NIST Technical Note 1337: Characterization of Clocks and Oscillators

ports of the pair of double balanced mixers. This the cycle ambiguity. It is only important to knowcommon oscillator's frequency is offset by a k if the absolute time difference is desired; fordes i red amount from the other two osci 11 ators. frequency and frequency stability measurements andThen two di fferent beat frequenci es come out of for time fluctuation measurements, k may be assumedthe two mi xers as shown. These two beat frequen­ zero unless one goes through a cycle during a setcies will be out of phase by an amount proportional of measurements. The fractional frequency can beto the time difference between osci 11 ator 1 and derived in the normal way from the time fluctua­2--exc1uding the differential phase shift that may tions.be inserted. Further, the beat frequencies differinfrequency by an amount equa 1 to the. frequency"l(i, t) - "Z(i, t)difference between oscillators 1 and Z. " 0This measurement technique is very usefulxCi + 1) - xCi)Y1,Z(i, t) = (1. 6)where one has oscillator 1 and oscillator 2 on thetsame frequency. This is typical for atomic stan­ ~t(1 + 1) - at(i)t 2dards (cesium, rubidium, and hydrogen frequencyb "0standards).Illustrated at the bottom of figure 1.5 is In eqs (1.5) and (1.6), assumptions are madewhat mi ght represent the beat frequenci es out of that the transfer (or common) oscillator is set atthe two mi xers. A phase shifter may be inserted a lower frequency than osci 11 ators 1 and Z, andas illustrated to adjust the phase so that the two that the voltage zero crossing of the beat "I - "cbeat rates are nominally in phase; this adjustment starts and that "2 - "c stops the time intervalsets up the nice condition that the noise of the counter. The fractional frequency difference maycommon osci 11 ator tends to cance 1 (for certai n be averaged over any integer multiple of to:types of noise) when the time difference is determined. After amp 1ifyi ng these beat signals, thestart port of a time interval counter is triggered= xCi + m) - xCi)mt b(1.7)with the positive zero crossing of one beat andthe stop port with the positive zero crossing of where m is any positive integer. If needed, tbthe other beat. Taki ng the time di fference be­ can be made to be very small by having very hightween the zero crossings of these beat frequencies, beat frequencies. The transfer (or common) oscilonemeasures the time difference between oscillator lator may be replaced with a low phase-noise1 and oscillator 2, but with a precision which has frequency synthesizer, which derives its basicbeen amp 1ified by the ratio of the carri er fre­ reference frequency from osci 11 ator Z. In thi squency to the beat frequency (over that normally set-up the nominal beat frequencies are simplyachievable with this same time interval counter). gi ven by the amount that the output frequency ofThe time difference x( i) for the i th measurement the synthesi zer 15 offset from "2' Sample timesbetween oscillators 1 and Z is given by eq (1.5). as short as a fE'll milliseconds are easiliy obtained.Logging the data at such a rate can be a(1. 5)problem without special equipment. The latest NBStime scale measurement system is based on the DMTDand is yielding an excellent cost benefit ratio.where ~t(i) is the i th time difference as read onthe counter, tbis the beat peri od, "0 is the C. Loose ohase lOCK loop methodnominal ~arrier frequency, ~ is the phase delay in This first type of phase lock loop method isradians added to the signal of oscillator I, and k illustrated in figure 1.6. The signal from anis an integer to be determi ned in order to remove oscillator under test is fed into one port of a5TN-18