fundamentals on continuousfundamentals ... - ieee web...
TRANSCRIPT
CASS: 2013-2014 DLP Program UT Austin
Fundamentals on ContinuousFundamentals on Continuous--TimeTimeFundamentals on ContinuousFundamentals on Continuous Time Time SigmaSigma--Delta ModulatorsDelta Modulators
DLP Series, sponsored by the CASSDLP Series, sponsored by the CASS
Jose SilvaJose Silva--MartinezMartinezAmesp02.tamu.edu/~jsilvap / j
[email protected] of ECE
Texas A&M Universitye as & U e s ty
September 19, 2013
- 1 -J. Silva-Martinez
CASS: 2013-2014 DLP Program UT Austin
Outline Introduction Fundamentals of Sigma-Delta Modulators
Stability Clock Jitter Blockers Tolerance
Quantizer Issues Voltage Mode versus Current Mode
Improving blocker tolerancep g Efficient non-invasive filtering Saturation detector
Current Research Projects Current Research Projects TAMU: Where we are? Conclusions
- 2 -J. Silva-Martinez
CASS: 2013-2014 DLP Program UT Austin
Introduction: ConnectivityWLANWLAN
UWB, 802.15.3a
WI-FIMAN
Wi-Max, 802.16a,e
GPS
802.16a,e
BLUETOOTH
Connectivity to EVERYTHING
PAN, 802.15BLUETOOTH
Increasing number of wireless standards
Support of multiple-standards on the same chip
- 3 -J. Silva-Martinez
Advances in Integrated RF design towards universal devices
Software Radio: easy addition of new standards
CASS: 2013-2014 DLP Program UT Austin
Direct Conversion Multi-Standard Receiver
System issues in broadband systems:High frequency filtering is especially critical in broadband applicationsRejection of Blockers: ADC filtering must be complemented by LP filteringRejection of Blockers: ADC filtering must be complemented by LP filteringNeighbor channels are quite relevant even if heavy filtering is used
Trade‐offs:
- 4 -J. Silva-Martinez
Light filtering in front demands an ADC with higher SNR and higher SDRHigher SNDR‐ADC implies more power and more circuit complexity
CASS: 2013-2014 DLP Program UT Austin
State of the artState of the art
1.E+06
1.E+07ISSCC 2010ISSCC 1997-2009VLSI 1997 2009
State of the artState of the arts)
1 E+03
1.E+04
1.E+05VLSI 1997-2009FOM=100fJ/conv-stepFOM=10fJ/conv-step
10mW @10MHz10mW @10MHzs(p
Joul
es
1.E+01
1.E+02
1.E+03
VideoVideo
10mW @10MHz10mW @10MHz
Pow
er/f
s
1.E-01
1.E+00
10 20 30 40 50 60 70 80 90 100 110 120
VideoVideo
B. Murmann, http://www.stanford.edu/~murmann/adcsurvey.html.SNDR (dB)SNDR (dB)
Challenge 1: Jitter Tolerance
- 5 -J. Silva-Martinez
Challenge 1: Jitter Tolerance
CASS: 2013-2014 DLP Program UT Austin
Challenge 2: Co-existenceSimultaneous Usage of Radio Bands
Bm
)Po
wer
(dB
Outpu
t
(G )
RF Interference: Frequency Overlap, O t f B d E i i R i S t ti
* Plot courtesy of Camille Chen/IntelFrequency (GHz)
- 6 -J. Silva-Martinez
Out‐of‐Band Emissions, Receiver Saturation
CASS: 2013-2014 DLP Program UT Austin
SigmaSigma--Delta Modulators:Delta Modulators:
Practical Design IssuesPractical Design Issues
- 7 -J. Silva-Martinez
CASS: 2013-2014 DLP Program UT Austin
Conventional (Conventional (NyquistNyquist) ADC) ADC
SQNR=6 02*n+1 76SQNR=6 02*n+1 76
- 8 -J. Silva-Martinez
SQNR=6.02*n+1.76SQNR=6.02*n+1.76
CASS: 2013-2014 DLP Program UT Austin
Basic concepts in Basic concepts in ModulatorsModulators
Vin
T 1H(s) A/D
Data Out
DACs
s FT
)s(HSTF NTF
1
- 9 -J. Silva-Martinez
)s(H1 )s(HNTF
1
CASS: 2013-2014 DLP Program UT Austin
Fundamentals of Oversampled A/D ConversionFundamentals of Oversampled A/D ConversionqEFilter
1
1
1
ZZ
q
ZY ZX
Filter
Ch k th i t t t t j t i
DAC
11 2221
T
eeeeZzH
T
TjTjTjTje
Check the input-output trajectoriesOriginal E(z) is shaped by NTF
O i i l E( ) i Spot SQNR
222 TsinjezH
Tj
e Original E(z) is
amplified here!
21
zHzEZzXSQNR
Spot SQNR
22 TsinzHe
2
2 1
T*zEzXSQNR
zHzE e
- 10 -J. Silva-Martinez
2
2
TsinzE
CASS: 2013-2014 DLP Program UT Austin
Oversampled A/D Conversion Oversampled A/D Conversion qEFilter
OSR***.SQNR N2
32 31251
1
1
1
ZZ
q
ZY ZX
DAC
Filter
Si l t Q ti ti 221 zXZzX
SQNR
dBOSRlog*.N*..SQNR 103025026761 DAC
Signal to Quantization
Noise Ratio (SQNR)
0
2
0
2 dfzHzEdfzHzESQNR
f
e
f
e
bb
N=number of bits
OSR=fs/2fb
SQNR improves by 30dB when
0
2
2
22
1
thffbIf
dfTsin*
zEzXSQNR
f b
SQNR improves by 30dB when
OSR increases by 10
Or 9dB SQNR improvement
2
3
2
2
26
2
212
21
f*f
f**
V
VSQNR
then,fsfbIf
S
b
Sq
N
- 11 -J. Silva-Martinez
when doubling OSR
2
32 31251
212
OSR***.SQNR
ff
*V
N
S
S
q
CASS: 2013-2014 DLP Program UT Austin
System Design ConsiderationsSystem Design Considerations
150
B) Filter order=5Dashed lines are for 2 bit quantizer
125
e R
atio
(dB
Filter order=4Quantizer Bits=1
Quantizer Bits=1Dashed lines are for 2-bit quantizerDashed=Solid+6.02
75
100
tion
Noi
se
Filter order=3Quantizer Bits=1
Q
50
o Q
uant
iza
Filter order=2Quantizer Bits=1
0
25
Sign
al to
- 12 -J. Silva-Martinez
0
0 10 20 30 40
Oversampling Ratio (OSR)
CASS: 2013-2014 DLP Program UT Austin
Design Considerations: 3Design Considerations: 3rdrd order looporder loop
20
) 3rd order loop
16
18
ts (E
NO
B)
4-bit quantizer
3-bit quantizer
3rd order loop
12
14
ber o
f Bit 3-bit quantizer
1-bit Quantizer
8
10
tive
Num
b
2-bit Quantizer
761)dB(SNDR
4
6
Effe
ct
02.676.1)dB(SNDRENOB
- 13 -J. Silva-Martinez
4 8 12 16 20 24 28 32 36 40
Oversampling Ratio (OSR)
CASS: 2013-2014 DLP Program UT Austin
Oversampled A/D ConversionOversampled A/D ConversionFeedforwardFeedforward Architecture Architecture
Eq stands for the quantization noise
Ed stands for DAC non-idealities (jitter + thermal noise) Ed stands for DAC non idealities (jitter + thermal noise)
Filter’s thermal noise is accounted in Eh
The modulator’s output becomes
E*NTFEEX*STFY sZOH*sH
The error signal (Filter’s input) is qhd E*NTFEEX*STFY
1
1
1
Z*sZOH*sHSTF
1
- 14 -J. Silva-Martinez
11
Z*sZOH*sHNTF qhde E*ZEEXZOH)s(H*NTFV 1
CASS: 2013-2014 DLP Program UT Austin
Oversampled A/D Conversion: EffectsOversampled A/D Conversion: Effects of ZOHof ZOH
S
Tj
TzPhaseeZ S
1
Notice that
S
fdB
0
‐5
0
fi
SffcsinfZOH
dB
‐15
‐10
‐5
‐20
‐15
‐10
S
Sff
sin
f
fcsin
1 dB4 dB
‐30
‐25
‐20
‐30
‐25
20
Sf
f20 dB
- 15 -J. Silva-Martinez
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10.1 1
Sf
f
CASS: 2013-2014 DLP Program UT Austin
Stability Issues:Stability Issues: ModulatorsModulators NotNot very difficult to stabilize the loop if very difficult to stabilize the loop if
the unity gain frequency is below Fs/4! the unity gain frequency is below Fs/4! (not very Low OSR); e.g. around (not very Low OSR); e.g. around 100Mhz if clock frequency is 400MHz100Mhz if clock frequency is 400MHz
NoticeNotice that larger OSR allow you to that larger OSR allow you to reduce the filter gain at highreduce the filter gain at high--frequency frequency
OutOut ofof Band noise is distributedBand noise is distributed inin OutOut--ofof--Band noise is distributedBand noise is distributed in in wider bandwidth, hence smaller noise wider bandwidth, hence smaller noise density but same integrated noisedensity but same integrated noise
ADC
Freq0
L(f)
fH
H(f) fS0.5fS
Loop Hard toHard toL(f)
2
3
Loop Phase
Hard to Hard to implement H(s) implement H(s)
SplitSplit the filter in the filter in 2 parts2 parts
- 16 -J. Silva-Martinez
Sf
f2
22 parts2 parts
Loop filter and Loop filter and fast pathfast path
CASS: 2013-2014 DLP Program UT Austin
Effect of the blockersEffect of the blockers on loop operationon loop operation Remarks on Filter’s operation: Filter’s input signalRemarks on Filter’s operation: Filter’s input signal
E*ZEEXV
qhde
1
sLe
1
X(f)
InbandInband signal is usually very small: Nice propertysignal is usually very small: Nice property
ADC
H(f)
1/(1+L) InbandInband signal is usually very small: Nice propertysignal is usually very small: Nice property
TransitionTransition band is more critical for filter’s linearity band is more critical for filter’s linearity (neighbor channels); (neighbor channels);
Filter must be designed for the blockersFilter must be designed for the blockers
0 dBFsNTF Freq
( )
- 17 -J. Silva-Martinez
Blocker tolerance is a major issue in broadband Blocker tolerance is a major issue in broadband applications (LTE)applications (LTE)
ZOH(f)
CASS: 2013-2014 DLP Program UT Austin
System Optimization: Tuning Filter ParametersSystem Optimization: Tuning Filter Parameters
Input Signature
Amplitude/Phase Comparators Filter tuning
schemeFilter tuning scheme
Calibration Algorithm
l
ADC
DAC
cal
op
cal
High-Order LP Filter
DAC
0 f0f
Effect of PVT VariationsEffect of PVT Variations FilterFilter TuningTuning;; reconfigurationreconfiguration oror
Calibration of (standalone) building blocks:Calibration of (standalone) building blocks:Does not guarantee loop stability (excess loop delay)Does not guarantee loop stability (excess loop delay)
MasterMaster slaveslave techniquestechniques couldcould bebe usedused
- 18 -J. Silva-Martinez
Does not guarantee loop stability (excess loop delay) Does not guarantee loop stability (excess loop delay) Does notDoes not guarantee best NTF (Coefficient adjustments)guarantee best NTF (Coefficient adjustments)
CASS: 2013-2014 DLP Program UT Austin
System Optimization: Global Tuning SchemeSystem Optimization: Global Tuning Scheme
Linear model Linear model used to findused to findused to findused to findNTFNTF
Model used to Model used to calibrate NTFcalibrate NTFcalibrate NTFcalibrate NTF
Loop response Loop response to calibration to calibration tones istones istones is tones is analyzed in analyzed in digital domaindigital domain
- 19 -J. Silva-Martinez
F. Silva-Rivas, et. al., “Digital Based Calibration Technique for Continuous-Time Bandpass Sigma-Delta Analog-to-Digital Converters,” Analog Integrated Circuits and Signal Processing, April-09.
C.Y. Lu, et. al., “A Sixth-Order 200MHz IF Bandpass Sigma-Delta Modulator With over 68dB SNDRin 10MHz Bandwidth,” IEEE J. Solid-State Circuits, June 2010.
CASS: 2013-2014 DLP Program UT Austin
Global Tuning Scheme for NTF Parameters: Global Tuning Scheme for NTF Parameters: DigitalDigital
A set of strategic tones are usedA set of strategic tones are used to optimize:Loop StabilityBest NTF
- 20 -J. Silva-Martinez
Bandwidth
CASS: 2013-2014 DLP Program UT Austin
Digitally assisted calibration schemeDigitally assisted calibration scheme
f
Test Tones
N-channel
Wideband Programmable Filter
Dout
Quantizer
-1/2Vi
Calibration
ff0
Freq calibration
Software Platform
N channel digital
receiver
z-1/2
Programmablez-1/2
Vin
Filter's control
DSPNRZ DACArray of DACs
DACs Control
Calibration Algorithm
Data control
- 21 -J. Silva-Martinez
A 6th-Order 200MHz IF Bandpass Sigma-Delta Modulator With over 68dB SNDR in 10MHz Bandwidth, C.Y. Lu, et.al., IEEE J. Solid-State Circuits, June 2010.
CASS: 2013-2014 DLP Program UT Austin
ADC Calibration: Experimental resultsADC Calibration: Experimental results
- 22 -J. Silva-Martinez
CASS: 2013-2014 DLP Program UT Austin
Design Examples
- 23 -J. Silva-Martinez
CASS: 2013-2014 DLP Program UT Austin
ADC Architecture Employing a TDC as ADC Architecture Employing a TDC as QuantizerQuantizerQQ
Dout
P(t)
Conventional Time domain Quantizer-DAC
• Multi-level quantizer and Digital to Analog Converter (DAC) are replaced by PWM generator and Time to Digtal Converter (TDC)
• Width of p(t) is proportional to the amplitude of the signal in a given• Width of p(t) is proportional to the amplitude of the signal in a given clock period
• Output code (Dout) represents “quantized pulse” edges with a
- 24 -J. Silva-Martinez
quantization step size = TQ
ISSCCISSCC--09, JSSC09, JSSC--2011 (March)2011 (March)
CASS: 2013-2014 DLP Program UT Austin
Output Spectrum: Output Spectrum: --5dB Input5dB Input
Output spectrum for 5dB 4MHzfor -5dB 4MHzInput signal
• SNR=62dB• SNR=62dB • THD=-65dB
- 25 -J. Silva-Martinez
CASS: 2013-2014 DLP Program UT Austin
Second ADC:Second ADC:System Architecture:System Architecture: Injection LockingInjection LockingSystem Architecture: System Architecture: Injection LockingInjection Locking
Φ1
Φ2
Φ3
RF
RQ
C C C
RC
5th d 3 bitOUT Φ4
Φ5
Φ6
Φ
+-
+-
RF
RF
RQ
C C
C C+- +
-
RF
Rin
Rin
OUTBP
OUTLP+-
RgC
C
+-
Rg
RC
IN IN• 5th-order 3-bit feedforward architecture
b5 3-Bit Two-StepQuantizer
Vin
Di it lBP
LP
2nd-Order Filter
Φ7
T s0
T s/7
2Ts/7
3Ts/7
4Ts/7
5Ts/7
6Ts/7
+-
BP
LP
2nd-Order Filter LP
1st-Order
• Local feedback is to compensate the excess loop delay
b1
b2
b4
Σb3
DigitalOutput
3-Bit NRZ DAC
Complementary Injection-4
excess loop delay
• LC-VCO+CILFD are used to generate
Level-to-PWM
VCO
1-B
it PW
M D
AC
Complementary InjectionLocked Frequency Divider
Programmable Delay
CILFD
gclean reference clocks
- 26 -J. Silva-Martinez
Converter11
“25MHz Bandwidth (BW) Continuous-Time Lowpass ΣΔ Modulator with Time-Domain3-bit Quantizer and DAC” Cho-Ying Lu, et.al., Sept 2010, JSSC
CASS: 2013-2014 DLP Program UT Austin
33--bit Algorithmic bit Algorithmic QuantizerQuantizer
The output is composed by 1 MSB + 3LSB+ 3LSB The MSB is determined first
- 27 -J. Silva-Martinez
CASS: 2013-2014 DLP Program UT Austin
LCLC--VCO + CILFDVCO + CILFDb5Vin 2nd-Order + 2nd-Order 1st-Order
b1
b2
b4
b5
Σb3
VCO
3-Bit Two-StepQuantizer
Vin
DigitalOutput
3-Bit NRZ DAC
BPFilter
Complementary Injection-Locked Frequency Divider
-BP
Filter
4
LP
Level-to-PWM Converter
VCO
Programmable Delay
11
CILFD
Ph i f VCO iPh i f VCO i 119dB /H @ 1MH119dB /H @ 1MH Jitter among phases
- 28 -J. Silva-Martinez
Phase noise of VCO is Phase noise of VCO is --119dBc/Hz @ 1MHz119dBc/Hz @ 1MHz CILFD phase noise is CILFD phase noise is --136dBc/Hz @ 1MHz136dBc/Hz @ 1MHz
Jitter among phases is highly correlated
CASS: 2013-2014 DLP Program UT Austin
Output Spectrum of the Modulator-2.2dBFS @ 5.08MHz
Peak SNR= 68.5dB @25MHz BW@ Peak SNDR= 67.7dB @25MHz BW SFDR>70dB
- 29 -J. Silva-Martinez
CASS: 2013-2014 DLP Program UT Austin
*Current-mode Quantizer
C. J. Park, **M. Onabajo, H. M. Geddada, J. Silva-Martinez, and A. I. KarsilayanMartinez, and A. I. Karsilayan
Paper under evaluation
*Partially sponsor by SRCPartially sponsor by SRC
** Currently with Northeastern University
- 30 -J. Silva-Martinez
CASS: 2013-2014 DLP Program UT Austin
Typical Typical QuantizerQuantizer: : Flash Flash ArchitectureArchitectureypyp QQ S/H operates at clock rate
Huge input capacitance if N>6 Huge input capacitance if N>6Kick back noise
Requires a precise low- Requires a precise lowimpedance resistive ladder: Power-accuracy-Speed tradeoff
Limited by comparatorSpeed and accuracyOff t ltOffset voltage
Hard to improve its resolution
- 31 -J. Silva-Martinez
State of the art: State of the art: ~ >2.4 ~ >2.4 GS/s 6 bits resolution GS/s 6 bits resolution
CASS: 2013-2014 DLP Program UT Austin
Conventional ADC ArchitectureConventional ADC Architecture
A CTSD modulator with feed-forward (FF) compensation. Summing Stage + Quantizer
DA
C2
DA
C1
Summing stage and quantizer in the conventional continuous- Summing stage and quantizer in the conventional continuoustime sigma-delta modulator with feed-forward compensation.
Excess loop delay (in addition to z-1) should be minimized. The direct path around the quantizer should be very fast
- 32 -J. Silva-Martinez
The direct path around the quantizer should be very fast. The summing amplifier must have a high unity-gain frequency.
CASS: 2013-2014 DLP Program UT Austin
Voltage Mode Voltage Mode QuantizerQuantizer
Performance summary of prior Adder-Quantizer in CT ΣΔ Modulators
Flash ADC architecture generally achieve the highest sampling rate and is usedas integrated quantizers in sigma-delta modulator
[23] [29] Technology
(nm) 180nm CMOS 180nm CMOS
Supply voltage 1.8V 1.8V
QuantizerQuantizer Resolution 3 bits 4 bits
Sampling rate 400MHz 800MHz
Input range 400mVpp 3Vpp
PowerAdder* 10mW Adder* 8.5mW
* Power consumption of the summing amplifier only.** 3-bit two-step Flash ADC.
[23] C.-Y. Lu, M. Onabajo, V. Gadde, Y.-C. Lo, H.-P. Chen, V. Periasamy, and J. Silva-Martinez, “A 25MHz bandwidth 5th-ordercontinuous-time lowpass sigma-delta modulator with 67 7dB SNDR using time-domain quantization and feedback ” IEEE J Solid-State
Power Flash ADC** 24mW Flash ADC N/A
continuous time lowpass sigma delta modulator with 67.7dB SNDR using time domain quantization and feedback, IEEE J. Solid StateCircuits, vol. 45, no. 9, pp. 1795-1808, Sep. 2010.[29] V. Singh, N. Krishnapura, S. Pavan, B. Vigraham, D. Behera, and N. Nigania, “A 16MHz BW 75dB DR CT ΣΔ ADC compensated formore than one cycle excess loop delay,” IEEE J. Solid-State Circuits, vol. 47, no. 8 pp. 1884-1895, Aug. 2012.
Demand for low power consumption
- 33 -J. Silva-Martinez
p p
Make the architecture robust to process-voltage-temperature (PVT) variations
CASS: 2013-2014 DLP Program UT Austin
Conventional VoltageConventional Voltage--Mode Summing and Flash ADCMode Summing and Flash ADC
12
14
16
mS)
CL=50fFCL=100fFCL=150fFCL=200fF
4
6
8
10
Tran
scon
duct
ance
(m
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
2
4
Beta
The loop gain is determined by the resistive feedback as well as the load TheThe minimumminimum transconductancetransconductance requiredrequired
AVloop ωGm
1RF
RLeq1 sRLeq CL
The loop gain is determined by the resistive feedback as well as the loadcapacitance.
RFF1 RFF2 … RFF RFRF
Th f th i lifi t DAC t i th
TheThe minimumminimum transconductancetransconductance requiredrequiredisis usuallyusually dictateddictated byby thethe valuevalue ofof DCDC looploopgaingain requirementrequirement
The response of the summing amplifier to a DAC step requires more thanfive loop time constants for a settling accuracy of 0.5%.
GBWGm
1RF AV DC 4
AsAs anan example,example, thethe casecase ofof 2626dBdB DCDC looploopgaingain withwith RRFFFF11||||....||R||RFFNFFN ==500500ΩΩ demandsdemands
- 34 -J. Silva-Martinez
GBW FCL
_
RLeq CL Tclk /2gg FFFF11|||| |||| FFNFFNGGmm>>4040mA/VmA/V
CASS: 2013-2014 DLP Program UT Austin
Conventional VoltageConventional Voltage--Mode Flash ADCMode Flash ADC
Conventional 3-bit voltage-mode flash ADC
The input signal (Vin) iscompared to seven referencevoltage levels using sevencomparators followed by latches.
To minimize the impact of PVT, large area resistors (R) and intricatelayout matching techniques are required
To minimize the effects of the kickback noise relative small values are
Larger transistor dimensions would be counterproductive with regards tothe maximum achievable speed
- 35 -J. Silva-Martinez
To minimize the effects of the kickback noise, relative small values arepreferred for the resistors (R) at the expense of larger static powerconsumption
CASS: 2013-2014 DLP Program UT Austin
Proposed CurrentProposed Current--Mode Flash Mode Flash ADC Architecture ADC Architecture
Incoming signals are added at the source of a common-gate stage
Input signals are replicated through NMOS current mirrors and compared with reference currents at high impedance nodeswith reference currents at high impedance nodes
The difference between input signal replicas and reference currents are converted to voltages at high impedance nodes
- 36 -J. Silva-Martinez
The different voltages are processed by seven comparator-latch combinations
CASS: 2013-2014 DLP Program UT Austin
CurrentCurrent--Mode AdderMode Adder--QuantizerQuantizer Minimum input impedance
Easy to interface it with loop filter and easy to implement the addition of filter coefficients
Two relevant parasitic poles limit its frequency response: Still can operate at GHz Resolution is limited by current mirror mismatches
Main design challenges:
Iq_REF1+IDC
VDD
Iq_REF2+IDC Iq_REF7+IDC
IBIAS2VBP0 VBP1
MP1 MP2 MP7MP0
Freset
Fsampleq1
Freset
Fsampleq2
Freset
Fsampleq7
VBP2
+-AV(s)
VCM
MXMCP1 MCP2 MCP7
MCP0B1 B2 B7
RFF1
RFF2
VFF1
V Comparator & Latch
VDD/2
Yout1
VDD/2
Comparator & Latch Yout2
VDD/2
Comparator & Latch Yout7
+CM
VBN2
MCN1 MCN1 MCN1MCN1MCN0
A
RFFN
VFF2
VFFN
IIn_AC
- 37 -J. Silva-Martinez
IBIAS1 IIn_AC+IDC IIn_AC+IDC IIn_AC+IDC IIn_AC+IDC
VBN1
VBN2
IDC=IIBIAS2-IIBIAS1
MN1 MN1 MN1MN1MN0
IDAC2
37
CASS: 2013-2014 DLP Program UT Austin
VoltageVoltage--Mode Summing Mode Summing vs. Currentvs. Current--Mode SummingMode Summing
Table II. Voltage-Mode Summing vs. Current-Mode Summing.
Voltage-Mode Current-Mode
vs. Currentvs. Current Mode SummingMode Summing
Voltage Mode Summing
Current ModeSumming
Technology (nm) 90nm CMOS 90nm CMOS Power Consumption(mW) 6.6mW* 3.1mW**
Input referred integrated noise 31 2 V 39 Vp g(in 20MHz) 31.2µV 39µV
Delay (@VLSB/4) 0.396ns 0.35ns * Power consumption of the summing amplifier only. ** Power consumption includes common-gate (CG) stage, Gm-boosting block and current
The most remarkable advantages of the Current-Mode approach: Superior performance with <50% power consumption. L ili
mirroring stages.
Less silicon area
In the Voltage-Mode case: Additional power should be added to account for the reference voltages generator
- 38 -J. Silva-Martinez
voltages generatorThe dual differential pair required at the input of each comparator.
CASS: 2013-2014 DLP Program UT Austin
Chip MicrophotographChip Microphotograph
The ADC area including bias gcircuitry: 0.0276mm2
The thermometer-to-binary encoder: 0.0072mm2.
m
Decoupling Capacitor
ESDESDESD ESD ESD
0.23mm
0.12
mm
Encoder
- 39 -J. Silva-Martinez 39
CASS: 2013-2014 DLP Program UT Austin
State of the artState of the art
Performance summary of the Proposed Current-Mode Adder-Quantizer and comparison with prior Quantizer in CT ΣΔModulators
[23] [29] This Work Technology
(nm) 180nm CMOS 180nm CMOS 90nm CMOS
Supply voltage 1.8V 1.8V 1.2VSupply voltage 1.8V 1.8V 1.2V
Quantizer Resolution 3 bits 4 bits 3 bits
Sampling rate 400MHz 800MHz up to 2GHz
Input range 400mVpp 3Vpp ±40App (equivalent to 400mVpp) * *
* Power consumption of the summing amplifier only.** 3-bit two-step Flash ADC.
Power Adder* 10mW Adder* 8.5mW Adder 1.1mW
Flash ADC** 24mW Flash ADC N/A Flash ADC 3.04mW
[23] C.-Y. Lu, et.al., “A 25MHz bandwidth 5th-order continuous-time lowpass sigma-deltamodulator with 67.7dB SNDR using time-domain quantization and feedback,” IEEE J. Solid-StateCircuits, vol. 45, no. 9, pp. 1795-1808, Sep. 2010.
[29] V Si h N K i h S P B Vi h D B h d N Ni i “A 16MH BW
- 40 -J. Silva-Martinez 40
[29] V. Singh, N. Krishnapura, S. Pavan, B. Vigraham, D. Behera, and N. Nigania, “A 16MHz BW75dB DR CT ΣΔ ADC compensated for more than one cycle excess loop delay,” IEEE J. Solid-StateCircuits, vol. 47, no. 8 pp. 1884-1895, Aug. 2012.
CASS: 2013-2014 DLP Program UT Austin
State of the artState of the art[20] * [21] ** This Work
Technology CMOS 0.13µm CMOS 0.18µm CMOS 90nm
Supply voltage 1.2VAnalog: 1.8V
1.2VDigital: 2.1-2.5V
Resolution 5 bits 4 bits 3 bits
Sampling rate up to 3.2GS/s up to 4GS/s up to 2GS/s
Input range 400mVpp ± 460mVpp ±40µApp
Power 120mWAnalog 78mW Analog 3.34mW
Digital (Including Clock Buffer) 530mW Digital (Including Clock Buffer) 8.94mW
DNL (After Calibration) -0.24LSB~ 0.18LSB (@3.2GS/s) - 0.14LSB ~ 0.15LSB (@4GS/s) - 0.206LSB ~ 0.138LSB (@1.48GS/s)
INL (After Calibration) -0.29LSB~ 0.39LSB (@3.2GS/s) - 0.24LSB ~ 0.20LSB (@4GS/s) - 0.056LSB ~ 0.206LSB (@1.48GS/s)PowerPower consumptionconsumption ifif extrapolatedextrapolated toto55 bitsbits isis lessless thanthan 1616mWmW forfor currentcurrent modemode
FoM 3.07pJ/conversion-step (2GS/s)
4.3pJ/conversion-step (3.2GS/s)
10.25pJ/conversion-step
(fin = 10MHz, 4GS/s)
1.37pJ/conversion-step
(fin = 10MHz, 1.48GS/s)
ENOB4.44 bits (2GS/s)
4 54 bits (3 2GS/s)
3.89 bits (4GS/s, 10MHz input)
3.48 bits (4GS/s, 100MHz input)
2.6 bits (1.48GS/s, 10MHz input)
2.6 bits (1.48GS/s, 44.4MHz input)
55 bitsbits isis lessless thanthan 1616mWmW forfor currentcurrent--modemoderealizationrealization
* Power consumption of output buffer was not included. ** Includes power consumptions for analog circuitry and digital encoder For the FoM calculation the digital power
4.54 bits (3.2GS/s)3.47 bits (3.4GS/s, 800MHz input) 2.54 bits (1.48GS/s, 118.4MHz input)
Area 0.18mm20.88mm2
(excluding resistor ladder)
ADC: 0.0276mm2
(excluding encoder)
- 41 -J. Silva-Martinez 41
Includes power consumptions for analog circuitry and digital encoder. For the FoM calculation, the digital power consumption for 4GS/s.
[20] Y.-Z. Lin, et. al, "A 5-bit 3.2-GS/s flash ADC with a digital offset calibration scheme," IEEE Trans. Very Large Scale Integration Systems, vol. 18, no. 3, pp. 509-513, March 2010. [21] S. Park, et. al, “A 4-GS/s 4-bit Flash ADC in 0.18μm CMOS,” IEEE J. Solid-State Circuits, vol. 42, no. 9, pp.1865-1872, Sep. 2007.
CASS: 2013-2014 DLP Program UT Austin
A Blocker Tolerant Continuous-Time ADC f B db d A li i * ADC for Broadband Applications*
H. M. Geddada, C. J. Park, H.-J. Jeon**, J. Silva-Martinez, and A. I. Karsilayan
*Sponsored by SRC
- 42 -J. Silva-Martinez
CASS: 2013-2014 DLP Program UT Austin
Effect of the blockersEffect of the blockers on loop operationon loop operation Remarks on Filter’s operation: Filter’s input signalRemarks on Filter’s operation: Filter’s input signal
E*ZEEXV
qhde
1
sLe
1
X(f)
InbandInband signal is usually very small: Nice propertysignal is usually very small: Nice property
ADC
H(f)
1/(1+L) InbandInband signal is usually very small: Nice propertysignal is usually very small: Nice property
TransitionTransition band is more critical for filter’s linearity band is more critical for filter’s linearity (neighbor channels); (neighbor channels);
Filter must be designed for the blockersFilter must be designed for the blockers
0 dBFsNTF Freq
( )
- 43 -J. Silva-Martinez
Blocker tolerance is a major issue in broadband Blocker tolerance is a major issue in broadband applications (LTE)applications (LTE)
ZOH(f)
CASS: 2013-2014 DLP Program UT Austin
Clock Jitter Sensitivity (SJNR)Clock Jitter Sensitivity (SJNR) The main effects of clock jitter is present at the input of the
quantizer and DAC.
Jitter induced noise at DAC output is processed according to the STF, which is a serious problem for continuous-time sigma-delta modulators
Clock Jitter in quantizer
is not very relevant:
processed by NTF.
Clock jitter introduce Clock jitter introduce
uncertainty at the DAC
output (jitter induced
- 44 -J. Silva-Martinez
noise) processed by STF.
CASS: 2013-2014 DLP Program UT Austin
Jitter Issues: High Clock FrequencyJitter Issues: High Clock FrequencyJitt d
Vin
Loop FilterDout
Jittered Clock
QuantizerVDAC
Re-timed Data
Multi-Bit DAC
Flip-Flops
The DAC error due to clock jitter:
Error FunctionError FunctionJittered Clock
DAC
The DAC error due to clock jitter:
In the frequency domain: Convolution of Differential of Dout and Jn()
TtTnDnDnE outoutj
1
noutS
noutj JDT
sinJDZE2
21 1
- 45 -J. Silva-Martinez
In-band signal is shaped by , then it is not very criticalOut-of-Band quantization noise and blockers convolve with the clock jitter
11 Z
CASS: 2013-2014 DLP Program UT Austin
Jitter tone around Fsconvolve with images ofinband signal and producei b d i tin-band noise tones
Jitter and blocker convolves, producing in-b d i tband noise tones.
Jitter also convolves with OOB quantization noise
0
No jitter tone 20
0
fj =80.26Mhz, f
b =75.27Mhz
and downconverts into in-band noise.
-60
-40
-20
SNRJitter = 67.2dB
dBFS
)
jJitter tone at 497.43Mhz
-60
-40
-20
SNR = 39.3dB
SNR = 71 2dBdBFS
)
Ideal
-100
-80 SNRIdeal = 71.2dB
Pow
er (
d
-100
-80
SNR = 71.2dB
Pow
er (
d
- 46 -J. Silva-Martinez10-1 100 101 102-140
-120
Frequency (MHz)
10-1 100 101 102-140
-120
Frequency (MHz)
CASS: 2013-2014 DLP Program UT Austin
Blocker Sensitivity: Feedforward ADC
0.6
0.7
V in (v
)
0.1 0.2 0.3 0.4 0.5 0.60.5
0.5
1
0.1 0.2 0.3 0.4 0.5 0.6-1
-0.5
0
DO
UT
Ti ( )
Loop filter, ADC, DAC Inherent antialiasing
Time ( sec)
STF=FF/(1+LG) 1(in-band); NTF=1/(1+LG)(Noise shaping) Quantizer can have relaxed specifications FF architecture is power and area efficient
- 47 -J. Silva-Martinez
Dynamically more stable
OOB peaking in STF
CASS: 2013-2014 DLP Program UT Austin
Another Major Issue: Another Major Issue: Loop SaturationLoop Saturation
10
20
30
-8
-4
r (dB
FS)
-20
-10
0
10
Gai
n (d
B)
BP 16
-12
able
blo
cker
pow
er
60
-50
-40
-30
G BP1LP1BP2LP2
-20
-16
Max
imum
allo
wa
- 48 -J. Silva-Martinez
107 108-60
Frequency (Hz)
Blocker Tolerance
0 20 40 60 80 100 120 140 160 180-24Frequency (MHz)
CASS: 2013-2014 DLP Program UT Austin
- 49 -J. Silva-Martinez
CASS: 2013-2014 DLP Program UT Austin
Blocker and Jitter Tolerant Wideband ModulatorsH. M. Geddada, et.al.H. M. Geddada, et.al.Re-evaluation TVLSI
TransientTransient responseresponse ofof thetheADCADC toto thethe 88..77dBFSdBFS blockerblockeratat 100100MHzMHz inin thethe presencepresence
OutputOutput spectrumspectrum ofofthethe ADCADC withwith andandwithoutwithout thethe 99dBFSdBFS
0 5
1
VIN
40
‐20
0
Vin = -6dBFS @ 5MHzNo blocker
ppofof aa --66dBFSdBFS inin--bandband signalsignalatat 55MHzMHz
withoutwithout thethe --99dBFSdBFSblockerblocker atat 4040MhzMhz
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2x 10-6
0
0.5V
1
T ‐120
‐100
‐80
‐60
‐40
Pow
er (
dBFS
)SNR = 75.1dB
0 8
0.9
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2x 10-6
0
0.5VDET
=0.5u sec
10‐1
100
101
102
120
Frequency (MHz)
-60
-40
-20
0
(dB
FS)
Vin = -6dBFS @ 5MHzBlocker = -9dBFS @ 40MHzAttenuation factor = 12dBStable Operation
- 50 -J. Silva-Martinez0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2x 10-6
0.4
0.5
0.6
0.7
0.8
Time (sec)
VOUT
10-1 100 101 102
-120
-100
-80
-60
Frequency (MHz)
Pow
er (
SNR = 46.5dB
CASS: 2013-2014 DLP Program UT Austin
Proposed minimally invasive LPF
0
-15
-10
-5
)
STF (Vx/Vin)NTF (Vx/Vn,r1)NTF( Vx/Vn,op)
10dB
Signal & Noise transfer functions
-30
-25
-20G
ain(
dB) , p
10dB attenuation at 60MHz In band noise 15 µV rms which is
- 51 -J. Silva-Martinez 100 101 102 103-40
-35
Frequency (MHz)
Signal bandwidth In-band noise 15 µV-rms which is
well below the ADC noise level
CASS: 2013-2014 DLP Program UT Austin
Measurement results
40
-20
0
With input signalWithout Input signal
RBW = 15.2KhzVin: -2.86dBFS @2.75 MHzHD2: -78.3dBFS @5.5MHzHD3: -73dBFS @8.25MHz
-80
-60
-40
Pow
er (d
BFS
)
HD3HD25th Order CT LP ADC,
10-1
100
101
102-120
-100
60
70
SNRSNDR
Active area < 0.43mm2
In 20MHz BW, Input of -2.86dBFS at 2.75 MHz
P k SNR 66dB P k SNDR
Frequency (MHz)
30
40
50
SN
DR
(dB
)
SNDR
Peak SNR= 66dB, Peak SNDR= 64 dB
HD3= -73 dBFS, HD2= -78.3 dBFS
10
20
30
SN
R &
S
- 52 -J. Silva-Martinez
dBFS DR=69dB-70 -60 -50 -40 -30 -20 -10 0
0
Input Power (dBFS)
Dynamic Range = 69dB
CASS: 2013-2014 DLP Program UT Austin
Summary: Blocker Tolerance LPF provides blocker rejection especially at the critical
frequencies. Overload detector and PGA prevents the ADC saturation
during transients. Blocker reduction techniques Blocker reduction techniques
Power overhead < 6% Area overhead < 20 %
Advantages L Less power Loop parameters are not disturbed so fast settling times Protects the ADC from instantaneous/temporary blockers
- 53 -J. Silva-Martinez
Protects the ADC from instantaneous/temporary blockers Moderate SNR is better than no communication or no SNR
CASS: 2013-2014 DLP Program UT Austin
Performance Summary of the ADC
[9] [10] [11] This work
Fs [Mhz] 250 64 160 500Fs [Mhz] 250 64 160 500BW [MHz] 10 1 5 20
SNDR/SNR [dB] 65/68 NA/75.5 69.5 64/66Dynamic Range [dB] 71 65# 76 69Dynamic Range [dB] 71 65# 76 69
Blocker reduction [dB]adjacent/alternate channels
8/15 9.5/20+ 10 15/18
Settling time (µsec)* 51 - - < 0.3
Power consumption [mW] 18 4.1 6 17.1
Area(mm2) 1 35 0 14 0 56 0 43Area(mm ) 1.35 0.14 0.56 0.43Technology (nm) 130 180 130 90
- 54 -J. Silva-Martinez
# fixed input resistor * switching input resistors* * Blocker adaptation time+ Extracted from a plot comparing measured and simulated performance, no in-band signal
CASS: 2013-2014 DLP Program UT Austin
Fundamentals on MASHFundamentals on MASH architecturesarchitecturesE
Analog Intensive
1qE
DAC
ZY1 ZX ZH1
ZH2 Analog Intensive MASH topology
DAC ZH2
2qE
ZY
Schreier-Temesq
ZY2
ADC
Leslie and SinghISCAS-1990
22413
2413 0
qENTF*HXSTF*HzYThen
STF*HNTF*HIf
2
1
2
2
2
1
22413
q
q
X*STFSTFESQNR
- 55 -J. Silva-Martinez
1122
qq E*NTFNTFESQNR
CASS: 2013-2014 DLP Program UT Austin
Fundamentals on MASHFundamentals on MASH architecturesarchitectures
Analog Intensive Pseudo-MASH
High-Resolution Pipeline ADC
N-bitsPseudo MASH topology
Software Pl tf
RIN
kLP2
kBP24-Bits
_Active-RC
4th order
PlatformVin
kBP1
Filter
NR
ZA
C2
Clock Generator
DA
C
kLP1CLK
DELAYDigital signals (D1-D7)
N DD
Todd Brooks, et. al., JSSC D 1997
- 56 -J. Silva-Martinez
JSSC, Dec 1997
CASS: 2013-2014 DLP Program UT Austin
Target: SNDR Improvement by 3 bits0
O t t With C l ti
-40
-20X: 5.005e+006Y: -6.02
Output With CancelationOutput Without Cancelation
80
-60
-40
dBFS
) SNR = 79.516dB SNR = 105.803dB
-100
-80
Pow
er (
d
-140
-120
SQNR improvement should be around 18dB (3 bits)
105 106 107 108-160
Frequency (Hz)
- 57 -J. Silva-Martinez
SQNR improvement should be around 18dB (3 bits)
Power overhead, area overhead < 20%
CASS: 2013-2014 DLP Program UT Austin
ConclusionsConclusions Minimization of filtering functions in the receiver
h i d d i ti ADC l tichain demands innovative ADC solutions;
SNDR > 80 dB will be frequently needed in wireless applications;applications;
Jitter and Blocker tolerant architectures are needed;;
DAC calibration techniques for fast and high-resolution applications;
Global tuning strategies (for stability)
Advanced LTE applications require high-resolutionf i l BW hi h 200MH
- 58 -J. Silva-Martinez
for signal BW as high as 200MHz
CASS: 2013-2014 DLP Program UT Austin
Recommended Readings
S. Norsworthy, R. Schreier and G. Temes, Delta-Sigma Data Converters: Theory, Design and Simulation, IEEE Press
R. Schreier and G. Temes, Understanding Delta-Sigma Data Converters, IEEE Press
J. Cherry, M. Snelgrove,Theory, Practice, and Fundamental Performance Limits of High-Speed Data Conversion Using Continuous-Time Delta-Sigma Modulators, Kluwer Academic Press.
J. M. de la Rosa, Sigma-Delta Modulators: Tutorial Overview, Design Guide, and State-of-the-Art Survey, TCAS-I, Jan 2011.
Other tutorials: IECE-2012
- 59 -J. Silva-Martinez