Self-bias and differential structure based PLL with fast lockup circuit and current range calibration for process variation

[0002] The present invention relates generally to phase locked loop (PLL) circuits. More particularly, this invention relates to a PLL circuit which is based on a voltage controlled differential oscillator and an advanced common biasing technique, which tolerates process variations and calibrates current ranges for operational request frequency that provides frequency stability with temperature change without the use of a bandgap reference bias circuit.

[0003] Phase-locked loops are often used in the I/O interfaces of digital integrated circuits in order to hide clock distribution delays and to improve overall system timing. The maintenance of the timing throughout a circuit is important. Timing becomes particularly critical for applications requiring high-speed processing of information, such as with video processors.

[0004] The timing throughout a circuit deviates from the system clock when noise is introduced by various system components and as a result of capacitive effects due to system interconnections. In recent years, the demand has risen for devices capable of high-speed processing. As a result, the demand for PLL circuits that quickly compensate for electronic noise and capacitive delays has also risen. The problem is that the amount of phase shift produced as a result of the supply, substrate noise and capacitor load is directly related to how quickly the PLL can correct the output frequency.

[0005] One type of design used by those skilled in the art to eliminate the noise present in the circuit at the required speed is a self-bias signal technique. Referring to FIG. 1, this prior art PLL circuit is a self-biasing configuration which is composed of a phase comparator, charge pump, loop filter, bias generator and a voltage-controlled oscillator (VCO). This PLL circuit also uses an additional charge pump current to the bias generator V_bpoutput. For a typical PLL, the charge pump current and the loop filter resistance are constant. These conditions give rise to a constant damping factor and a constant loop bandwidth. A constant loop bandwidth can constrain the achievement of a wide operating frequency range and low input tracking jitter. If the frequency is disturbed, the phase error that results from each cycle of the disturbance will accumulate for many cycles until the loop can compensate for the frequency error. The error will be accumulated for a number of cycles, which is proportional to the operating frequency divided by the loop bandwidth. Thus the loop bandwidth would have to be positioned as close as possible to the reference frequency bandwidth to minimize the total phase error. The result is that the frequency bandwidth must be conservatively set for stability at the lowest operating frequency with worst case process variations, rather than set for optimized jitter performance. The self-biased PLL also exhibits much faster locking times only when locking from similar or higher operating frequencies. If, however, the self-biased PLL is started at a very low operating frequency, it will exhibit very slow locking times.

[0006] Accordingly, there is a need for a PLL circuit which provides a fast lock-up, improved jitter performance, tolerates process variations, and extends the PLL operating frequency range.

[0007] The present invention is a phase locked loop (PLL) which is based on a common bias technique, comprising a voltage controlled differential oscillator, fast lock-up circuit, self-calibration current range setting circuit and a high-speed phase frequency detector. This design provides improved speed in locking to the frequency of an incoming signal, extends the PLL operating frequency range, improves PLL jitter performance, and provides greater immunity to environmental noise, which results in improved power supply rejection ratio (PSRR).

[0008] FIG. 1 is a typical PLL block diagram;

[0009] FIG. 2 is a block diagram of the circuit in accordance with the preferred embodiment of the present invention;

[0010] FIG. 3A is a logic gate diagram of the speed-up circuit in accordance with the preferred embodiment of the present invention;

[0011] FIG. 3B is a state diagram of the speed-up circuit inputs in accordance with the preferred embodiment of the present invention;

[0012] FIG. 3C is a logic table for the speed-up circuit in accordance with the preferred embodiment of the present invention;

[0013] FIG. 4A is a diagram of the bias generator;

[0014] FIG. 4B is a diagram of a delay element of the VCO in accordance with the preferred embodiment of the present invention (only one is shown though there are a plurality of delay cells 20_a. . . 20_nused);

[0015] FIG. 5 is an example of a graph of four I-V curves in accordance with the preferred embodiment of the present invention;

[0016] FIG. 6 is a flow diagram in accordance with the preferred embodiment of the present invention.

[0017] The first embodiment of the present invention will be described with reference to the drawing figures wherein like numerals represent like elements throughout.

[0018] Referring to FIG. 2, a PLL circuit 10 made in accordance with the present invention is shown. The PLL circuit 10 includes a phase frequency detector (PFD) 11, a speed-up circuit 12, a charge pump 13, a charge pump control circuit 14, a loop filter 15, a bias generator 16, a current range control circuit 17, a voltage-controlled oscillator (VCO) 18, and a frequency divider 19. The PLL circuit 10 receives a reference frequency F_reffrom an outside source. Along with the reference frequency F_ref, the phase frequency detector (PFD) 11 receives a reset signal PLL_resetand the divided PLL 10 output frequency signal F_O/N. Coupled to the phase frequency detector 11 are the speed-up circuit 12 and the charge pump 13. As those skilled in the art should know, the PFD 11 determines the phase and frequency difference between the reference frequency F_refand the divided PLL 10 output signal F_O/N. If the PFD 11 detects a difference between the two input signals F_ref, F_O/N, the phase error signals U, D are sent to the speed-up circuit 12 and the charge pump 13. The duration of the output signals U, D pulse widths depend on the amount of phase and frequency error that is detected by the PFD 11.

[0019] The speed-up circuit 12, shown in FIG. 3A, is controlled by the phase error signals U, D output from the PFD 11. Coupled to the PFD 11 and the charge pump control switch 14, the speed-up circuit 12 receives the phase error signals U, D and a reset signal PLL_reset. The purpose of the speed-up circuit 12 is to monitor the crossing of the PLL 10 output frequency F_O/N and the reference frequency F_ref, as will be disclosed hereinafter. The speed-up circuit 12 comprises a plurality of invertors 3, a plurality of AND gates 5, two S-R flip flops 7, a NAND gate 8, and a delay stage 9. Initially, when the PLL circuit 10 receives a reset signal PLL_reset, the speed-up circuit 12 signals the charge pump control switch 14 to adjust the charge pump 13 current to the maximum. When the speed-up circuit 12 detects the crossing of the two frequencies F_O, F_ref, the speed-up circuit 12 signals the charge pump control switch 14 to reduce the charge pump 13 current. The output of the speed-up circuit 12, as the PLL circuit 10 attempts to match the output frequency F_O/N with the reference frequency F_ref, is illustrated in the state diagram of FIG. 3B and the logic table of FIG. 3C.

[0020] Referring to FIGS. 3B and 3C, when the speed-up circuit 12 receives the reset signal PLL_resetand the reset PFD 11 outputs U, D, the speed-up circuit 12 signals the charge pump 13 to output its maximum current, thus discharging the loop filter capacitor C1 to 0V. This triggers the VCO 18 to output its maximum frequency F_O. The PLL 10 output frequency F_O/N will decrease as the capacitor C1 charges and will eventually be equal to the reference frequency F_ref. This condition is identified by U, D changing from 00,01,11 to 00,10,11. As the PLL 10 comes out of reset, the signals U, D are in the state 0,0. Since the PLL 10 output frequency F_O/N is faster than the reference frequency F_ref, the next state of U, D is 0, 1. This indicates to the charge pump 13 to charge the capacitor C1, reducing the PLL 10 output frequency F_O/N. On the next rising edge of the reference frequency F_ref, the signals U, D go to 1, 1. This state causes the PFD 11 to reset and return U, D to its neutral state 0, 0. This process continues until the PLL 10 output frequency F_O/N is slower than the reference frequency F_ref. This condition is indicated by the U, D signals going to a state 1, 0. This condition signals the speed-up circuit 12 to output a logical zero (0) Q₃to the charge pump control circuit 14, indicating that the charge pump 13 should reduce its current to the level prescribed by the charge control signal CC_signalto be disclosed hereinafter.

[0021] The speed-up circuit 12 operates to quickly detect the matching of the reference and output frequencies F_ref, F_O/N, and then output a control signal Q₃to the charge pump control circuit 14 to reduce the current of the charge pump 13 in order to find the optimal bias current range I_bias, as will be disclosed hereinafter, to be output to the VCO 18. This speed-up circuit 12 provides a fast frequency lock by signaling the initialization of the charge pump 13 current output to its maximum and signaling for the reduction of this current when the frequencies F_refand F_O/N are equal.

[0022] Referring back to FIG. 2, the control signal Q₃output from the speed-up circuit 12 is a logical one or a logical zero and is received by the charge pump control circuit 14, which is coupled to the speed-up circuit 12 and the charge pump 13. The charge pump control circuit 14, using a variation of switches, converts the signal from the speed-up circuit 12 and the charge control signal CC_signalto logic signals S1, S2, S3, which will adjust the amount of current the charge pump 13 outputs. Initially, when the speed-up circuit 12 indicates that the charge pump 13 should output its maximum current, by outputting a logical one (1) signal Q₃, the logic signals S1, S2, S3 will be equivalent to 1, 1, 1, respectively. When the speed-up circuit 12 outputs a logical zero (0) signal Q₃, indicating that the charge pump 13 should reduce its current, the logic signals S1, S2, S3 output from the charge pump control circuit 14 will be equivalent to 1, 1, 0 or 1, 0, 1, respectively, for example. This output S1, S2, S3 is set by the charge control signal CC_signal, an outside input signal whose value depends on the operation for which the PLL 10 output is to be used. The charge pump control circuit 14 converts the charge control signal CC_signalto the logic signals S1, S2, S3. This instructs the charge pump 13 to switch out an internal current source (not shown) when the logic signal S1, S2, S3 associated with the current source is zero (0), thereby dividing the current of the charge pump 13 by a number m (e.g., 3, 6, or 9).

[0023] The charge pump control circuit 14 provides the PLL 10 with the ability to slew the output frequency F_Otoward lock at the fastest rate possible, instead of at a constant rate, using the maximum charge pump 13 output current. As should be well known to those having skill in the art, the faster damping of the PLL 10 output frequency F_Ois a result of the high frequency response to the error signals U, D from the PFD 11. Once the PLL circuit 10 outputs the desired frequency, the charge pump control circuit 14 reduces the charge pump 13 output current, which reduces the output frequency response of the PLL 10 and improves jitter performance.

[0024] The charge pump 13 outputs a current that charges or discharges the capacitors C1, C2 of the loop filter 15 to a voltage level VLPF. The charge pump 13 receives the error signals U, D, a voltage input V_bpfrom the common bias generator 16, and the logic signals S1, S2, S3 from the charge pump control circuit 14. V_bpis a reference bias voltage which controls the charge pump output current. It is well known to those skilled in the art that the charging and discharging of the loop filter 15 capacitors C1, C2 create a voltage change V_LPFacross the loop filter 15. As V_LPFincreases, the charge pump 13 output current decreases. This voltage change V_LPFis a reference for the common bias generator 16 and the current control circuit 17 that generates the reference signals for controlling the amount of delay created by each delay element 20_a. . . 20_nof the VCO 18.

[0025] The loop filter 15 comprises a first capacitor C1 and a second capacitor C2 with a variable resistor 22. The variable resistor 22 comprises a symmetric load. The loop filter 15 is coupled to the common bias generator 16 as well as the charge pump 13. This loop filter 15 receives a current output generated by the charge pump 13 using the bias voltage V_bp, and an input voltage V_dd. As the current from the charge pump 13 shifts to adjust the frequency of the delay cells 20_a. . . 20_n, the variable resistor 22 is also adjusted. As those skilled in the art should know, the use of the variable resistor 22 in the loop filter 15 allows the bias generator 16 to maintain stability over a wide bandwidth.

[0026] The current range control circuit 17, coupled to the loop filter 15 and the bias generator 16, comprises two voltage comparators (not shown) and provides two (2) single bit outputs R0, R1. The current range control circuit 17 receives the loop filter output voltage V_LPFand adjusts the level of the bias current I_biasgenerated in the common bias generator 16 through the output leads R0, R1. The voltage comparators in the current range control circuit 17 monitor the voltage V_LPFagainst an internal reference voltage V_refwhich is indicative of the point where a small change in the voltage V_LPFcauses a large change in the bias current I_bias, as disclosed hereinafter. The reference voltage V_ref, which can be fixed or variable, is illustrated in the FIG. 5 I_biasI-V curves 60-66. As those skilled in the art should know, each of these curves 60-66 includes an approximate narrow linear region, illustrated by the curves up to the voltage level V_refin FIG. 5, and a nonlinear region. The nonlinear region, where a small increase in the voltage creates a large decrease in the current, produces a large change in the delay elements 20_a. . . 20_nof the VCO 18, causing increased jitter within the PLL circuit 10. If the PLL circuit 10 is operated past a certain voltage point V_refon the curves, into the nonlinear region, the amount of jitter will increase and the overall performance of the PLL circuit 10 will decrease. Therefore, when the voltage V_LPFis greater than the reference voltage V_ref, PFD 11, speed-up circuit 12, and the charge pump control circuit 14 receive a local reset signal L_resetfrom the current range control circuit 17 which resets the voltage V_LPFto zero. The current range control circuit 17 outputs R0, R1 switch, adjusting the bias current I_biasto a lower level, as will be described hereinafter.

[0027] Initially, when the PLL circuit 10 is reset, the current range control circuit 17 signals the bias generator 16 to switch on all four currents sources I₁, I₂, I₃, I₄, which is indicated by the output 0, 0 for the two outputs R0, R1, respectively (shown in FIG. 4A). When the speed-up circuit 12 detects the crossing of the reference and output frequencies F_ref, F_O/N, the current control circuit 17 is adjusted to find the optimal bias current I_biascurve to be utilized by the bias generator 16 for output to the VCO 18. As disclosed above, when the voltage comparators in the current range control circuit 17 detect that V_LPFis too high, the current range outputs R0, R1 switch by one (1). For example, after the resetting of the PLL circuit 10, the current range outputs R0, R1 are equivalent to 0, 0. When the comparator detects the high voltage V_LPF, the current range outputs R0, R1 switch to 0, 1, respectively, which indicates that the common bias generator 16 should switch off current source I₄. The current range control circuit 17 also outputs an internal reset signal L_resetto the PFD 11, speed-up circuit 12, and the charge pump control circuit 14, which resets the loop filter 15 voltage V_LPFto zero (0). Each time the voltage comparators of the loop filter 15 detect this condition, the current range outputs R0, R1 increase by one (1), and another current source is switched off. This process continues until R0, R1 is equivalent to 1, 1, where the only remaining current source is I₁. At this point, the current range control circuit 17 signals to the common bias generator 16 to remain at the lowest current level I₁.

[0028] If the reference voltage V_refis greater than the loop filter voltage V_LPF, the current range control circuit 17 indicates to the bias generator 16 to remain at the present current level, which is considered the optimal operating point for the process utilizing the output frequency signal F_O. As those skilled in the art should know, even though the current range control circuit 17 is illustrated utilizing two (2) single bit outputs R0, R1, a single two-bit output may also be utilized, or any signaling scheme which provides a selective control output.

[0029] Referring to FIGS. 4A and 4B, the present invention utilizes a single common bias generator 16 comprising a bias current generator 56, a bias voltage generator 58 and a differential amplifier 54. The differential amplifier 54 reduces the noise from the power supply, which improves the power supply rejection ratio performance of the PLL circuit 10. The bias current generator 56 includes four symmetric loads 70, 72, 74, 76 and a switching circuit 57. Each of the symmetric loads 70, 72, 74, 76 (which may or may not have the same device widths), has as its voltage source V_ddand the low pass filter 15 output signal V_LPFas its gate voltage. The four symmetric loads 70, 72, 74, 76 act as current sources I₁, I₂, I₃, I₄. These current sources I₁, I₂, I₃, I₄are switched on and off by the switching circuit 57. The current range control circuit 17 outputs R0, R1 dictate to the switching circuit 57 which of the current sources I₁, I₂, I₃, I₄should be on and which should be off. As should be known to those skilled in the art, there are numerous switching circuits that are responsive to a digital input and can be used as described herein. In this manner, the current range control circuit 17 provides four (4) discrete levels of bias current I_biasto control the delay elements 20_a. . . 20_nwithin the VCO 18. This allows for better control of the PLL circuit 10 by providing the flexibility of varying the current slope without having to vary the delay cell itself. The bias current I_biasis output from the bias current generator 56 to the bias voltage generator 58.

[0030] The bias voltage generator 58 comprises two n channel transistors 51, 52 and one p channel transistor 53. The gate voltage to n channel transistors 51, 52 is connected to a differential amplifier 54. This differential amplifier 54 eliminates the noise generated by power supply voltage V_dd. The n channel transistors 51, 52 are configured such that the bias current I_biasthrough transistor 51 is mirrored onto transistor 52 and reflected up to transistor 53. When the bias current I_biasis mirrored in this way, two reference voltages are created. The two reference voltages are the bias voltages V_bnand V_bpwhich determine the amount of delay for each delay element 20_a. . . 20_nwithin the VCO 18. Using this common bias generator 16 and a selectable number of delay cells will provide better linearity in the overall delay.

[0031] The voltage controlled oscillator (VCO) 18 is coupled to the common bias generator 16 and a frequency divider 19. Although only one delay element 20_ais shown for clarity, it should be understood that the VCO 18 includes a plurality of differential delay elements 20_a. . . 20_n. The delay elements 20_a. . . 20_nare configured in such a way that the voltage inputs V₊, V₋of the delay elements come from the voltage outputs V_O⁺, V_O⁻of the preceding delay elements. The voltage outputs V_O⁺, V_O⁻from the last delay element 20_nare feedback to the voltage inputs V⁺, V³¹of the first delay element 20_a. This configuration generates the desired output frequency F_O.

[0032] Each delay cell 20_a. . . 20_ncontains two p channel transistors 21 and 22 and three n channel transistors 23, 24, 25. Transistors 23, 24 act as switches in the delay cell 20_aand determine the actual delay for each element based on the bias current I_biasvia the voltages V_bnand V_bp. The p channel transistors 21, 22 act as current sources for the transistors 23, 24. Transistor 25 acts as a current source as well. The p channel transistors 21, 22 are biased by the voltage V_bp. Since the amount of bias current I_biasdetermines the voltage level of V_bp, the delay element delay time changes with the V_bp. Transistors 23 and 24 receive a voltage input V⁺and V⁻. The current supplied by transistor 22 does not pass through transistor 24 when transistor 24 is “off” (or not conducting). Likewise, when transistor 23 is not conducting, the current provided by transistor 21 does not pass through transistor 23. There are parasitic capacitances at the inputs of transistors 23 and 24 that charge and discharge to affect the voltages V⁺and V⁻, which rise and fall. When transistors 23 and 24 are on and off, respectively, the charge on the parasitic capacitors at the input of transistors 23 and 24 on the subsequent delay cell will be affected. When transistor 23 is turned on, it discharges the parasitic capacitances of the next delay cell and V_O−changes from (V_dd−Vds 21) to (0V+V_ds25+V_ds23). Likewise, when transistor 24 is off, transistor 22 charges the capacitance of the following delay cell and V_O⁺changes from (0V+V_ds25+V_ds24) to (V_dd−Vds 22), Vds 22 at saturation. As is well known to those skilled in the art, the delay provided by the delay cell is equivalent to the duration between turning on transistor 23 and turning off transistor 24, and when the voltages V⁺and V⁻are equal. When this point is reached, the transistors in the next delay cell are activated. V_O⁺and V_O⁻are the output voltages of each delay cell that provide the input voltages V⁺, V⁻to the next delay cell.

[0033] The frequency output from the VCO 18 is then input to a frequency divider 19. Since the output frequency F_O, is a multiplied version of the reference frequency F_ref, by a factor of N times, the frequency divider 19 eliminates this N factor for comparison to the reference frequency F_ref.

[0034] The flow diagram in accordance with the present invention is illustrated in FIG. 6. The PLL 10 receives a reset signal PLL_resetfrom an external source and resets all PLL circuit 10 components (step 700). Upon receipt of this reset signal PLL_reset, the PFD 11 resets the up and down output signals U, D (step 701). The speed-up circuit 12 receives the reset signal PLL_resetas well as the output signals of the PFD 11 and sends a signal to the charge pump control circuit 14 to boost the current to maximum (step 702). The charge pump control circuit 14, upon receipt of the signal from the speed-up circuit 12, outputs logic signals S1, S2, S3 to the charge pump 13 to adjust the current in the charge pump 13 to its maximum (step 703). The VCO 18 is set to the maximum output frequency (step 704). The maximum output frequency F_Ois then forwarded to the frequency divider 19 where it outputs F_O/N to the phase frequency detector 11 for comparison to the reference frequency F_ref(step 709). If F_O/N is equal to F_ref, the current range control circuit 17 locks at the present current level (step 710) and the charge pump control circuit 14 reduces the charge pump 13 current output (step 711). Once the current output of the charge pump 13 is reduced, the PLL circuit 10 is in the lock range position (step 712). If the two frequencies F_O/N, F_refare not equal, the PFD 11 outputs signals U, D whose duration depend on the amount of phase and frequency error that is detected between the two frequency signals F_ref, F_O/N (step 705). The charge pump 13 receives the control signals U, D, S1, S2, S3 from both the PFD 11 and the charge pump control circuit 14 and outputs a current, which sources or sinks the loop filter capacitors C1, C2 (step 706). As a result of the charging or discharging of the loop filter capacitors C1, C2, the control voltage V_LPFis generated (step 707). If the loop filter voltage V_LPFis greater than V_refand the current range control circuit 17 is not at the minimum current level, the current control circuit 17 outputs a local reset signal L_resetto the loop filter 15, which resets the voltage V_LPFto zero (0) (step 707a). If the current range control circuit 17 is at the minimum current level, the common bias generator 16 outputs the bias current I_biasand bias voltages V_bp, V_bnto the VCO 18 (step 708), which generates the output frequency F_O(step 709). Since the current range control circuit 17 is at its minimum current level I₁, the PLL circuit is in the lock range position (step 712). If the current range control circuit 17 is not at its minimum current level I₁, the current range control outputs R0, R1 switch to the next lower level (step 707b). Because of the local reset signal of the current range control circuit 17, the VCO is set to its maximum frequency output (step 704). The described process continues until F_refis equal to F_O/N.

[0035] This design of the PLL circuit 10, in accordance with the preferred embodiment, will achieve a wide operating frequency range with a fast lock up circuit and good jitter performance over a wide power supply voltage range and short lock in time. The differential VCO 18, operating in the biasing current mode, provides a much wider operating frequency range with high common-mode noise immunity. The common biasing technique provides the necessary bias with less sensitivity to temperature and process variations. It also provides better power supply rejection ratio and current range calibration regulation when the power supply droops or when process variations change.

[0036] While a specific embodiment of the present invention has been shown and described, many modifications and variations can be made by one skilled in the art without departing from the spirit and scope of the invention. The above description serves to illustrate and not limit the particular form in any way.