NIST Technical Note 1337: Characterization of Clocks and Oscillators

Weean make the following general remarks aboutpower-law noise processes:1. Random walk FM (1/f4) noise is difficultto measure since it is usua 11 y veryclose to the carrier. Random walk FMusually relates to the OSCILLATOR'SPHYSICAL ENVI RONMENT. If random lola 1k FMis a predominant feature of the spectraldensity plot then MECHANICAL SHOCK,VIBRATION, TEMPERATURE, or other environmentaleffects lIlay be causing "random"shifts in the carrier frequency.2. Flicker FM (1/f 3 ) is a noise whosephysical cause is usually not fullyunderstood but may typically be relatedto the PHYSICAL RESONANCE MECHANISM OFAN ACTIVE OSCI LLATOR or the DESIGN ORCHOICE OF PARTS USED FOR THE ELECTRONICS,or ENVIRONMENTAL PROPERTIES. Flicker FMis common in high-quality oscillators,but may be masked by white FM (1/f 2 ) orfliCker PM (1/f) in lower-quality oscillators.3. White FM (1/f2) noise is a common typefound in PASSIVE-RESONATOR FREQUENCYSTANDARDS. These contain a slave oscillator,often quartz, which is locked toa resonance feature of another devi cewhich behaves much like a high-Q filter.Cesium and rubidium standards have whiteFM noise characteristics.4. Fl iclter PM (1/t) noi se may relate to aphysical resonance mechanism in anoscillator, but it usually is added byNOISY ELECTRONICS. Thi s type of noi seis common, even in the highest qualityoscillators, because in order to bringthe signal amplitUde up to a usablelevel, amplifiers are used after thesignal source. Flicker PM noise may beintroduced in these stages. It may alsobe introduced in a frequency multiplier.• See Appendix Note /I 7*38Flicker PM can be reduced with gDodlow-noise amplifier design (e.g., usingrf negative feedback) and hand-selectingtransistors and other electronic components.5. Whi te PM (f0) noi se is broadband phasenoi se and has 1itt1e tD do ....ith theresonance mechanism. It is probablyproduced by similar phenomena as flickerPM (1/f) noise. STAGES OF AMPLIFICATIONare usually responsible for ....hite PMnoise. This noise can be kept at a very10.... value ....ith good amp 1ifier design,hand-selected components, the additionof narrowband fi lteri ng at the output.or increasing, if feasible, the pDwer ofthe primary frequency sDurce. 1312.2 Other types of nDiseA common1y encDuntered type of no i se from asignal source or measurement apparatus is thepresence of 60 Hz A.C. line noise. Sho....n infigure 12.2 is a CDnstant ....hite PM noise source....i th 60 Hz, 120 Hz and 180 Hz components added.This kind of nDise is usually caused by AC powergetti ng into the measurement system or the. sourceunder test. In the plot of S(j)(f) , one observesdiscrete line spectra. Although S$(f) is a measureof spectral density, one can interpret the linespectra ....ith no loss of general i ty. although oneusually does not refer to spectral densities whenCharacterizing discrete lines. Figure 12.3 is thetime domain representation of the same white phasemodulation level ....ith 60 Hz noise. Note that theamplitUde of ay(t) varies up and down depending onsampling time. This is because in the time domainthe sensitivity to a periodic ....ave varies directlyas the sampling interval. This effect (which isan alias effect) is a very powerful tool forfiltering out a periOdic wave imposed on a signalsource. By sampling in the time domain at integerperiods, one is virtually insensitive to theperiodic (discrete line) term.IN-51