Duty-cycle correction circuit

[0001] This application claims the benefit of the filing date of U.S. provisional application No. 60/571,733, filed on May 17, 2004, the teachings of which are incorporated herein by reference.

[0002] 1. Field of the Invention

[0003] The present invention relates to electronics and, in particular, to clock-signal generating circuits.

[0004] 2. Description of the Related Art

[0005] FIG. 1 shows a block diagram of a phase-locked loop (PLL) 100 of the prior art. PLL 100 includes a phase detector 102, a loop filter 104, a voltage-controlled oscillator (VCO) 106, and a feedback path having a frequency divider 108. A periodic reference signal 110 of frequency F_ref is fed to phase detector 102 together with feedback signal 112 (the output of frequency divider 108). The output of phase detector 102 is a pulse that is related to the phase difference between reference signal 110 and feedback signal 112. The output of phase detector 102 is filtered through loop filter 104 and fed to VCO 106. Due to the feedback in the PLL, the frequency F_out of output signal 114 of VCO 106 is driven to equal the reference frequency F_ref multiplied by the division factor of frequency divider 108, thereby resulting in a relatively high frequency of the output signal. Output signal 114 is typically a differential signal having two differential components.

[0006] One application of PLL 100 is to provide a reference clock signal, e.g., for a microprocessor or communication circuit. For this particular application, it is often desirable to design PLL 100 such that output signal 114 has a 50% duty cycle. In a 50% duty-cycle clock signal, the time interval between a positive transition edge and a negative transition edge is equal to the time interval between that negative transition edge and the next positive transition edge. In other words, all transition edges, regardless of the transition-edge direction, are equally spaced, which enables a half-rate (double-edge) clocking system. In contrast, in a clock signal having a duty cycle different from 50%, the time interval between a positive transition edge and a negative transition edge is different from the time interval between the negative transition edge and the next positive transition edge. Consequently, only one transition edge per cycle can be utilized as a valid reference point, which is known as a full-rate (single-edged) clocking system. Advantageously, a half-rate clocking system eases circuit design constraints compared to those of a full-rate clocking system because it effectively doubles the clock rate without having to double the corresponding VCO frequency.

[0007] PLL 100 is usually incorporated into a relatively large integrated circuit (chip) and manufactured using a suitable fabrication process, e.g., CMOS. However, technological limitations of the fabrication process often cause the duty cycle of PLL 100 to vary from chip to chip and deviate from the intended 50%. Due to relatively strict tolerances of certain half-rate clocking systems, a significant percentage of manufactured chips falls outside the acceptable duty-cycle range and has to be discarded.

[0008] Problems in the prior art are addressed, in accordance with the principles of the present invention, by a duty-cycle correction (DCC) circuit adapted to adjust the duty cycle of a differential clock signal to conform it to the requirements of a half-rate clocking system. In a representative embodiment, the DCC circuit has a buffer circuit adapted to generate a differential output clock signal by adding offset voltage to a differential input clock signal. A feedback loop coupled to the buffer circuit processes the output clock signal to evaluate deviation of its duty-cycle value from 50% and, based on the evaluation, configures the buffer circuit to adjust the offset voltage such that the duty-cycle deviation is reduced. The feedback loop and the buffer circuit are controlled by a duty-cycle calibration engine, e.g., a digital logic circuit adapted to determine an appropriate value for the offset voltage, which causes the duty-cycle value in the output clock signal to be substantially 50% regardless of the duty-cycle value in the input clock signal. As a result, technological limitations in the circuit-fabrication process do not significantly reduce the yield of chips for half-rate clocking systems.

[0019] Reference herein to “one embodiment” or “an embodiment” means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the invention. The appearances of the phrase “in one embodiment” in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.

[0020] FIG. 2 graphically shows representative differential output signal 114 of PLL 100 (FIG. 1). Signal 114 is characterized by a period T₀ and has two components 114p and 114n. Within each cycle of signal 114, signal 114p is greater than signal 114n for a time interval T_p, and signal 114n is greater than signal 114p for a time interval T_n, where T_p+T_n=T₀. The duty cycle of signal 114 is determined, e.g., by the value of T_p/T₀, and is 50% when T_p/T₀=T_n/T₀=0.5. In signal 114 shown in FIG. 2, the duty cycle is about 62% and the deviation of the duty cycle from 50% is about 95 ps. As explained above, the deviation may be caused by imperfections in the fabrication process.

[0021] FIG. 3 shows a block diagram of a clock circuit 300 according to one embodiment of the present invention. Clock circuit 300 has a duty-cycle correction (DCC) circuit 302 coupled to the output of PLL 100 of FIG. 1. DCC circuit 302 receives differential signal 114, e.g., that shown in FIG. 2, and generates a differential signal 314 having a substantially 50% duty cycle, which can then be used in a half-rate clocking system. DCC circuit 302 is designed such that it is capable of correcting the duty cycle of signal 114 falling within a relatively wide range, e.g., from 30% to 70%. As a result, technological limitations in the fabrication process of clock circuit 300 have insignificant adverse effect on the duty cycle of signal 314.

[0022] FIG. 4 graphically shows a representative differential signal 314 generated by DCC circuit 302 based on differential signal 114 shown in FIG. 2. Similar to signal 114, signal 314 is characterized by the same period T₀ and has two components 314p and 314n. However, in signal 314, the values of T_p and T_n are adjusted to become T′_p and T′_n, respectively, such that T_p′/T₀≈T_n′/T₀≈0.5. Consequently, the duty cycle of signal 314 is substantially 50%.

[0023] In one embodiment, DCC circuit 302 performs duty-cycle correction as follows. Suppose that, for signal 114 received by DCC circuit 302, T_p>T_n as illustrated in FIG. 2. DCC circuit 302 then generates signals 314n and 314p by adding an offset voltage between signals 114n and 114p and transferring the offset signals to the output. As can be seen in FIG. 2, a positive offset (i.e., an upward shift) of signal 114n with respect to signal 114p will cause an increase in the value of T_n while causing a corresponding decrease in the value of T_p. DCC circuit 302 selects the offset voltage such that the adjusted values T′_p and T′_n become substantially equal to each other as shown in FIG. 4. Similarly, when T_p<T_n, DCC circuit 302 negatively offsets (i.e., shifts downward) signal 114n with respect to signal 114p until T′_p≅T′_n.

[0024] FIG. 5 shows a block diagram of a DCC circuit 502 that can be used as DCC circuit 302 according to one embodiment of the present invention. DCC circuit 502 has a correction buffer 510 and a feedback loop 520. Based on a control signal 530 generated by feedback loop 520, correction buffer 510 offsets the components of a differential input signal, e.g., of differential signal 114 applied to DCC circuit 502, with respect to each other to generate a differential output signal 514 having a substantially 50% duty cycle. A portion of output signal 514 is processed by feedback loop 520, which then configures correction buffer 510, via control signal 530, to reduce possible deviations from 50% in the duty cycle of the output signal. In one embodiment, control signal 530 is a digital control signal representing, in a digital format, the offset voltage added to differential signal 114 by correction buffer 510.

[0025] In one embodiment, DCC circuit 502 is adapted to operate in two different operational modes, a calibration mode and a tracking mode. In the calibration mode, DCC circuit 502 performs an initial calibration of differential signal 114, e.g., at the circuit power-up, and then locks the offset voltage value determined during that initial calibration for use, without any further changes, during the subsequent circuit operation, e.g., until the circuit power-down. This mode is useful, for example, when differential signal 114 is relatively stable, i.e., has a relatively stable value of its duty cycle. In contrast, in the tracking mode, DCC circuit 502 performs an initial calibration of differential signal 114, e.g., similar to that performed during the calibration mode, but does not lock the offset voltage value determined during the initial calibration. Instead, DCC circuit 502 continues to track differential signal 514 for possible deviations of its duty cycle from 50%. Based on the tracking, control signal 530 and, thus the offset voltage value, are continuously (or intermittently/periodically) adjusted to maintain the duty-cycle value of differential signal 514 at substantially 50%. The tracking mode is useful, for example, when the duty cycle of differential signal 114 is subject to fluctuations, e.g., due to a thermal drift in PLL 100.

[0026] Feedback loop 520 includes a voltage-to-current (V/I) converter 522, a signal filter 524, a slicer 526, and a duty-cycle calibration engine 528, a functional description of each of which is provided below in the context of FIG. 6. In a preferred embodiment, V/I converter 522, signal filter 524, and slicer 526 are implemented using analog circuitry, while calibration engine 528 is implemented in a digital logic circuit adapted to control the operations of both feedback loop 520 and correction buffer 510. More specifically, calibration engine 528 receives a signal 538 from slicer 526 and generates control signals 530 and 536. Signal 538 provides a measure of duty-cycle deviation from 50% in differential signal 514 and control signal 530 configures correction buffer 510 to change the offset voltage such that the duty-cycle deviation from 50% in differential signal 514 is reduced. Control signal 536 controls operations of signal filter 524 and slicer 526 as described in more detail below.

[0027] FIGS. 6A–F graphically illustrate the principles of operation of feedback loop 520 (FIG. 5) according to one embodiment of the present invention. More specifically, FIGS. 6A–C show various signals in feedback loop 520 at the beginning of a duty-cycle correction process, and FIGS. 6D–F show those various signals after the duty cycle has been corrected.

[0028] FIGS. 6A–B graphically show the initial input and output signals of V/I converter 522, e.g., at the circuit power-up. The offset voltage corresponding to FIGS. 6A–B is substantially zero and, consequently, signals 514p and 514n are substantially the same as signals 114p and 114n, respectively. As a result, T_p≠T_n and the duty cycle of signal 514 deviates from 50%. Based on signals 514p and 514n, V/I converter 522 generates a signal 532 shown in FIG. 6B. Signal 532 is a square-wave alternating current signal, which switches between a positive rail +i₀ and a negative rail −i₀. More specifically, signal 532 switches polarity when the difference between signals 514p and 514n changes sign. For example, when signal 514p is greater than signal 514n, signal 532 is at the positive rail. Similarly, when signal 514n is greater than signal 514p, signal 532 is at the negative rail.

[0029] Signal filter 524 processes signal 532 as follows. Signal filter 524 includes an integrator (not shown in FIG. 5) adapted to generate an output signal 534 (see FIG. 5) by integrating signal 532. As seen in FIG. 6C, signal 534 is an alternating ramp signal, which ramps up or down as signal 532 alternates between the positive and negative rails, respectively. Before the integration begins, signal filter 524 is reset by control signal 536 such that the value of signal 534 is zero. After the integration begins, signal 534 goes up and down as indicated in FIG. 6C. However, because T_p≠T_n, at the end of each cycle, signal 534 accumulates an increment of ΔV=i₀(T_p−T_n). In the particular situation illustrated in FIGS. 6A–C, T_p>T_n, the value of ΔV is positive, and signal 534 on average ramps up. Similarly, when T_p<T_n, signal 534 will on average ramp down. Control signal 536 (FIG. 5) sets the length of the integration period in signal filter 524. In general, if the value of ΔV is relatively small, a relatively long integration period is typically needed for a sufficiently large cumulative increment to accrue. At the end of the integration period, control signal 536 configures slicer 526 to evaluate signal 534 and provide the evaluation result to calibration engine 528 via signal 538. The evaluation result may, for example, be the polarity of signal 534. Based on signal 538, calibration engine 528 generates control signal 530, which configures correction buffer 510 (FIG. 5) to add a corresponding offset voltage to signal 114.

[0030] FIGS. 6D–F graphically show the operation of feedback loop 520 after a proper offset voltage V₀ has been introduced into differential signal 114. More specifically, the dashed traces in FIGS. 6D–F indicate the initial signals in feedback loop 520 shown in FIGS. 6A–C, respectively, and the solid traces in FIGS. 6D–F show the updated signals in the feedback loop after the proper offset voltage has been added. Referring to FIG. 6D, signals 514p and 514n are shifted with respect to each other by offset voltage V₀, which results in new values for T_p and T_n labeled T′_p and T′_n, respectively, where T′_p is substantially equal to T′_n. Referring to FIG. 6E, due to the change in the values of T_p and T_n shown in FIG. 6D, time for signal 532 at the positive rail decreases compared to that shown in FIG. 6B while time at the negative rail increases by the corresponding amount. As a result, ΔV is substantially zero and signal 534 on average does not ramp up or down (see FIG. 6F). Consequently, calibration engine 528 generates control signal 530, which configures correction buffer 510 to keep the offset voltage at V₀. As can be appreciated by one skilled in the art, signal 514 shown in FIG. 6D has the desired duty-cycle value of substantially 50%.

[0031] FIG. 7 shows a flowchart of a method 700 that is used to operate calibration engine 528 according to one embodiment of the present invention. More specifically, method 700 corresponds to a tracking mode of operation of DCC circuit 502. In step 702, when DCC circuit 502 is turned on, calibration engine 528 resets correction buffer 510 such that the offset voltage is zero. In step 704, calibration engine 528 configures signal filter 524 to set signal 534 to zero and to start integration of signal 532. Calibration engine 528 preferably proceeds to step 706 after allowing signal filter 524 to integrate signal 532 for N signal cycles, where N>1. In a representative configuration, N=256. In step 706, calibration engine 528 configures slicer 526 to slice signal 534 to determine the sign of ΔV (see also FIG. 6C) and to provide that information to the calibration engine via signal 538. In one implementation, signal 538 is a binary signal generated as follows. When the value of signal 532 is greater than or equal to zero, signal 538 is a binary “one”. Similarly, when the value of signal 532 is less than zero, signal 538 is a binary “zero”. In step 708, calibration engine 528, via control signal 530, configures correction buffer 510 to increment the offset voltage by an increment amount, v_inc, in the direction corresponding to the sign of ΔV determined in step 706. For example, when signal 538 is a binary “one”, control signal 530 configures correction buffer 510 to increment the offset voltage by +v_inc, thereby reducing the duty cycle of signal 514 by a corresponding amount (see also FIG. 6D). Similarly, when signal 538 is a binary “zero”, control signal 530 configures correction buffer 510 to increment the offset voltage by −v_inc, thereby increasing the duty cycle of signal 514. Preferably, the value of v_inc is relatively small, e.g., 2 mV, to permit fine-tuning of the duty cycle. After step 708, the processing of method 700 returns to step 704. Steps 704–708 form a processing loop 710, which is used to dynamically control the duty cycle of signal 514.

[0032] FIG. 8 graphically shows signals 534 and 530 when DCC circuit 502 is configured to operate in accordance with method 700. More specifically, FIG. 8 shows a time interval during which calibration engine 528 cycles through processing loop 710 of method 700 about 170 times. About 60 initial processing loops 710 correspond to the initial conversion of DCC circuit 502 toward the desired duty-cycle value of 50% from an initial duty-cycle value that has a relatively large positive deviation from 50%. Until such conversion is achieved at time t_c, the accumulated increment at the end of each processing loop 710 (i.e., step 708) in signal 534 is relatively large (see the top panel of FIG. 8). Consequently, the value of signal 530 (see the bottom panel of FIG. 8) at the end of each processing loop 710 is such that correction buffer 510 keeps adding positive increments to the offset voltage, thereby decreasing the duty-cycle value in signal 514 and eventually decreasing the accumulated increment in signal 534 as indicated in the top panel of FIG. 8. At time t_c, the offset voltage becomes large enough to cause the duty-cycle value to cross the 50% mark, i.e., to become less than 50%. As a result, calibration engine 528 changes the value of signal 530 (see the bottom panel of FIG. 8) to configure correction buffer 510 to subtract an increment amount from the offset voltage, thereby increasing the duty-cycle value in signal 514 and causing it to become again greater than 50%. After that point in time, the value of control signal 530 begins to alternate between the two values (see the bottom panel of FIG. 8) and, as a result, the offset voltage begins to alternate between two levels, e.g., V₊ and V₋, where |V₊−V₋|=v_inc. offset voltage has a value of V₊, the duty-cycle value in signal 514 is slightly larger than 50%. Similarly, when the offset voltage has a value of V₋, the duty-cycle value in signal 514 is slightly smaller than 50%. When v_inc has an appropriately small value, deviations of the duty-cycle in signal 514 from 50% are relatively small, e.g., such that signal 514 is suitable for use in a half-rate clocking system.

[0033] FIG. 9 shows a diagram of a DCC circuit 902 that can be used in the clock circuit of FIG. 3 according to another embodiment of the present invention. DCC circuit 902 is analogous to DCC circuit 502 of FIG. 5 and includes a correction buffer 910, a V/I converter 922, a signal filter 924, a slicer 926, and a duty-cycle calibration engine 928. Operation of each of these elements is described in more detail below.

[0034] Correction buffer 910 has a differential transistor pair T1 coupled to a voltage-offset generator (VOG) G1. Differential transistor pair T1 is configured to substantially repeat a differential input signal labeled CLOCK-IN, while VOG G1 is configured to shift the components of the repeated signal with respect to one another to produce a differential output signal labeled CLOCK-OUT. VOG G1 has a pair of current sources labeled IP and IM, which are controlled by a control signal 930 provided by calibration engine 928 at a port labeled DCAL(6:0). Control signal 930 is a 7-bit digital signal, whose value (d) determines the currents, I_p and I_m, generated by current sources IP and IM, respectively, in accordance with Eq. (1) as follows:

[0035] $\begin{array}{cc}{I}_p=I_0+\frac{d}{2}{i}_{\mathrm{inc}}& \left(1a\right)\\ I_m=I_0-\frac{d}{2}{i}_{\mathrm{inc}}& \left(1b\right)\end{array}$
where I₀ is a constant and i_inc is an increment amount. A difference between the currents causes correction buffer 910 to generate an offset voltage value of dv_inc, which serves to offset the components of signal CLOCK-OUT, where v_inc=Ri_inc and R is the resistance of each of the two resistors coupled to differential transistor pair T1.

[0036] V/I converter 922 has a differential transistor pair T2, the gates of which are configured to receive signal CLOCK-OUT. The effect of signal CLOCK-OUT on the transistors is such that only one of the transistors is in the “on” state at a time. Consequently, current flowing through each of the transistors alternates between 0 and I_c, where I_c is the current generated by a current source labeled IC.

[0037] Signal filter 924 has two integrators, each having a transistor coupled to two current sources as shown in FIG. 9. In addition, the source of each transistor is coupled to a corresponding transistor of differential transistor pair T2 in V/I converter 922, while the drain of each transistor is coupled to a corresponding one of two capacitors labeled CINTP and CINTM. Each integrator substantially works by accumulating in the corresponding capacitor (CINTP or CINTM) the amount of charge corresponding to the current flowing through that transistor of differential transistor pair T2, to which the integrator is coupled. Calibration engine 928 provides a control signal (not shown in FIG. 9), which resets capacitors CINTP and CINTM at the beginning of each integration period. As a result, at the end of each integration period, the voltage across each of capacitors CINTP and CINTM is proportional to the integral of the current that passed through the corresponding transistor of differential transistor pair T2 during the integration period. A common-mode feedback (CMFB) circuit coupled to the current sources in the integrators of signal filter 924 serves to adjust common-mode voltages in the signal filter to maximize the filter's dynamic range as known in the art. Representative CMFB circuits that can be used in signal filter 924 are disclosed, for example, in U.S. Pat. Nos. 4,533,876, 4,906,943, and 4,933,644, the teachings of all of which are incorporated herein by reference.

[0038] Slicer 926 is a comparator circuit configured to compare the voltages of capacitors CINTP and CINTM at the end of each integration period, which end is indicated by a control signal 936 provided by calibration engine 928 at a port labeled SLICE_CLK. The output of slicer 926 is a binary signal generated, for example, as follows. When the voltage across capacitor CINTP is greater than the voltage across capacitor CINTM, slicer 926 outputs a binary “one”. Similarly, when the voltage across capacitor CINTM is greater than the voltage across capacitor CINTP, slicer 926 outputs a binary “zero”. The output signal generated by slicer 926 is applied to calibration engine 928 at a port labeled VERROR. Based on that signal, the calibration engine adjusts the value of d provided to correction buffer 910 via control signal 930, thereby changing the duty-cycle value in output signal CLOCK-OUT.

[0039] While this invention has been described with reference to illustrative embodiments, this description is not intended to be construed in a limiting sense. For example, circuits of the invention can be used to adjust a clock signal to achieve a duty-cycle value other than 50% for applications other than that in a half-rate clocking system. The clock signal being adjusted may be generated by any appropriate clock-signal generator other than a PLL, e.g., a delay-locked loop (DLL). Various modifications of the described embodiments, as well as other embodiments of the invention, which are apparent to persons skilled in the art to which the invention pertains are deemed to lie within the principle and scope of the invention as expressed in the following claims.

[0040] Although the steps in the following method claims, if any, are recited in a particular sequence with corresponding labeling, unless the claim recitations otherwise imply a particular sequence for implementing some or all of those steps, those steps are not necessarily intended to be limited to being implemented in that particular sequence.

[0041] The present invention may be implemented as circuit-based processes, including possible implementation on a single integrated circuit. As would be apparent to one skilled in the art, various functions of circuit elements may also be implemented as processing steps in a software program. Such software may be employed in, for example, a digital signal processor, micro-controller, or general-purpose computer.