Switched-capacitor circuitry with reduced loading upon reference voltages

[0002] This application claims priority under 35 USC § 119(e)(1) of provisional application No. 60/098,334, filed Aug. 28, 1998.

[0003] Not applicable.

[0004] Not applicable.

[0005] This invention is in the field of analog circuitry, and is more specifically directed to switched-capacitor circuits.

[0006] In the field of analog circuits, such as implemented in modem analog and mixed-signal (i.e., containing both digital and analog functions) integrated circuits, switched-capacitor techniques are utilized in many applications. In general, switched-capacitor circuits operate by periodically connecting an input voltage to a capacitor, which in turn stores a charge corresponding to the applied voltage. The capacitor is subsequently connected to an input of an amplifier or other circuit to communicate the input voltage to downstream circuitry. As such, sample-and-hold functions are commonly realized by way of switched-capacitor techniques.

[0007] One important application of switched-capacitor techniques is the comparison of an input signal voltage to a reference voltage. In this application, the sampling capacitor is initially connected to a reference voltage, such as may be generated by a band-gap voltage reference circuit or similar circuit for generating a reference voltage that is relatively stable over variations in power supply voltage, temperature, and manufacturing process parameters. The capacitor is then switchably connected to receive a signal input voltage, such that the resulting charge on the capacitor corresponds to the differential between the signal input voltage and the reference voltage. This resulting charge can then be sensed by an amplifier, which in turn generates a signal corresponding to that differential. The reference voltage is then again connected to the sampling capacitor in preparation for the next sample of the input voltage.

[0008] This switched-capacitor comparator function is commonly used in many integrated circuit applications. One important use of this function is in pipelined analog-to-digital converter circuits (ADCs), an example of which is presented, in block diagram form, as ADC 10 of FIG. 1. As illustrated therein, ADC 10 receives an analog input voltage on line V_INand generates, at its output, an n-bit digital word on lines V_OUTthat correspond to the input analog voltage. To accomplish this function, ADC 10 includes a series of analog stages 4₀through 4_k. First stage 4₀receives the analog input voltage on line V_INand, based on the amplitude of this input voltage, generates an m-bit digital output on lines D₀and a residue output on lines RES₀. Typically, the number of digital bits m generated by each stage 4 is 2 or more. The residue voltage on line RES₀corresponds to a remainder of the “division” carried out in the digitization operation. As shown in FIG. 1, each subsequent, lower-order, stage 4_j(from the set of stages 4₁through 4_k) receives the residue voltage on lines RES_j−1from the previous stage 4_j−1, and similarly generates m digital output bits on lines D_j, and a residue analog voltage on line RES_jthat is forwarded to the next stage 4_j+1in the pipeline.

[0009] The digital results of each stage 4_jare stored in corresponding latch 6_j, the contents of which are summed, by adder 8_j, with the digital result on lines D_j+1from the next stage 4_j+1in the sequence; the output of the final adder 8_kis the n-bit digital output word on lines V_OUT. The pipelining effect of ADC 10 is enabled by the operation of latches 6; once a stage 4_jderives a digital result and residue, it can begin conversion of the next sample in time while the next stage 4_j+1operates on the result from the previous sample. As such, each of the multiple stages 4 in ADC 10 can be operating on different samples of the input signal voltage V_IN, with a sequence of digital results generated by final adder 8_k.

[0010] FIG. 2 illustrates the functional construction of exemplary analog stage 4_jin ADC 10 of FIG. 1. As shown in FIG. 2, input voltage V_inis applied to sub-ADC 5, which produces the digital output on lines D_j; these digital outputs are also applied to sub-DAC 7, which presents an analog signal to subtractor 9. Subtractor 9 subtracts the output from sub-DAC 7 from the value of input voltage V_in, sampled and held by sample-and-hold circuit 3; the output of subtractor 9 is amplified by gain amplifier 11 to generate analog residue voltage on line RES_jthat is within a voltage range suitable for use by a next, downstream, analog stage 4_j+1.

[0011] In typical modern pipelined ADCs, certain of the functions of each stage 4 are combined into single circuits, which may be realized according to switched-capacitor techniques. An example of such a conventional modern pipelined ADC may be found in Lewis, et al., “A 10-b 20 Msample/s Analog-to-Digital Converter”, J. Solid State Circ., Vol. 27, No. 3 (IEEE, March 1992), pp. 351-58. As described therein, the functions of sample-and-hold 3, subtractor 9, sub-DAC 7, and amplifier 11 may be combined into a switched-capacitor amplifier circuit that operates upon differential input voltages, and a comparison against two reference voltage levels. In the ten-bit case of the Lewis, et al. article, each of nine stages receives two reference voltage levels at switched-capacitor inputs. As exemplified in the Lewis, et al., article, and as is known in the art, switched-capacitor pipelined ADC circuits provide excellent resolution at extremely high conversion rates, considering the pipelined architecture of the ADC.

[0012] Of course, as is known in the art, ADCs of 14-bit and 16-bit precision are now commonly used, with even higher precision ADCs expected in the near future. According to the typical switched-capacitor architecture, as described in the Lewis et al. article and as will now be described relative to FIG. 3, each switched-capacitor stage of the ADC performs a comparison of sampled differential input voltages, precharging sample nodes to reference voltages before each sample of the differential input signal. Accuracy in this comparison requires, of course, reference voltages that are not only stable over variations in power supply voltage, temperature, and manufacturing process parameters, but is also stable considering the switching operation of the ADC.

[0013] FIG. 3 illustrates an example of conventional differential switched-capacitor stage 15_jas may be used in modem pipelined ADCs and other circuits that utilize switched-capacitor comparisons. For purposes of clarity, conventional devices that are commonly included to minimize charge injection current and apply precharge voltages at certain circuit nodes are not shown in FIG. 3. In this example, switched-capacitor stage 15_jreceives a differential input voltage on lines V_in⁺ and V_in⁻, and two reference voltage levels V_refpand V_refn. In operation, switched-capacitor stage 15_jcompares the voltage on line V_in⁺ with the voltage on line V_in⁻, after precharging respective capacitor inputs to reference voltages V_refp, V_refn. Reference voltages V_refpand V_refnare generated by reference voltage circuit 20 which includes, in this conventional example, bandgap circuit 12 for generating a voltage on line V_rthat is stable over variations in power supply voltage, temperature, and manufacturing process parameters. Various implementations of bandgap circuit 12 are well known in the art. Line V_r, in this example, is applied to an inverting input of operational amplifier 14, which has its non-inverting input biased to ground. According to conventional circuit techniques, the resistive input and feedback arrangement of amplifier 14, in combination with the capacitive coupling of the inverting and non-inverting outputs as illustrated, provide reference voltages V_refpand V_refnto switched-capacitor stage 15_jand other similar stages (not shown) which remain stable over various parameter variations as noted above.

[0014] Switched-capacitor stage 15_jincludes a differential operational amplifier 16 which, as will now be described, generates a differential voltage on lines V_out⁺ and V_out⁻ in response to a switched-capacitor comparison of the voltages on lines V_in⁺ and V_in⁻. In this regard, input voltage V_in⁺ is coupled to node V_Aat sample-and-hold capacitor CIN+ by way of switch 17+, and is coupled to feedback capacitor CFB+ by way of switch 18+. Each of switches 17+, 18+, as well as the other switches 17, 18, 19, 21 in switched-capacitor stage 15_jare conventional passgates, such as n-channel MOS transistors, or parallel CMOS transistors receiving complementary signals at their gates, as well known in the art.

[0015] As shown in FIG. 3, each of switches 17+, 18+ are controlled by clock φ₁, which is the sample clock in this arrangement. Reference voltage V_refpis coupled to node V_Aat sample-and-hold capacitor CIN+ by way of switch 19+, which is controlled by amplify clock φ₂. Clock φ₂also controls, via switch 21+, the coupling of feedback from the non-inverting output of amplifier 16 to the inverting input thereof through feedback capacitor CFB+.

[0016] Similarly, the inverting input of amplifier 16 is connected to sample-and-hold capacitor CIN−, the opposite plate of which is coupled, at node V_B, to input voltage V_in⁻ through switch 19−, and to reference voltage V_refnthrough switch 19−. The inverting input of amplifier 16 is also connected to one plate of feedback capacitor CFB−. The opposite plate of feedback capacitor CFB− is coupled to input voltage V_in⁻ via switch 18− and, through switch 21−, to line V_out⁻ at the inverting output of amplifier 16. Switches 17− and 18− are controlled by sample clock φ₁while switches 19− and 21− are controlled by amplify clock φ₂.

[0017] Sample clock φ₂and amplify clock φ₂are non-overlapping clock phases of the same frequency. In operation, sampling is effected by switched-capacitor stage 15_jupon sample clock φ₁being driven active (amplify clock φ₂inactive at this time); during this phase, nodes V_A, V_Breceive input voltages V_in⁺, V_in⁻, through switches 17+, 17−, respectively; these input voltages V_in⁺, V_in⁻ are also respectively applied to feedback capacitors CFB+, CFB− through switches 18+, 18−. During amplify clock phase φ₂becoming active (sample clock φ₁being inactive), switches 21+, 21− apply feedback from output lines V_out⁺, V_out⁻, respectively, to their respective feedback capacitors CFB+, CFB−, so that amplification is carried out by amplifier 16 according to the appropriate desired characteristics. In preparation for the next sample-and-hold operation, reference voltages V_refp, V_refnare applied to nodes V_A, V_B, respectively, through respective switches 19+, 19− during this active phase of amplify clock φ₂, such that the next samples of input voltages V_in⁺, V_in⁻ are accurately obtained (and their differential accurately compared against the reference voltage differential V_refp−V_refn), independently from the voltage of the prior sample. The operation then repeats, in the next cycles of sample clock φ₁and amplify clock φ₂, to obtain and amplify the next sample of input voltages V_in⁺, V_in⁻.

[0018] It is contemplated that conventional switched-capacitor circuits, such as switched-capacitor stage 15_jof FIG. 3, present substantial load upon the reference voltages generated by reference voltage circuit 20 of FIG. 3. This load is, of course, exacerbated in circuits such as pipelined ADCs, where multiple switched-capacitor stages (up to sixteen such stages, in typical high-resolution ADC examples) receive the reference voltages and simultaneously switch their inputs. Because of this load, and the switching operations that such conventional circuits perform, the stability of the reference voltage can be compromised.

[0019] Referring now to FIG. 4, a worst-case example of reference voltage variations, as caused by switched-capacitor circuits such as switched-capacitor stage 15_jof FIG. 3, will now be described. In this example, the circuit of FIG. 3 was simulated, in a case where the total load presented by the switched capacitors in switched-capacitor stage 15_jwas on the order of tens of picofarads. In the example illustrated in FIG. 4, an initial condition was established (i.e., prior to time t=0) by simulated operation of switched-capacitor stage 15_jfor several cycles at a clock frequency (φ₁, φ₂) of 5 MHz using a minimum power supply voltage of 3 volts, and with input voltage V_in⁺ set equal to reference voltage V_refn, and input voltage V_in⁻ set equal to reference voltage V_refp(i.e., each input voltage at its opposite state). This operation corresponds to the largest capacitor voltage swings at sample-and-hold capacitors CIN+, CIN−. At time t₀, new samples are obtained with input voltage V_in⁺ set equal to reference voltage V_refp, and input voltage V_in⁻ set equal to reference voltage V_refn. As evident from FIG. 4, the reference voltage differential V_refp−V_refnmodulates, in this condition, to a lower voltage from that at time t=0, differing therefrom by about 90 microvolts. After the reference voltage differential V_refp−V_refnstabilized, this simulation reverted the input voltages to their initial state, with input voltage V_in⁺ set equal to reference voltage V_refn, and input voltage V_in⁻ set equal to reference voltage V_refp. The reference voltage differential V_refp−V_refndoes not remain at the lower voltage, however, but instead returns to its higher level, approximately equal to its level at time t=0. The actual reference voltage differential V_refp−V_refnwill, in practice, vary within these worst case limits of FIG. 4.

[0020] As evident from this simulation, the reference voltage differential V_refp−V_refnvaries according to the value of the input voltages V_in⁺, V_in⁻, due to the loading presented by switched-capacitor stage 15_j, and the charge sharing between sample-and-hold capacitors CIN+, CIN− and the capacitors in reference voltage circuit 20. This dependence on input voltage is, of course, undesirable in applications such as ADCs, where accuracy in measurement of the input voltage is of highest importance. While variations of on the order of 90 microvolts are within the tolerance of some applications, such variations are not acceptable for high precision ADCs; indeed, the stability of reference voltages can limit the precision (i.e., number of bits) of the ADC itself. Conversely, improved stability of reference voltages can enable the construction and accurate operation of ADC circuits with additional bits of precision.

[0021] It is therefore an object of the present invention to provide a switched-capacitor circuit that presents reduced switching load to the reference voltages applied thereto.

[0022] It is a further object of the present invention to provide such a circuit that may be readily implemented in switched-capacitor circuits with minimal overhead.

[0023] It is a further object of the present invention to provide a pipelined ADC including multiple ones of such switched-capacitor circuits, such an ADC having improved precision over conventional ADC circuits.

[0024] Other objects and advantages of the present invention will be apparent to those of ordinary skill in the art having reference to the following specification together with its drawings.

[0025] The present invention may be implemented in a switched-capacitor circuit having an amplifier with at least one input coupled, by way of a sample-and-hold capacitor, to switches through which an input voltage and a reference voltage are applied in non-overlapping sample and reference clock phases, respectively. The circuit also includes another switch through which a midlevel voltage is applied to the sample-and-hold capacitor during a third non-overlapping phase between the sample and reference clock phases. The application of the midlevel voltage limits the worst case loading presented to the reference voltage, thus limiting variations in the reference voltage as a result of input switching. When implemented into a pipelined ADC, the reduced reference voltage variations enable additional bits of precision in the analog-to-digital conversion.

[0026] FIG. 1 is an electrical diagram, in block form, of a conventional pipelined analog-to-digital converter (ADC).

[0027] FIG. 2 is an electrical diagram, in block form, of a conventional stage in the pipelined ADC of FIG. 1.

[0028] FIG. 3 is an electrical diagram, in schematic form, of a conventional differential switched-capacitor stage, and accompanying voltage reference circuit, such as used in the conventional pipelined ADC of FIG. 1.

[0029] FIG. 4 is a voltage plot over time of a simulation of the operation of the differential switched-capacitor stage of FIG. 3.

[0030] FIG. 5 is an electrical diagram, in schematic form, of a differential switched-capacitor stage according to the preferred embodiment of the invention.

[0031] FIG. 6 is a timing diagram illustrating the operation of the differential switched-capacitor stage of FIG. 5, constructed according to the preferred embodiment of the invention.

[0032] FIG. 7 is a voltage plot over time of a simulation of the operation of the differential switched-capacitor stage of FIG. 5 according to the preferred embodiment of the invention.

[0033] FIG. 8 is an electrical diagram, in block form, of a pipelined ADC incorporating the differential switched-capacitor stage of FIG. 5 according to the preferred embodiment of the invention.

[0034] As will become apparent to those skilled in the art having reference to the following description, the present invention may be implemented according to many alternative realizations and applications. In this regard, while the preferred embodiment of the invention will be described hereinbelow relative to a differential switched-capacitor circuit, it is contemplated that the present invention may be realized in switched-capacitor circuits having single-ended inputs. Furthermore, while it is contemplated, as described hereinbelow, that the present invention is of particular benefit in circuits having several to many switched-capacitors stages, such circuits including pipelined analog-to-digital converters (ADCs), other applications of the present invention may also benefit from the present invention. It is therefore contemplated that such alternative realizations and applications will be within the scope of the present invention, as claimed hereinbelow.

[0035] Referring first to FIG. 5, the construction of differential switched-capacitor stage 35_jaccording to the preferred embodiment of the present invention will now be described in detail. According to the preferred embodiment of the invention, differential switched-capacitor stage 35_jincludes differential operational amplifier 36, having inverting and non-inverting inputs, and inverting and non-inverting outputs, and which is capacitively coupled to input voltages (and with capacitive feedback) so as to operate as a differential sarnple-and-hold amplifying circuit stage.

[0036] The non-inverting input of amplifier 36 is connected to one plate of sample-and-hold capacitor CIN+, and to one plate of feedback capacitor CFB+. The opposite plate of sample-and-hold capacitor CIN+ at node V_Ais coupled to input voltage V_in⁺ through switch 37+, and to reference voltage V_refpthrough switch 39+. The plate of feedback capacitor CFB+ opposite that of the non-inverting input of amplifier 36 is coupled to input voltage V_in⁺ through switch 38+, and to the non-inverting output of amplifier 36 at line V_out⁻ through switch 41+. Switches 37+, 38+, 39+, 41+ are all implemented as conventional passgates (e.g., CMOS or NMOS passgates). In this embodiment of the invention, switches 37+, 38+ are controlled by sample clock φ₁, and switch 41+ is controlled by amplify clock φ₂. Further, switch 39+ is controlled by reference voltage clock φ₂″ which, as will be described in detail below, is driven active during a latter portion of the active phase of amplify clock φ₂.

[0037] The inverting input of amplifier 36 is similarly connected to one plate of sample-and-hold capacitor CIN−, and to one plate of feedback capacitor CFB−. The opposite plate of sample-and-hold capacitor CIN− at node V_Bis coupled to input voltage V_in⁻ through switch 37−, and to reference voltage V_refnthrough switch 39−, while the opposite plate of feedback capacitor CFB− is coupled to input voltage V_in⁻ through switch 38−, and to the inverting output of amplifier 36 at line V_out⁺ through switch 41−. Identically as noted above for switches 37+, 38+, 39+, 41+, switches 37−, 38−, 39−, 41− are all conventional passgates, implemented either as single transistors are as CMOs passgates. Also similarly to the switches 37+, 38+, 39+, 41+ associated with the non-inverting input of amplifier 36, switches 37−, 38− are controlled by sample clock φ₁, switch 39− is controlled by reference clock φ₂″, and switch 41− is controlled by amplify clock φ₂.

[0038] According to the preferred embodiment of the invention, reference voltages V_refp, V_refndefine the extremes of the voltage range over which the input voltages V_in⁺, V_in⁻ may vary. For example, in the case where input voltages V_in⁺, V_in⁻ may vary from 0.5 volts to 2.5 volts, reference voltage V_refpwill be at approximately 2.5 volts, and reference voltage V_refnwill be at approximately 0.5 volts. Of course, the particular voltage range of input voltages V_in⁺, V_in⁻ and reference voltages V_refp, V_refnwill vary according to the particular application.

[0039] According to the preferred embodiment of the invention, nodes V_Aand V_Bare each also coupled to a third voltage V_mid, by way of switches 40+, 40−, respectively. Voltage V_midis preferably a voltage at a level between reference voltages V_refp, V_refn; however, this voltage V_midmay be unregulated, as its specific level is not critical to operation. In this example, voltage divider 42 is provided to generate voltage V_midfrom reference voltages V_refp, V_refnthemselves, by way of a conventional resistor divider connected in series between reference voltages V_refp, V_refn, with a stabilizing capacitor connected between its output and ground. Alternatively, voltage V_midmay be generated from a voltage division of power supply voltage V_ddand ground, or as a third reference voltage from a voltage generator circuit as known in the art. As will become apparent from the following description, however, the actual level of voltage V_midneed not be is not critical, and as such need not be precisely generated or regulated.

[0040] According to the preferred embodiment of the invention, voltage V_midis selected to be about the midlevel voltage between the reference voltage received at a particular input, and the opposite extreme of the input signal applied to that same input. Of course, in the differential switched-capacitor case described herein, voltage V_midmay be selected to a voltage that is approximately midway between the extremes of the range of input voltages V_in⁺, V_in⁻, these extremes defined by reference voltages V_refp, V_refn. For the above example in which input voltages V_in⁺, V_in⁻ may vary from 0.5 volts to 2.5 volts, voltage V_midis preferably at about 1.5 volts. Of course, the particular voltage of midlevel V_midwill vary according to the particular application, and according to the particular voltages utilized therein. Further in the alternative, it is contemplated, in the differential switched-capacitor case such as described relative to FIG. 5, that the application of different voltages V_mid⁺, V_mid⁻ to the differential sample nodes V_A, V_Bmay be beneficial.

[0041] Switches 40+, 40− are each controlled by precharge clock φ₂′. According to this preferred embodiment of the invention, precharge clock φ₂′ is driven active during an early portion of the active phase of amplify clock φ₂. In this way, during such time as precharge clock φ₂′ is active, both node V_Aand also node V_Breceive the voltage on line V_mid. According to the preferred embodiment of the invention, the active phase of precharge clock φ₂′ does not overlap either of the active phases of sample clock φ₁or reference clock φ₂″, and preferably occurs, in each cycle, after the active phase of sample clock φ₁and during an initial portion of the active phase of amplify clock φ₂. As will become apparent from the following description, the duration of the active phase of precharge clock φ₂′ need not be selected or maintained with a high degree of precision.

[0042] Referring now to FIG. 6, an example of the operation of switched-capacitor stage 35_jwill now be described relative to a cycle in which input voltage V_in⁺ is relatively low (near low reference voltage V_refn) and in which input voltage V_in⁻ is relatively high (near high reference voltage V_refp). Prior to time t₀, node V_Ais at high reference voltage V_refp, while node V_Bis at low reference voltage V_refn; these voltages result from the operation of the previous cycle (as will be apparent from the following description). Briefly prior to time to, all of clocks φ₁, φ₂, φ₂′, φ₂″ are inactive, and as such all switches 37, 38, 39, 40, 41 in stage 35_jare open. At time t₀, an active phase of sample clock φ₁begins. At this time, switches 37+, 37−, 38+, 38− in switched-capacitor stage 35_jare closed, thus applying input voltage V_in⁺ to node V_Aat sample capacitor CIN+ and input voltage V_in⁻ to node V_Bat sample capacitor CIN−. Additionally, input voltages V_in⁺, V_in⁻ are applied to feedback capacitors CFB+, CFB−, respectively, through respective switches 38+, 38−. As illustrated in FIG. 6, nodes V_A, V_Bdischarge and charge, respectively, toward their respective input voltages V_in⁺, V_in⁻ after time t₀, reaching the level of these input voltages during the active phase of sample clock φ₁. During this time, amplifier 36 begins generating its output at lines V_out⁻, V_out⁺, based on the samples of the input voltages obtained at its inputs via capacitors CIN+, CIN−. After the desired duration, the active phase of sample clock φ₁ends shortly prior to time ti in the example of FIG. 6, responsive to which switches 37+, 38+, 37−, 38− open.

[0043] According to the preferred embodiment of the present invention, amplify clock φ₂and precharge clock φ₂′ begin active phases at time t₁. The active phase amplify clock φ₂effects amplification via amplifier 36, while the active phase of precharge clock φ₂′ prepares switched-capacitor stage 35_jfor the next cycle. Specifically, the active phase of amplify clock φ₂closes switches 41+, 41−, so that feedback from output lines V_out⁻, V_out⁺ are respectively applied to feedback capacitors CFB+, CFB−. During this phase, amplifier 36 continues its amplification operation, generating the appropriate differential voltage at output lines V_out⁻, V_out⁺ according to the sampled differential input voltages V_in⁺, V_in⁻ and according to the negative feedback applied via feedback capacitors CFB+, CFB−.

[0044] The transition of precharge clock φ₂′ at time t₁closes switches 40+, 40−, such that voltage V_mid, which in this case is generated by voltage divider 42, is applied to nodes V_A, V_B. As illustrated in FIG. 6, the voltages at nodes V_A, V_Bcharge and discharge, respectively, toward voltage V_mid. According to the preferred embodiment of the invention, the duration PW of the active pulse of precharge clock φ₃may be quite short, for example on the order of 10 nsec for a 5 MHz application, considering that it is not essential that nodes V_A, V_Bactually reach voltage V_midin order to provide the benefits of the present invention. Furthermore, as noted above, the specific voltage V_midtoward which nodes V_A, V_Bare driven during the active phase of precharge clock φ₃is not of critical importance. These relatively loose constraints upon pulse width PW and voltage V_midresult from the consideration that switches 40+, 40− and precharge clock φ₂′ are intended simply to bring the voltages at nodes V_A, V_Baway from the worst case extremes of their opposite states (as in the example of FIG. 6), and toward a voltage that is somewhat within the middle of the expected range of input voltages V_in⁺, V_in⁻. Upon the completion of the active phase of precharge clock φ₂′, prior to time t₂, switches 40+, 40− are then opened.

[0045] At time t₂, reference clock φ₂″ enters its active phase, closing switches 39+, 39− to apply reference voltages V_refp, V_refnto nodes V_A, V_B, respectively. This closing of switches 39+, 39− charges and discharges nodes V_A, V_Bto reference voltages V_refp, V_refn, respectively, in preparation for the next cycle. Prior to time t₃, both amplify clock φ₂and reference clock φ₂″ return to an inactive state, prior to initiation of the next sample cycle.

[0046] It is contemplated, according to the present invention, that the operation of precharge clock φ₂′ in applying voltage V_midto sample nodes V_A, V_Bwill result in significant improvement in the stability of reference voltages V_refp, V_refnover time, primarily by presenting a constant impedance to the circuitry generating the reference voltages. This improvement may be explained by consideration of the distribution of charge between the reference voltage circuit and the sample capacitors of the switched-capacitor stages, as will now be described.

[0047] Conventional reference voltage circuits (such as reference voltage circuit 20 of FIG. 3) include large capacitors at their outputs, relative to the sample capacitors V_A, V_Bof switched-capacitor stages 35. For a given reference voltage V_ref, one may thus consider a total charge Q_TOTas the sum of the charge on the corresponding voltage reference capacitor of capacitance C_refplus the charge on the sample capacitor of capacitance C_sample. Of course, the charge on a capacitor is simply the product of the capacitance of the capacitor with the voltage thereacross. For an initial condition where a reference voltage V_refis generated but is not yet connected to sample capacitance C_sample, such that the voltage at the sample capacitor is at a voltage V_pre, the charge equation corresponds to the following:

[0048] Q_TOT=V_refC_ref+V_preC_sample

[0049] Once a switch is closed to apply the reference voltage to the sample node (e.g., switch 19+ of FIG. 3 doses), the voltages across each of the sample and reference capacitors equalize to voltage V_fin, changing the charge equation to:

[0050] Q_TOT=V_finC_ref+V_finC_sample

[0051] Combining these two equations to solve for the voltage V_fin: $V_{\mathrm{fin}}=V_{\mathrm{ref}}\frac{C_{\mathrm{ref}}}{C_{\mathrm{ref}}+C_{\mathrm{sample}}}+V_{\mathrm{pre}}\frac{C_{\mathrm{sample}}}{C_{\mathrm{ref}}+C_{\mathrm{sample}}}$

[0052] Even though, as noted above, the capacitance C_sampleof the sample capacitor is very small compared to the reference capacitance C_ref(e.g., on the order of one thousand times smaller), variations in the initial voltage V_preat the sample capacitor, before the reference voltage is applied thereto, will necessarily result in variations in the final voltage V_finafter the reference voltage is applied. These variations are described above, in the worst case, relative to FIG. 4. While quite small in the absolute sense, the variations in the final reference voltage V_fincan be quite limiting in some applications of switched-capacitor circuitry, such as high-precision ADCs.

[0053] Relative to the above discussion, the present invention is directed to eliminating the variations of the reference voltage by ensuring that sample capacitors CIN+, CIN+ are at the same voltage, in each cycle, immediately prior to the application of reference voltages V_refp, V_refnthereto. As described above, this is accomplished by way of switches 40+, 40− which apply voltage V_midto sample nodes V_A, V_Bduring the active phase of precharge clock φ₂′, prior to the active phase of reference clock φ₂″ as shown in FIG. 6.

[0054] Referring now to FIG. 7, a simulation of the operation of switched-capacitor stage 35_jaccording to the preferred embodiment of the invention will now be described, to provide a comparison of the contemplated performance of the present invention relative to the conventional switched-capacitor stage 15_jdescribed relative to FIG. 3. In this example, the initial condition of switched-capacitor stage 35_jprior to time t=0 is established by simulated operation for several cycles at a clock frequency (φ₁, φ₂, φ₂′, and φ₂″) of 5 MHz using a minimum power supply voltage of 3 volts; for these initializing cycles, input voltage V_in⁺ was set equal to reference voltage V_refn, and input voltage V_in⁻ was set equal to reference voltage V_refp, establishing each input voltage at its opposite state. At time t=0, new samples are obtained with input voltage V_in⁺ set equal to reference voltage V_refp, and input voltage V_in⁻ set equal to reference voltage V_refn, similarly as in the example of the operation of the conventional switched-capacitor stage 15_jillustrated in FIG. 4. As illustrated in FIG. 7 for this simulation, beginning at time t=0, the reference voltage differential V_refp−V_refnthus modulates to a lower voltage, settling, at about 20 microseconds, to a lower voltage.

[0055] However, the reference voltage differential V_refp−V_refnaccording to the preferred embodiment of the invention remains stable once it has settled. Similarly as in the simulation illustrated relative to FIG. 4, switched-capacitor stage 35_j, the simulation illustrated in FIG. 7 reverted the input voltages to their initial state, with input voltage V_in⁺ set equal to reference voltage V_refn, and input voltage V_in⁻ set equal to reference voltage V_refp, again occurring at about 60 μsec after time t=0. As evident from FIG. 7, the reference voltage differential V_refp−V_refnremains at this lower voltage despite this switching operation, because of the application of voltage V_midto sample nodes V_A, V_Bas described above. As such, the variation of the reference voltage differential V_refp−V_refnas a result of switching of the switched-capacitor input stages is greatly reduced through implementation of the preferred embodiment of the invention, with variations of less than 10 μvolts, after 40 cycles at 5 MHz, as evident in FIG. 7. This stability is desired in many applications, especially in pipelined ADCs as will now be described relative to FIG. 8.

[0056] FIG. 8 illustrates pipelined ADC 50, constructed according to the preferred embodiment of the present invention and as such including a series of switched-capacitor stages 35₀through 35_k, constructed as described hereinabove relative to FIG. 5. As illustrated in FIG. 8, switched-capacitor stages 35 are all associated with a digital latch and adder circuitry 44, which carries out the necessary digitization of the results of the switched-capacitor stages 35. In this regard, switched-capacitor stage 35₀receives a differential voltage on lines V_IN⁺ and V_IN⁻. In response, switched-capacitor stage 35₀generates output that is applied to digital latch and adder circuitry 44; switched-capacitor stage 35₀also generates a differential residue signal on lines RES₀, which corresponds to an analog remainder from the digitization operation for this stage, amplified to voltage levels suitable for operation by the next combination of switched-capacitor stage 35₁to which lines RES₀are applied as shown in FIG. 8.

[0057] Switched-capacitor stage 35₁then, in turn, generates a digital output that is applied to digital latch and adder circuitry 44, along with a residue signal on lines RES₁for application to the next stage 35₂(not shown). This sequence of stages 35 continues, as shown in FIG. 8, through k^thswitched-capacitor stage 35_k; switched-capacitor stage 35_kproduces no residue signal, as no further stages are present in ADC 50. Digital latch and adder circuitry 44 produces, in a pipelined fashion, a digital output of pipelined ADC 50 on bus DATA OUT. In modern pipelined ADCs, such as those generating a digital output of n-bit precision (where n may be 14, 16, or higher), it is contemplated that the number k+1 of switched-capacitor stages 35 can number up to as many as n₃₁1 (i.e., fifteen or more such stages 35).

[0058] Clock generation circuitry 60 is also provided in pipelined ADC 50 according to the preferred embodiment of the invention, and is constructed according to conventional clock circuit techniques to generate, from an external clock signal or crystal oscillator connected to line CLIN, non-overlapping phases of sample clock φ₁, amplify clock φ₂, precharge clock φ₂′, and reference clock φ₂″ for application to each of switched-capacitor stages 35₀through 35_k. The relative timing of the active phases of sample clock φ₁, amplify clock φ₂, precharge clock φ₂′, and reference clock φ₂″, in this preferred embodiment of the invention, correspond to that described above relative to FIG. 6.

[0059] Also as illustrated in FIG. 8, pipelined ADC 50 includes voltage reference circuit 20, which may be constructed in the manner described hereinabove relative to FIG. 3. Voltage reference circuit 20 generates reference voltages V_refp, V_refnand applies these voltages, in parallel, to each of switched-capacitor stages 35₀through 35_k. Voltage reference circuit 20 also applies reference voltages V_refp, V_refnto voltage divider 42, which in turn generates voltage V_midthat is applied to each of switched-capacitor stages 35o through 35_k. Considering that each of switched-capacitor stages 35₀through 35_kreceive reference voltages V_refp, V_refnin parallel, and are clocked simultaneously by sample clock φ₁, amplify clock φ₂, precharge clock φ₂′, and reference clock φ₂″, voltage reference circuit 20 is presented with a cumulative load due to the switching at the respective sample-and-hold capacitors CIN+, CIN− of each switched-capacitor stages 35₀through 35_k. According to the preferred embodiment of the invention described hereinabove, however, in which each of switched-capacitor stages 35₀through 35_kapplies voltage V_midto its sample-and-hold capacitors CIN+, CIN− under the control of precharge clock φ₂′, a constant impedance is presented to voltage reference circuit 20 over time, resulting in much improved stability of reference voltages V_refp, V_refnover the switching operation of ADC 50. This improved stability, as noted above, enables the implementation of additional bits of precision in the analog-to-digital conversion by way of pipelined ADCs, and thus additional usability of digital techniques for processing of analog signals.

[0060] As described above relative to FIG. 8, many circuit applications of the present invention such as ADC 50 will implement several to many switched-capacitor stages, each of which utilize reference voltages generated from an on-chip reference voltage circuit. In the operation of such applications, it is contemplated that the stabilization of reference voltages may require many cycles. For example, the simulation of switched-capacitor stage 35_jarranged according to the preferred embodiment of the invention has determined that the reference voltages applied thereto can stabilize, after operation for 40 cycles, to an accuracy of less than 10 microvolts. The actual reference voltage stability has been found to depend upon the latency that can be afforded in executing the sample, hold, and amplify operations. Considering that high-resolution pipelined ADCs, such as the example of ADC 50 described above, generally require many cycles for proper operation, such latency can often be afforded, especially with improved precision as a benefit.

[0061] According to the present invention, therefore switched-capacitor circuits may be constructed to operate at relatively high frequencies, but with minimal loading upon reference voltages applied thereto, such that the stability of the reference voltage is greatly improved, even in integrated circuits in which several to many switched-capacitor circuits each receive the reference voltages, and are simultaneously switched. The present invention is thus contemplated to provide important benefits, particularly in enabling precision functionality of modern high performance circuits, such as pipelined analog-to-digital converters.

[0062] While the present invention has been described according to its preferred embodiments, it is of course contemplated that modifications of, and alternatives to, these embodiments, such modifications and alternatives obtaining the advantages and benefits of this invention, will be apparent to those of ordinary skill in the art having reference to this specification and its drawings. It is contemplated that such modifications and alternatives are within the scope of this invention as subsequently claimed herein.