A proposed system for simultaneous characterization of the instability of several precise, low-noise oscillators of nominally equal frequency would be built around a commercially available time-tag counter. One of the oscillators would be deemed to be a reference oscillator, and each of the other oscillators would be compared with it by operation of a combination of hardware and software. In addition, without further modification of the hardware, any two nonreference oscillators could be compared with each other via software.
The design of the proposed stability analyzer is of a type called "dual mixer" in the precise-time-and-frequency-measurement art because the comparison of any two nonreference oscillators would involve the outputs of two mixers. There are also single-mixer stability analyzers. Single-mixer analyzers exhibit low measurement noise, but an offset-frequency reference oscillator is needed for each pair of nonreference oscillators to be compared. A prior dual-mixer analyzer contains only one offset-frequency reference oscillator, but exhibits noise greater than that of a single-mixer analyzer. The proposed system would offer both the convenience and low cost of a dual-mixer analyzer and measurement noise about as low as that of the best single-mixer analyzer.
A typical prior dual-mixer stability analyzer utilizes interpolation or extrapolation to convert several incoherent channels of beat-note zero crossings into phase residuals at a predetermined grid of times, so that the residuals of any two channels i and j can be subtracted to give an i-vs.-j comparison. This measurement is contaminated by uncanceled noise from the offset-frequency reference oscillator. The proposed system would take advantage of a modern high-rate time-tag counter to collect zero-crossing times of beat notes, the nominal frequency of which must be much greater than the desired data rate. Then the system would effect a combination of interpolation and averaging to process the time tags into low-rate phase residuals at the desired grid times. The advantage over prior art would be greater cancellation of the reference noise.
The figure schematically depicts the system. The oscillators to be compared would be of nominal frequency nr. The frequency of the reference oscillator would be offset by an amount nb. The offset reference signal would be mixed with the signal from each of the nonreference oscillators, and the mixer outputs would be low-pass filtered, thereby generating beat notes of nominal frequency nb. By use of zero-crossing detectors, the beat notes would be converted to square-wave signals. The time-tag counter would capture the zero-crossing time tags of all the beat notes on a common time axis.
In software, the time tags would be converted to phase residuals that would be averaged over sequential intervals of duration τs. These intervals would be the same for all channels. The averages thus computed would constitute one of the sets of output data of the system. An essential feature of the design is that ts must be much greater than the beat period τb = 1/nb.
Each beat note would yield phase residuals for one pair-channel [e.g., the ith channel, defined with respect to the ith oscillator (a nonreference oscillator) vs. the zeroth oscillator (the reference oscillator)]. Because the averaging intervals would be the same for all pair-channels, the data for two pair channels could be differenced to give a synthesized dual-mixer (i-vs.-j) channel. The ability of this system to suppress the noise of the reference oscillator would depend on the relation τs >> τb.
This work was done by Charles Greenhall of Caltech for NASA's Jet Propulsion Laboratory. For further information, access the Technical Support Package (TSP) free on-line at www.nasatech.com/tsp under the Electronics & Computers category.