US007129874B2
`
`a2) United States Patent
`US 7,129,874 B2
`(0) Patent No.:
`Bjornsen
`Oct. 31, 2006
`(45) Date of Patent:
`
`(54) METHOD AND APPARATUS FOR
`OPERATING A PIPELINED ADC CIRCUIT
`
`5,982,313 A *
`6,556,158 B1*
`
`11/1999 Brooks et al... 341/143
`4/2003 Steensgaard-Madsen .... 341/131
`
`(75)
`
`Inventor:
`
`Johnny Bjornsen, Trondheim (NO)
`
`(73) Assignee: Nordic Semiconductor ASA,Tiller
`(NO)
`
`(*) Notice:
`
`Subject to any disclaimer, the term of this
`patent is extended or adjusted under 35
`US.C. 154(b) by 0 days.
`
`(21) Appl. No.: 11/149,899
`
`(22)
`
`Filed:
`
`Jun. 10, 2005
`
`(65)
`
`Prior Publication Data
`
`US 2005/0275577 Al
`
`Dec. 15, 2005
`
`OTHER PUBLICATIONS
`
`Siragusa et al., “A Digitally Enhanced 1.8-V 15-bit 40-MSample/s
`CMOSPipelined ADC,” IEEE Journal of Solid-State Circuits, vol.
`39, No. 12, Dec. 2004.
`Siragusa et al., “Gain Error Correction Technique for Pipelined
`Analogue-to-Digital Converters,” Electronics Letters, vol. 36, No.
`7, Mar. 30, 2000.
`Qin etal., “Sigma-Delta ADC with Reduced Sample Rate Multibit
`Quantizer,” IEEE Transactions on Circuits and Systems-II Analog
`and. Digital Signal Processing, vol. 46, No. 6, Jun. 1999, 824-828.
`
`* cited by examiner
`
`Primary Examiner—Jean Bruner Jeanglaude
`(74) Attorney, Agent, or Firm—Marshall, Gerstein & Borun
`LLP
`
`Related U.S. Application Data
`
`(57)
`
`ABSTRACT
`
`(60) Provisional application No. 60/579,016, filed on Jun.
`10, 2004.
`
`(51)
`
`Int. CL
`(2006.01)
`H03M 3/00
`(52) U.S. Ch eee 341/143; 341/120; 341/119;
`341/118
`
`(58) Field of Classification Search ................ 341/155,
`341/143, 131
`See application file for complete search history.
`
`(56)
`
`References Cited
`U.S. PATENT DOCUMENTS
`
`5,153,593 A * 10/1992 Walden et al. ......0.. 341/143
`3/1996 Ribner et al. oe. 341/143
`5,500,645 A *
`
`5,889,482 A *
`3/1999 Zarubinsky et al.
`........ 341/131
`5,949,361 A *
`9/1999 Fischer et al. wc. 341/143
`
`An analog-to-digital converter (ADC) circuit that converts
`an analog input signal into a digital output circuit includes
`a noise shapingfirst stage cascaded with a pipelined second
`stage. The first stage includes a sample-and-hold circuit and
`a first order modulator, where the first order modulator
`includes a noise shaping filter, a FLASH ADC and a
`feedback DAC. A digital dither generator is used to provide
`a dither signal to the ADC circuit. The second stage includes
`a switching circuit and an ADC. A calibration filter con-
`nected to the second stage calibrates the ADC circuit. A first
`reconstruction filter and a second reconstruction filter are
`
`used to recombine outputs of the first stage and the second
`stage of the ADC circuit. The ADC circuit allows high
`resolution analog-to-digital conversion at a low over-sam-
`pling rate and low powerdissipation levels.
`
`18 Claims, 13 Drawing Sheets
`
`
`
`Xilinx Exhibit 1007
`
`Page 1
`
`

`

`U.S. Patent
`
`Oct. 31, 2006
`
`Sheet 1 of 13
`
`US 7,129,874 B2
`
`3(NPDP+pop
`
`(NYPFy()"'p7-Cy'p(y)aasjon/aupauia(pt))/$e9¢92Z||8é0c
`
`
`(+)+ay)(+)‘<esayySujdeysESJeyi4a()3
`(z)'H
`13NIEpzy2|(yl|H/S
`
`s10j19ua8
`
`sy4(wopBUQqHeD
`6viveOl
`
`b“Sid
`
`Page 2
`
`
`
`
`

`

`U.S. Patent
`
`Oct. 31, 2006
`
`Sheet 2 of 13
`
`US 7,129,874 B2
`
`FIG.2
`
`Page 3
`
`

`

`U.S. Patent
`
`Oct. 31, 2006
`
`Sheet 3 of 13
`
`US 7,129,874 B2
`
`OrTTOT00TTLOord100000v9
`J24A+asaJtg/t-JAgie-‘AsisAsiePAA3Agsasic™Ait
`
`JaA+5gisasiciAgiJAgniJAgie-AsisJA
`
`
`}§-—+——_}+—_+++.
`TITOOTTOTOTO00rd1100or0010000000agai
`
`€“Sls09
`
`89!
`
`adjul7sa-joxuta-
`
`atjar.atA-
`
`OZ
`
`Page 4
`
`
`
`

`

`U.S. Patent
`
`Oct. 31, 2006
`
`Sheet 4 of 13
`
`US 7,129,874 B2
`
`(pa
`
`
`
`pndjnos10;ejnpoyJe4At
`
`T/PAAT
`
`
`
`
`
` S4ABISSMAPIESMAUII-SIAPIE-
`
`IMAZTSMARTSMADIT-
`
`
`
`
`
`
`
`yndjno20;e(npoWPAAt
`
`0g
`
`TAP4At
`
`cB
`
`
`
`
`
`
`
`JMAP/I-PMAUI-JUABE-JaaA-
`
`SAADdIS-
`
`Page 5
`
`
`
`
`
`
`

`

`U.S. Patent
`
`Oct. 31, 2006
`
`Sheet 5 of 13
`
`US 7,129,874 B2
`
`Vot
`
`}>
`
`100
`
`
`XL
`
` 92
`
`wo
`
`Oo
`LL.
`
`oOao
`
`
`
`wi
`a
`as)
`"2
`o
`<a
`
`Page 6
`
`.o
`
`Vint
`
`a
`©
`
`+
`oO
`
`Vin-@
`
`

`

`U.S. Patent
`
`Oct. 31, 2006
`
`Sheet 6 of 13
`
`US 7,129,874 B2
`
`9‘Sid
`
`06
`
`No
`
`va
`
`m o
`
`youqpeay
`
`‘'"4'ova'6Aad4i'1|bes!+e
`’P/D;'i:j264’‘t§1’‘4o
`i.+4A1t‘'‘'(‘'§‘‘8‘‘t't'‘1
`—e-LA
`
`’
`
`Page 7
`
`
`
`

`

`U.S. Patent
`
`Oct. 31, 2006
`
`Sheet 7 of 13
`
`US 7,129,874 B2
`
`
`
`LAdLNOSOTIVNV
`
`YOLVINGOWdO
`
`
`
`OLANGONIMSTWNOILIGAV
`
`
`
`L5S44d0YOLVEVdNOO
`
` NOUNSNOLLWZLLWNO“LOdvOL3ndONIMS| bodvOLANGONIMS|
`
`—ZNYA-oz
`
`
`
`
`
`OLANGONIMSTWNOILIGGV
`
`
`
`15S440YOLVYEVdNOO
`
`Page 8
`
`

`

`U.S. Patent
`
`Oct. 31, 2006
`
`Sheet 8 of 13
`
`US 7,129,874 B2
`
`rerep(NPDSi
`
`
`
`4931]UOIEWU]IIG
`
`8“Old Ort
`
`Page 9
`
`

`

`U.S. Patent
`
`Oct. 31, 2006
`
`Sheet 9 of 13
`
`Ort
`
`US 7,129,874 B2
`
`6‘Old
`
`Page 10
`
`

`

`U.S. Patent
`
`Oct. 31, 2006
`
`Sheet 10 of 13
`
`US 7,129,874 B2
`
`JayyoHAUpeq
`
`Ob“Sid
`
`Ort
`
`Page 11
`
`
`

`

`U.S. Patent
`
`Oct. 31, 2006
`
`Sheet 11 of 13
`
`US 7,129,874 B2
`
`
`
`13}WornwUNDag
`
`Lb“Old
`
`Ori
`
`Page 12
`
`
`

`

`U.S. Patent
`
`Oct. 31, 2006
`
`Sheet 12 of 13
`
`O8t
`
`
`
`Jaq)UopEUNIIaq
`
`US 7,129,874 B2
`
`Zb“Sls
`
`Page 13
`
`

`

`+
`
`
`
`43q[JUOPRUPIIG
`
`Oct. 31, 2006
`
`Sheet 13 of 13
`
`U.S. Patent (NDP
`
`US 7,129,874 B2
`
`€l‘Sls
`
`08}
`
`Page 14
`
`
`

`

`US 7,129,874 B2
`
`1
`METHOD AND APPARATUS FOR
`OPERATING A PIPELINED ADC CIRCUIT
`
`CROSS-REFERENCES TO RELATED
`APPLICATIONS
`
`
`
`This application claimspriority to U.S. Provisional Appli-
`cation Ser. No. 60/579,016, entitled, “Method and Apparatus
`for Operating a Pipelined ADCCircuit,” filed Jun. 10, 2004,
`the disclosure of which is hereby expressly incorporated
`herein by reference.
`
`TECHNICAL FIELD
`
`This patent relates generally to analog-to-digital convert-
`ers, and more specifically to an apparatus and a method for
`operating a pipelined analog-to-digital converter.
`
`BACKGROUND
`
`Analog-to-digital converters (ADCs) are employed in a
`variety of electronic systems including computer modems,
`wireless telephones, satellite receivers, process control sys-
`tems, etc. Such systems demandcost-effective ADCs that
`can efficiently convert an analog input signal to a digital
`output signal over a wide range of frequencies and signal
`magnitudes with minimal noise and distortion.
`An ADCtypically converts an analog signal to a digital
`signal by sampling the analog signal at pre-determined
`sampling intervals and generating a sequence of binary
`numbers via a quantizer, wherein the sequence of binary
`numbers is a digital representation of the sampled analog
`signal. Some of the commonly used types of ADCs include
`integrating ADCs, Flash ADCs, pipelined ADCs, successive
`approximation register ADCs, Delta-Sigma (AZ) ADCs,
`two-step ADCs, etc. Of these various types, the pipelined
`ADCs and the AX ADCsare particularly popular in appli-
`cations requiring higher resolutions.
`A pipelined ADC circuit samples an analog input signal
`using a sample-and-hold circuit to hold the input signal
`steady and a first stage flash ADC to quantize the input
`signal. The first stage flash ADC then feeds the quantized
`signalto a digital-to-analog converter (DAC). The pipelined
`ADCcircuit subtracts the output of the DAC from the analog
`input signal to get a residue signalofthe first stage. The first
`stage of the pipelined ADC circuit generates the most
`significant bit
`(MSB) of the digital output signal. The
`residue signalofthe first stage is gained up by a factor and
`fed to the next stage. Subsequently, the next stage of the
`pipelined ADC circuit further quantizes the residue signal to
`generate further bits of the digital output signal.
`On the other hand, a AX ADC employs over-sampling,
`noise-shaping, digital filtering and digital decimation tech-
`niques to provide high resolution analog-to-digital conver-
`sion. One popular design of a AX ADC is multi-stage noise
`shaping (MASH) Az ADC. A MASH Ad ADCis based on
`cascading multiple first-order or second-order AX ADCs to
`realize high-order noise shaping. An implementation of a
`MASH Az ADCis well known to those of ordinary skill in
`the art. While both pipelined ADCs and AX ADCsprovide
`improved signal-to-noise ratio, improvedstability, etc., AX
`ADCs generally provide higher levels of resolution and
`therefore are preferred in applications involving asynchro-
`nousdigital subscriber lines (ADSL), very high speeddigital
`subscriber lines (VDSL), etc. Highly-linear, high-resolution
`and wide-bandwidth ADCs are required for VDSL systems.
`
`2
`However, AX ADCstypically employ higher over-sam-
`pling ratios (OSRs) to achieve such higher resolutions,
`normally of the range of OSRs over 50. Such high OSR
`results in higher level of power dissipation. Moreover, AZ
`ADCs also require anti-alias filters for inputting analog
`signals into a first stage of the AX ADC. Suchanti-aliasing
`filters also result
`in higher power dissipation. Therefore
`there is a need for an ADC circuit that can provide high
`resolution digital output at a lower OSRsand/or lowerlevels
`of power dissipation.
`
`BRIEF DESCRIPTION OF THE DRAWINGS
`
`The present patent is illustrated by way of examples and
`not limitations in the accompanying figures, in which like
`references indicate similar elements, and in which:
`FIG. 1 is a block diagram of an ADC circuit operating at
`lower sampling rates;
`FIG.2 is a block diagram ofa first order modulator used
`in the ADC circuit of FIG. 1;
`FIG. 3 is an exemplary signal diagram of reference
`voltages of various flash ADCs used in the first order
`modulator of FIG. 2;
`FIG. 4 describes exemplary transfer functions ofthefirst
`order modulator of FIG.2;
`FIG. 5 is an exemplary circuit diagram of a multiplying
`and integrating DAC used in the first order modulator of
`FIG.2;
`FIG. 6 is another exemplary circuit diagram of a multi-
`plying and integrating DAC usedinthefirst order modulator
`of FIG. 2;
`FIG.7 is an exemplary signal diagram of an output signal
`of the first order modulator of FIG. 2;
`FIG.8 is a block diagram of a 10 bit MASH ADC with
`a first order sinc decimation filter;
`FIG. 9 is a block diagram of a 10 bit MASH ADC with
`a first order sinc decimationfilter in front of a recombination
`adder;
`FIG. 10 is a block diagram of a 10 bit MASH ADC with
`a combined first order sinc decimation filter and recombi-
`nationsfilter;
`FIG. 11 is a block diagram of a 10 bit MASH ADC with
`a reduced sampling rate in a second stage;
`FIG. 12 is a block diagram of a 10 bit MASH ADC with
`a general decimation filter; and
`FIG. 13 is a block diagram of a 10 bit MASH ADC with
`a general decimation filler combined with a recombination
`filter.
`
`
`
`DETAILED DESCRIPTION OF THE EXAMPLES
`
`An analog-to-digital converter (ADC) system that con-
`verts an analog input signal
`into a digital output signal
`includesa noise shapingfirst stage cascaded with a pipelined
`second stage. The first stage includes a sample-and-hold
`circuit and a first order modulator, where the first order
`modulator includes a noise shaping filter, a FLASH ADC
`and a feedback DAC.A digital dither generator is used to
`provide a dither signal to the ADCcircuit. The second stage
`includes a switching circuit and an ADC. A calibrationfilter
`connected to the second stage calibrates the ADC circuit. A
`first reconstruction filter and a second reconstruction filter
`
`are used to recombine outputs of the first stage and the
`second stage of the ADC circuit. The ADC circuit allows
`high resolution analog-to-digital conversion at a low over-
`sampling rate and low power dissipationlevels.
`
`10
`
`15
`
`20
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`Page 15
`
`

`

`US 7,129,874 B2
`
`10
`
`15
`
`20
`
`25
`
`30
`
`35
`
`3
`4
`An alternate embodiment of such an analog-to-digital
`An embodimentof the analog-to-digital converter system
`converter system further comprises an offset correction
`for converting an analog input signal to a digital output
`circuit adapted to correct the offset introduced by the quan-
`signal comprises a first stage analog-to-digital converter
`tizer, wherein the second stage analog-to-digital converter
`circuit adapted to convert the analog input signal intoafirst
`circuit includes a calibration filter adapted to correct gain
`stage digital output; a combiner coupled to thefirst stage
`errors and pole errors in the first stage digital output.
`analog-to-digital converter circuit and adapted to combine a
`An alternate embodiment of the analog-to-digital con-
`digital test signal with the first stage digital output; a residue
`verter system includes a method of converting an analog
`signal generator adapted to receive the output of the com-
`input signalto a digital output signal, the method comprising
`biner and to generate a residue signal using the output of the
`generating a first stage digital output based on the analog
`combiner andthe analog input signal; a noise shaping circuit
`input signal; combining the first stage digital output with a
`adapted to noise shape the residue signal; a sampling circuit
`digital
`test signal; generating a residue signal from the
`coupled to the output of the noise shaping circuit and
`combination of the first stage digital output and the digital
`adapted to sample the noise shaped residue signal; a second
`test signal; noise-shaping the residue signal; sampling the
`stage analog-to-digital converter circuit coupled to the out-
`noise shapedresidue signal at a second operating frequency;
`put of the sampling circuit and adapted to generate a second
`generating a second stage digital output signal based on the
`stage digital output based on the sampledresidue signal; and
`sampled residue; and combiningthe first stage digital output
`a digital correction circuit coupled to the output ofthefirst
`and the second stage digital output in a manner so as to
`stage analog-to-digital convertercircuit and to the output of
`cancel the effect of the digital test signal.
`the second stage analog-to-digital converter circuit and
`adapted to combine the first stage digital output and the
`In an alternate embodiment of such an analog-to-digital
`second stage digital output using a reconstruction filter.
`converter system generating the first stage digital output
`includes generating thefirst stage digital output using a first
`In an embodiment of such an analog-to-digital converter
`stage analog-to-digital converter circuit operating at a first
`system the first stage analog-to-digital converter circuit
`frequency, and the second operating frequency is a fraction
`operates at a first operating frequency and the second stage
`of the first operating frequency. In an alternate embodiment
`analog-to-digital converter circuit operates at a second oper-
`of such an analog-to-digital converter system the second
`ating frequency, wherein the second operating frequency is
`operating frequency is equal to a Nyquist frequency of a
`a fraction of the first operating frequency. In yet another
`sampling circuit used for sampling the noise shaped residue
`embodiment of such an analog-to-digital converter system
`signal andthe first operating frequency is at least twice the
`the second operating frequency is equal to a Nyquist fre-
`Nyquist frequency of a sampling circuit. In an alternate
`quency of the sampling circuit and the first operating fre-
`embodiment of such an analog-to-digital converter system
`quency is at
`least
`twice the Nyquist frequency of the
`noise-shaping the residue ofthe first stage analog-to-digital
`sampling circuit.
`converter circuit further includes providing a gain to the
`An embodimentof the analog-to-digital converter system
`residue ofthe first stage analog-to-digital converter circuit.
`further comprises a digital signal generator adapted to
`Yet another embodiment of the analog-to-digital con-
`generate the digital test signal, wherein the digital signal
`verter system further comprises changing the amplitude of
`generator is further adapted to change an amplitude of the
`the digital test signal based on one of a polarity or an
`digital test signal based on oneofa polarity or an amplitude
`amplitude of the analog input signal. An alternate embodi-
`of the analog input signal. In yet another embodimentof the
`ment of such an analog-to-digital converter system further
`analog-to-digital converter system the digital signal genera-
`comprises changing the amplitudeofthedigital test signal in
`tor is further adapted to change an amplitude ofthe digital
`a manner so as to prevent the second stage analog-to-digital
`test signal in a manner so as to prevent the second stage
`converter circuit from saturating.
`analog-to-digital converter circuit from saturating. In yet
`another embodimentof the analog-to-digital converter sys-
`Referring now to the accompanying drawings, FIG. 1
`tem the digital correction circuit further includes a decima-
`illustrates an ADC circuit 10 that converts an analog input
`tion filter adapted to cancel the effect of the digital test signal
`signal g(t) into a digital output signal dg(k/N). The ADC
`from the combinationofthe first stage digital output and the
`circuit 10 includes a noise shaping first stage 12 cascaded
`second stage digital output.
`with a pipelined second stage 14. The ADC circuit 10 also
`includes a sample-and-hold circuit 16 to input the analog
`In yet another embodiment of the analog-to-digital con-
`input signal g(t), a digital dither generator 18 to generate a
`verter system the second stage analog-to-digital converter
`dither signal dt(k), and a calibration filter 19. The second
`circuit
`is one of a pipelined analog-to-digital converter
`stage 14 is represented by an ADC labeled ADC2 14. The
`circuit or a cyclic analog-to-digital converter circuit and
`ADC2 14 mayhave a resolution close to the overall reso-
`wherein the first stage analog-to-digital converter circuit is
`lution of the ADC circuit 10. In this example, the ADC2 is
`a delta sigma analog-to-digital converter circuit.
`In yet
`a 12 bit ADC with a signal-to-noise ratio (SNR) of 74 dB.
`another embodimentof the analog-to-digital converter sys-
`tem,thefirst stage analog-to-digital convertercircuit further
`The first stage 12 includes a noise shaping filter 20, a
`FLASH ADC labeled ADC1 22 and a feedback DAC 24.
`includes a multiplier coupled to the output of the noise
`shaping circuit and adapted to multiply the output of the
`noise shaping circuit, a feed-forward path adapted to feed-
`forward the analog input signal to a summercircuit, the
`summercircuit adapted to combine the analog input signal
`and the output of the multiplier, and a quantizer adapted to
`generate a quantized output signal based on the combined
`output of the summercircuit. In an alternate embodiment of
`such an analog-to-digital converter system the quantizer is
`further adapted to introduce an offset into the quantized
`output signal.
`
`The noise shaping filter 20 provides a gain, noise shaping
`and dithering. The ADC1 22 converts an analog input signal
`v(k) into a digital signal d1(&). A summation circuit 26
`combines the digital signal d1(4) with the dither signal d(t)
`to generate a combined signal d11(4). The signal d1z(k) is
`input into the feedback DAC 24, which converts it into an
`analog signal. The output of the feedback DAC 24 is
`subtracted from the analog input signal g(k) to generate a
`residue signal e(k). The noise shaping filter 20 provides gain,
`noise shaping, and dithering to the residue signal e(k).
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`Page 16
`
`

`

`US 7,129,874 B2
`
`5
`The noise shaping filter 20 is used to increase a sampling
`frequency of the ADC circuit 10 while keeping the sampling
`frequency of ADC2 14 fixed at the Nyquist frequency. For
`example,
`in a very high speed digital subscriber lines
`(VDSL) system having a 12 MHz bandwidth, using the ADC
`circuit 10, the S/H circuit 16 can operate at 100 MHz and the
`ADC2 14 can operate at 25 MHz. This feature reduces the
`overall powerdissipation of the ADC circuit 10 significantly
`compared to a case when the ADC2 14is operating at 100
`MHz. Normally when the ADC2 14 is operating at 100
`MHz, a decimation is applied to the output of the ADC
`circuit 10, resulting in increased power dissipation. The
`reason for the significant reduction in powerdissipation is
`the design ofthefirst stage 12, which may include only one
`integrator in the noise shaping filter 20, one simple 3.1 bit
`FLASH ADC as ADC 1 22, and only one feedback DAC 24.
`The total powerdissipation of running these components of
`the first stage 12 at 100 MHz andthe second stage 14 at 25
`MHzis significantly less than running the second stage 14
`at 100 MHz, as it would be necessary without the noise-
`shaping filter 20 of the first stage 12.
`The noise-shapingfilter 20 also provides an analog gain,
`thereby improving the signal-to-noise ratio (SNR) of the
`ADC2 14 with an amount equal to the analog gain. How-
`ever, the improvement of SNR of the ADC2 14 depends on
`an over-sampling ratio (OSR). At OSRs of two and four the
`improvementin an overall SNR of the ADCcircuit 10 is less
`than the analog gain provided by the noise-shapingfilter 20.
`Moreover, the noise-shaping filter 20 provides analog noise
`shaping, thus increasing the SNR of the ADC circuit 10 at
`high OSRs. The noise shaping becomeseffective for OSRs
`above four. For instance, at an OSRofeight, the achievable
`SNRis 96 dB compared to and SNR of 87 dB at an OSR of
`four. Finally, the noise-shaping filter 20 dithers the input to
`the ADC2 14, thus improving the overall linearity of the
`ADCcircuit 10.
`
`Thefirst stage 12 of the ADC circuit 10 operates like a AX
`ADCcircuit due to the noise shaping provided by the noise
`shapingfilter 20. However, due to the analog gain provided
`by the noise shaping filter 20, the first stage 12 also behaves
`like a stage of a pipelined ADC circuit. This is possible due
`to a forward path 28 from the sampled input signal g(k) to
`a summation circuit 30 before ADC1 22, and due to scaling
`of the output r(k) of the noise shaping filter 20 with an
`analog gain component 32 providing a gain of al. Alternate
`implementationsofthe first stage 12 using multiple outputs
`from the noise shaping filter 20 are possible. For example,
`the noise shaping filter 20 can have n outputs, each of them
`can be scaled with factors al .. . an, and a sum ofthe scaled
`outputs may be added to the input signal g(k) to get a
`combined signal v(k), which is fed to ADC1 22.
`The output of the noise shaping filter 20 r(k) is fed to the
`ADC214 for further quantization. The output of the ADC2
`14 may be gain calibrated by the calibration filter 19.
`However, the calibration filter 19 is optional. The purpose of
`the calibration filter 19 is to digitally compensate for any
`errors introducedin the first stage 12. The gain calibration
`filter 19 performs such calibration by measuring the output
`dg(k/N) and correlating it against the dither signal dt(k). The
`ADC circuit 10 uses a switching device 34 that reduces the
`sampling by the ADC2 14 bya ratio of 1/N. Recombination
`filters 36 and 38filters the outputs of the first stage 12 and
`the ADC2 14. A recombination circuit 40 combines the
`
`30
`
`35
`
`40
`
`45
`
`outputs of the reconstructionfilters 36 and 38 to generate the
`digital output signal dg(k/N).
`FIG. 2 is a block diagram ofa first order modulator 50
`used in the ADC circuit of FIG. 1. The first order modulator
`
`65
`
`6
`50 includes an integrator 52 with a built-in gain of 4. An
`ADC1 54 can be a single-bit or a multi-bit quantizer. In this
`example, the ADC1 54 is shown by a 3.1-bit FLASH ADC.
`The ADC1 54is configured with /4 LSB systematic offset in
`its comparator, as it would be in a pipeline stage, with the
`benefit of improvedlinearity of a AX modulator. A feedback
`DAC 56 converts the output of the ADC1 54 into an analog
`signal. Here the feedback DAC 56 is implemented as a 3.25
`bit DAC, where the term 3.25 bit is referred to denote that
`the DAC 56 hasten reference levels.
`
`FIG. 3 is an exemplary signal diagram 60 of reference
`voltages for the ADC1 54 of FIG. 2 implementedas a 3.1-bit
`FLASH ADC. (Here the term 3.1 bit ADCis used to denote
`an ADC having nine output levels. Generally, an x.y bit
`ADCis said to have 2°x+y output levels. However, please
`note that a 1.5 bit ADC has three output levels and a 2.5 bit
`ADC hassix output levels). In FIG.3, the signal diagram 62
`shows reference voltages for a regular 3 bit FLASH ADC,
`while the signal diagram 64 shows a configuration of a 2.5
`bit FLASH ADC(six reference voltage levels) when used in
`a pipeline ADC stage with an analog gain of four. The signal
`diagram 66 showsthe reference voltages for the ADC1 54 of
`FIG. 2 implementedas a 3.1 bit FLASH ADC. Compared to
`the signal diagram 64 the signal diagram 66 has two extra
`comparators, one 68 at the top and another 70 at the bottom
`of the input range. This feature allows for full input swing
`to the modulator 50, which is a major benefit compared to
`a regular modulators that usually saturate at an input voltage
`level 6 dB below full scale reference voltages.
`The 3.1 bit configuration of the ADC1 54 uses two extra
`comparators than a comparable pipeline stage to accommo-
`date the memorized output from the integrator 52. This
`configuration of the ADC1 54 allowsfordigital correction of
`offsets in ADC1 54 by means of redundancy. Such offset
`
`correction suppresses first order nonlinearity in the ADC
`
`
`
`
`circuit 10 compared to a regular
`modulator without
`correction of offsets. With the systematic offset in the ADC1
`54 and the subsequent digital correction, only second order
`effects of the offset contribute to nonlinearity through the
`noise shaping filter 20, thus enhancing the overall linearity
`of the ADC circuit 10.
`
`FIG. 4 describes exemplary transfer functions ofthefirst
`order modulator 50 of FIG. 2. Specifically, FIG. 4a describes
`a transfer function 80 of the modulator 50 in response to an
`output v(k) of a summing node 58 before the ADC1 54. As
`shown, amplitudes up to +/-1.25 Vref are allowed at this
`node. This is to fitin a full scale +/-1 V input while featuring
`a worst case memory of +/-1 V at the integrator output. After
`scaling and summing, the worst case sum becomes +/—1.25
`Ver FIG. 46 describes one possible transfer function of the
`first order modulator 50 as a function of the amplitude of the
`input signal g(k). Because the integrator 52 dithers the
`transfer function 82 to a seemingly nonlinear function, the
`shapeofthe transfer function 82 will depend on the level and
`shape of the input signal g(k).
`FIGS. 5 and 6 are exemplary circuit diagrams of a
`multiplying and integrating DAC (MDAC) 90 used in the
`modulator 50. The MDAC 90 consists of twoparts, a regular
`feedback DAC 92 and a dither DAC 94. The regular
`feedback DAC 92 includes unit capacitors 96, each having
`a size of C. The dither DAC 94 includes dither capacitors 98,
`each having a size of C/4. The MDAC 90 also includes a
`comparator 100 and integrating capacitors 102.
`FIG. 5 shows the MDAC 90 being used in a sampling
`phase of the modulator 50. The dither DAC 94 sums in a
`digital dither signal dt(k) into the outputs of the ADC1 54.
`
`Page 17
`
`

`

`US 7,129,874 B2
`
`7
`In the sampling phase, the MDAC 90 includes sampling
`capacitors where the size of the sampling capacitors is 4C.
`FIG. 6 shows the MDAC 90 being used in a feedback
`phase of the modulator 50.
`In the feedback phase,
`the
`MDAC 90 is shown to have feedback capacitors 104, each
`having a size of C. Thus the MDAC 90 achieves a gain of
`four by keeping the feedback capacitors 104 equal to 4 the
`size of the sampling capacitor 100. In the feedback phase the
`feedback DAC 92 matches thermometer outputs of the
`ADC1 54.
`
`5
`
`10
`
`15
`
`8
`ADC154 in orderto create a digital representation dg of the
`analog input g(t). This is made possible by combining a
`decimation filter with reconstruction filters, resulting in
`combined decimation and reconstructionfilters H1(z) 36 and
`H2(z) 38. The combined decimation and reconstruction
`filters H1(z) 36 and H2(z) 38 operate as a digital correction
`circuit that combines the outputs of the ADC2 14 and the
`ADC1 22 and the digital dither signal in a mannerso as to
`cancel the effect of the digital dither signal from the com-
`bined outputs of the ADC2 14 and the ADC1 22.
`An example of a design procedure for reducing the
`Thus the effective sampling capacitor is 4C while the
`
`effective feedback capacitor is C. The unit capacitors of the sampling rate of ADC2 14 usingafirst order sincfilter is
`MDAC 90 are C, with a maximum of 4C connected to a
`shown in FIGS. 8-12. FIG.8 is a block diagram of a 10 bit
`MASH ADC circuit 140 with a first order sinc decimation
`reference voltages when the ADC1 54 sees an input close to
`one of the rails. Nine reference levels are obtained from
`filter 142. The ADC circuit 140 includes a first recombina-
`connecting, the four differential unit capacitors to the posi-
`tion filter 144 to filter a digital output d1(A) ofa first stage
`tive reference voltage Vr+, the negative reference voltage
`ADC, a second recombination filer 146 to filter a digital
`Vr-, and/or to an analog ground. A maximum reference is
`output d2(4) of the second stage ADC, and a recombination
`adder 148.
`attained by connecting 4C at the positive input to Vr+, while
`4C at the negative input are connected to Vr-. A zero
`reference is obtained by connectedall the DAC capacitors to
`analog ground.
`The magnitude of the dither dt(k) is 4 of the magnitude
`of the reference voltages. This is achieved by using the
`dither capacitor 98 equal to 4 thesize of the unit capacitor
`96 in the MDAC 90. Thus, from FIGS. 5 and6 it is obvious
`that a 5 bit DACis not necessary in order to account for the
`digital dither with 4 the magnitude of an LSB from the 3.1
`bit ADC1 54. The feedback DAC 92is a regular 3 bit DAC,
`while the dither DAC 94 constitutes an extra 4 unit capaci-
`tor. This configuration relaxes the complexity of switches 97
`for the feedback DAC 92 compared to a full 5 bit DAC.
`Since the dither DAC 94 features dither capacitors 98 equal
`to % of the unit capacitors 96 of the feedback DAC 92, the
`minimum size of the sampling capacitors is limited some-
`what. Dueto the offseted ADC1 54 and the inherentlinearity
`gain from the multi-bit feedback DAC 92, mismatch shaping
`of the MDAC 90 is not necessary.
`FIG. 7 is an exemplary signal diagram 120 of an output
`signal d1(4) of the first order modulator 50 of FIG. 2. The
`diagram 120 shows that the total output swing of d1(4)is
`+/-1 Vref. Of this total output swing, a swing of +/-0.5 Vref
`is dueto the quantization noise of the 3.1 bit ADC1 54. This
`will be the total output swing in absence of dither and when
`the comparators are ideal. The dither dt(k) introduces an
`additional +/-0.25 Vref swing when the dither DAC 94 uses
`a 4% unit capacitor. In this configuration,
`the maximum
`tolerable comparatoroffset is +/-4 LSB ofthe 3.1 bit ADC
`1 54,
`resulting in additional +/-0.25 Vref swing. For
`example, when Vref equals 1V, the LSB of the ADC1 54
`equals 250 mV, and the tolerable comparator offset
`is
`+/-62.5 mV. Thus, to fit dither dt(k) in a residue voltage
`budget, the comparators of the ADC1 54 must be designed
`for 4 bit accuracy. In this design, the dither dt(k) is con-
`trolled by a 1 bit pseudo-random signal that is interpreted as
`positive dither if equal to ‘1’, and negative dither if equal to
`‘0’. In order to ensure maximum swing of the input signal,
`the dither changes magnitude when the ADC1 54 sees an
`input signal that is close to one ofthe rails. With an input
`near positive rail,
`the dither is only allowed to switch
`between maximum negative dither and zero. Similarly, with
`an input near negative rail, the dither is only allowed to
`switch between maximum positive input and zero.
`The pipelined ADC2 14 operates at a reduced samplerate,
`clocked with a frequency fs/N, where fs is the system clock
`frequency. Every Nth digital residue voltage d2 at the input
`of the ADC2 14 is combined with a filtered output of the
`
`FIG. 11 is a block diagram of the 10 bit MASH ADC 140
`of FIG.8, with the switching node 154 movedin front of the
`ADC2 160, to provide a reduced sampling rate in a second
`stage of the 10 bit MASH ADC 140.
`A sine decimation filter is not appropriate for a VDSL
`application, because ripple or band bending in the signal
`bandis too high near the edge of a band using a first order
`sinc filter. An alternate decimation filter must be used instead
`of the sinc decimation filter, with the requirement that the
`denominator of such alternate decimation filter is exactly
`D=1-z-1, while the nominator is a function of f(z-N). The
`decimation filter should be designed to match the require-
`ments of an application, yielding proper band-pass ripple
`and stop-band attenuation, and fulfilling requirements for
`the numerator and the denominator given above. An alter-
`nate implementation using such a decimation filter that has
`such a transfer function is illustrated in FIGS. 12 and 13.
`Specifically FIG. 12 is a block diagram of a 10 bit MASH
`ADC 180 with a general decimation filter 182.
`FIG. 13 is a block diagram of the 10 bit MASH ADC 180
`of FIG. 12 with the general decimation filter 182 combined
`with a recombination filter, as shown by the recombined
`decimation filter 184.
`
`20
`
`25
`
`30
`
`35
`
`40
`
`45
`
`50
`
`55
`
`60
`
`65
`
`FIG. 9 is a block diagram of the 10 bit MASH ADC 140
`of FIG.8, with a first