NIST Technical Note 1337: Characterization of Clocks and Oscillators

fluctuations prior to the bias box (biased slightlyaway from zero) may be fed to a voltage to frequencyconverter whi ch in turn ;s fed to a frequencycounter where one may read out the frequencyfluctuations with great amp1ific3tion of theinstabi 1ities bet..een this pair of oscillators.The frequency counter data are logged with a datalogging device. The coefficient of the varicapand the coefficient of the voltage to frequencyconverter are used to determine the fractionalfrequency fluctuations, Yi' between the oscillators,where i denotes the i th measurement asshown in figure 1.7. It is not difficult toachieve a sensitivity of a part in 1014 per Hzresa1uti on of the frequency counter. so one hasexcellent precision capabilities with this system.E. Time difference methodThe last measurement method we will illustrateis very commonly used, but typically does not havethe measurement precision more readily availablein the first four methods illustrated above. Thismethod is called the time difference method, andis shown in figure 1.8. Because of the wideI,all. 312.'" UIIIf·I...·-----'-".ocS1IlII '.tt81 ! +N ~ I IiG~FIGURE 1.8bandwidth needed to measure fast rise-time pulses,this method is limited in signa1-to-noise ratio.However, some counters are commercially availableallowing one to do signal averaging or to doprecision rise-time comparison (precision of timedi fference measurements in the range of 10 ns to10 piS are now available). Such a method yields adirect measurement of x(t) without any translation,conversion, or multiplication factors. Cautionshould be exercised in using this technique evenif adequate measurement precision is availablebecause it is not uncommon to have significantinstabi 1ities in the frequency dividers shown infigure 1. a--of the order of severa1 nanoseconds.The techno 1ogy exi sts to build better frequencydividers than are commonly available, but manufacturershave not yet availed themselves of state-ofthe-arttechniques in a cost beneficial manner. Atrick to by-pass divider problems is to feed theoscillator signals directly into the time intervalcounter and' observe the zero voltage crossing intoa well matched impedance. (In fact, in all of theabove methods one needs to pay attention to impedancematchi ng, cab 1e 1engths and types, and connectors).The divided signal can be used toresolve cycle ambiguity of the carrier, otherwisethe carrier phase at zero volts may be used as thetime reference. The slope of the signal at zerovolts is 2TcV h: 1 • where t 1 = l/Ulp(the period ofoscillation). For Vp= 1 volt and a 5 MHz signal,this slope is 3m vo1ts/ns, whiCh is a very goodsensitivity.II. MEASUREMENT METHODS COMPARISONWhen maki n9 measurements between a pair offrequency standards or clocks, it is desirable tohave less noise in the measurement system than thecomposite noise in the pair of standards beingmeasured. This places stringent requirements onmeasurement systems as the state-of-the-art ofprecision frequency and time standards has advancedto its current 1eve1. As wi 11 be shown, perhapsone of the greatest areas of di spari ty betweenmeasurement system noise and the noise in currentstandards is in the area of time difference measurements.Commercial equipment can measure timedifferences to at best 10_ 11 S, but the time fluctuationssecond to second of state-of-the-artstandards is as good as 10_ 13 S.The disparity is unfortunate because if timedifferences between two standards could be measuredwith adequate precision then one may also know thetime fluctuations, the frequency differences, andthe frequency fluctuations. In fact, one can set7TN-20