Another way to implement a folding ADC
Transcript of Another way to implement a folding ADC
![Page 1: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/1.jpg)
Another way to implement a folding ADC
J. Van Valburg and R. van de Plassche, An 8-b 650 MHz Folding ADC, IEEE JSSC, vol 27, #12, pp. 1662-6, Dec 1992
![Page 2: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/2.jpg)
Coupled Differential Pair
J. Van Valburg and R. van de Plassche, An 8-b 650 MHz Folding ADC, IEEE JSSC, vol 27, #12, pp. 1662-6, Dec 1992
![Page 3: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/3.jpg)
1 bit ADC with 2 CDPs
J. Van Valburg and R. van de Plassche, An 8-b 650 MHz Folding ADC, IEEE JSSC, vol 27, #12, pp. 1662-6, Dec 1992
![Page 4: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/4.jpg)
2-b ADC with 4 CDPs
Folding rate = 44x fIN
J. Van Valburg and R. van de Plassche, An 8-b 650 MHz Folding ADC, IEEE JSSC, vol 27, #12, pp. 1662-6, Dec 1992
![Page 5: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/5.jpg)
3-b ADC with 8 CDPs
J. Van Valburg and R. van de Plassche, An 8-b 650 MHz Folding ADC, IEEE JSSC, vol 27, #12, pp. 1662-6, Dec 1992
Folding rate = 88x fIN
Practical limit due to bandwidth needed
![Page 6: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/6.jpg)
I-Q folding; offset by ½ LSBResolution can be increased by 1 bit using parallelismwithout increasing folding rate. 4 folding blocks neededto get to 5 bits by folding alone.
You could build all 8 bits this way, but the complexity would be as greatAs a flash converter. So, use interpolation to get the last 3 bits.
J. Van Valburg and R. van de Plassche, An 8-b 650 MHz Folding ADC, IEEE JSSC, vol 27, #12, pp. 1662-6, Dec 1992
![Page 7: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/7.jpg)
Fine bit generation through interpolation
Interpolation rate = 8. provides last 3 bits
J. Van Valburg and R. van de Plassche, An 8-b 650 MHz Folding ADC, IEEE JSSC, vol 27, #12, pp. 1662-6, Dec 1992
![Page 8: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/8.jpg)
Resistive interpolator
Interpolation rate = 8. provides last 3 bits
J. Van Valburg and R. van de Plassche, An 8-b 650 MHz Folding ADC, IEEE JSSC, vol 27, #12, pp. 1662-6, Dec 1992
![Page 9: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/9.jpg)
Now, the MSFF comparators are all referenced to zero diff input. They are sensing zero crossings
Total number of comparators is reduced (from flash) by folding rate: 256/8=32
J. Van Valburg and R. van de Plassche, An 8-b 650 MHz Folding ADC, IEEE JSSC, vol 27, #12, pp. 1662-6, Dec 1992
![Page 10: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/10.jpg)
Pipeline architectureTypically used up to about 100 MS/s
(Imaging, digital receiver, base station, HDTV, xDSL, cable modems, ethernet)
12 bits
Maxim AN383
![Page 11: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/11.jpg)
Pipeline example
• Increased sampling rate in exchange for latency• First 4 stages: S/H; 3 bit flash; DAC; residue
multiplied by 4x; 4 bit accuracy.• 2 bits output per stage. Digital error correction
uses 3rd bit to allow extra range in next stage. This can correct for offset errors.
• Final stage must have >4 bit accuracy.• Shift registers needed for digital time alignment
![Page 12: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/12.jpg)
14 bits; 105 MS/s
Analog Devices AD6645 6/2003
![Page 13: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/13.jpg)
20 Gs/s 8-b Pipelined – Interleaved ADCHigh bandwidth oscilloscope application - Agilent
K. Poulton, et al.”A 20Gs/s 8b ADC with a 1MB Memory in 0.18 um CMOS,” ISSCC 2003, paper 18.1.
![Page 14: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/14.jpg)
Design details
• BiCMOS buffer chip used to drive the 4pF input capacitance of the ADC
• ADC is organized in 80 slices – parallelism used to increase throughput– Each block works at 250 MS/s– T/H, V-I converter, current-mode pipelined 1-bit ADC– Residue is amplified by 1.6X– Radix 1.6 to binary conversion– Data capture in 1MB on-chip SRAM
![Page 15: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/15.jpg)
L-C-L t-lineAdjustable gm to control damping
K. Poulton, et al.”A 20Gs/s 8b ADC with a 1MB Memory in 0.18 um CMOS,” ISSCC 2003, paper 18.1.
![Page 16: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/16.jpg)
Clock Generation
4 GS/s example
K. Poulton, et al.”A 4Gs/s 8b ADC in 0.35 um CMOS,” ISSCC 2002, paper 10.1.
![Page 17: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/17.jpg)
Clock generation
• The 20 GS/s ADC requires 80 250 MHz clocks, each offset by 50 ps with error < 1ps
• 1 GHz clock; 5 DLL stages => 5 diff clocks• Interpolate to get 20 clocks• Divide by 4 to get 80 clocks• Each has digital time adjustment
![Page 18: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/18.jpg)
Track-Hold slice
Differential transconductor
K. Poulton, et al.”A 4Gs/s 8b ADC in 0.35 um CMOS,” ISSCC 2002, paper 10.1.
![Page 19: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/19.jpg)
Cgs provides hold functionWidth scaling provides gain
ThresholdDetection
Bit outresidue
12 stages of radix 1.6 produces 12 bits; convert to 8 bits binary
K. Poulton, et al.”A 20Gs/s 8b ADC with a 1MB Memory in 0.18 um CMOS,” ISSCC 2003, paper 18.1.
![Page 20: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/20.jpg)
Extensive calibration
• Sawtooth input– Per slice gain and offset correction – 160
DACs on-chip– Gain coefficients loaded into radix converter
• Pulse input– Fourier analysis used to set timing
adjustments
![Page 21: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/21.jpg)
K. Poulton, et al.”A 20Gs/s 8b ADC with a 1MB Memory in 0.18 um CMOS,” ISSCC 2003, paper 18.1.
![Page 22: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/22.jpg)
K. Poulton, et al.”A 20Gs/s 8b ADC with a 1MB Memory in 0.18 um CMOS,” ISSCC 2003, paper 18.1.
![Page 23: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/23.jpg)
5x any other CMOS ADC
K. Poulton, et al.”A 20Gs/s 8b ADC with a 1MB Memory in 0.18 um CMOS,” ISSCC 2003, paper 18.1.
![Page 24: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/24.jpg)
Analog Devices 4/2005
![Page 25: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/25.jpg)
![Page 26: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/26.jpg)
Analog Devices 11/2004
![Page 27: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/27.jpg)
![Page 28: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/28.jpg)
![Page 29: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/29.jpg)
Maxim Integrated Products MAX108
![Page 30: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/30.jpg)
Maxim Integrated Products MAX108
![Page 31: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/31.jpg)
Successive Approximation ADC (SAR ADC)Up to 5 MS/s
8 to 18 bitsLow power
Binary search algorithmStart at midscale: MSB = 1Works down to LSB onebit per cycle
Is VIN < VDAC orVIN > VDAC ?
Maxim AN387
![Page 32: Another way to implement a folding ADC](https://reader031.fdocuments.us/reader031/viewer/2022013009/61ce3c743f39b30bb30ec353/html5/thumbnails/32.jpg)
Maxim AN387