NIST Technical Note 1337: Characterization of Clocks and Oscillators

36th Annual Frequency Control Symposium· 1982SummaryPERFORMANCE OF AN AUTOMATED HIGH ACCURACYPHASE MEASUREMENT SYSTEMS. Stein, D. Glaze, J. Levine, J. Gray. D. Hilliard, D. HoweTime and Frequency DivisionNational Bureau of StandardsBoulder, Colorado 80303A fully automated measurement system hasbeen developed that combines many propertiespreviously realized with separate techniques.This system is an extension of the dual mixertime difference technique, and maintains itsimportant features: zero dead time, absolutephase difference measurement, very high precision,the ability to measure oscillators ofequal frequency and the ability to make me1surementsat the time of the operator's choice. Forone set of des i gn parameters, the theoreti ca1resolution is 0.2 ps, the measurement noise is 2ps rms and measurements may be made within 0.1 sof any selected time. The dual mixer techniquehas been extended by adding scalers which removethe cycle ambiguity experienced in previousreal izations. In this respect, the systemfunctions like a divider plus clock, storing theepoch of each device under test in hardware.The automation is based on the ANSI/IEEE­583 (CAMAC) interface standard. 2 Each measurementchannel consists of a mixer, zero-crossingdetector, scaler and time interval counter.Four channels fit in a double width CAMAC modulewhich in turn is installed in a standard CAMACcrate. Controllers are available to interfacewith a wide variety of computers as well as anyIEEE-488 compatible device. Two systems havebeen in operation for several months. Oneoperates 24 hours a day, taki ng data from 15clocks for the NBS time scale, and the other isused for short duration laboratory experiments.Review of the Dual MixerTime Difference TechnigueIt is advantageous to measure time di rectlyrather than time fluctuations, frequency orfrequency fl uctuationns. These measurementsconstitute a hierarchy in which the subsequentlylisted quantities may always be calculated fromthe previous ones. However, the reverse is nottrue when there are gaps in the measurements.In the past, frequency was usually not derivedfrom time measurements for short sample timesbecause time interval measurements could not beperformed with adequate precision. The dualandL. ErbER8TEC Engineering Inc.Boulder, Coloradomixer technique, illustrated in Figure I, madeit possible to realize the precision of the beatfrequency technique in time interval measurements.The signals from two oscillators (clocks) areapplied to two ports of a pair of double balancedmixers. Another signal synthesized from oneof the oscillators is applied to the remainingtwo ports of the mixer pair. The input signalsmay be represented in the usual fashionVI(t) = V 10sin [2nu 10 t + ~l(t)],V 2(t) = V 20sin [2nu 20t + ~2(t)]andVs(t) =V cos [2nu t + ~s(t)]so sowhere u = u (l-l/R) and R is a constantusually eRlled th~ heterodyne factor.areThe low passed outputs of the two mixersV BI= V BIOsin [~l(t) -~s(t)] andV B2=V S20sin [~2(t)-~s(t)] where$(t) = 2nu t + ~(t).oThe time interval counter starts at time t M whenV Rlcrosses zero in the positive directitJ"n andstops at time tN' the time of the very nextpositive zero croS'sing of V B2. Thus$l(t M) - ~s(tM) = 2Mrr and$2(t N) - $s(t N) = 2Nn whereNand Mare integers.SUbtracting the two equations in order to com­pare the phases of osc; 11 ators 1 and 2, oneobtains$2(t N)-$1(t M) =$s{t N )-$s(t M )+2(N-M)n.The phase of an asci 11 ator at time t Nmaybe written in terms of its phase at t Mand it~314TN-241