Lucix - Apps1

Phase noise and jitter are two related performance parameters of oscillators ( i.e., clock references, local oscillators, timing references) which describe the oscillators' inherent noise performance. Phase noise is the frequency domain representation of this noise performance. Random jitter is the time domain representation.

An ideal, i.e., noise free, oscillator would exhibit a singularity in the frequency domain as measured by a spectrum analyzer. A singularity in the frequency domain is essentially an infinitely narrow line (zero bandwidth) at the oscillator's operating frequency. Figure 1 represents the frequency response of a perfect oscillator operating a 10 GHz.
In actuality, the "perfect" oscillator does not exist. In fact, the operating frequency will vary in the short term (second) around the desired operating frequency. This variation when examined in the frequency domain looks similar to low rate, low amplitude phase modulation, hence the name, phase noise. The net result of this effective phase modulation is a broadening of the output spectrum of the oscillator which represents a variation in the output phase or frequency of the oscillator over a short time period. Figure 2 depicts this frequency broadening.

This broadening of the oscillator output can have various shapes and is characterized in dBc versus frequency offset from the nominal operating frequency. The close-in slope of the phase noise can typically be between -25 and -30 dBc/decade. Other parts of the phase noise spectrum typically have lower slopes beyond 10's of kHz offsets from the carrier. The net result of this random variation in frequency is a time variation of the carrier frequency or, in other words, a fast, low rate frequency (or phase) modulation of the carrier.

The effect of this modulation is shown n Figure 3 as a highly exaggerated variation of the time period of the original clock signal. The noise energy contained in the phase noise side bands of a typical oscillator might modulate a 10 GHz clock signal, in a peak-to-peak manner, by less than 0.1% of the nominal time period of the oscillator, i.e., 0.1 ps for a 10 GHz oscillator.

However, this small, but real, variation in clock period can lead to difficulties when trying to detect bits in a high speed digital communication system. If the rising (or falling) edge of transmitted bits vary by more than one half a bit period, far end sampling circuitry may begin to misinterpret the receive bits due to timing variation between the received bits edges and the sampling circuit trigger controlled by the far end timing reference. This timing uncertainty in the received bits boundaries is called jitter.

Jitter is usually measured in the time domain using sequential sampling or real time scopes, depending on the speed (bandwidth) of the signal to be examined. Jitter in the received bit stream arises from many sources, but includes the intrinsic jitter of the clock source which created the bits, timing errors in digital regeneration circuitry, the presence of spurious signals and clock harmonics, pattern dependent clock recovery issues and system noise. Typically, jitter is separated into two general components: random and deterministic. Deterministic jitter is jitter which is repeatable in nature and due to specific component performance within the system. By definition, deterministic jitter is not random in nature. In high speed communication equipment, one component of deterministic jitter is pattern dependent jitter. Pattern dependent jitter arises when a clock recovery circuitry has difficulty responding to long series of ones and zeroes in a received pattern. The inherent clock frequency present in a random collection of one and zeroes in a digital bit stream is momentarily lost and the clock recovery circuitry's output frequency may drift. The result is output bits with varying bit periods, i.e., jitter.

Random jitter is just that, random, and typically assumed to be gaussian in nature. Random jitter arises from system noise, detector noise and phase noise in reference clocks used to detect or to create bits in a digital system. As a result, system designers need to use system clocks (oscillators) with the best inherent phase noise performance in order to reduce intrinsic system jitter.

The jitter performance of a clock source, i.e., an oscillator, can be either measured using a sampling scope or calculated from the phase noise performance of the oscillator. Both methods yield marginal results for extremely low phase noise oscillators. Sampling scopes have inherent jitter in their sampling circuitry which may exceed the jitter of the oscillator to be measured. Calculations of random jitter performance from phase noise plots require decisions to be made over what bandwidth the calculation will be made and what is the actual phase noise performance of the oscillator at a specified clock frequency offset. What can be unequivocally stated is that the oscillator which has lowest total sideband phase noise over in a given sideband offset frequency range of interest will exhibit the lowest random jitter performance. If accurate determination of phase noise or oscillator jitter is required, Lucix would recommend the use of the Agilent Technologies E5500 Phase Noise Measurement System (or equivalent) which has been specifically designed to accurately quantify oscillator phase noise and can calculate intrinsic jitter from phase noise integration