Delay Locked Loops (DLLs) can be used to generate equally spaced multiple clock phases. The phase shifts in a DLL are generated using a delay line that includes a cascade of delay stages or elements where each stage delays the phase a defined amount (e.g., 22.5 degrees, 45 degrees, 90 degrees). The delay provided by each stage is created by an active voltage, current, or digitally control delay element. Most commonly used delay elements have higher amplitude gain at DC compared to its gain at the operating clock frequency. This difference between DC gain and operating frequency gain (or amplitude gain roll-off) amplifies jitter and duty cycle error. Larger delay per delay stage implies higher amplitude gain roll-off and hence higher jitter amplification. Since the delay stages in a delay line are cascaded, the aggregate amplitude gain roll-off of the whole delay line is much higher compared to a single delay element, which results in even more jitter amplification through the delay line.
Reducing the delay provided by each delay element in the delay line reduces the amplitude roll-off of each delay element and the resulting jitter amplification. However, reducing the amount of delay provided by each delay element requires additional delay elements (stages). Increasing the total number of stages required to generate a 180 degree phase shift (which effectively means reducing the amount of delay per stage) increases the power consumption of the DLL. Accordingly, there is a trade-off between jitter amplification and power consumption in the design of DLLs.
Regardless of the number of stages, the lowest jitter amplification that can be achieved with conventional DLLs is 1 (e.g., output jitter is at least equal to the input jitter). Conventional DLLs cannot attenuate incoming jitter on the input clock.
The features and advantages of the various embodiments will become apparent from the following detailed description in which:
Reducing the phase shift (e.g., from 45° to 30° or a lower number) of each delay element reduces the amplitude roll-off at the clock frequency and associated jitter amplification. However, reducing the phase shift may result in additional delay elements being used to generate the 0°-360° clocks (e.g., from 4 to 6). Increasing the number delay elements (which effectively means reducing the amount of delay per stage) may increase the power of the DLL. Accordingly, there may be a tradeoff between jitter amplification and power consumption. Furthermore, the outgoing jitter will at least equal the incoming jitter (jitter amplification factor will be at least 1) as the delay lines 200 do not attenuate incoming jitter on the input clock.
The first delay element 310 may receive at a first pair of inputs (P_a, N_a) the incoming clock signal (ck_in, ck_in#) and may receive at it's a second pair of inputs (P_b, N_b) feedback from end of the delay line (after a 180 degree phase shift). The positive output (180 degrees) of the delay line 300 (last delay element 310) may be provided to a negative input (N-b) of the first delay element 310 and the negative output (360 degrees) of the delay line 300 may be provided to a positive input (P-b) of the first delay element 310.
The two positive and two negative inputs may be shorted together for each of the other delay elements 310. The outputs of the first delay element 310 may be provided as inputs to the second delay element 310 and so on. The positive output of a proceeding delay element 310 may be provided to both positive inputs of succeeding delay element 310 and the negative output of a proceeding delay element 310 may be provided to both negative inputs of succeeding delay element 310. Using the same delay element 310 for each phase shift and shorting the inputs may provide uniformity in the design of the delay elements and reduces variability in operation that may be caused by utilizing different designed delay elements (e.g., using single input delay element for all delay elements except first delay element).
The feedback in the delay line 300 configures the delay line 300 into a regenerative amplifier, the frequency response of which has peaking or resonance at the input frequency when the loop is delay locked. Peaking/resonance at the input frequency results in jitter filtering. The amount of regeneration (or positive feedback) is determined by relative size of devices 410, 430 (a inputs) to the devices 420, 440 (b inputs). The devices 410, 420 may control the strength of the ck_in (0°) with respect to the feedback (360°) and the devices 430, 440 may control the strength of the ck_in# (180°) with respect to the feedback (180°).
The ratio of strength of the “a” inputs (clock signals) to the “b” inputs (feedback signals) controls the amount of regeneration. The strength of the feedback signal may be considered β and the strength of the input signal may be considered 1−β. The larger the value of β the stronger the feedback signal is. Increasing β implies increased regeneration and increased jitter filtering. Increasing β past a certain level (e.g., above 0.5) may configure the delay line 300 into an injection locked oscillator that is tuned by the DLL loop. For β<1, resonant frequency (with or without oscillations) of the delay line 300 may automatically be tuned to the incoming clock frequency by the DLL control loop and jitter filtering may be achieved.
Jitter filtering may be achieved with or without oscillations in the delay line 300. The delay line 300 may work over a wide range of frequencies (e.g., 2 GHz-5.5 GHz) limited only by the lock range of the DLL. The feedback/regeneration may need to be shut-off during startup and lock of a regenerative DLL and enabled only after the DLL has locked.
Although the disclosure has been illustrated by reference to specific embodiments, it will be apparent that the disclosure is not limited thereto as various changes and modifications may be made thereto without departing from the scope. Reference to “one embodiment” or “an embodiment” means that a particular feature, structure or characteristic described therein is included in at least one embodiment. Thus, the appearances of the phrase “in one embodiment” or “in an embodiment” appearing in various places throughout the specification are not necessarily all referring to the same embodiment.
The various embodiments are intended to be protected broadly within the spirit and scope of the appended claims.