Sampled-data circuits using zero crossing detection

[0001] The present application claims priority, under 35 U.S.C. §119(e), from U.S. Provisional Patent Application Ser. No. 60/595,414, filed on Jul. 1, 2005. Also, the present application claims priority, under 35 U.S.C. §119(e), from U.S. Provisional Patent Application Ser. No. 60/595,493, filed on Jul. 11, 2005. The entire contents of U.S. Provisional Patent Application Ser. No. 60/595,414, filed on Jul. 1, 2005, and U.S. Provisional Patent Application Ser. No. 60/595,493, filed on Jul. 1, 2005, are hereby incorporated by reference.

[0002] The present invention relates generally to a switched capacitor circuit and, more particularly, to a switched capacitor circuit for integrated circuits manufactured according to scaled semiconductor processes.

[0003] Most sampled-data analog circuits such as switched-capacitor filters, analog-to-digital converters, and delta-sigma modulators require operational amplifiers to process a signal. Consider the switched-capacitor integrator example shown in FIG. 2. First, the switches S₁₁ and S₁₃ are closed so that the input voltage v_in is sampled on the sampling capacitor C_S1. Next, the switches S₁₁ and S₁₃ are opened and S₁₂ and S₁₄ are closed. This operation transfers the charge in the sampling capacitor C_S1 to the integrating capacitor C_I1. The output voltage, v_out, of a first integrator 1100 is typically sampled by another sampled-data circuit, for example, another switched-capacitor integrator. In the circuit shown in FIG. 2, the circuit consisting of switches S₂₁, S₂₂, S₂₃, S₂₄, and a second sampling capacitor C_S2 comprise a part of the second switched-capacitor integrator. The output voltage, v_out, of the first integrator 1100 is sampled by the second sampling capacitor C_S2 by closing switches S₂₁ and S₂₃.

[0004] An example of a timing diagram is shown in FIG. 3. The clock signal has two non-overlapping phases φ₁ and φ₂. The phase φ₁ is applied to switches S₁₁, S₁₃, S₂₁, and S₂₃, and phase φ₂ is applied to switches S₁₂, S₁₄, S₂₂, and S₂₄. With this timing, the circuit performs non-inverting discrete integration with full clock delay. The waveforms at the output of the integrator, v_out, and at the virtual ground node 100, v₁ are also shown in FIG. 3. Different clock phasing arrangements yield different responses from the integrator. For example, if φ₁ is applied to switches S₁₁, S₁₃, S₂₂, and S₂₄, and phase φ₁ is applied to switches S₁₂, S₁₄, S₂₁, and S₂₃, the circuit performs non-inverting integration with half-clock delay.

[0005] For an accurate integration of the input signal, v₁ must be driven as close to ground as possible. In order to accomplish this, the operational amplifier must provide sufficient open-loop gain and low noise. In addition, for fast operation, the operational amplifier 10 of FIG. 2 must settle fast.

[0006] In FIG. 3, the voltage v₁ is shown to settle back to ground after a disturbance when the sampling capacitor C_S1 is switched to Node 100 by closing S₁₂ and S₁₄. In addition to high open-loop gain and fast settling time, operational amplifiers must provide large output swing for high dynamic range. As the technology scales, it becomes increasingly difficult to achieve these characteristics from operational amplifiers. The primary factors that make the operational amplifier design difficult are low power supply voltages and low device gain.

[0007] As noted above, accurate output voltage can be obtained if Node 100 in FIG. 2 is maintained precisely at ground. However, in sampled-data circuits, the only point of time that an accurate output voltage is required is at the instant the output voltage is sampled by another sampling circuit. Thus, it is not necessary to maintain the voltage at Node 100 at ground all the time.

[0008] Therefore, it is desirable to provide a sampled-data circuit that maintains the proper level at the virtual ground node at the instant the output voltage is sampled by another sampling circuit. Moreover, it is desirable to provide a sampled-data circuit that maintains the proper level at the virtual ground node at the instant the output voltage is sampled by another sampling circuit and provides differential signal paths for sampled-data circuits. Furthermore, it is desirable to provide a sampled-data circuit that reduces the effect of power supply, substrate, and common-mode noise by symmetric differential signal processing. Also, it is desirable to provide a sampled-data circuit that increases the signal range by incorporating differential signal paths.

[0009] One aspect of the present invention is a switched-capacitor circuit. The switched-capacitor circuit includes a level-crossing detector to generate a level-crossing detection signal when an input signal crosses a predetermined level; a plurality of capacitors; and a sampling switch, operatively coupled to the level-crossing detector. The sampling switch turns OFF when the level-crossing detection signal indicates a level-crossing.

[0010] Another aspect of the present invention is a switched-capacitor circuit. The switched-capacitor circuit includes a level-crossing detector to generate a level-crossing detection signal when an input signal crosses a first predetermined level; a second level-crossing detector to generate a second level-crossing detection signal when the input signal crosses a second predetermined level; a plurality of capacitors; and a sampling switch, operatively coupled to the second level-crossing detector. The sampling switch turns OFF when the second level-crossing detection signal indicates the input signal crosses the second predetermined level.

[0011] Another aspect of the present invention is a switched-capacitor circuit. The switched-capacitor circuit includes a first switched capacitance network having an input terminal to receive a circuit input voltage; a level-crossing detector, having an output terminal and an input terminal, to generate a level-crossing detection signal when a signal at the input terminal crosses a predetermined level; and a second switched capacitance network operatively coupled to the level-crossing detector. The second switched capacitance network turns OFF when the difference between the first and second input signals crosses the predetermined level.

[0012] Another aspect of the present invention is a method for sampling an analog signal. The method samples an input voltage using the switched capacitance, network; determines when a node voltage of the switched capacitance network crosses a predetermined level; and provides a sample of an output voltage of the switched capacitance network when it is determined that the node voltage of the switched capacitance network crosses the predetermined level.

[0013] Another aspect of the present invention is a method for sampling an analog signal. The method samples an input voltage using the switched capacitance network; determines when a node voltage of the switched capacitance network crosses a first predetermined level; determines when the node voltage of the switched capacitance network crosses a second predetermined level; and provides a sample of an output voltage of the switched capacitance network when it is determined that the node voltage of the switched capacitance network crosses the second predetermined level.

[0014] Another aspect of the present invention is a differential switched-capacitor circuit. The differential switched-capacitor circuit includes a level-crossing detector to generate a level-crossing detection signal when a difference between first and second input signals crosses a predetermined level; a plurality of capacitors; and a sampling switch, operatively coupled to the level-crossing detector. The sampling switch turns OFF when the level-crossing detection signal indicates a level-crossing.

[0015] Another aspect of the present invention is a differential switched-capacitor circuit. The differential switched-capacitor circuit includes a level-crossing detector to generate a level-crossing detection signal when a difference between first and second input signals crosses a first predetermined level; a second level-crossing detector to generate a second level-crossing detection signal when a difference between first and second input signals crosses a second predetermined level; a plurality of capacitors; and a sampling switch, operatively coupled to the level-crossing detector. The sampling switch turns OFF when the second level-crossing detection signal indicates the difference between first and second input signals crosses the second predetermined level.

[0016] Another aspect of the present invention is a differential switched-capacitor circuit. The differential switched-capacitor circuit includes a first switched capacitance network having an input terminal to receive a first input voltage; a first switched capacitance network having an input terminal to receive a second input voltage; a level-crossing detector, having an output terminal and an input terminal, to generate a level-crossing detection signal when a difference between the first and second input signals crosses a predetermined level; and a third switched capacitance network operatively coupled to the level-crossing detector. The third switched capacitance network turns OFF when the difference between the first and second input signals crosses the predetermined level.

[0017] Another aspect of the present invention is a method for sampling an analog signal. The method samples input voltages; determines when a difference between first and second signals crosses a predetermined level; and provides output voltage samples when it is determined that the difference between the first and second signals crosses the predetermined level.

[0018] Another aspect of the present invention is a method for sampling an analog signal. The method samples input voltages; determines when a difference between first and second signals crosses a first predetermined level; determines when a difference between the first and second signals crosses a second predetermined level; and provides a sample of the output voltage when it is determined that the difference between the first and second signals crosses the second predetermined level.

[0042] The present invention will be described in connection with preferred embodiments; however, it will be understood that there is no intent to limit the present invention to the embodiments described herein. On the contrary, the intent is to cover all alternatives, modifications, and equivalents as may be included within the spirit and scope of the present invention, as defined by the appended claims.

[0043] For a general understanding of the present invention, reference is made to the drawings. In the drawings, like reference have been used throughout to designate identical or equivalent elements. It is also noted that the various drawings illustrating the present invention may not have been drawn to scale and that certain regions may have been purposely drawn disproportionately so that the features and concepts of the present invention could be properly illustrated.

[0044] It is noted that, in the various Figures, the earth symbol indicates the system's common-mode voltage. For example, in a system with 2.5 V and −2.5 V power supplies, the system's common-mode voltage may be at ground. In a system with a single 2.5 power supply, the system's common-mode voltage may be at 1.25 V.

[0045] As noted above, accurate output voltage can be obtained if Node 100 in FIG. 2 is maintained precisely at ground. However, in sampled-data circuits, the only point of time accurate output voltage is required is at the instant the output voltage is sampled by another sampling circuit. Thus, it is not necessary to maintain the voltage at Node 100 at ground all the time.

[0046] FIG. 4 illustrates a non-inverting integrator according to the concepts of the present invention. More specifically, as an example, a non-inverting integrator with half-clock delay is illustrated in FIG. 4.

[0047] As illustrated in FIG. 4, a clock phase φ₁ is applied to switches S₁₁, S₁₃, S₂₂, and S₂₄, and another phase φ₂ is applied to switches S₁₂, S₁₄, and S₂₁. A zero crossing detector 30 is used to detect the point of time at which Node 100 crosses ground. The switch S₂₃ is controlled by the output of the zero crossing detector 30. The output of the zero crossing detector 30 is used to determine the time point to take the sample of the output voltage v_out. A waveform generator 20 generates a voltage waveform as the output voltage v_out in such way the voltage at Node 100 crosses zero if the charge in capacitors C_S1 and C_I1 is within a normal operating range.

[0048] In the timing diagram shown in FIG. 5, the waveform generated by the waveform generator 20 is shown as a ramp. When v₁, the voltage at Node 100, crosses zero at time t₁, the output v_zc of the zero crossing detector 30 goes low, turning the switch S₂₃ OFF. At that instant, the output voltage v_out is sampled on C_S2.

[0049] Since v₁ is very close to zero when the sample of v_zc is taken, an accurate output voltage is sampled on C_S2. A similar operation repeats during the next clock cycle, and the sample of the output voltage is taken at time t₂.

[0050] It is noted that the zero crossing detector 30 may optionally have an overflow detection feature that determines when the charge in capacitors C_S1 and C_I1 is outside the normal range of operation. It can be implemented by a logic circuit that makes the output v_zc of the zero-crossing detector 30 go low when φ₂ goes low. In the event v₁ fails to cross zero, the sample is taken on the falling edge of φ₂. At the same time, the logic circuit produces a flag indicating overflow.

[0051] In the embodiment described above and in the various embodiments described below, a zero crossing detector is utilized in lieu of a comparator. Typically, a comparator is designed to compare two arbitrary input voltages. A comparator may be implemented as cascaded amplifiers, a regenerative latch, or a combination of both. A comparator may be used to detect a zero voltage level or a predetermined voltage level crossing.

[0052] It is noted that the input waveform of the various described embodiments is not arbitrary, but deterministic and repetitive. Thus, the various described embodiments determine the instant the zero voltage level or the predetermined voltage level is crossed than relative amplitudes of the input signals. For such a deterministic input, a zero crossing detector is more efficient.

[0053] An example of a zero-crossing detector for the detection of a positive-going input signal is shown in FIG. 1. Initially, node 1 and node 2 are precharged to V_DD and ground, respectively. The ramp input voltage V_IN is applied according to the zero crossing circuit. At the time the input node crosses the threshold, node 1 is discharged rapidly, and node 2 is pulled up to V_DD. Since the zero crossing detector in FIG. 1 is a dynamic circuit, there is no DC power consumption, allowing extremely low power and fast operation. For the detection of a zero-crossing of a negative-going signal, a complementary circuit with a PMOS input transistor can be utilized.

[0054] As illustrated in FIG. 6, the non-inverting integrator includes a waveform generator which is a current source 200. As illustrated in FIG. 6, a clock phase φ₁ is applied to switches S₁₁, S₁₃, S₂₂, and S₂₄, and another phase φ₂ is applied to switches S₁₂, S₁₄, and S₂₁. A zero crossing detector 30 is used to detect the point of time at which Node 100 crosses ground. The switch S₂₃ is controlled by the output of the zero crossing detector 30. The output of the zero crossing detector 30 is used to determine the time point to take the sample of the output voltage v_out.

[0055] The current source 200 charges the capacitors C_S2 and the series connected capacitors C_S1 and C_I1, generating a ramp. At the start of φ₂, the output is briefly shorted to a known voltage V_NEG, the value of which is chosen to ensure the voltage v₁ at Node 100 crosses zero with signals in the normal operating range.

[0056] As illustrated in FIG. 7, the non-inverting integrator includes a waveform generator 20 that produces, preferably, a plurality of segments in the waveform with varying rate of change of the output voltage. The first segment may be controlled so as to have the highest rate of change, with subsequent segments having progressively lower rate of change. The detection of zero crossing by the zero crossing detector 30 causes the waveform to advance to the next segment. An output signal v_zc2 of the zero crossing detector 30 remains high until the zero crossing is detected in the last segment of the waveform.

[0057] One clock cycle of the timing diagram is shown in FIG. 8. At the start of φ₂, the waveform generator 20 produces an up ramp. The voltage v₁ is shown to cross zero at time t₁. One output, v_zc1, of the zero crossing detector 30 changes its state after a finite delay t_d1.

[0058] The delay t_d1 represents finite delay of a typical zero crossing detector 30. This change of state advances the waveform to the next segment.

[0059] Due to the t_d1 of the zero crossing detector 30, the voltage v₁ overshoots by a small amount above ground. The second segment of the waveform generator is a down ramp to permit another zero crossing at time t₂. After a second delay t_d2, the output v_zc2 of the zero crossing detector 30 goes low, causing the switch S₂₃ to turn OFF, locking the sample of the output voltage v_out.

[0060] The delay t_d2 of the second zero crossing is not necessarily the same as the delay associated with the first zero crossing t_d1. The delay t_d2 contributes a small overshoot to the sampled output voltage. The effect of the overshoot can be shown to be constant offset in the sampled charge. In most sampled-data circuits, such constant offset is of little issue.

[0061] The zero crossing detector 30 preferably becomes more accurate in detecting the zero crossing as the segments of the waveform advances. The first detection being a coarse detection, it doesn't have to be very accurate. Therefore, the detection can be made faster with less accuracy. The last zero crossing detection in a given cycle determines the accuracy of the output voltage. For this reason, the last zero crossing detection must be the most accurate.

[0062] The accuracy, speed, and the power consumption can be appropriately traded among progressive zero crossing detections for the optimum overall performance. For example, the first detection is made less accurately and noisier but is made faster (shorter delay) and lower power. The last detection is made more accurately and quieter while consuming more power or being slower (longer delay).

[0063] An example of a two-segment waveform generator constructed of two current sources (210 and 220) is shown in FIG. 9. As illustrated in FIG. 9, a clock phase φ₁ is applied to switches S₁₁, S₁₃, S₂₂, and S₂₄, and another phase φ₂ is applied to switches S₁₂, S₁₄, and S₂₁. A zero crossing detector 30 is used to detect the point of time at which Node 100 crosses ground. The switch S₂₃ is controlled by the output of the zero crossing detector 30. The output of the zero crossing detector 30 is used to determine the time point to take the sample of the output voltage v_out.

[0064] Current sources 210 and 220 charge the capacitors C_S2 and the series connected C_S1 and C_I1 generating two segments of a ramp waveform. At the start of φ₂, the output is briefly shorted to a known voltage V_NEG, the value of which is chosen to ensure the voltage v₁ crosses zero with signals in the normal operating range. During the first segment, the current source 210 is directed to the output, while during the second segment, the current source 220 is directed to the output, generating two different slopes of ramp.

[0065] As illustrated in FIG. 10, the non-inverting integrator includes a level crossing detector 300 having plurality of thresholds. As illustrated in FIG. 10, a clock phase φ₁ is applied to switches S₁₁, S₁₃, S₂₂, and S₂₄, and another phase φ₂ is applied to switches S₁₂, S₁₄, and S₂₁. A level crossing detector 300 is used to detect the point of time at which Node 100 crosses one of plurality of predetermined levels as discussed below. The switch S₂₃ is controlled by the output of the level crossing detector 300. The output of the level crossing detector 300 is used to determine the time point to take the sample of the output voltage v_out.

[0066] The thresholds are predetermined voltage levels. The thresholds of the level crossing detector 300 can be adjusted to minimize overshoot.

[0067] For example, the threshold for the first detection may be made negative by a slightly smaller amount than the expected overshoot in the first segment. This minimizes the ramp-down time in the second segment. Also, the threshold for the second segment may be made more positive by the amount of the overshoot in the second segment in order to cancel the effect of the overshoot. Alternatively, the threshold for the first segment may be made more negative than the expected overshoot during the first segment. This permits the second segment to be a positive ramp rather than a negative ramp as shown in FIG. 11.

[0068] It is advantageous to make the detection during the last segment to be the most accurate detection. The accuracy of the detection during the last segment is made higher than during other segments. This can be achieved by making the delay longer or making the power consumption higher during the last segment.

[0069] As illustrated in FIG. 12, the non-inverting integrator includes a level crossing detector having two zero-crossing detectors, Zero Crossing Detector 1 (310) and Zero Crossing Detector 2 (320). As illustrated in FIG. 12, a clock phase φ₁ is applied to switches S₁₁, S₁₃, S₂₂, and S₂₄, and another phase φ₂ is applied to switches S₁₂, S₁₄, and S₂₁. Zero Crossing Detector 1 (310) and Zero Crossing Detector 2 (320) are used to detect the point of time at which Node 100 crosses one of plurality of predetermined levels as discussed below. The switch S₂₃ is controlled by the output of the Zero Crossing Detector 2 (320). The output of the Zero Crossing Detector 2 (320) is used to determine the time point to take the sample of the output voltage v_out.

[0070] The thresholds of the Zero Crossing Detector 1 (310) and Zero Crossing Detector 2 (320) are selected to minimize overshoot. For example, the threshold for Zero Crossing Detector 1 (310) may be made negative by a slightly smaller amount than the expected overshoot in the first segment. This minimizes the ramp-down time in the second segment. Also, the threshold for Zero Crossing Detector 2 (320) may be made more positive by the amount of the overshoot in the second segment in order to cancel the effect of the overshoot. Alternatively, the threshold for Zero Crossing Detector 1 (310) may be made more negative than the expected overshoot during the first segment. This permits Zero Crossing Detector 2 (320) to be a positive ramp rather than a negative ramp.

[0071] In other words, Zero Crossing Detector 1 (310) makes a coarse detection, whereas Zero Crossing Detector 2 (320) makes a fine detection. Thus, it is advantageous to make Zero Crossing Detector 2 (320) to have a higher accuracy.

[0072] As illustrated in FIG. 13, the non-inverting integrator includes a level crossing detector having two zero-crossing detectors, Zero Crossing Detector 1 (310) and Zero Crossing Detector 2 (320). As illustrated in FIG. 13, a clock phase φ₁ is applied to switches S₁₁, S₁₃, S₂₂, and S₂₄, and another phase φ₂ is applied to switches S₁₂, S₁₄, and S₂₁. Zero Crossing Detector 1 (310) and Zero Crossing Detector 2 (320) are used to detect the point of time at which Node 100 crosses one of plurality of predetermined levels as discussed below. The switch S₂₃ is controlled by the output of the Zero Crossing Detector 2 (320). The output of the Zero Crossing Detector 2 (320) is used to determine the time point to take the sample of the output voltage v_out.

[0073] Both detectors, Zero Crossing Detector 1 (310) and Zero Crossing Detector 2 (320), have nominally zero thresholds. The detection thresholds are determined by voltages V_tr1 and V_tr2 applied to the inputs of Zero Crossing Detector 1 (310) and Zero Crossing Detector 2 (320), respectively. Zero Crossing Detector 1 (310) makes a coarse detection, whereas Zero Crossing Detector 2 (320) makes a fine detection. Thus, it is advantageous to make Zero Crossing Detector 2 (320) to have a higher accuracy.

[0074] It is noted that the above-described embodiment may operate as a self-timed system. In this configuration, Rather than supplying constant frequency clock phases φ₁ and φ₂, the clock phases are derived from the outputs of Zero Crossing Detector 1 (310) and Zero Crossing Detector 2 (320). FIG. 14 illustrates a self-timed operation.

[0075] As illustrated in FIG. 14, the end of the phase φ₂ is defined by the output of the detection during the last segment. The beginning of the clock phase φ₁ is defined by a short delay, such as logic delays, after the end of φ₂ The short delay is generally necessary to ensure non-overlapping clock phases. The end of the clock phase φ₁ is determined by the zero crossing detection of the previous stage or the following stage in the similar manner.

[0076] It is noted that the various embodiments described above can be utilized in a pipeline analog-to-digital converter, an algorithmic analog-to-digital converter, a switched-capacitor amplifier, a delta-sigma modulator, or a self-timed algorithmic analog-to-digital converter.

[0077] It is further noted that the various embodiments described above have signal paths that are single-ended, thus, it is desirable to provide differential signal paths. The various embodiments described below provide differential signal paths.

[0078] FIG. 15 illustrates another example of a non-inverting integrator with half-clock delay shown. In FIG. 15, there are two signal paths, the true path and the complementary path. The true signal path comprises capacitors (C_S1p, C_I1p, and C_S2p) and switches (S_11p, S_12p, S_13p, S_14p, S_21p, S_22p, S_23p, and S_24p). The complementary signal path comprises capacitors (C_S1n, C_I1n, and C_S2n) and switches (S_11n, S_12n, S_13n, S_14n, S_21n, S_22n, S_23n, and S_24n). A clock phase φ₁ is applied to switches S_11p, S_13p, S_22p, S_24p, S_11n, S_13n, S_22n, and S_24n, and another clock phase φ₂ is applied to switches S_12p, S_14p, S_21p, S_12n, S_14n, and S_21n. A zero crossing detector 300 is used to detect the point of time at which Node 110 and Node 120 voltages cross each other. The sampling switches S_23p and S_23n are controlled by the output of the zero crossing detector 300. The output of the zero crossing detector 300 is used to determine the time point to take the samples of the output voltages v_outp and v_outn

[0079] The waveform generator 20 generates a differential voltage waveform at the output nodes (130 and 140) in such way the difference of voltages at Node 110 and Node 120 (v_1p−v_1n) cross zero if the charge in capacitors C_S1p, C_I1p, C_S1n, and C_I1n is within a normal operating range. In the timing diagram shown in FIG. 16, the waveforms for v_outp and v_outn generated by the waveform generator 20 are shown as a positive going ramp and a negative going ramp, respectively.

[0080] The waveform generator 20 optionally includes a common-mode-feedback circuit to keep the common-mode output voltage (v_1p+v_1n)/2 substantially constant. When v_1p−v_1n crosses zero at time t₁, the output v_zc of the zero crossing detector 300 goes low, turning the switches S_23p and S_23n OFF. At that instant, the output voltages v_outp and v_outn are sampled on C_S2p and C_S2n, respectively. Since v_1p−v_1n is very close to zero when the sample of v_outp and v_outn are taken, accurate output voltages are sampled on C_S2p and C_S2p. The operations are repeated for the next clock cycle, and the samples of the output voltages are taken at time t₂.

[0081] The zero crossing detector 300 optionally has an overflow detection feature that determines the charge in capacitors C_S1p, C_I1p, C_S1n, and C_I1n is outside the normal range of operation. It can be implemented by a logic circuit that makes the output v_zc of the zero-crossing detector 300 to go low when φ₂ goes low.

[0082] In the event v₁ fails to cross zero, the sample is taken on the falling edge of φ₂. At the same time, the logic circuit produces a flag indicating overflow.

[0083] FIG. 17 illustrates another example of a non-inverting integrator with half-clock delay shown. In FIG. 17, there are two signal paths, the true path and the complementary path. The true signal path comprises capacitors (C_S1p, C_I1p, and C_S2p) and switches (S_11p, S_12p, S_13p, S_14p, S_21p, S_22p, S_23p, and S_24p). The complementary signal path comprises capacitors (C_S1n, C_I1n, and C_S2n) and switches (S_11n, S_12n, S_13n, S_14n, S_21n, S_22n, S_23n, and S_24n). A clock phase φ₁ is applied to switches S_11p, S_13p, S_22p, S_24p, S_11n, S_13n, S_22n, and S_24n, and another clock phase φ₂ is applied to switches S_12p, S_14p, S_21p, S_12n, S_14n, and S_21n. A zero crossing detector 300 is used to detect the point of time at which Node 110 and Node 120 voltages cross each other. The sampling switches S_23p and S_23n are controlled by the output of the zero crossing detector 300. The output of the zero crossing detector 300 is used to determine the time point to take the samples of the output voltages v_outp and v_outn.

[0084] The waveform generator 20 generates a differential voltage waveform at the output nodes (130 and 140) in such way the difference of voltages at Node 110 and Node 120 (v_1p−v_1n) cross zero if the charge in capacitors C_S1p, C_I1p, C_S1n, and C_I1n is within a normal operating range.

[0085] The waveform generator may comprise a pair of current sources (210 and 220), as shown in FIG. 17. The current source 220 charges the capacitors C_S2p and the series connected capacitors C_S1p and C_I1p, generating a positive going ramp. The current source 210 charges the capacitors C_S2n and the series connected capacitors C_S1n and C_I1n, generating a negative going ramp. The currents 210 and 220 are nominally equal in magnitude and opposite in sign. At the start of φ₂, the outputs v_outp and v_outn are briefly shorted to a known voltage V_NEG and V_POS, respectively, the values of which are chosen to ensure the voltage v_1p−v_1n crosses zero with signals in the normal operating range.

[0086] FIG. 18 illustrates another example of a non-inverting integrator with half-clock delay shown. In FIG. 18, there are two signal paths, the true path and the complementary path. The true signal path comprises capacitors (C_S1p, C_I1p, and C_S2p) and switches (S_11p, S_12p, S_13p, S_14p, S_21p, S_22p, S_23p, and S_24p). The complementary signal path comprises capacitors (C_S1n, C_I1n, and C_S2n) and switches (S_11n, S_12n, S_13n, S_14n, S_21n, S_22n, S_23n, and S_24n). A clock phase φ₁ is applied to switches S_11p, S_13p, S_22p, S_24p, S_11n, S_13n, S_22n, and S_24n, and another clock phase φ₂ is applied to switches S_12p, S_14p, S_21p, S_12n, S_14n, and S_21n. A zero crossing detector 300 is used to detect the point of time at which Node 110 and Node 120 voltages cross each other. The sampling switches S_23p and S_23n are controlled by the output of the zero crossing detector 300. The output of the zero crossing detector 300 is used to determine the time point to take the samples of the output voltages v_outp and v_outn.

[0087] The waveform generator 20 generates a differential voltage waveform at the output nodes (130 and 140) in such way the difference of voltages at Node 110 and Node 120 (v_1p−v_1n) cross zero if the charge in capacitors C_S1p, C_I1p, C_S1n, and C_I1n is within a normal operating range.

[0088] In FIG. 18, the waveform generator 20 may produce a plurality of segments in the waveforms with varying rate of change of the output voltage. For optimum operation, the first segment may be controlled to have the highest rate of change, with subsequent segments having progressively lower rate of change. The waveform generator 20 may optionally include a common-mode-feedback circuit to keep the common-mode output voltage (v_1p+v_1n)/2 substantially constant. The detection of zero crossing by the zero crossing detector 300 causes the waveform to advance to the next segment. An output signal v_zc2 of the zero crossing detector 300 remains high until the zero crossing is detected in the last segment of the waveform.

[0089] One clock cycle of the timing diagram for the multi-segment circuit of FIG. 18 is shown in FIG. 19. At the start of φ₂, the waveform generator 20 produces an up ramp for v_outp and a down ramp for v_outn. The voltage v_1p−v_1n is shown to cross zero at time t₁. The output, v_zc1, of the zero crossing detector 300 changes its state after a finite delay t_d1. The delay t_d1 represents the finite delay associated with the zero crossing detector 300. This change of state advances the waveform to the next segment.

[0090] Due to the t_d1 of the zero crossing detector 300, the voltage v_1p−v_1n overshoots by a small amount above zero. The second segment of the waveform generator 20 is a down ramp for v_outp and an up ramp for v_outn to permit another zero crossing at time t₂. After a second delay t_d2, the output v_zc2 of the zero crossing detector 300 goes low, causing the switch S₂₃ to turn OFF, locking the samples of the output voltages v_outp and v_outn.

[0091] The delay t_d2 of the second zero crossing may not be the same as the delay associated with the first zero crossing t_d1. The delay t_d2 contributes a small overshoot to the sampled output voltage. The effect of the overshoot can be shown to be constant offset in the sampled charge. In most sampled-data circuits, such constant offset is of little issue.

[0092] The zero crossing detector 300 preferably becomes more accurate in detecting the zero crossing as the segments of the waveform advances. The first detection being a coarse detection, it doesn't have to be very accurate. Therefore, the detection can be made faster with less accuracy. The last zero crossing detection in a given cycle determines the accuracy of the output voltage. For this reason, the last zero crossing detection must be the most accurate.

[0093] The accuracy, speed, and the power consumption can be appropriately traded among progressive zero crossing detections for the optimum overall performance. For example, the first detection is made less accurately and noisier but is made faster (shorter delay) and lower power. The last detection is made more accurately and quieter while consuming more power or being slower (longer delay).

[0094] FIG. 20 illustrates another example of a non-inverting integrator with half-clock delay shown. In FIG. 20, there are two signal paths, the true path and the complementary path. The true signal path comprises capacitors (C_S1p, C_I1p, and C_S2p) and switches (S_11p, S_12p, S_13p, S_14p, S_21p, S_22p, S_23p, and S_24p). The complementary signal path comprises capacitors (C_S1n, C_I1n, and C_S2n) and switches (S_11n, S_12n, S_13n, S_14n, S_21n, S_22n, S_23n, and S_24n). A clock phase φ₁ is applied to switches S_11p, S_13p, S_22p, S_24p, S_11n, S_13n, S_22n, and S_24n, and another clock phase φ₂ is applied to switches S_12p, S_14p, S_21p, S_12n, S_14n, and S_21n. A zero crossing detector 300 is used to detect the point of time at which Node 110 and Node 120 voltages cross each other. The sampling switches S_23p and S_23n are controlled by the output of the zero crossing detector 300. The output of the zero crossing detector 300 is used to determine the time point to take the samples of the output voltages v_outp and v_outn.

[0095] The waveform generator 20 generates a differential voltage waveform at the output nodes (130 and 140) in such way the difference of voltages at Node 110 and Node 120 (v_1p−v_1n) cross zero if the charge in capacitors C_S1p, C_I1p, C_S1n, and C_I1n is within a normal operating range.

[0096] In FIG. 20, the waveform generator may comprise plurality of current sources. Current sources 210 and 220 charge the capacitors generating two segments of a ramp waveform on the true output v_outp. Current sources 230 and 240 charge the capacitors generating two segments of a ramp waveform on the complementary output v_outn.

[0097] At the start of φ₂, the outputs v_outp and v_outn are briefly shorted to known voltages V_NEG and V_POS, respectively, the values of which are chosen to ensure the voltage v_1p−v_1n crosses zero with signals in the normal operating range.

[0098] During the first segment, the current sources 210 and 230 are directed to the outputs v_outp and v_outn, respectively, while during the second segment, the current source 220 and 240 are directed to the outputs generating two different slopes of ramp. This is accomplished by throwing switches S_16p, S_17p, S_16n, and S_17n appropriately controlled by the zero-crossing detector 300.

[0099] FIG. 21 illustrates another example of a non-inverting integrator with half-clock delay shown. In FIG. 21, there are two signal paths, the true path and the complementary path. The true signal path comprises capacitors (C_S1p, C_I1p, and C_S2p) and switches (S_11p, S_12p, S_13p, S_14p, S_21p, S_22p, S_23p, and S_24p). The complementary signal path comprises capacitors (C_S1n, C_I1n, and C_S2n) and switches (S_11n, S_12n, S_13n, S_14n, S_21n, S_22n, S_23n, and S_24n). A clock phase φ₁ is applied to switches S_11p, S_13p, S_22p, S_24p, S_11n, S_13n, S_22n, and S_24n, and another clock phase φ₂ is applied to switches S_12p, S_14p, S_21p, S_12n, S_14n, and S_21n. A zero crossing detector 300 is used to detect the point of time at which Node 110 and Node 120 voltages cross each other. The sampling switches S_23p and S_23n are controlled by the output of the zero crossing detector 300. The output of the zero crossing detector 300 is used to determine the time point to take the samples of the output voltages v_outp and v_outn.

[0100] FIG. 21 further illustrates a common-mode feedback circuit 400. The common-mode feedback circuit is similar as those in conventional fully-differential operational amplifiers. The common-mode amplifier 400 amplifies the difference between the output common-mode voltage (v_outp+v_outn)/2 and the desired output common-mode voltage. The output of the common-mode amplifier 400 provides negative feedback to controls the current sources 210 and 220 to keep the output common-mode voltage constant.

[0101] Alternatively, the output of the common-mode amplifier 400 may control the current sources 230 and 240. The common-mode feedback can be engaged during all or any of the segments. It is preferred that the common-mode feedback be engaged during the first segment only while keeping current source 220 constant and matched to current source 240.

[0102] FIG. 22 illustrates another example of a non-inverting integrator with half-clock delay shown. In FIG. 22, there are two signal paths, the true path and the complementary path. The true signal path comprises capacitors (C_S1p, C_I1p, and C_S2p) and switches (S_11p, S_12p, S_13p, S_14p, S_21p, S_22p, S_23p, and S_24p). The complementary signal path comprises capacitors (C_S1n, C_I1n, and C_S2n) and switches (S_11n, S_12n, S_13n, S_14n, S_21n, S_22n, S_23n, and S_24n). A clock phase φ₁ is applied to switches S_11p, S_13p, S_22p, S_24p, S_11n, S_13n, S_22n, and S_24n, and another clock phase φ₂ is applied to switches S_12p, S_14p, S_21p, S_12n, S_14n, and S_21n. A zero crossing detector 300 is used to detect the point of time at which Node 110 and Node 120 voltages cross each other. The sampling switches S_23p and S_23n are controlled by the output of the zero crossing detector 300. The output of the zero crossing detector 300 is used to determine the time point to take the samples of the output voltages v_outp and v_outn.

[0103] In FIG. 22, capacitors C_CMp and C_CMn produce the common-mode voltage v_oc of v_outp and v_outn at the input of the common mode amplifier 10. MOS transistors M₁ and M₂ function as current sources to produce currents I_1p and I_2p, respectively. The output of the common-mode amplifier 10 controls the gates of the transistors M₁ and M₂. The resulting negative feedback forces the common-mode output voltage v_oc to the desired common-mode voltage.

[0104] Alternatively, the output of the common-mode amplifier 10 may control the current sources which produce currents I_1n and I_2n, respectively. The common-mode feedback can be engaged during all or any of the segments. It is preferred that the common-mode feedback be engaged during the first segment only while keeping current I_2p constant and matched to current I_2n.

[0105] While various examples and embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that the spirit and scope of the present invention are not limited to the specific description and drawings herein, but extend to various modifications and changes.