Politecnico di Torino Electronic Eng. Master Degree Analog and Telecommunication Electronics D6 - High speed A/D converters» Spectral performance analysis» Undersampling techniques» Sampling jitter» Interleaving ADC» Dithering AY 2015-16 04/05/2016-1 ATLCE - D6-2012 DDC 2012 DDC 1
Lesson D6: high speed ADC Spectral performance analysis Undersampling techniques Sampling jitter Interleaving ADC Dithering References: Application Report SLAA510 January 2011 ADC Input Noise: Is No Noise Good Noise? Analog Dialogue, - Febr 2006 04/05/2016-2 ATLCE - D6-2012 DDC 2012 DDC 2
SNR t and ENOB Each unit in the ADC chain introduces errors and noise Aliasing, quantization, sampling jitter Other errors (amplifier, mux, ) Actual accuracy depends from all these elements Key parameter: total Signal/Noise ratio: SNR t Not just the bit number N of the A/D ENOB = (SNR t - 1,76)/6 = SNR/6-0,3 Represents the number of actually useful bits of the ADC (sys) ENOB is always less then N ENOB = N-1, N-2.. good system design ENOB < N-3.. bad system design 04/05/2016-3 ATLCE - D6-2012 DDC 2012 DDC 3
AD systems glossary (similar to ampl.) SNR: Signal-to-Noise Ratio. Ps/Pn, excluding DC and first five harmonics (sometime first 9). SFDR: Spurious Free Dynamic Range. Ps/Ph (Ph is the highest spur). THD: Total Harmonic Distortion. Ps/Pd (Pd is the power of the first five (or 9) harmonics) SINAD: SIgnal to Noise And Distortion (SNR T for ADC). Ps/(Pn+Pd) (no DC) Can be specified in dbc(db to carrier, reference is the fundamental), or dbfs(db to full scale, fundamental extrapolated to full-scale). 04/05/2016-4 ATLCE - D6-2012 DDC 2012 DDC 4
Aliasing: Spectrum folding 04/05/2016-5 ATLCE - D6-2012 DDC 2012 DDC 5
Folding of harmonics SFDR Fundamental Largest spurious (Ph) Folded harmonics spurious Harm 2 Harm 3 Harm 6 04/05/2016-6 ATLCE - D6-2012 DDC 2012 DDC 6
Spectral view of ADC parameters 04/05/2016-7 ATLCE - D6-2012 DDC 2012 DDC 7
Spectral view of SFDR Fundamental 3 H2 3 4 5 6 Fs/2 Fs SFDR Fundamental Largest spurious Folded harmonics Spurious Harm 2 Harm 3 Harm 6 04/05/2016-8 ATLCE - D6-2012 DDC 2012 DDC 8
Oversampling Sampling at a rate far higher than the Nyquist limit Example: 3 khz audio signal (Nyquist = 6 ks/s) 8 ks/s Nyquist sampling; 1 MS/s Oversampling Oversampling sends aliased spectra far from baseband Reduced aliasing noise, folded from first alias Relaxed specifications on the anti-alias input filter Quantization noise is spread over a wider band (0 - Fs) Reduced spectral density of quantization noise Higher bit rate (more samples/s) Can be reduced with digital filtering Move complexity from analog digital domain 04/05/2016-9 ATLCE - D6-2012 DDC 2012 DDC 9
Oversampling vs. Nyquist Nyquist Main spectrum (baseband) First alias Second alias X(ω) f 0 F S1 2F S1 Quantization noise (0-Fs1 band) Oversampling X(ω) First alias f 0 Quantization noise (0-Fs2 band) F S2 04/05/2016-10 ATLCE - D6-2012 DDC 2012 DDC 10
Oversampling vs. Nyquist filtering Nyquist X(ω) Steep filter f Oversampling F S1 0 2F S1 Different filters: same quantization noise power (after reconstruction filter) X(ω) Smooth filter 0 F S2 f 04/05/2016-11 ATLCE - D6-2012 DDC 2012 DDC 11
Oversampling vs. Nyquist noise Nyquist X(ω) Steep filter f F S1 0 2F S1 Oversampling X(ω) Steep filter Same filter: reduced quantization noise power (after reconstruction filter) Removed quantization noise 0 F S2 f 04/05/2016-12 ATLCE - D6-2012 DDC 2012 DDC 12
Which is the actual limit? Actual Nyquist rule: A signal must be sampled at least twice the signal BANDWIDTH Example: a 1 GHz carrier, 100 khz BW signal can be safely sampled at Fs > 200 ks/s Spectrum is folded around K Fs/2 Less stringent specs for RF A/D converters Sampling rate related with bandwidth, not carrier Tight specs for the S/H sampling jitter related with carrier, not bandwidth 04/05/2016-13 ATLCE - D6-2012 DDC 2012 DDC 13
Filter for Nyquist sampling NYQUIST X(ω) Steep antialias filter, to limit aliasing noise Spectrum segment folded to baseband (aliasing noise) f F S 0 2F S F S /2 A/D Complex analog LP filter 04/05/2016-14 ATLCE - D6-2012 DDC 2012 DDC 14
Oversampling: more simple filter Complex, steep digital filter: - reduce noise - reduce bit rate (decimation) Alias is far away; antialias analog filter can be simple X( ) f 0 F S2 04/05/2016-15 ATLCE - D6-2012 DDC 2012 DDC 15
Filters with oversampling NYQUIST Complex analog LP filter A/D OVERSAMPLING Simple analog filter A/D Move complexity from the analog to the digital domain Complex digital filter Can reduce the bit rate (decimation) 04/05/2016-16 ATLCE - D6-2012 DDC 2012 DDC 16
Oversampling: noise shaping Oversampling X(ω) Reconstruction filter Flat quantization noise 0 F S2 f Noise shaping X(ω) Shaped quantization noise 0 In ΣΔ ADC noise power is moved to HF, with lower power density in baseband F S2 f 04/05/2016-17 ATLCE - D6-2012 DDC 2012 DDC 17
Standard sampling Sample at (at least) 2 x signal frequency Keeps aliases out of useful band Standard technique: signal rebuilt with low-pass filter 04/05/2016-18 ATLCE - D6-2012 DDC 2012 DDC 18
Undersampling Sample at 2 x signal bandwidth (can be far less than signal frequency) Aliases arise, but out of useful band Signal can be rebuilt with bandpass filter; no informatin loss 04/05/2016-19 ATLCE - D6-2012 DDC 2012 DDC 19
Undersampling - correct 04/05/2016-20 ATLCE - D6-2012 DDC 2012 DDC 20
Undersampling not correct 04/05/2016-21 ATLCE - D6-2012 DDC 2012 DDC 21
ADC example 04/05/2016-22 ATLCE - D6-2012 DDC 2012 DDC 22
Block diagram 04/05/2016-23 ATLCE - D6-2012 DDC 2012 DDC 23
Electrical characteristic 04/05/2016-24 ATLCE - D6-2012 DDC 2012 DDC 24
Output signal spectrum Sampling rate 500 Ms/s (MSPS) SFDR 04/05/2016-25 ATLCE - D6-2012 DDC 2012 DDC 25
Undersampling Sampling: 500 Ms/s < Nyquist In-band Alias (500-300) 04/05/2016-26 ATLCE - D6-2012 DDC 2012 DDC 26
Intermodulation Input signal: Fin1: 65,1 MHz Fin2: 70,1 MHz Sideband at 65,1 5 70,1 + 5 Higher sideband at 65,1 10 70,1 + 10 04/05/2016-27 ATLCE - D6-2012 DDC 2012 DDC 27
Differential and integral nonlinearity 04/05/2016-28 ATLCE - D6-2012 DDC 2012 DDC 28
Spurious Free Dynamic Range dbfullscale: error referred to full scale (SNR independent from signal level) dbcarrier: error referred to carrier (SNR depends on signal level) 04/05/2016-29 ATLCE - D6-2012 DDC 2012 DDC 29
Clock jitter High speed ADC need precise sampling low clock jitter differential clock Effect of clock jitter: SNRj = -20 log 10 2π Fin Tj 04/05/2016-30 ATLCE - D6-2012 DDC 2012 DDC 30
Aperture jitter specification 04/05/2016-31 ATLCE - D6-2012 DDC 2012 DDC 31
Interleaving ADCs 04/05/2016-32 ATLCE - D6-2012 DDC 2012 DDC 32
Spectrum folding with interleaved ADC 04/05/2016-33 ATLCE - D6-2012 DDC 2012 DDC 33
Averaging ADCs 04/05/2016-34 ATLCE - D6-2012 DDC 2012 DDC 34
Averaging ADC benefits 04/05/2016-35 ATLCE - D6-2012 DDC 2012 DDC 35
Dithering 04/05/2016-36 ATLCE - D6-2012 DDC 2012 DDC 36
Adding sampling jitter (dither) 04/05/2016-37 ATLCE - D6-2012 DDC 2012 DDC 37
Dither effect 04/05/2016-38 ATLCE - D6-2012 DDC 2012 DDC 38
Lesson D6 final test Describe the limits of undersampling techniques. Plot spectrum of undersampled sinewave, with no distortion and with some distortion. Which parameters contribute to total sampling jitter? Describe the interleaving ADC technique and related benefits. Describe the averaging ADC technique and related benefits. Why dithering can improve ADC performance? 04/05/2016-39 ATLCE - D6-2012 DDC 2012 DDC 39