Charge pump phase locked loop with improved power supply rejection

[0001] This application claims priority under 35 USC § (e)(1) of European Application Number 03292140.5 filed Aug. 29, 2003.

[0002] Not Applicable

[0003] Not Applicable

[0004] 1. Technical Field

[0005] This invention relates in general to electronic circuits and, more particularly, to phase locked loop circuits.

[0006] 2. Description of the Related Art

[0007] A phase locked loop (PLL) synchronizes an output signal to a reference signal. Phase locked loops are used in a variety of electronic circuits for frequency synthesizing, frequency and phase modulation and demodulation, clock recovery and clock synchronization, among other uses.

[0008] FIG. 1a illustrates a basic diagram of a PLL. A phase frequency detector (PFD) 12 receives two signals, F_R, a reference signal, and F_V, a divided-down version of the output signal. The PFD generates UP and DN (down) signals, indicating whether the frequency of the output signal needs to increase (UP high) or decrease (DN high). The UP and DN signals are input into a loop filter 16. When UP is high and DN is low, V_C rises, thereby increasing the frequency of the output of the VCO 18. Similarly, when DN is high and UP is low, V_C falls, thereby decreasing the frequency of the output of the VCO 18. In many cases, a divider 20 is used to divide the frequency to a lower frequency by a factor of N; in this case F_out=N*F_R.

[0009] FIG. 1b illustrates a schematic diagram of a prior art active filter that may be used as the loop filter 16 in the PLL of FIG. 1a. The DN signal from the PFD 12 is coupled to the inverting input of operational amplifier (op-amp) 22, via resistor 24. The UP signal from the PFD 12 is coupled to the non-inverting input of op-amp 22 via resistor 26. The output of op-amp 22 is coupled to its inverting input via capacitor 28 and resistor 30. The non-inverting input is coupled to ground via resistor 32 and capacitor 34.

[0010] This type of filter is not often used in integrated applications, mainly because it requires a dual voltage supply and needs two tightly matched filters. Also, since both inverting and non-inverting inputs are driven by the PFD 12, this filter exhibits common mode problems.

[0011] FIG. 1c illustrates a schematic of a second active filter that may be used as the loop filter 16 in the PLL of FIG. 1a. The UP signal from the PFD 12 is coupled to the inverting input of operational amplifier (op-amp) 40, via resistor 42. The DN signal from the PFD 12 is coupled to the non-inverting input of op-amp 40 via resistor 44. The output of op-amp 40 is coupled to its inverting input via resistor 46. The non-inverting input of op-amp 40 is coupled to ground via resistor 48. The output of op-amp 40 is coupled to the inverting input of op-amp 50 via resistor 52. The output of op-amp 50 is coupled to its inverting input via capacitor 54 and resistor 56. The non-inverting output is coupled to a DC voltage, V_DC.

[0012] This design also is not particularly useful for integrated designs, since it requires two op-amps, which results in larger area requirements and higher noise and power consumption. Once again, matching the filters is difficult to achieve optimum performance. Also, op-amp 40 does not present a high impedance state to op-amp 50 and, thus, all of the noise from resistors 42, 46, 44 and 48 and PFD 12, and the DC mismatch between op-amp 40 and the inverting input of op-amp 50 will introduce spurs (feed-through).

[0013] A different type of PLL is the charge-pump phase locked loop (CP-PLL). A CP-PLL pumps current in and out of a loop filter in response to detected deviations between the output frequency and the reference frequency. Among other factors, CP-PLLs are considered superior with regard to frequency range and cost. However, these devices may be difficult to integrate onto silicon die, which is highly desirable in many applications.

[0014] A basic block diagram of a CP-PLL 58 is shown in FIG. 2. In this design, a phase frequency detector (PFD) 12 receives two signals, F_R, a reference signal, and F_V, a divided-down version of the output signal. The PFD 12 generates UP and DOWN pulses, indicating whether the frequency of the output signal needs to increase (UP pulsed) or decrease (DOWN pulsed). The UP and DOWN pulses cause a charge pump 14 to either source current into a loop filter 16 or sink current from the loop filter 16. As current is sourced to the loop filter 16, V_C rises, thereby increasing the frequency of the output of the VCO 18. Similarly, as the charge pump 14 sinks current from the loop filter 16, V_C falls, thereby decreasing the frequency of the output of the VCO 18.

[0015] The noise floor of a digital circuit and the noise current of the analog charge pump 14 represent the major noise contributors in a CP-PLL within the PLL bandwidth. In a traditional analog charge pump design used in CP-PLL applications, the noise contributions mainly come directly from the active devices used in the analog charge pump to sink and source current. In order to improve the signal-to-noise ratio, a higher reference current can be used in the charge pump; however, the higher reference current may necessitate a higher capacitor value in the loop filter 16, preventing a possible integration into a silicon die. In addition, mismatches between sink and source devices increase the spurious level that must be filtered out with a lower bandwidth filter (again requiring a larger capacitor) and a consequently longer settling time.

[0016] Another critical issue for the traditional charge pump is the need of a complex and low noise biasing circuit and the requirements in terms of supply headroom.

[0017] Therefore, a need has arisen for a compact charge-pump phase locked loop with low noise characteristics.

[0018] In a first aspect of the present invention, a filter comprises a digitally controlled switch, a voltage source for providing a predetermined voltage signal, a capacitor coupled to the voltage source through the digitally controlled switch for storing a charge at the predetermined voltage, and an operational amplifier having an input node coupled to the capacitor and the switch. The node and the capacitor are charged by the voltage source when a connection through the digitally controlled switch is enabled and the node is charged by the capacitor when the connection through the digitally controlled switch is disabled.

[0019] Accordingly, the isolation of the circuit from the voltage signal protects the circuit from variations on the voltage source, providing and improved power supply rejection ratio of the filter.

[0020] In a second aspect of present invention, a phase locked loop circuit comprises a phase detector, a charge pump for generating a charge on a first node responsive to the output of the phase detector, a voltage generating circuit for generating a predetermined voltage on a second node responsive to a supply voltage provided by a power supply, an active filter for generating an output responsive to voltages on the first and second nodes, and a voltage controlled oscillator for generating an output frequency responsive to the output of the active filter. A digitally controlled switch coupled between the second node and the active filter selectively provides a high impedance state between the second node and the active filter during inactive periods of the charge pump to isolate the filter from the power supply.

[0021] Accordingly, the frequency output of the phase locked loop circuit is less affected by variations on the power supply.

[0033] The present invention is best understood in relation to FIGS. 3–7a–b of the drawings, like numerals being used for like elements of the various drawings.

[0034] FIG. 3 illustrates a schematic diagram of a PLL 60 that can be fully integrated on a semiconductor die, with low noise characteristics, low spur levels and high linearity. A reference signal F_R is input to PFD 62. PFD 62 outputs UP and DN pulses to a digital charge pump 64. The UP node of PFD 62 is coupled to a non-inverting buffer 66 and the DN node of PFD 62 is coupled to an inverting buffer 68. The output of the non-inverting buffer 66 is coupled to both gates of p-channel transistor 70 and n-channel transistor 72 that produce a logical inversion of the signal at the gates of the transistors. A first source/drain of p-channel transistor 70 is coupled to V_DD and a second source/drain of p-channel transistor 70 is coupled to a first source drain of n-channel transistor 72. The second source/drain of n-channel transistor 72 is coupled to ground. The node connecting the second source/drain of p-channel transistor 70 and the first source/drain of n-channel transistor 72 is coupled to a first terminal of resistor 74 (having resistive value R1). The second terminal of resistor 74 is coupled to the input of transmission gate 76. Transmission gate 76 is controlled by dynamic digital control 78.

[0035] The output of the inverting buffer 68 is coupled to both gates of p-channel transistor 80 and n-channel transistor 82 that produce another logical inversion. A first source/drain of p-channel transistor 80 is coupled to V_DD and a second source/drain of p-channel transistor 80 is coupled to a first source drain of n-channel transistor 82. The second source/drain of n-channel transistor 82 is coupled to ground. The node connecting the second source/drain of p-channel transistor 80 and the first source/drain of n-channel transistor 82 is coupled to first terminal of resistor 84 (having resistive value R2). The second terminal of resistor 84 is also coupled to the input of transmission gate 76. The current through resistor 74 is designated as I1, the current through resistor 84 is designated as I2 and the current through transmission gate 76 to the inverting input of op-amp 86 is designated as Icp.

[0036] The output of transmission gate 76 is coupled to the inverting input of op-amp 86 of the active filter 87. The non-inverting input of op-amp 86 is coupled to a DC voltage source. This voltage source comprises p-channel transistor 90 and n-channel transistor 92. The gate of p-channel transistor 90 is coupled to ground and the gate of n-channel transistor 92 is coupled to V_DD. A first source/drain of p-channel transistor 90 is coupled to V_DD and a second source/drain of p-channel transistor 90 is coupled to a first terminal of resistor 91. The first source/drain of n-channel transistor 92 is coupled to a first terminal of resistor 93. The second source/drain of n-channel transistor 92 is coupled to ground. The second terminals of resistors 91 and 93 are coupled to the non-inverting input of op-amp 86.

[0037] The output of op-amp 86 is coupled to its inverting input via capacitor 94 and resistor 96. The output of op-amp 86 is also coupled to a first terminal of resistor 98. The second terminal of resistor 98 is coupled to integrated VCO 100. A capacitor 102 is coupled between the input to VCO 100 and ground. The output of VCO 100 is input to divider 104. The output of divider 104 (F_V) is coupled to PFD 62.

[0038] The value of the resistors is designed such that R1=R2=n*R3=n*R4. Since R1 will be proximate R2 and R3 will be proximate R4, the matching of the ratios should be highly accurate, even in the face of process variations (better than 1% matching using standard processing techniques). In operation, non-inverting buffer 66 and inverter 68 operate to sink or source current to the inverting input of op-amp 86 as follows. Assuming transmission gate 76 is open (i.e. in a high impedance state), if UP and DN are both high, resistors 84 and 74 will be series connected between V_DD and ground. If UP and DN are both low, resistors 74 and 84 will be series connected between V_DD and ground. In either case, no current will be sourced to or drawn from the inverting node. Assuming transmission gate 76 is closed (i.e. in a low impedance state) if UP and DN are both high, resistors 84 and 74 will be series connected between V_DD and ground. If UP and DN are both low, resistors 74 and 84 will be series connected between V_DD and ground. In either case, no current will be sourced to or drawn from the inverting node due to the matching between the ratio R1/R2 and R3/R4.

[0039] If UP is high and DN is low, resistor 74 will be between ground and the inverting input and resistor 84 will be between the ground and the inverting input. As a result, current will sink from the inverting input and the voltage output of the active filter will increase. Accordingly, the frequency output from the VCO 100 will increase.

[0040] On the other hand, if UP is low and DN is high, resistor 74 will be between V_DD and the inverting input and resistor 84 will be between V_DD and the inverting input. As a result, current will be drawn to the inverting input and the voltage output of the active filter will decrease. Therefore, the frequency output from the VCO 100 will decrease.

[0041] The operation of the present invention, as described thus far, greatly improves on the noise characteristics in the prior art. With the traditional analog charge pump used for an integrated PLL, the noise contribution comes largely from the active devices used in the analog charge pump itself and from the mismatch between sinking and sourcing devices. The noise floor in the PLL bandwidth of the prior art PLL is limited by the analog charge pump. The charge pump uses large transistors and works with relatively high current in order to improve the signal-to-noise ratio.

[0042] In the explanation above, it is assumed that the transmission gate 76 is open. The transmission gate 76, under control of dynamic digital control 78, is used in the preferred embodiment to provide additional digital noise filtering. The dynamic digital control circuitry 78 operates responsive to the UP and DN signals to isolate the inverting input of op-amp 86 from the rest of the circuitry during time periods when the charge pump 64 should not be sinking or sourcing current into the inverting input. Specifically, the dynamic digital control circuit 78 provides a current path to the inverting input from a time period just before either the UP or DN signals transition to an active voltage level and isolates the inverting input just after the UP or DN signal transition to an inactive voltage level (although exact timing of the opening and closing of the transmission gate 76 is generally not critical). In this way, much of the noise due to the power supply and the resistors is decoupled from the inverting input, causing a reduction in the digital noise floor and the spurs due to any slight voltage mismatch between the inverting and non-inverting inputs.

[0043] Because of the attenuation of noise, the resistance value of R1, R2, R3 and R4 may be increased, with low impact on the phase noise (up to the level it is negligible compared to the input noise of the op-amp 86), with the consequential reduction of the charge pump current in such a way as to allow reduction of the capacitance of capacitor 94. With a smaller capacitance, it may be possible to integrate this capacitor onto the silicon die with the rest of the PLL.

[0044] FIG. 4 illustrates the phase noise contributors for the embodiment shown in FIG. 3. The noise is dominated by the reference clock F_R phase noise for low frequency offset (frequency<1 kHz) by the digital noise floor in the close-in bandwidth (as shown, the PLL exhibits a very low noise floor), by the filter and the op-amp 86 at the limit of the PLL loop bandwidth, and by the VCO noise over the PLL loop bandwidth.

[0045] The advantage of the embodiment of FIG. 3 is that the active device (op-amp 86) has a relatively low effect on noise contribution because it contributes only at the limit of the bandwidth like the filter contributors (resistors 96 and 98), an op-amp with standard characteristics is sufficient to reach high performance levels. Resistors 91 and 93 have a very low effect on noise contribution because they are connected to the non-inverting input of op-amp 86 and thus have a unity gain. Resistors 74 and 84, the inverter/non-inverting buffers 68 and 66, and the PFD 62 contribute to the close-in noise floor of the PLL, with very low level of noise due to two reasons: (1) the input current to the op-amp 86 is not generated by active current generators, but rather by resistors 74 and 84 and gates (i.e., inverter/non-inverting buffers 68 and 66), which generate minimal noise and (2) the transmission gate 76 provides a high impedance state with fast switching characteristics, which isolates the op-amp 86 from the digital inverter/non-inverting buffers 68 and 66 and the resistors 74 and 84 for the vast majority of the time and make the op-amp 86 working as a follower.

[0046] FIGS. 5a–c better illustrate the benefits of the high impedance state provided by the transmission gate 76. FIG. 5a illustrates the generally low noise and spur characteristics of the PLL 60. As can be seen in FIGS. 5b and 5c the UP signal transitions high responsive to an upward transition of the F_R reference and the DN signal transitions high responsive to an upward transition of the F_V signal. The UP and DN signal overlap for 3 ns before both transitioning to a low voltage state. By providing a high impedance at transmission gate 76 the digital charge pump 64 is electrically isolated from the op-amp 86 for all time other than the period where UP or DN signal is at an active logic level—for a 400 KHz example as shown in FIG. 3b, the transmission gate is in a high impedance state for (2.5 μs–3 ns)/2.5 μs or 99.88% of the time (when in a locked state). The dynamically-controlled transmission switch 76 greatly improves the signal-to-noise ratio, because it allows the complete signal from the charge pump 64 to charge the filter with high linearity during the time the transmission switch is in a low impedance state due to the fast switching of the gates (inverter/non-inverting buffers 68 and 66) and the speed of the transmission switch 76, without a dead-zone, and it isolates noise from entering the op-amp 86 during the remainder of the time.

[0047] The noise and spurious performance are related to the matching between the p-channel transistors 70, 80 and 90, between the n-channel transistors 72, 82, and 92, between the ratios R1/R2 and R3/R4 of resistors 74, 84, 91 and 93. Accordingly, the devices that must be matched are similar devices (rather than complementary devices) and a better dynamic matching behavior can be achieved. Thus, mismatching of components has a negligible effect on noise in this structure. Whatever noise is generated through mismatching is largely isolated from propagation to the op-amp 86 due to the transmission switch, which is normally in a high impedance state. The op-amp is in follower mode during the high impedance state and cannot integrate any mismatch or offset error. Additionally, no biasing blocks are needed, as in prior art charge pump devices, minimizing the size of the analog section. Also, a DC voltage is applied to the non-inverting input of op-amp 86, eliminating common mode problems.

[0048] A standard analog current source charge pump, on the other hand, needs an overlap time between UP and DN signals that is relatively big, on the order of 10 ns for a 400 kHz reference signal, as compared to the 3 ns signal required by the present invention (also for a 400 kHz reference signal). This is due to the difficulty of the analog charge pump to cut the current source and it needs time to settle, as opposed to fast switching gates. Additionally, the setting of the current on a p-type source is difficult to match with an n-type source; therefore a certain time period is required to allow the output to null the charge at the filter when the signal is locked without generating excessive spurs due to the mismatch on the current spikes generated by cutting the current source. The large overlap time requirement degrades the noise performance of the standard approach.

[0049] FIG. 6 illustrates a schematic of a PLL 110 which includes circuitry for reducing the PSRR (power supply rejection ratio) of the active filter 87. To reduce the PSRR, a second transmission gate 112 is placed between the resistors 91 and 93 and the non-inverting input of operational amplifier 86. In addition, a small capacitor 114 is placed between the non-inverting input of operational amplifier 86 and ground.

[0050] The PSRR is a measure of the effect of a variation in the power supply voltage on the output of operational amplifier 86. More precisely, PSRR is the ratio between the variation on a certain output (in this case the voltage at the input of the VCO 100) and the variation on the power supply source (V_DD) that caused the variation on the output. As an example, if a 1 Vpp variation on V_DD causes a 0.5 Vpp variation on the output, the PSRR will be equal to 20 log(0.5/1)=−6.02 dB. If a 1 Vpp variation on V_DD causes a 0.25 Vpp variation on the output, the PSRR will be equal to 20 log(0.25/1)=−12.04 dB. Hence, a low PSRR indicates that the circuit is less sensitive to small variations in the power supply output.

[0051] In the case of PLL 60 of FIG. 3, a variation in voltage supply VDD, coupled to p-channel transistor 90, will directly cause a variation in voltage at the non-inverting input of operational amplifier 86. A variation in the voltage at the non-inverting input of the operational amplifier 86 will cause a variation in the output voltage of the operational amplifier 86. Since the output of operational amplifier 86 drives VCO 100, the variation in the power supply voltage will thus cause an error in VCO_—out. Consequently, in a PLL with a low PSRR, the frequency output will be more stable relative to noise on the power supply than PLL with higher PSRR values.

[0052] Accordingly, for stable PLL operation, it is beneficial to reduce the effect of variations in the power supply voltage as much as possible. With R3=R4, the voltage swing of a variation at the non-inverting input will be one-half of the variation at the V_DD power supply. Accordingly, with the non-inverting input of the operational amplifier at unity gain, there will be only −6 dB rejection to power supply variations. While this is satisfactory with a high quality voltage regulator, it could cause problems with less precise voltage regulators.

[0053] Referring again to FIG. 6, in PLL 110, transmission gate 112 selectively isolates the non-inverting input of operational amplifier 86 from the power supply under control of dynamic digital control 78. While the transmission gate is in an closed state (low impedance), the non-inverting input of operational amplifier 86 is coupled to the V_DD power supply through the voltage divider of resistors 91 and 93. During this time, capacitor 114 is charged to the voltage at the output of transmission gate 112 (i.e., V_DD/2 with R3=R4). While the transmission gate is in an open state (high impedance), the capacitor maintains the voltage at the non-inverting input.

[0054] In the preferred embodiment, both transmission gates 76 and 112 are similarly controlled—the dynamic digital control circuit 78 places transmission gates 76 and 112 in an closed state from a time period just before either the UP or DN signals transition to an active voltage level and places transmission gates 76 and 112 in an open state just after the UP or DN signal transition to an inactive voltage level (once again, the exact timing of the opening and closing of the transmission gate 76 is generally not critical).

[0055] Accordingly, while transmission gate 112 is in an open state, variations in the power supply voltage are isolated from the non-inverting input and, therefore, cannot affect the output frequency of the VCO 100. During these times, the voltage at the non-inverting input of operational amplifier 86 is maintained by capacitor 114. Capacitor 114 can be on the order of 50 pF. The operational amplifier 86 draws only a negligible amount of current from the non-inverting input; therefore, the voltage across the capacitor 114 should remain stable. Any minor change in voltage across the capacitor will be corrected during periods when transmission gate 112 is in a closed state.

[0056] FIGS. 7a and 7b illustrates the benefits of the present invention using a simulated implementation. FIG. 7a illustrates a noise transfer function showing the effect of noise on the input of VCO 100 as a function of the frequency of the noise. As can be seen in FIG. 7a, this transfer function has its peak at approximately 30 kHz (the cut-off frequency of the closed loop response of the entire PLL 110). Accordingly, the range of frequencies around 30 kHz are critical for improving the PSRR, since noise at this frequency range will have the most effect on the VCO frequency output.

[0057] FIG. 7b is a graph of the PSRR of the active loop filter 87 as a function of frequency. The PSRR of the loop filter 87 of FIG. 3, represented by line 120, has a PSRR of about −6 dB at 30 kHz. By contrast, the circuit of FIG. 6, represented by line 122, has a PSRR of about −28 dB at 30 kHz, representing a significant improvement of 22 dB.

[0058] Accordingly, in addition to the improvements in signal-to-noise ratio provided by the circuit of FIG. 3, the circuit of FIG. 6 improves upon the power supply rejection ratio of the circuit, improving the accuracy of the output frequency of the PLL.

[0059] While the present invention has been described in relationship to a specific filter, using a digitally controlled switch to selectively isolate and connect a voltage source with the non-inverting input to a operational amplifier, the invention could be used in a number of circuits where a node needs to be maintained at a certain voltage, but does not draw significant current for a part of the time. The digitally controlled switch can enable a connection between a voltage source and the node when necessary to provide current, or to recharge the capacitor and can isolate the voltage source from the node during periods when the desired voltage can be maintained by the capacitor.

[0060] The present invention provides significant advantages over the prior art. By isolating the power supply from the operational amplifier during periods where the power supply is not needed to charge the capacitor or an input node to the operation amplifier, the power supply rejection ratio is greatly improved.

[0061] Although the Detailed Description of the invention has been directed to certain exemplary embodiments, various modifications of these embodiments, as well as alternative embodiments, will be suggested to those skilled in the art. The invention encompasses any modifications or alternative embodiments that fall within the scope of the claims.