14-Jul-2005

Introduction

As you can see elsewhere in the museum, oscillators and frequency standards come in many shapes and sizes: from 19" rack-mount to surface mount, from square to round. Below are representative photos (some not to scale) of a dozen oscillators that I have measured, ranging in accuracy or stability from 10^-5 to 10^-15.

1. JK Knights type JKST -- thermostatically controlled 6.3 V 55 ºC oven oscillator
2. Bliley type BG9D -- 8-pin octal 100 KC quartz crystal in vacuum tube
3. CTS 10 MHz TO-5 quartz crystal, with ceramic trimmer, on HP 53132A circuit board
4. 5 MHz gold plated crystal resonator in clear glass vacuum
5. Oscilloquartz 8607 -- BVA quartz oscillator in evacuated thermal shield (wine cooler)
6. Earth -- gravitationally suspended high-Q rotating spherical planet in vacuum
7. General Radio type 376J -- 100.00 ± 0.001 KC quartz plate
8. HP type 10811-60109 -- 10 MHz oven controlled quartz oscillator (OCXO)
9. Rakon TX0690 -- 10.000 MHz TCXO, surface mount
10. Sigma Tau type MHM-2010 -- active hydrogen maser, with backup batteries and physicists
11. FEI / Hughes type FE-2089A -- space-qualified OCXO, 2 lbs, 3.83 oz
12. NTSC 3.579545 MHz colorburst crystal oscillator, 12 VDC, 115 VAC

Photo credits; oscillator 6 by NASA (1972); oscillators 1 - 5, 7 - 12 by tvb (2005).

Today, we'll have a good look of the 6th oscillator in the above gallery: the large round blue & white one.

First Impressions

I couldn't find the make or model so we'll just call it earth in this report. The earth oscillator appears old; the photo given to me by NASA was dated December 7, 1972, so it was certainly made prior to that. I have been trying to contact the Manufacturer, but as yet there is no operation or service manual. There is no visible release date or serial number and the original specifications are unknown. It's missing a power cord. It is used and shows signs of wear. In short, very similar to most oscillators one finds on eBay.

Typically, oscillators are housed in cylindrical or rectangular enclosures; this one is spherical and spins freely in a high vacuum with gravitational suspension. High mass and low friction makes for very high Q. All in all, a neat design. Like other space-qualified oscillators I have seen, this one contains iron, silicon, aluminum, and titanium.

Earth appears somewhat dirty, dusty, and wet in places. Since I don't own it (on approval only) it was decided to test is in as-is condition even though one could obtain better performance with thorough cleaning.

A close-up photo of Earth (not to scale)

The Earth rotates freely on its axis; this makes it a frequency standard. The nominal frequency is approximately 11.5 µHz, or more exactly, 11.5740740e-6 Hz (a period of 86400 s). Unlike a 10811 which delivers a 10 MHz RF output the 11.5 µHz frequency output of Earth is typically detected optically: either shine light onto earth or detect light from earth.

As delivered the average measured frequency, over 40 years, was about 2.2x10^-8 low. A primary component of this earth is quartz (SiO₂). The daily drift rate is 1.3x10^-12 making this one of the better quartz oscillators I have measured.

Measured Performance

The rotation rate can be measured two ways: from the nearest point of light (sun) or from a distant point of light (star). These rates are not the same due to the combined effect of the rotation of the planet and its revolution around the sun.

Specs were not available so we'll just look at measured performance. Daily phase samples were collected for 40 years; 7/1/1965 to 7/14/2005. Following are phase, frequency, and stability plots for Earth.

In the raw phase plot below note the oscillator slowly losing time. Some instability can also be seen, especially near the end of the run.

Both short and long-term variations in frequency clearly can be seen. In addition, the Earth was on-frequency several times recently.

Frequency error, yearly averages.

The frequency varies by approx ½ to 3½ parts in 10^-8 over decades.

A mix of imperfections cause several humps in the Allan Deviation plot. It would appear long-term frequency drift limits the noise floor to 2e-9.

The Allan Deviation for Earth is a little disappointing. It should be able to do much better. Clearly there are some problems with Earth.

Indeed, the power spectrum shows a number of interesting peaks ranging from 35 and 65 nHz to 1.15 µHz.

MTIE plot

It looks like the time error would approach 1 hour in 3000 to 4000 years.

Extrapolating MTIE plot for tau past 1011th seconds (ca 3600 years)

Using Earth as a Time & Frequency Standard

Earth spins in a vacuum enclosure supported by gravitational attraction; a clever design resulting in very high-Q. It has the potential of an excellent time-keeper. But a rate error, internal instabilities, and no frequency adjustment makes earth less attractive as a frequency standard.

The Manufacturer could have a better time product with the following changes:

• The Earth is dirty, dusty, and damp in places. In fact, a significant portion is beyond damp; it's wet. Performance would improve if it were thoroughly cleaned and dried with a towel, removing all surface water and dust. Fine polishing the surface is probably not necessary.
• Although enclosed in high vacuum environment there is a layer of air surrounding the surface. It would spin more regularly if this layer were removed.
• Earth, as delivered, is seriously misaligned. It appears to be some 23 ½° off axis. It would suffer fewer thermal effects if this tilt were corrected.
• This Earth also has a slight wobble which contributes to mid-term instability. Suggest that it be re-aligned more precisely and not disturbed during shipment.
• This prototype Earth was not sufficiently left to cool. As a result parts of the core are still warm which appears to contribute to long-term instability. Suggest keeping Earth models in the factory longer before they are shipped to customers. This pre-aging would greatly improve performance.
• The nearby moon has a strong negative impact on Earth performance. Suggest removing moon from the vicinity of the timing laboratory. After measurement, nearby planets should also be suspect.
• Earth should be better protected from evaporation into space; likewise debris (meteors, comets) coming from space should not be deposited onto Earth. Protective shield would reduce inflow of mass. Foil or plastic wrap might help outflow.
• Some form of temperature stabilization should be considered. Diurnal temperature effects are significant.
• An older, colder planet, further from the sun, without atmosphere, and no or smaller moon would make a much better frequency standard.

In general, a better prototype would be preferred; one that is cleaned, polished, re-aligned, re-stabilized, and rate re-calibrated. I suspect timing performance would improve by several orders of magnitude.

Conclusion

Earth is a vintage, quality constructed, high-Q oscillator, composed of SiO₂ Al Fe Ti, spherical, although not perfectly shaped (flat by about 1/298), and off-axis (about 23 °). During the 40 year test period, the average frequency of the oscillator was about a few parts in 10⁸th below the nameplate value.

There appears to be no fine or coarse frequency adjustment on earth. In cases like this one must either use an external frequency synthesizer, or phase microstepper, to correct for the frequency error. Or one can post-process the data and apply corrections in software. The lack of electronic frequency adjustment prevents it from being used in a GPSDO.

If used as a time standard it may be more convenient to make one-second corrections every several hundred cycles. These "leap second" time steps would create a discontinuous time-scale, requiring tables (rather than algorithms); a pain for all users, especially high-precision ones. But this solution is required when the oscillator is designed with no rate adjustment.

No power cord could be found for the oscillator; but neither is it subject to power failure. It has moderately stability, with poor accuracy, but excellent drift. It may be too large and heavy for many applications.

So the Earth makes a decent low-cost frequency standard; all in all, not a bad oscillator for the price. If size is not a problem you could incorporate it into your system designs. However, the inherent inaccuracy, the lack of frequency tuning, and its relatively high instability would suggest one look elsewhere for a precise time reference. On the other hand, because it requires no batteries the advantages may outweigh the disadvantages for many applications.