations on square, sawtooth, and other signals. "White" noise. Shaped excitations.

22 LLRF4 Example Memory block of 2048 samples. Short causes coarse frequency quantization on sinewaves. Still provides a flexible tool.

23 LLRF4 Example Memory block of 2048 samples. Short causes coarse frequency quantization on sinewaves. Still provides a flexible tool.

24 LLRF4 Example Memory block of 2048 samples. Short causes coarse frequency quantization on sinewaves. Still provides a flexible tool.

25 Outline 1 Introduction 2 Signal synthesis Arbitrary Waveform Generation CORDIC Direct Digital Synthesis 3 Up- and Down-conversion I&Q Processing Frequency Conversion 4 Signal Blocks Phase Shifting

26 A Great Tool COordinate Rotation DIgital Computer first described by Jack E. Volder in Performs iterative coordinate rotations. Two basic modes: Rotate an input vector by an arbitrary angle (rotation mode); Rotate an input vector to align with x-axis (vectoring mode). Applications: Sine and cosine generation; Cartesian to polar transformation; Arctangent computation; Arcsine, arccosine; Extensions to linear and hyperbolic functions.

27 A Great Tool COordinate Rotation DIgital Computer first described by Jack E. Volder in Performs iterative coordinate rotations. Two basic modes: Rotate an input vector by an arbitrary angle (rotation mode); Rotate an input vector to align with x-axis (vectoring mode). Applications: Sine and cosine generation; Cartesian to polar transformation; Arctangent computation; Arcsine, arccosine; Extensions to linear and hyperbolic functions.

28 A Great Tool COordinate Rotation DIgital Computer first described by Jack E. Volder in Performs iterative coordinate rotations. Two basic modes: Rotate an input vector by an arbitrary angle (rotation mode); Rotate an input vector to align with x-axis (vectoring mode). Applications: Sine and cosine generation; Cartesian to polar transformation; Arctangent computation; Arcsine, arccosine; Extensions to linear and hyperbolic functions.

29 A Great Tool COordinate Rotation DIgital Computer first described by Jack E. Volder in Performs iterative coordinate rotations. Two basic modes: Rotate an input vector by an arbitrary angle (rotation mode); Rotate an input vector to align with x-axis (vectoring mode). Applications: Sine and cosine generation; Cartesian to polar transformation; Arctangent computation; Arcsine, arccosine; Extensions to linear and hyperbolic functions.

30 CORDIC in DDS advance φ[n] φ cos Z 1 amp A CORDIC sin Run in rotation mode. New phase angle every clock sample. Get sine and cosine every clock sample.

31 Outline 1 Introduction 2 Signal synthesis Arbitrary Waveform Generation CORDIC Direct Digital Synthesis 3 Up- and Down-conversion I&Q Processing Frequency Conversion 4 Signal Blocks Phase Shifting

32 Topology A phase accumulator followed by a wave shape generator. Accumulator advance per clock cycle is adjustable: Changes the frequency; Advance can be modulated as well. Wave shape generator memory or CORDIC. With a 30-bit accumulator (MSB=π) frequency quantization is f s /10 9. Efficient accumulators (Bresenham).

33 Topology A phase accumulator followed by a wave shape generator. Accumulator advance per clock cycle is adjustable: Changes the frequency; Advance can be modulated as well. Wave shape generator memory or CORDIC. With a 30-bit accumulator (MSB=π) frequency quantization is f s /10 9. Efficient accumulators (Bresenham).

34 Outline 1 Introduction 2 Signal synthesis Arbitrary Waveform Generation CORDIC Direct Digital Synthesis 3 Up- and Down-conversion I&Q Processing Frequency Conversion 4 Signal Blocks Phase Shifting

35 I&Q Definition I&Q Representation x(t) = I(t) cos(ωt) + Q(t) sin(ωt) Narrowband technique. As you move away from ω, signals are further from quadrature. Obvious transition to polar coordinates. Analog domain processing historically favored polar coordinates. Involves amplitude and phase detectors. Phase shifters, VGAs, variable attenuators.

36 Outline 1 Introduction 2 Signal synthesis Arbitrary Waveform Generation CORDIC Direct Digital Synthesis 3 Up- and Down-conversion I&Q Processing Frequency Conversion 4 Signal Blocks Phase Shifting

37 Downconversion 0 I NCO 90 Q Project the input signal into quadrature components. Is something missing on this block diagram?

38 Downconversion 0 I NCO 90 Q Project the input signal into quadrature components. Is something missing on this block diagram?

39 Optimized Demodulation Idea from Larry Doolittle. Carrier signal with θ phase advance per sample. y n = I cos nθ + Q sin nθ y n+1 = I cos(n + 1)θ + Q sin(n + 1)θ Rewrite as ( yn y n+1 ) = ( cos nθ sin nθ ) ( ) I cos(n + 1)θ sin(n + 1)θ Q To reconstruct I and Q, take the these two samples and multiply by the inverse matrix: ( ) I = 1 ( ) ( ) sin(n + 1)θ sin nθ yn Q D cos(n + 1)θ cos nθ y n+1

40 Upconversion 0 NCO I Q Σ 90 Start from I and Q (baseband or IF). Frequency translation. Phase and frequency shifting, modulations.

41 Outline 1 Introduction 2 Signal synthesis Arbitrary Waveform Generation CORDIC Direct Digital Synthesis 3 Up- and Down-conversion I&Q Processing Frequency Conversion 4 Signal Blocks Phase Shifting

42 Gain and Phase Block x[n] a 0 y[n] Two-tap FIR filter. Z 1 a 1 θ is the IF phase advance per sampling period. Works well near θ = π/2. Coefficients [ 1 cos θ 0 sin θ ] [ a0 a 1 ] [ cos φ = G sin φ ]

43 Gain and Phase Block x[n] a 0 y[n] Two-tap FIR filter. Z 1 a 1 θ is the IF phase advance per sampling period. Works well near θ = π/2. Coefficients [ 1 cos θ 0 sin θ ] [ a0 a 1 ] [ cos φ = G sin φ ]

44 Gain and Phase Block x[n] a 0 y[n] Two-tap FIR filter. Z 1 a 1 θ is the IF phase advance per sampling period. Works well near θ = π/2. Coefficients [ 1 cos θ 0 sin θ ] [ a0 a 1 ] [ cos φ = G sin φ ]

Design of NCO by Using CORDIC Algorithm in ASIC-FPGA Technology

Advance in Electronic and Electric Engineering. ISSN 2231-1297, Volume 3, Number 9 (2013), pp. 1109-1114 Research India Publications http://www.ripublication.com/aeee.htm Design of NCO by Using CORDIC