Typical wireless DSP system. Lecture 2 Data Conversion. Typical hard disk DSP system. Typical PCM voiceband DSP system.

Transcription

1 Lecture Data Converion Typical wirele DSP ytem Objective: Review ignal converion in context of DSP ytem Important iue relating to ignal converion including: Sampling and aliaing Signal to quantization noie ratio Harmonic ditortion Sampling clock jitter Overampling converter Reference: DSP Sytem Deign uing the TMS30C6000 by N. Kehtarnavaz & M. Keramat, Prentice Hall 00 Two data converter required: low peed (8 ksps) and high reolution (3bit) for voice high peed (0 MSPS) and low reolution (8-bit) for RF DSP function integrated - mixture of ASIC & DSP proceor Lecture - Data converion - Lecture - Data converion - Typical PCM voiceband DSP ytem Typical hard dik DSP ytem Voiceband proceing often ue a low cot converter baed around a igmadelta ADC/DAC ytem with a reaonable amount of DSP hardware. Hard dik interface electronic alo rely heavily on digital ignal proceing. Data converion i normally done at very high frequency (00 MSPS or more) and low reolution (6 bit). Complex feedback ytem between DSP, ADC and analogue circuit. Very demanding example: high performance, low power and low cot! Lecture - Data converion - 3 Lecture - Data converion - 4

2 Typical motor control DSP ytem Aliaing in Sampling Sampling theorem: given a ignal which i bandlimited to f max ampling frequency f mut be at leat *f max. *f max i called the Nyquit frequency. Sampling the ignal below thi rate will caue ignal corruption - aliaing DSP proceor are often ued in control application. Motor control i an example - low peed and moderate reolution. Data converion requirement tend to be modet. Control algorithm can be very complex. Lecture - Data converion - 5 Lecture - Data converion - 6 Harmonic Ditortion and ampling Real life circuit are not perfectly linear Output of linear (analogue) circuit may have ditortion Subequent ampling (e.g. aturation of the input tage of a ADC ytem) will caue puriou ignal appearing in the baeband Quantization effect of A/D converter V ref N The minimum tep that an A/D can reolve i: LSB Quantization noie n q i aumed to be ignal independent and i uniformly ditributed over -0.5 LSB and 0.5 LSB, then the quantization noie variance i: σ q Ε[ nq ] nq dnq Effect on pectrum of a inewave: Lecture - Data converion - 7 Lecture - Data converion - 8

3 Signal-to-Noie Ratio of A/D converter For a inuoidal ignal with an amplitude of A m : P 0log P n 0log ( A ) m V Vref N SNR 3 max 0log 0log 6.0N +.76dB V ref N ref N Signal-to-Noie Ratio for Gauian ignal Aume a Gauian ignal with zero mean and tandard deviation σ uch that: V Kσ SNR max σ 0 log 6N log σ q For a linear time-invariant ytem (e.g. FIR, IIR filter), the noie variance at the output caued by input quantization i: σ where h[n] impule repone of the ytem ref o σ q n h [] n 0 K db Lecture - Data converion - 9 Lecture - Data converion - 0 Effect of ampling time jitter Jitter aociated with ampling clock contribute to additive noie For inuoidal ignal, maximum allowable time jitter which reult in le than ½LSB i given by: tmax N + π The following graph provide a ueful guideline: f in Signal recontruction with D/A converter Perfect recontruction can be achieved by filtering the ampled ignal with a brickwall filter, which i the ame a convoluting the ampled ignal with a inc function: v a () t v ( kt ) a k t kt inc T Lecture - Data converion - Lecture - Data converion -

4 Zero-order hold effect of D/A converter Thi i difficult to achieve, therefore ue approximation of ample-and-hold function with a D/A converter. The tranfer function of a D/A converter i: H ( jω) jω e jω jωt in ωt ( ωt / ) jωt f / jπf / f / e inc f e How to reduce thi D/A error? Two approache to reduce thi error: inc correction: pre-ditort ignal to compenate for thi error over-ampling Lecture - Data converion - 3 Lecture - Data converion - 4 Sigma-Delta A/D converter: why? Two type of A/D converter: Nyquit rate: flah A/D, ucceive approximation A/D Overampling: igma-delta Advantage of igma-delta converter: Inherently linear High reolution (6-4 bit) Good for mixed-ignal IC procee (e.g. CMOS) No ample-and-hold circuit required Diadvantage: Limited to voiceband and audio Difficult to multiplex one ADC to multiple channel Almot all audio ADC ue igma-delta technqiue Overampling A/D If the overampling ratio OSR i: f OSR f max SNR can be improved according to: SNRoverampli ng 6.0N log( OSR) db Lecture - Data converion - 5 Lecture - Data converion - 6

5 Eay antialiaing filter Sigma-delta modulator for noie-haping Overampling A/D require imple lowpa filter at input Lecture - Data converion - 7 Ue igma-delta modulator to puh the quantization noie power toward high frequency end, which i then filtered out Lecture - Data converion - 8 Block diagram of a igma-delta converter Overampled D/A converter Lecture - Data converion - 9 Lecture - Data converion - 0

6 Overampled D/A converter TLC30AD535 igma-delta codec Switch-Cap comb filter t order Σ modulator Typical 6-bit Σ DAC Lecture - Data converion - Lecture - Data converion -