Single-edge clock adjustment circuits for PLL-compatible, dynamic duty-cycle correction circuits

1. Field of the Invention

The invention relates generally to a clock scheme in a digital circuit and, more particularly, to a dynamic duty-cycle correction scheme using a single-edge clock adjustment technique.

2. Description of the Related Art

Some high-speed circuit techniques in digital-circuit design, such as separated-latch design, render chip performance sensitive to clock duty cycle. That is, duty cycle sometimes improves or degrades maximum chip frequency. Whether a particular asymmetry yields an improvement or degradation is a function of the asymmetry type and the critical timing path. The asymmetry type, for example, is determined by whether the clock remains at a logical 1 longer than a logical 0 , or vice versa. If reversible, a particular clock asymmetry that results in chip performance degradation can yield a performance improvement when reversed.

The ideal clock distribution, therefore, should permit fine adjustments for the clock duty cycle via either primary input control or BIOS control bits. Furthermore, these adjustments should be invariant to changes in process, temperature, and supply voltage. For example, settings on a first chip that yield a 10-picosecond adjustment should yield the same 10-picosecond adjustment on a second chip regardless of temperature, process, and supply voltage of the second chip.

Because this degree of invariance is only achievable via active feedback control, another design restriction has to be mentioned. The active feedback loop employed in the duty cycle correction circuit should not interfere with the feedback loop employed in a phase-locked loop (PLL). At best, such interference between the two feedback loops would increase clock jitter. At worst, such interference might render one or both circuits unstable.

Therefore, there is a need for a mechanism that permits fine adjustments of the clock duty cycle, achieves the desired invariance to process, temperature, and supply voltage, and does not interfere with the operation of the PLL.

A duty cycle correction circuit has a delay-control circuit coupled to a clock-inverter circuit. The delay-control circuit includes first and second voltage dividers coupled to a differential amplifier coupled to a stability circuit. The first voltage divider receives a first clock signal. The first and second voltage dividers generate first and second output voltage, respectively. The differential amplifier has first and second input terminals coupled to the first and second output voltages, respectively, and outputs a third output voltage. The stability circuit receives the third output voltage and outputs a delay-control voltage. The clock inverter circuit receives a second clock signal and is coupled to the delay-control voltage. The delay-control voltage has first, second, third, and fourth field effect transistors (FETs).

The first FET has a source, a gate, and a drain. The gate of the first FET is coupled to the delay-control voltage, and the source of the first FET is coupled to the first bias voltage.

The second FET has a source, a gate, and a drain. The source of the second FET is coupled to the first bias voltage. The gate of the second FET receives the second clock signal, and the drain of the second FET outputs the third clock signal.

The third FET has a source, a gate, and a drain. The source of the third FET is coupled to the drain of the first FET, and the gate of the third FET is coupled to the gate of the second FET. The drain of the third FET is coupled to the drain of the second FET.

The fourth FET has a source, a gate, and a drain. The source of the fourth FET is coupled to the second bias voltage, and the gate of the fourth FET is coupled to the gates of the second and third FETs. The drain of the fourth FET is coupled to the drains of the second and third FETs.

For a more complete understanding of the present invention and the advantages thereof, reference is now made to the following descriptions taken in conjunction with the accompanying drawings, in which:

FIG. 1 depicts a block diagram of a VLSI clock distribution network according to the present invention;

FIG. 2 depicts a schematic circuit diagram of one embodiment of a duty cycle correction circuit of FIG. 1;

FIG. 3 depicts another embodiment of a second voltage divider of FIG. 2 according to the present invention; and

FIG. 4 depicts various embodiments of a first clock-inverter circuit of FIG. 2 according to the present invention.

The principles of the present invention and their advantages are best understood by referring to the illustrated operations of the embodiments depicted in FIGS. 1-4.

In FIG. 1, a reference numeral 100 indicates a block diagram of a VLSI clock distribution network embodying the present invention. The VLSI clock distribution network 100 includes a clock receiver 102 A, a duplicate clock receiver 102 B, a phase-locked loop (PLL) 104 , a duty cycle correction circuit 106 , a first inverter stage 108 , a second inverter stage 110 , and a local clock generator (LCG) stage 112 . The number of inverter stages required in a clock distribution network is a function of the size of the VLSI chip and the number and quality of clock signals required. Therefore, the number of inverter stages of the VLSI clock distribution network 100 may vary depending on a specific implementation without departing from the spirit of the present invention. The first inverter stage 108 includes a plurality of inverters such as a first inverter 120 and a second inverter 122 . Similarly, the second inverter stage 110 includes a plurality of inverters such as a third inverter 130 , a fourth inverter 132 , a fifth inverter 134 , and a sixth inverter 136 . The LCG stage 112 includes a plurality of LCGs such as a first LCG 140 , a second LCG 142 , a third LCG 144 , a fourth LCG 146 , a fifth LCG 148 , a sixth LCG 150 , a seventh LCG 152 , and an eighth LCG 154 . As indicated in FIG. 1, the number of inverters and LCGs in each stage may vary depending on specific implementations without departing from the spirit of the present invention.

The clock receiver 102 A receives an external clock signal EXT_CLK. Preferably, the EXT_CLK comes from an external clock generator (now shown). The clock receiver 102 A is connected to the PLL 104 and drives the PLL 104 with a copy of the EXT_CLK. The PLL 104 is connected to the duty cycle correction circuit 106 and drives the duty cycle correction circuit 106 with a PLL-derived clock signal PLL_CLK. Typically, while the PLL_CLK may differ from the EXT_CLK in frequency, some integer multiple of the EXT_CLK has a constant phase relationship to the PLL_CLK. The duty cycle correction circuit 106 is connected to the first inverter stage 108 , which is connected to the second inverter stage 110 . The second inverter stage 110 is connected to the LCG stage 112 . The duty cycle correction circuit 106 is also connected to the LCG stage 112 to receive a local clock LOC_CLK from the LCG stage 112 .

Specifically, the duty cycle correction circuit 106 is connected to the first inverter 120 and the second inverter 122 . The first inverter 120 is connected to the third inverter 130 and the fourth inverter 132 , whereas the second inverter 122 is connected to the fifth inverter 134 and the sixth inverter 136 . The third inverter 130 is connected to the first LCG 140 and the second LCG 142 . The fourth inverter 132 is connected to the third LCG 144 and the fourth LCG 146 . The fifth inverter 134 is connected to the fifth LCG 148 and the sixth LCG 150 . The sixth inverter 136 is connected to the seventh LCG 152 and the eighth LCG 154 . The eighth LCG 154 is connected to the duty cycle correction circuit 106 to provide the LCG_CLK as a feedback signal to the duty cycle correction circuit 106 .

Once the duty cycle correction circuit 106 receives the PLL_CLK from the PLL 104 , the duty cycle correction circuit 106 initially adjusts the PLL_CLK and outputs an output clock CLK_OUT to the first inverter stage 108 . The CLK_OUT travels through the second inverter stage 110 and the LCG stage 112 , and is provided to the duplicate clock receiver 102 B and the duty cycle correction circuit 106 as the LOC_CLK.

The duty cycle correction circuit 106 actively adjusts the transitional delay in a single edge (i.e., either a rising edge or a falling edge) of the PLL_CLK, thereby fine-tuning the duty cycle of the PLL_CLK without interfering the operation of the PLL 104 . This single-edge adjustment scheme is an intentional effort not to affect the behavior of the PLL 104 . Details on how to control the duty cycle are discussed further below in reference to FIGS. 2 and 3.

Typically, PLLs such as the PLL 104 use digital phase comparators (not shown) to detect phase errors between an external reference signal such as the EXT_CLK and an internal feedback signal such as the LOC_CLK. Such digital phase comparators employ edge-triggered latches (not shown) and thus are sensitive to transitions in only a single clock edge of each clock pulse. That is, the digital phase comparators sample either the arrival time difference between two rising clock edges or two falling clock edges. If the duty cycle correction circuit 106 can be restricted to making adjustments on a single clock edge that is not sampled by the digital comparators, then its adjustments will not interfere with the behavior of the PLL 104 . For example, if the digital comparators in the PLL 104 sample a rising clock edge of each clock pulse input to the PLL 104 , then the duty cycle correction circuit 106 can be restricted to making adjustments on a falling clock edge of each clock pulse input to the PLL 104 . In FIG. 1, the duty cycle correction circuit 106 is configured to adjust the transitional delay in a single clock edge of the PLL_CLK with duty cycle control bits CTRL 1 and CTRL 2 . Preferably, the CTRL 1 and CTRL 2 are input from a configuration register (not shown). A more detailed configuration of a duty cycle correction circuit 106 is discussed further below in reference to FIG. 2 .

As mentioned above, each LCG in the LCG stage 112 generates a local clock signal LOC_CLK, which is provided to a plurality of local logic elements (not shown). The LOC_CLK is also provided to the duplicate clock receiver 102 B and the duty cycle correction circuit 106 . For example, the eighth LCG 154 is connected to the duplicate clock receiver 102 B and the duty cycle correction circuit 106 .

Now referring to FIG. 2, a preferred embodiment of the duty cycle correction circuit 106 of FIG. 1 is shown to include a delay-control circuit 202 and a first clock-inverter circuit 204 . The delay-control circuit 202 receives the LOC_CLK from the eighth LCG 154 of FIG. 1 and provides a delay-control voltage Vy to the first clock-inverter circuit 204 through a node y. The delay-control circuit 202 comprises a first voltage divider 206 , a second voltage divider 208 , a differential amplifier 210 , and a stability circuit 212 .

The first voltage divider 206 comprises transistors p 1 and n 1 , a resistor R 1 , and capacitors C 1 a and C 1 b . Preferably, the transistors p 1 and n 1 are respectively p-channel and n-channel metal-oxide-semiconductor field effect transistors (MOSFETs). The transistors p 1 and n 1 are connected in series between ground and a supply voltage Vdd to form a first transistor inverter 220 . Typically, a MOSFET has three terminals, i.e., a gate, a drain, and a source. The biasing scheme of the terminals of a MOSFET is well known in the field of the invention, and thus is mentioned herein without numbering such terminals in the drawings.

In the first voltage divider 206 , the source of the transistor p 1 is connected to the supply voltage Vdd, whereas the source of the transistor n 1 is connected to ground. The drain of the transistor p 1 is connected to the drain of the transistor n 1 . The gates of the transistors p 1 and n 1 are connected together to the eighth LCG 154 of FIG. 2 to receive the LOC_CLK from the eighth LCG 154 . The drains of the transistors p 1 and n 1 are connected to one terminal of the resistor R 1 . The other terminal of the resistor R 1 is connected to the node x. The capacitor C 1 a is connected between the node x and the supply voltage Vdd, whereas the capacitor C 1 b is connected between the node x and ground. Therefore, the node x carries an output voltage Vx of the first voltage divider 206 . The output voltage Vx is determined as follows:

Vx=Vdd *(1-duty cycle of LOC_CLK)

The second voltage divider 208 comprises transistors p 2 and n 2 , resistors R 3 a , R 3 b , R 4 a , and R 4 b , and capacitors C 3 a and C 3 b . The source of the transistor p 2 is connected to the supply voltage Vdd, whereas the source of the transistor n 2 is connected to ground. The drain of the transistor p 2 is connected to the resistor R 4 a , whereas the drain of the transistor n 2 is connected to the resistor R 4 b . The gate of the transistor p 2 receives the CTRL 1 , whereas the gate of the transistor n 2 receives the CTRL 2 . As mentioned above, the CTRL 1 and CTRL 2 are preferably input from a configuration register (not shown). The resistor R 4 a is connected to the resistor R 4 b at a node r. The resistor R 3 a is connected between the supply voltage Vdd and the node r, whereas the resistor R 3 b is connected between ground and the node r. Similarly, the capacitor C 3 a is connected between the supply voltage Vdd and the node r, whereas the capacitor C 3 b is connected between ground and the node r. Therefore, the node r carries an output voltage Vr of the second voltage divider 208 .

The differential amplifier 210 has an inverting input terminal 210 a and a non-inverting input terminal 210 b and an output terminal 210 c . The inverting input terminal 210 a has an inverting input voltage V − , whereas the non-inverting input terminal 210 b has a non-inverting input voltage V + . The output voltage Vo is determined by a differential gain DG of the differential amplifier 210 , the inverting input voltage V − , and the non-inverting input voltage V + :

Vo=DG *( V + −V − )

Thus, an output voltage Vo of the output terminal 210 c is proportionate to (V + −V − ). As shown in FIG. 2, the inverting input terminal 210 a is connected to the node x, whereas the non-inverting input terminal 210 b is connected to the node r. This connection is based on the assumption that LOC_CLK is logically equal to the CLK_OUT. If LOC_CLK is the logical inverse of CLK_OUT, the opposite connections should be made. That is, the inverting input terminal 210 a should be connected to the node r, and the non-inverting terminal 210 b should be connected to the node x. A detailed analysis of this assumption is provided further below.

The stability circuit 212 comprises a resistor R 2 and capacitors C 2 a and C 2 b . The resistor R 2 is connected between the output terminal 210 c and the node y. The capacitor C 2 a is connected between the supply voltage Vdd and the node y, whereas the capacitor C 2 b is connected between ground and the node y.

The first clock-inverter circuit 204 comprises transistors p 3 , n 3 , n 4 , and n 5 . The source of the transistor p 3 is connected to the supply voltage Vdd. The drain of the transistor p 3 is connected to the drains of the transistors n 3 and n 5 , and provides the CLK_OUT to the first inverter stage 108 of FIG. 1 . The source of the transistor n 3 is connected to the drain of the transistor n 4 , whereas the sources of the transistors n 4 and n 5 are connected to ground. The gates of the transistors p 3 , n 3 , and n 5 are connected to the PLL 104 of FIG. 1 to receive the PLL_CLK from the PLL 104 . The gate of the transistor n 4 is connected to the node y. The first clock-inverter circuit 204 is configured to adjust the transitional delay in the falling edge of each clock pulse of the CLK_OUT, which subsequently affects the duty cycle of LOC_CLK, without changing the transitional delay in the rising edge of each clock pulse of the CLK_OUT. This is because the delay-control voltage Vy controls the transistor n 4 . If the delay-control voltage Vy is above the threshold voltage of the transistor n 4 , the transistor n 4 is turned on, thereby providing an additional path to pull down the CLK_OUT when the PLL_CLK transitions from a logical 0 to a logical 1 . This additional path comprising the transistors n 3 and n 4 decreases the transitional delay in the falling edge of the CLK_OUT. On the other hand, if the delay-control voltage Vy is below the threshold voltage of the transistor n 4 , the transistor n 4 is turned off, thereby eliminating the additional path to pull down the CLK_OUT when the PLL_CLK transitions from a logical 0 to a logical 1 . The absence of an additional path comprising the transistors n 3 and n 4 increases the transitional delay in the falling edge of the CLK_OUT. However, the delay-control voltage does not affect the rising edge delay of the CLK_OUT, because only the transistor p 3 determines the rising edge delay of the CLK_OUT.

Therefore, the first clock-inverter circuit 204 is used in the duty cycle correction circuit 106 , when the PLL 104 of FIG. 1 adjusts the rising edge of each clock pulse of the PLL_CLK. This ensures that the duty cycle correction circuit 106 does not affect the functions performed by the PLL 104 .

The following conditions are critical to the successful operation of the delay-control circuit 202 :

1. The resistor R 1 has resistance R 1 , and two capacitors C 1 a and C 1 b each have capacitance C 1 . The resistors R 3 a and R 3 b each have resistance R 3 . The resistors R 4 a and R 4 b each have resistance R 4 . The capacitors C 3 a and C 3 b each have capacitance C 3 . The capacitors C 2 a and C 2 b each have capacitance C 2 . The resistor R 2 has resistance R 2 .

2. When employed in the VLSI clock distribution network 100 , the LOC_CLK logically equals the CLK_OUT. That is, the LOC_CLK is not an inverted version of the CLK_OUT. If the LOC_CLK happens to be an inverted version of the CLK_OUT, the aforementioned assumption that the output voltage Vx is proportionate to the output voltage Vo is incorrect. In that case, the node x should be connected to the non-inverting input terminal 210 b , whereas the node r should be connected to the inverting input terminal 210 a.

3. To avoid a PLL lock-up condition, the maximum delay variation of the first clock-inverter circuit 204 is less than one half the clock cycle time of the PLL_CLK.

4. Transistors n 1 and p 1 have approximately equal device strengths. This guarantees that the first transistor inverter 220 to switch at 0.5 Vdd

5. The resistance R 1 is much larger than the resistances of the transistors n 1 and p 1 . This guarantees that the time required to charge and discharge the capacitors C 1 a and C 1 b is dominated by the resistance R 1 rather than the device characteristics of the transistors n 1 and p 1 .

6. The time constant R 1 *C 1 is much larger than the clock period of the LOC_CLK. This ensures that, after the duty cycle correction circuit 106 reaches a steady state, the average dc value of the output voltage Vx is equal to the supply voltage Vdd times the duty cycle of the LOC_CLK, and furthermore that the ac ripples on the node x are small.

7. The differential amplifier 210 has a high differential gain (DG) and a small common-mode error.

8. To help ensure stability of the duty cycle correction circuit 106 , the time constant R 2 *C 2 is much larger than the time constant R 1 *C 1 .

9. To filter power supply noise, the capacitors C 1 a , C 1 b , C 2 a , C 2 b , C 3 a , and C 3 b are symmetrically terminated to both the supply voltage Vdd and ground.

10. To make fine duty cycle adjustments, the resistance R 4 is much greater than the resistance R 3 , and the transistors n 2 and p 2 are much stronger than the resistance R 4 .

Given these conditions, the duty cycle correction circuit 106 has three settings depending on the combination of the CTRL 1 and the CTRL 2 . When the CTRL 1 and the CTRL 2 have different bits, both the transistors p 2 and n 2 are either off or on. In this case, the duty cycle correction circuit 106 produces a 50-percent duty cycle. When both the transistors p 2 and n 2 are off or on, the transistors p 2 and n 2 and the resistors R 4 a and R 4 b do not affect the output voltage Vr. Therefore, the output voltage Vr is 0.5 Vdd. For example, if an initial duty cycle is 60 percent, the output voltage Vx has an average dc value of 0.4 Vdd. Therefore, the output voltage Vo at the output terminal 210 c outputs the supply voltage Vdd, because the inverting input terminal 210 a is lower in voltage than the non-inverting input terminal 210 b . This turns on the transistor n 4 , thereby decreasing the transitional delay in the falling edges of the CLK_OUT, which means decreased duty cycle of the CLK_OUT. Therefore, the duty cycle correction circuit 106 reaches a steady state at a 50-percent duty cycle.

Alternatively, if the initial duty cycle is 40 percent, the output voltage Vx has an average dc value of 0.6 Vdd. Therefore, the output voltage Vo at the output terminal 210 c outputs zero voltage, because the inverting input terminal 210 a is higher in voltage than the non-inverting input terminal 210 b . This turns off the transistor n 4 , thereby increasing the transitional delay in the falling edges of the CLK_OUT, which means increased duty cycle of the CLK_OUT. Therefore, the duty cycle correction circuit 106 reaches a steady state at a 50-percent duty cycle.

When both the CTRL 1 and the CTRL 2 are a logical 1 , the transistor p 2 is off, and the transistor n 2 is on. In this case, the duty cycle correction circuit 106 produces a slightly asymmetric LOC_CLK with logical- 1 pulse width slightly longer than logical- 0 pulse width in a given clock cycle. When the transistor p 2 is off and the transistor n 2 is on, the resistor R 4 a and the transistor p 2 are taken out. Thus, the output voltage Vr is determined by the resistors R 4 b , R 3 a , and R 3 b:

Vr =[( R 3 // R 4 )/( R 3 + R 3 // R 4 )]* Vdd , where R 3 // R 4 = R 3 * R 4 /( R 3 + R 4 ).

Since R 4 is much larger than R 3 , R 3 //R 4 is only slightly smaller than R 3 . Therefore, Vx is slightly lower than 0.5 Vdd. For example, let us assume Vr equals 0.45 Vdd. Now assuming that the initial LOC_CLK has a 50-percent duty cycle, Vx has an average dc value of 0.5 Vdd. Since the inverting input terminal 210 a has a higher voltage than the non-inverting input terminal 210 b , the output voltage Vo approaches to zero voltage, thereby increasing the transitional delay in the falling edges of the CLK_OUT. Since an increased transitional delay in the falling edges increases the duty cycle of the CLK_OUT, the output voltage Vx drops slightly below 0.5 Vdd and ultimately equals the output voltage Vr at the node r in a steady state. Therefore, the LOC_CLK reaches a steady state at slightly over a 50-percent duty cycle.

Alternatively, if the output voltage Vr equals 0.45 Vdd and the LOC_CLK initially has a 60-percent duty cycle, the output voltage Vx has an average dc value of 0.4 Vdd. Since the inverting input terminal 210 a has a lower voltage than the non-inverting input terminal 210 b , the output voltage Vo approaches to the supply voltage Vdd, thereby decreasing the transitional delay in the falling edges of the CLK_OUT. Since a decreased transitional delay in the falling edges decreases the duty cycle of the CLK_OUT, the output voltage Vx rises slightly above 0.4 Vdd and ultimately equals the output voltage Vr at the node r in a steady state. Therefore, the LOC_CLK reaches a steady state at slightly over a 50-percent duty cycle.

When both the CTRL 1 and the CTRL 2 are a logical 0 , the transistor p 2 is on, and the transistor n 2 is off. In this case, the duty cycle correction circuit 106 produces a slightly asymmetric LOC_CLK with logical- 1 pulse width slightly shorter than logical- 0 pulse width in a given cycle. When the transistor p 2 is on and the transistor n 2 is off, the resistor R 4 b and the transistor n 2 are taken out. Thus, the output voltage Vr is determined by the resistors R 4 a , R 3 a , and R 3 b:

Vr=[R 3 /( R 3 + R 3 // R 4 )]* Vdd , where R 3 // R 4 = R 3 * R 4 /( R 3 + R 4 ).

Since R 4 is much larger than R 3 , R 3 //R 4 is only slightly smaller than R 3 . Therefore, Vx is slightly higher than 0.5 Vdd. For example, let's assume Vr equals 0.55 Vdd. Now assuming that the initial LOC_CLK has a 50-percent duty cycle, Vx has an average dc value of 0.5 Vdd. Since the inverting input terminal 210 a has a lower voltage than the non-inverting input terminal 210 b , the output voltage Vo approaches to the supply voltage, thereby deceasing the transitional delay in the falling edges of the CLK_OUT. Since a decreased transitional delay in the falling edges decreases the duty cycle of the CLK_OUT, the output voltage Vx rises slightly above 0.5 Vdd and ultimately equals the output voltage Vr at the node r in a steady state. Therefore, the LOC_CLK reaches a steady state at slightly under a 50-percent duty cycle.

Alternatively, if the output voltage Vr equals 0.55 Vdd and the LOC_CLK initially has a 40-percent duty cycle, the output voltage Vx has an average dc value of 0.6 Vdd. Since the inverting input terminal 210 a has a higher voltage than the non-inverting input terminal 210 b , the output voltage Vo approaches zero voltage, thereby increasing the transitional delay in the falling edges of the CLK_OUT. Since an increased transitional delay in the falling edges of the CLK_OUT increases the duty cycle of the CLK_OUT, the output voltage Vx falls slightly above 0.6 Vdd and ultimately equals the output voltage Vr at the node r in a steady state. Therefore, the LOC_CLK reaches a steady state at slightly under a 50-percent duty cycle.

Now referring to FIG. 3, a reference numeral 208 A indicates a third voltage divider, which is another embodiment of the second voltage divider 208 of FIG. 2 . The third voltage divider 208 A has all the components of the second voltage divider 208 , and additionally has transistors p 2 a and n 2 a , and resistors R 4 c and R 4 d . The source of the transistor p 2 a is connected to the supply voltage Vdd, whereas the source of the transistor n 2 a is connected to ground. The drain of the transistor p 2 a is configured to receive a control bit CTRL 1 a , whereas the drain of the transistor n 2 a is configured to receive a control bit CTRL 2 a . As mentioned above, the CTRL 1 a and the CTRL 2 a are preferably stored in a configuration register (not shown). The resistor R 4 c is connected between the drain of the transistor p 2 a and the node r. The resistor R 4 d is connected between the drain of the transistor n 2 a and the node r. Preferably, the resistors R 4 c and R 4 d each have resistance R 4 .

Since the third voltage divider 208 A has the additional control bits CTRL 1 a and CTRL 2 b , it is possible to make a wider range of adjustments on the output voltage Vr with the third voltage divider 208 A than with the second voltage divider 208 . Therefore, the duty cycle correction circuit 106 of FIG. 2, with the third voltage divider 208 A replacing the second voltage divider 208 , can make a finer adjustment on the duty cycle of the CLK_OUT or the LOC_CLK. For example, assume that the CTRL 1 , the CTRL 1 a , the CTRL 2 , and the CTRL 2 a are respectively 1, 0, 1, and 1. Then, the transistors p 2 , p 2 a , n 2 , and n 2 a are respectively turned off, turned on, turned on, and turned on. In this case, the output voltage Vr is determined by the resistors R 3 a , R 3 b , R 4 b , R 4 c , and R 4 d :

Vr =[( R 3 //(0.5 *R 4 ))/(( R 3 //(0.5 *R 4 ))+ R 3 // R 4 )]* Vdd , where R 3 // R 4 = R 3 * R 4 /( R 3 + R 4 ) and R 3 //(0.5 *R 4 )=0.5 *R 3 * R 4 /( R 3 +0.5 *R 4 ).

It is noted that adding more control bit inputs with accompanying components in a similar fashion enables further finer adjustments of the duty cycle of the CLK_OUT or the LOC_CLK. Since such additional variations are straightforward and can be easily made by a person with ordinary skill in the art, they are considered to fall within the spirit of the present invention.

Now referring to FIG. 4A, a second clock-inverter circuit 204 A is another embodiment of the first clock-inverter circuit 204 of FIG. 2 . The second clock-inverter circuit 204 A has transistors p 4 , p 5 , p 6 , and n 6 . The source of the transistor p 4 is connected to the supply voltage Vdd, whereas the source of the transistor n 6 is connected to ground. The drain of the transistor p 4 is connected to the source of the transistor p 5 , whereas the drain of the transistor n 6 is connected to the drain of the transistor p 5 . The source and drain of the transistor p 6 are respectively connected to the supply voltage Vdd and the drain of the transistor n 6 . The gate of the transistor p 4 is connected to the node y. The gates of the transistor p 5 , p 6 , and n 6 are all connected together to receive the PLL_CLK. The CLK_OUT is taken at the drain of the transistor p 5 , p 6 , or n 6 .

The second clock-inverter circuit 204 A is configured to adjust the transitional delay in the rising edge of each clock pulse of the CLK_OUT, which subsequently affects the duty cycle of the LOC_CLK, without changing the transitional delay in the falling edge of each clock pulse of the CLK_OUT. This is because the delay-control voltage Vy controls the transistors p 4 . If the delay-control voltage Vy is below the threshold voltage of the transistor p 4 , the transistor p 4 is turned on, thereby providing an additional path to pull up the CLK_OUT when the PLL_CLK transitions from a logical 1 to a logical 0 . This additional path comprising the transistors p 4 and p 5 decreases the transitional delay in the rising edge of the CLK_OUT. On the other hand, if the delay-control voltage Vy is above the threshold voltage of the transistor p 4 , the transistor p 4 is turned off, thereby elimination the additional path to pull up the CLK_OUT when the PLL_CLK transitions from a logical 1 to a logical 0 . This lack of the additional path comprising the transistors p 4 and p 5 increases the transitional delay in the rising edge of the CLK_OUT. However, the delay-control voltage does not affect the falling edge delay of the CLK_OUT, because only the transistor n 6 determines the falling edge delay of the CLK_OUT.

Therefore, the first clock-inverter circuit 204 A is used in the duty cycle correction circuit 106 , when the PLL 104 of FIG. 1 is non-inverting and samples the rising edge of each clock pulse of the PLL_CLK. This ensures that the duty cycle correction circuit 106 does not affect the functions performed by the PLL 104 .

In FIG. 4B, a third clock-inverter circuit 204 B is presented. The third-clock inverter circuit 204 B has the first clock-inverter circuit 204 , a second transistor inverter 402 , and a third transistor inverter 404 . The second transistor inverter 402 comprises transistors p 7 and n 7 . The third transistor inverter 404 comprises transistors p 8 and n 8 . The second transistor inverter 402 is configured to receive the PLL_CLK through the gates of the transistors p 7 and n 7 . The second transistor inverter 402 inverts the PLL_CLK and outputs a clock signal PLL_CLK _I_ 1 , an inverted version of the PLL_CLK. The PLL_CLK_I_ 1 is provided to the first clock-inverter circuit 204 to control the gates of the transistors p 3 , n 3 , and n 5 . The first clock-inverter circuit 204 once again inverts the PLL_CLK_ 1 , and outputs a clock signal PLL_CLK_F. Thus, the PLL_CLK_F has the same polarity as the PLL_CLK with its “falling” edge adjusted by the delay-control voltage Vy applied at the node y. The PLL_CLK_F is provided to the third transistor inverter 404 , which outputs the CLK_OUT. Therefore, the CLK_OUT is an inverted signal of the PLL_CLK with its “rising” edge adjusted by the delay-control voltage Vy applied at node y.

In FIG. 4C, a fourth clock-inverter circuit 204 C is presented. The fourth clock-inverter circuit 204 C has the second clock-inverter circuit 204 A, a fourth transistor inverter 406 , and a fifth transistor inverter 408 . The fourth transistor inverter 406 comprises transistors p 9 and n 9 . The fifth transistor inverter 408 comprises transistors p 10 and n 10 . The fourth transistor inverter 406 is configured to receive the PLL_CLK through the gates of the transistors p 9 and n 9 . The fourth transistor inverter 406 inverts the PLL_CLK and outputs a clock signal PLL_CLK_I_ 2 , an inverted version of the PLL_CLK. The PLL_CLK_I_ 2 is provided to the second clock-inverter circuit 204 A to control the gates of the transistors p 5 , p 6 , and n 6 . The second clock-inverter circuit 204 A once again inverts the PLL_CLK_I 2 , and outputs a clock signal PLL_CLK_R. Thus, the PLL_CLK_R has the same polarity as the PLL_CLK with its “rising” edge adjusted by the delay-control voltage Vy applied at the node y. The PLL_CLK_R is provided to the fifth transistor inverter 408 , which outputs the CLK_OUT. Therefore, the CLK_OUT is an inverted signal of the PLL_CLK with its “falling” edge adjusted by the delay-control voltage Vy applied at node y.

It is noted that the second, third, and fourth clock-inverter circuits 204 A, 204 B, and 204 C shown in FIGS. 4A, 4 B, and 4 C are merely examples of some variations of the first clock-inverter circuit 204 , and that further variations of the first clock-inverter circuit 204 in the manner suggested in FIGS. 4A, 4 B, and 4 C should be considered within the spirit of the present invention.

It will be understood from the foregoing description that various modifications and changes may be made in the preferred embodiment of the present invention without departing from its true spirit. This description is intended for purposes of illustration only and should not be construed in a limiting sense. The scope of this invention should be limited only by the language of the following claims.