Phase Noise and Jitter Characterization in Oscillator Applications

Recent advancements in communication systems design place strict demand on system timing accuracy. As frequencies approaches 1GHz, and beyond designers are required to pay special attention to timing variations. Jitter has become a familiar parameter to designers as system speeds accelerate. Jitter is usually characterized in the time domain via metrics such as deterministic jitter, random jitter, cycle-cycle jitter, etc. An alternative and more thorough approach to quantifying timing accuracies is through frequency domain analysis techniques. Phase noise, a common metric for RF engineers now find applicability to common designs as the frequencies of operation extend into the gigahertz range. This article discusses the definition and relationship of jitter and phase noise in an oscillator application. Although jitter is the more prevalent metric used today there is a strong interest in being able to provide additional metrics such as phase noise for those applications where ‘noise’ is paramount. Data will be taken with Aeroflex’s PN9000 at 311MHz frequency in support of SONET/SDH OC-192 applications. Jitter and Phase Noise: Jitter and phase noise are just different ways of quantifying the same phenomenon. As an example of this phenomenon, one of the reference clocks of a SERDES to SONET framer is 311MHz. This clock signal at 311MHz will have a period of 3.21 nanoseconds for one complete cycle. Successive cycles of a ‘noise-free’ waveform will measure exactly 3.21 nanoseconds. Unfortunately, ‘noise-free’ waveforms do not exist. As shown in figure 1, there are variations in the ideal timing event of the period which causes uncertainty about when the next edge of the signal will occur. This uncertainty is referred to as jitter or phase noise. Jitter in Time Domain: Jitter is in the time domain and is defined by the ‘deviation of a timing event of a signal from its ideal position.’ Jitter is the convolution of all independent jitter components defined within a Probability Density Function (PDF). A PDF in simplest terms describes the feasibility of a given measurement relative to all other possible measurements to fall within the known region and is typically represented by a normalized histogram. Jitter includes contributions from both deterministic and random elements. Deterministic Jitter: Is a non-Gaussian PDF and is characterized by a bounded peakpeak value that cannot be statistically analyzed. Figure 1. Waveform Timing Variations There are various sources of deterministic jitter. The following are the most prominent; • Crosstalk; This occurs when incremental inductance from one conductor (signal line) converts induced magnetic field from an adjacent signal line into induced current. This induced current results in either an increase or decrease in voltage resulting in jitter. • Electromagnetic Interference (EMI) radiation; EMI sources include power supplies, AC power lines, and RF-signal sources. Like crosstalk, a noise current is induced on the timing signal path thereby modulating the voltage level of the time signal. • Inter-Symbol Interference (ISI) has two contributors; o Simultaneous switching; This can induce current spikes on power and ground planes, creating an opportunity for shifting of the threshold voltage levels. o Reflections in transmission lines; Causes ISI by changing where the edges occur. A reflection will cause energy to flow back through a conductor that sum with the original signal. This can cause an amplitude change that is different for each conductor in differential pairs creating a time variation in the crossing points, therefore introducing a certain amount of jitter. Random Jitter: Is characterized by a Gaussian distribution and assumed to be unbounded. As a result it generally affects long-term device stability. This long-term device stability is a result of thermal vibrations of semiconductor crystal structure causing mobility to vary depending upon the instantaneous temperature of the material. Additionally, imperfections to semiconductor process variation such as non-uniform doping density all are contributors to random jitter. Because random jitter is Gaussian in nature, the distribution is quantified by the standard deviation or 1σ and mean (μ) as shown in figure 2. This normal distribution yields two common jitter specifications; • Root Mean Squared (RMS) jitter; or the value of one standard deviation of the normal distribution. Since this changes very little as the number of samples increases, it is a more meaningful measurement. However, it is only valid in pure Gaussian distributions. If any deterministic jitter exists in the distribution, the use of 1-sigma based on the entire jitter histogram for the estimation of probability of occurrence is invalid. • Peak-to-peak jitter; or the distance from the smallest to the largest measurement on the normal curve. In most circuits, this value increases with the number of samples taken and can theoretically reach to infinity.