Field of the Invention
The present invention relates to phase detection circuits as may be used in phase locked loops.
Background of the Invention
Phase detection circuits are typically used to detect a phase difference between two signals. These circuits can be used in phase locked loops (PLL) to detect a difference between a reference signal generated by a reference oscillator and a loop feedback signal. The output of the phase detection circuit is then used to produce a fixed phase relationship between the PLL output signal and the input reference signal. PLLs are discussed in H. Krauss, C. Bostian, and F. Raab, Solid State Radio Engineering at Chapter 6 (1980).
A typical three state phase detector circuit is illustrated in FIG. 1. The D inputs to both flip-flops is tied to a logic high. The reference signal of reference frequency F.sub.r clocks one flip-flop and the loop feedback signal of variable frequency F.sub.v clocks the other flip-flop. The outputs of the flip-flops are input to an AND gate. The output of the AND gate is input to a delay element the output of which is connected to the reset inputs of the flip-flops. The delay element causes the outputs of the flip-flops to have a minimum pulse width, even during a zero phase error condition. The outputs of the flip-flops also each enable a charge pump. The charge pumps are connected at a common node that is the output of the circuit.
Assuming both flip-flops start in the zero state, the circuit operates by the F.sub.r and F.sub.v signals clocking in the logic high from the flip-flop inputs. If F.sub.r leads F.sub.v, as illustrated in FIG. 2A, the positive polarity charge pump is turned on by the output from that D flip-flop causing a positive pulse to be output from the circuit. The width of the net output pulse is the difference between the time the F.sub.r flip-flop is clocked to the time the F.sub.v flip-flop is clocked. Once the F.sub.v flip-flop is clocked, the inputs to the AND gate will both be a logic one causing the output of the AND gate to be a logic one, thereby reseting the flip-flops after a delay, turning off the charge pumps and ending the output pulses from the flip-flops.
If F.sub.r lags F.sub.v, as illustrated in FIG. 2B, the negative polarity charge pump is turned on by the output of the F.sub.v flip-flop, causing a negative pulse to be output from the circuit. The width of the net output pulse is the time between the F.sub.v flip-flop being clocked and the F.sub.r flip-flop being clocked. Once the F.sub.r flip-flop is clocked, both inputs to the AND gate will be a logic one causing the output of the AND gate to be a logic one and turning off the charge pumps. The logic one output from the AND gate will reset the flip-flops after a delay, thereby turning off the charge pumps and ending the output pulses from the flip-flops.
If F.sub.r and F.sub.v are in phase, as illustrated in FIG. 2C, both charge pumps will be turned on at the same time, producing a net output current of zero. Ideal charge pumps used in an ideal phase detector will produce no net charge in this state because both currents would be exactly balanced in phase and amplitude and have infinitely fast rise and fall times.
The output of this circuit is low pass filtered before driving a voltage controlled oscillator, the frequency divided output of which is the loop feedback signal. The positive or negative pulse from the phase detection circuit will cause the VCO to change frequency and phase until F.sub.v =F.sub.r both in frequency and phase. At this point, the output from the phase detection circuit will be a zero value and the VCO will be locked on that particular frequency.
Present phase detection circuits used in PLLs typically have a linear transfer function. Around the zero phase error region, however, there can exist a non-linear transfer function of phase to current or charge. This non-linearity degrades the spectral purity of the PLLs signal, especially when used with fractional division, by increasing the amplitude of the sidebands at subharmonics and harmonics of the output frequency resolution. See Martin, U.S. Pat. No. 4,816,774 assigned to Motorola, for a discussion on output frequency resolution. There is a resulting need for a phase detection circuit that has a more linear transfer function of phase to current or charge for small phase differences, thereby attenuating these sidebands.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide linear phase detection for small phase differences.
The present invention is comprised of two D type flip-flops, an AND gate, and a delay element. The inputs to the D flip-flops are tied to a logic high. The first flip-flop is clocked with a reference signal while the other flip-flop is clocked with a feedback signal, typically from a voltage controlled oscillator in a phase locked loop. The outputs of the flip-flops are ANDed together with the result of this operation going through a delay element before reseting one of the flip-flops. The other flip-flop is reset by the output of the AND gate without the delay element. Each flip-flop output enables a charge pump--one negative polarity and one positive polarity.
The present invention will maintain a lock condition in a phase locked loop by extending the DOWN pulse enabling the negative polarity charge pump to the same width as the UP pulse that enables the positive pump. This will create a net zero charge from the present invention.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 shows a prior art phase detection circuit.
FIGS. 2A, 2B, 2C show timing diagrams generated by the prior art phase detection circuit shown in FIG. 1.
FIG. 3 shows a block diagram of the present invention.
FIG. 4 shows a block diagram of the present invention as used in a phase locked loop.
FIGS. 5A, 5B, 5C, 5D, 5E, and 5F show timing diagrams generated by the present invention.
FIG. 6 shows an alternate embodiment of the present invention.
FIG. 7 shows another alternate embodiment of the present invention.
FIG. 8 shows yet another alternate embodiment of the present invention.
FIG. 9 shows yet another alternate embodiment of the present invention.
FIG. 10 shows a plot of the transfer function for the prior art and the present invention.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
The present invention detects phase differences between two signals, and is particularly useful in detecting small phase differences while maintaining a substantially linear transfer function of phase to current or charge. The present invention is used for phase detection in phase locked loops (PLL).
As illustrated in FIG. 3, the present invention is comprised of two D type flip-flops (301 and 302), an AND gate (303), and a delay element (304). The flip-flops (301 and 302), with their D inputs tied to a logic high, are clocked by two separate signals. One is clocked by a reference signal having a frequency F.sub.r and the other is clocked by a variable frequency signal F.sub.v. If the present invention is used in a PLL (400), as illustrated in FIG. 4, F.sub.r is generated by a reference oscillator and F.sub.v is the loop feedback output of the frequency divider (401). The outputs of the flip-flops (301 and 302) are ANDed and the resulting AND gate (303) output delayed by the delay element (304). The output of the delay element (304) is used to reset one of the flip-flops (301) while the output from the AND gate (303) prior to the delay element (304) resets the other flip-flop (302); it is irrelevant which flip-flop (301 or 302) gets reset by the delayed AND gate (303) output. Each of the flip-flop outputs is used to enable a charge pump--a negative polarity source (306) and a positive polarity source (305). The output of the present invention is the common node between the two charge pumps (305 and 306).
Assuming both flip-flops (301 and 302) start in the zero state, the present invention operates by the F.sub.r and F.sub.v signals clocking in the logic high from the flip-flop inputs. The F.sub.v signal clocks the F.sub.v flip-flop (302) to a high state causing a high pulse on the DOWN line. Similarly, the F.sub.r signal clocks the F.sub.r flip-flop (301) to a high state causing a high pulse on the UP line. These pulses enable the negative and positive polarity charge pumps (305 and 306) respectively. The width of any negative pulse will be the difference between the time the F.sub.v flip-flop (302) is clocked to the time the F.sub.r flip-flop (301) is clocked by the rising edge of F.sub.r. Once the F.sub.r flip-flop (301) is clocked, the inputs to the AND gate (303) will both be a logic one, thereby resetting the F.sub.v flip-flop (302) immediately and resetting the F.sub.r flip-flop (301) after a delay of T.sub.delay. A positive output current pulse is thus produced that will always be T.sub.delay wide due to the T.sub.delay before the Fr flip-flop (301) is reset.
When F.sub.v leads F.sub.r by exactly T.sub.delay, as illustrated in FIG. 5A, the width of the positive and negative current pulses are both equal to T.sub.delay so that the net current at the output, integrated over both pulses, is zero. This condition occurs when the PLL is in phase lock.
When F.sub.v leads F.sub.r by more than T.sub.delay, as illustrated in FIG. 5B, the negative output pulse is wider than the positive output pulse; the positive output pulse having a width of T.sub.delay. In this case, the net current produced by the present invention, integrated over both pulses, is negative.
When F.sub.v leads F.sub.r by less than T.sub.delay, as illustrated in FIG. 5C, the negative output pulse is narrower than the positive output pulse; the positive output pulse having a width of T.sub.delay. The net current produced by the present invention in this case, integrated over both pulses, is positive.
It is also possible for F.sub.v to lag F.sub.r, as illustrated in FIG. 5D. In this case, the PLL is greatly out of lock. The UP pulse produced is wider than T.sub.delay by the amount of time that F.sub.r leads F.sub.v, while the DOWN pulse is only as wide as the sum of the inherent delays of the circuit elements.
In an actual circuit, the inherent delay of the flip-flops (301 and 302) and AND gate (303) will make the width of the positive pulse slightly more than T.sub.delay and the negative pulse will be active slightly beyond when F.sub.r goes high. This does not result in behavior any different from the idealized case previously described except that the positive and negative pulses overlap for a time equal to the inherent delay of the flip-flops (301 and 302) and AND gate (303).
When the present invention is used in a PLL (400) to maintain a lock condition, the DOWN pulse extends to the same width as the UP pulse, creating a net zero charge from the pump. The DOWN pulse, however, will always lead the UP pulse in phase by T.sub.delay. Any phase modulation in the PLL is done by the leading edge (501) of the DOWN pulse, as seen by the moving leading edge (501), as long as the DOWN pulse width is maintained greater than PW.sub.minDOWN as shown in FIG. 5E. This, in effect, produces a fixed phase error between F.sub.r and F.sub.v. The remaining non-linearity in the present invention will be due to the DOWN path alone and will be small if the modulation is small compared to PW.sub.minUP.
The composite current output of the present invention (300) produced by summing the UP/DOWN pulses in opposite phase, as illustrated in FIG. 5F, will produce reference spurs that will modulate the VCO (402) in the PLL (400). While this will be acceptable in many applications, some applications will not accept this. This can be remedied by the alternate embodiments illustrated in FIGS. 6 and 7. A delay element (601 and 701) of delay T.sub.delay can be inserted in the path of the flip-flop that is being reset with the AND gate output that is not delayed. This delay will bring the two pulses from the flip-flops back into phase, thereby eliminating any induced VCO modulation caused by non-overlapping pulses. An alternate embodiment can also have a delay (801 and 802) in both reset signals, as illustrated in FIG. 8. Yet another alternate embodiment, illustrated in FIG. 9, utilizes a delay (901) of T.sub.delay in the DOWN path to bring the two pulses from the flip-flops into phase.
FIG. 10 illustrates both the prior art phase to charge transfer function (1001) and the present invention's phase to charge transfer function (1002). The prior art transfer function (1001) is non-linear at the zero phase error region while the present invention transfer function (1002) is linear in this region.
In summary, the present invention has been shown to detect small phase differences between two signals. The present invention, when used in a PLL having fractional division, will maintain a linear transfer function of phase to current or charge at and around the zero phase error region.