The Role of Jitter in Timing Signals
Total Page：16
File Type：pdf, Size：1020Kb
® WHITE PAPER The Role of Jitter in Timing Signals Timing signal jitter can have a profound impact on a wide variety of applications from analog radio frequency (RF) or audio-to-digital communications. While information in a communications system, for example, is extracted from serial data streams by sampling the data signal at specific points, the presence of small quantities of jitter can alter the edge positions enough to lead to data errors and high bit-error rates. Complicating the measurement of jitter is the way it is specified. Generally, timing signal jitter is specified in different ways from one application environment to another. For example, in digital data path synchronization applications, jitter is specified in the time domain as period or cycle-to-cycle jitter. In data communications applications, jitter is specified in the frequency domain as phase noise or “root mean square” (RMS) jitter. This paper provides a basic tutorial on timing signal jitter for designers building electronics systems. It defines this phenomenon and describes how it is measured in different applications. As part of that process, it explains the many ways timing jitter is specified and the differences between terms such as period jitter, cycle-to-cycle jitter and absolute period jitter. It then uses some common applications, such as the use of first-in/first-out (FIFO) memory devices in data acquisition subsystems, to describe the impact of jitter and basic error detection and recovery strategies. Defining jitter A timing signal is defined as a repetitive digital signal, as shown in figure 1. Typically, these signals are used to synchronize the transfer of data between two points, but their role varies widely from application to application. In digital chips, they are usually employed to sequence data through a set of processing stages in a pipeline. Analog applications, in contrast, may use a timing signal to sample a digital-to-analog converter (DAC) to digitize an audio signal. Communications systems designers use timing signals to synchronize the transmission of data over high-speed serial links, such as local area networks (LANs) and wide area networks (WANs). However, designers of RF systems use timing signals in the front end of their design to convert the modulated analog signal into the digital domain for digital signal processing (DSP). In each of these applications, the performance and reliability of the system can be dramatically affected by the quality of the timing signals used. IDT® THE ROLE OF JITTER IN TIMING SIGNALS 1 Figure 1. Perfect timing signal Cycle 1 Cycle 2 Cycle n B C Cycle 1 = 2 = n A A = Period, 1/A = Frequency, B = C, B + C = A A = xxx ns A = 1 unit interval A = 2 pi radians A = 360 degrees Figure 1. Perfect timing signal A perfect timing signal, such as the one illustrated in figure 1, has a fixed period and duty cycle that never varies over time and a starting point that is fixed in time. In reality, however, all timing signals exhibit small variations. These variations in phase position, period and duty cycle are generally called “jitter.” To measure jitter in the time domain, the timing signal is typically compared to a “perfect reference signal.” The variations of the timing signal’s risingexactly and falling edges as compared to the perfect reference are defined as “jitter” and are often specified as time measured from the ideal signal’s edge locations (see Figure 2. Timing signal with jitter figure 2). These measurements are normally specified in time units, such as picoseconds (10-12 seconds). Perfect reference timing signal Timing signal with jitter Jitter Jitter Jitter Note: Jitter around each edge can be different and may not be equally distributed around an edge. Also jitter in the next periods can be different. Figure 2. Timing signal with jitter IDT® THE ROLE OF JITTER IN TIMING SIGNALS 2 There are two ways to look at jitter in the time domain. The most intuitive method is to directly compare the reference signal to the jittering timing signal. The other way is in the frequency domain where the power level of the noise or jitter is compared to the power level of the fundamental signal’s power level. In practice, these preferred schemes are not often used because it is difficult to find a perfect reference signal at the desired frequency or access the equipment needed to perform precise spectrum analysis. Accordingly, most engineers do the next best thing. They look at the timing signal with an oscilloscope and observe the period-to-period or cycle-to-cycle variations. With the scope set up to trigger on every rising or falling edge, running in infinite persistence mode and allowed to trace sufficient cycles, it is possible to determine the maximum and minimum periods of the timing signal. Digital scopes can also store a finite number of period durations, and post-processing software can analyze the data to find the maximum and minimum periods. Designers using this scheme face several shortcomings related to detecting jitter: large unit values (due to finite sample time), accumulated jitter and phase shifts relative to the missing perfect reference. To address these limitations, engineers rely on basic time domain jitter measurements that are referred to as period jitter, cycle-to-cycle jitter or absolute period jitter. Period jitter is the maximum jitter observed at the end of a period’s edge when compared to the position of the perfect reference clock’s edges. The number of cycles used to look for the maximum jitter varies by application; the JEDEC specification is a minimum of 10,000 observed cycles. Cycle-to-cycle jitter is the maximum observed variation between two adjacent cycle periods over a defined number of observed cycles. The number of cycles observed is application dependent; the JEDEC specification is a minimum of 1,000 observed cycles. Absolute period jitter is the maximum jitter observed for the test signal when compared to the perfect reference edges in the application environment. Sometimes this is also called the 6-sigma jitter value. In many applications, these three measurements can provide all the data a designer needs to meet jitter requirements. In some cases, only a single measurement is required. For example, when a timing signal is used to drive a digital chip, an absolute period jitter specification may be sufficient to meet all requirements. Here, data must be processed and transferred between pipeline stages with the timing signal. If the timing signal exhibits too short a period due to jitter, there may not be sufficient time to process the data in the pipeline stage, and have it set up and ready for the next stage. Since these digital chips have no embedded error detection or correction capability, once a single stage is corrupted, all further processing stages will operate on false data and proceed without detection of the problem. Thus, meeting the minimum absolute period requirement is critical to driving these applications. When the timing signal period is too long, the system will still function, although it may experience a decline in performance. Clearly some jitter is acceptable as long as it stays within some absolute bounds. Jitter over time While, at first glance, jitter appears to be a relatively simple subject, many applications bring additional issues that complicate the measurement process. First, the designer must consider the effects of the jitter on the timing signals phase with respect to the perfect reference signal. Second, it is important to know if the jitter is equally distributed on the plus and minus sides over a span of time. These effects result in what is called “accumulated jitter,” “long-term jitter” or “phase jitter.” Consider what happens when several successive clock periods have a positive jitter value. At the end of this set of clock periods, the timing edges could be significantly displaced in time from their ideal locations (see figure 3). One way to compensate for this effect in the period or cycle-to-cycle jitter specification is to specify the number of cycles over which the jitter is measured. An example of how this is done is specified in the JEDEC Standard JESD65. Different applications may require the specification of a different number of cycles over which the jitter is measured. IDT® THE ROLE OF JITTER IN TIMING SIGNALS 3 Figure 3. Timing signal with accumulated jitter Perfect time reference signal Timing signal with positive accumulated jitter In this example the period continues to get larger due to accumulated jitter. The reverse is also possible where the period grows shorter. Spread spectrum modulation is an example of intentional accumulated jitter generation. Figure 3. Timing signal with positive accumulated jitter Consider a timing signal used to drive a pixel output on a CRT. The integration effects of the phosphor and the inherent limitations of the human eye will tolerate small variations in the pixel position. So this application will permit a fairly large cycle-to-cycle jitter value. If the jitter is not random, however, and it accumulates over a significant portion of the CRT raster, the user will see image distortion or swimming due to improper positioning of the pixels. Modulation frequency and period also have a major impact. If the jitter is modulating from a low frequency 60 Hz source and has a period of 16.7 ms, the jitter will increase for approximately one quarter of the period (~ 4 ms) and decrease in the remaining quarter of the period. Since most CRT raster displays use a screen refresh rate of 60 Hz or higher, the full effect of the jitter will be easily visible on the screen.