Fischer 08 1 Analog to Digital Converters Nyquist-Rate ADCs Flash ADCs Sub-Ranging ADCs Folding ADCs Pipelined ADCs Successive Approximation (Algorithmic) ADCs Integrating (serial) ADCs Oversampling ADCs Delta-Sigma based ADCs
Fischer 08 2 Conversion Principles
Fischer 08 3 ADC Architectures Flash ADCs: High speed, but large area and high power dissipation. Suitable for low-medium resolution (6-10 bit). Sub-Ranging ADCs: Require exponentially fewer comparators than Flash ADCs. Hence, they consume less silicon area and less power. Pipelined ADCs: Medium-high resolution with good speed. The trade- offs are latency and power. Successive Approximation ADCs: Moderate speed with medium- high resolution (8-14 bit). Compact implementation. Integrating ADCs or Ramp ADCs: Low speed but high resolution. Simple circuitry. Delta-Sigma based ADCs: Moderate bandwidth due to oversampling, but very high resolution thanks to oversampling and noise shaping.
Fischer 08 4 Performance Limitations 1 Thermal Noise LimitationClock Jitter (Aperture) Limitation Normalized Noise Powers: Limiting Condition: n-Bit ADC Sinusoidal Input Swing: ±1[V] f max = ½ f conv System Definitions Maximum Resolution: f in =½f conv
Fischer 08 5 Performance Limitations 2 Selection of ADC Architecture is driven by Application Displays Audio Sonar Ultra Sound Video Wireless Communications Seismology
Fischer 08 6 Parallel or Flash ADCs Conceptual Circuit
Fischer 08 7 Sub-Ranging ADCs Half-Flash or Two-Step ADC
Fischer 08 8 Folding ADCs Principle Configuration … 2 n1 Sub-Ranges
Fischer 08 9 Folding Processor Example: 2-Bit Folding Circuit ( 2 n-1 +1)Io for n-Bit 2Io
Fischer Successive Approx. ADCs Concept Implementation
Fischer DAC Realization 1 (Voltage Mode)
Fischer DAC Realization 2 Spread Reduction through R-2R Ladder
Fischer DAC Realization 3 Charge-Redistribution Circuit Pros Insensitive w.r.t. Op-amp Gain Offset (1/f Noise) compensated Cons Requires non-overlapping Clock High Element Spread Area Output requires S&H valid only during 2
Fischer DAC Realization 4 Spread Reduction through capacitive Voltage Division valid only during 2 Spread=2 n/2 Example: 8-Bit ADC
Fischer DAC Realization 5 Charge-Redistribution Circuit with Unity-Gain Amplifier Pros Voltage divider reduces spread Buffer low output impedance No clock required Cons Parasitic cap causes gain error High Op-amp common mode input required No amplifier offset compensation Amplifier Input Cap. Cp Gain Error: є G =-Cp/16C 16/15C Spread=½2 n/2 Example: 8-Bit ADC
Fischer DAC8 with Unity-Gain Amplifier Sub-range Output (4 LSBs) 00.5u2u1.5u2.5u3.5u4.5u5.5u6.5u3u4u5u6u1u Amplifier Output 6.5u00.5u2u1.5u2.5u3.5u4.5u5.5u3u4u5u6u1u
Fischer DAC Realization 6 Current Mode Implementation
Fischer Current Cell & Floor Plan Unit Current Cell Symmetrical Current Cell Placement Current summing RailIout Cascode Current Source Switching Devices Array of 256 Cells R
Fischer DAC Implementation Layout of 10-Bit Current-Mode DAC ( 0.5 m CMOS) Current summing Rails
Fischer Modified SA Algorithm 1 Idea: Replace DAC by an Accumulator Consecutively divide Ref by 2
Fischer Modified SA Algorithm 2 First cycle only Accumulator Idea: Maintain Comparator Reference (½ FS=Gnd) Double previous Accumulator Output
Fischer SC Implementation SC Implementation of modified SA ADC
Fischer Timing Diagram
Fischer Offset Compensated Circuit Offset Compensated SC Implementation
Fischer Building Blocks 1 DC Gain77 dB Gain- bandwidth 104 CL= 1.5 pF Power 1.3 mW Output Swing 4 V p-p Transconductance Amplifier
Fischer Building Blocks 2 Power 0.5 mW Resolution > 0.5 mV Settling Time 3 ns Latched CMOS Comparator
Fischer Layout of 8-Bit ADC 165 m (0.5 m CMOS)
Fischer Spice Simulation (Bsim3) 8-Bit ADC: fclk=10MHz fconv=1.25MHz
Fischer Pipelined ADCs Pipelined modified SA or Algorithmic ADC Pros Offset (1/f Noise) compensated Minimum C-spread One conversion every clock period Cons Matching errors digital correction for n>8 Clock feed-through very critical High amplifier slew rate required
Fischer Integrating or Serial ADCs Dual Slope ADC Concept Constant Ramp Prop. to Input Ramp Using 2 N /k samples requires Ref = FS/k reduced Integrator Constant (Element Spread) N represents digital equivalent of analog Input
Fischer SC Dual-Slope ADC 10-Bit Dual-Slope ADC
Fischer ADC Testing Types of Tests Static Testing Dynamic Testing In static testing, the input varies slowly to reveal the actual code transitions. Yields INL, DNL, Gain and Offset Error. Dynamic testing shows the response of the circuit to rapidly changing signals. This reveals settling errors and other dynamic effects such as inter-modulation products, clock-feed-trough, etc. Circuit Under Test Output Input Clock
Fischer Performance Metrics 1 Error Types Offset Gain DNL INL Missing Codes IDEAL ADC Static Errors
Fischer Performance Metrics 2 Frequency Domain Characterization Ideal n-Bit ADC: SNR = 6.02 x n [dB] fsig Amplitude
Fischer ADC Error Sources Static Errors Element or Ratio Mismatches Finite Op-amp Gain Op-amp & Comparator Offsets Deviations of Reference Dynamic Errors Finite (Amplifier) Bandwidth Op-amp & Comparator Slew Rate Clock Feed-through Noise (Resistors, Op-amps, switched Capacitors) Intermodulation Products (Signal and Clock)
Fischer Static Testing Servo-loop Technique Comparator, integrator, and ADC under test are in negative feedback loop to determine the analog signal level required for every digital code transition. Integrator output represents equivalent analog value of digital output. Transition values are used to generate input/output characteristic of ADC, which reveals static errors like Offset, Gain, DNL and INL.
Fischer Dynamic Testing Types of Dynamic Tests Histogram or Code-Density Test FFT Test Sine Fitting Test Test Set-up
Fischer Histogram or Code-Density Test DNL appears as deviation of bin height from ideal value. Integral nonlinearity (INL) is cumulative sum (integral) of DNL. Offset is manifested by a horizontal shift of curve. Gain error shows as horizontal compression or decompression of curve.
Fischer Histogram Test Pros and Cons of Histogram Test Histogram test provides information on each code transition. DNL errors may be concealed due to random noise in circuit. Input frequency must be selected carefully to avoid missing codes (f clk /f in must be non-integer ratio). Input Swing is critical (cover full range) Requires a large number of conversions (o 2 n x 1,000).
Fischer Simulated Histogram Test 8-Bit SA ADC with 0.5% Ratio Error and 5mV/V Comparator Offset
Fischer FFT Test Pros and Cons of FFT Test Offers quantitative Information on output Noise, Signal- to-Noise Ratio (SNR), Spurious Free Dynamic Range (SFDR) and Harmonic Distortion (SNDR). FFT test requires fewer conversions than histogram test. Complete characterization requires multiple tests with various input frequencies. Does not reveal actual code conversions
Fischer Simulated FFT Test 8-Bit SA ADC with 0.5% Ratio Error and 5mV/V Comparator Offset SFDR=60 dB SNDR=49 dB ENOB=7.85