GPS Disciplined Oscillator Performance

About this Article
This paper by Carmel Fleming and Mike Hutchins was presented the British Electromagnetic Measurements Conference in November 1997 when the Agilent's Test & Measurement business was part of Hewlett-Packard.

Foreword

Following the initial study carried out by NPL and industry in 1995 on the suitability of GPS disciplined oscillators (GPS-DO) as standards for Time and Frequency traceable to the UK National Timescale, UTC(NPL), NPL have now completed a further, more extensive study to determine GPS-DO performance in both ideal and non-ideal laboratory conditions. This report supplements the NPL paper on the study by outlining the measurements made in a working calibration laboratory with commercially available equipment. The measurements are the difference between the GPS-DO output and the laboratory house standard. The results are also compared with the measurements made at NPL.

Introduction

Coordinated Universal Time (UTC) is the official accepted standard for time and frequency. UTC is generated by the Bureau International des Poids et Mesures (BIPM) in France and defines time for the world. Timing centers around the world steer their own time scales to UTC, e.g. the UK timescale is referred to as UTC_(NPL).

The Global Positioning System (GPS), which was originally designed for military use, has rapidly grown in its commercial applications because of the availability of relatively inexpensive and easy to operate receivers.

GPS-based products are found in:-

• vehicle navigation systems (cars, aircraft and ships)
• electrical power control and fault analysis
• communication network systems
• emergency sensing systems
• calibration laboratory references

The Hewlett-Packard 58503A GPS receiver was designed to meet the timing and frequency standard needs of these users without investing in a primary frequency standard. It generates precise 10MHz and 1pps signals using HP's proprietary SmartClock^TM technology utilizing a precision quartz oscillator and incorporates an RS232 port for monitoring and control.

The use of GPS-DO's in calibration labs in the UK, however, is limited as they have not been recognized as standards traceable to the national timescale, UTC_(NPL). The NPL study aims to determine GPS-DO suitability as traceable standards and analyze some of the sources of uncertainty associated with their use.

Our Objective

Hewlett-Packard's Customer Support Center, like many other calibration laboratories, wish to replace their existing method of achieving traceability via reception of "MSF" Rugby (60kHz transmission) with a GPS receiver-based system.

GPS disciplined oscillators represent a significant improvement to the more conventional radio broadcast methods of time and frequency dissemination, in both increased accuracy and global coverage. However, any new method for traceability requires confidence and understanding by the relevant accreditation bodies. The United Kingdom Accreditation Service (UKAS), are aiming to use guidelines drafted by NPL when assessing laboratories for accreditation based on GPS-DO's A number of other countries already accept this traceability approach, including accredited capabilities.

Our intention in performing these measurements is that they may be used in support of the findings made by NPL and demonstrate the correlation between the two geographically close locations using different examples of the same GPS-DO model.

Method

Continuous measurements were made over two weeks in our environmentally controlled laboratory (temperature: 23±1șC, relative humidity: 50±10% and a.c. mains stabilization: ±1%). The 1pps output from the receiver was compared with the 1pps output from our own HP5071A Primary Frequency Standard, the equipment set-up is shown in Figure1. During the study period, NPL made similar measurements on several GPS-DO's which included another HP58503A. The NPL measurements were made in a one minute ON, three minutes OFF arrangement as they were switching between all the receivers in the group. Other data was also logged from each GPS-DO, (i.e. which satellites were being tracked, their elevation and azimuth, signal strength health status, dilution of precision (DOP) values, DAC corrections, GPS lock and other status information). This data will be analyzed later and may be useful in the accreditation guidelines.

Figure 1 -- Equipment set-up

Representative Results

Figure 2 shows some of the 1pps data taken by NPL from the unit loaned to them for the project. Each datapoint is the mean of 15x1 minute blocks of measurements. NPL's test set-up was principally the same as HP's except for a switching arrangement to allow 4 receivers to be tested simultaneously.

Figure 2

Figure 3 -- Shows frequency offset derived from NPL's phase measurements which were made using a different technique and on a separate output signal.

Figure 4 shows the daily mean of NPL's measurements and also that there was good correlation between the 1pps (circles) and 10MHz (squares) output signals. The results are consistent with the product's specified accuracy (±1x10^-12 averaged over 24 hours).

Figure 4

Figure 5 shows our results using another HP58503A, although a month later. NPL supplied software was used to control the test system and analyze the data. Unlike NPL, data was taken continuously.

Figure 5

Figure 6 -- Shows the daily mean of our data and again confirms the product's compliance with spec.

Figure 7 -- Our phase results were taken using a vector voltmeter (analogue phase comparator) and chart recorder.

Figure 8 -- Shows our determination of the daily mean offset frequency.

If using a GPS-DO as a local frequency standard, rather than as a means to monitor an atomic standard as at HP Winnersh, its accuracy during the period of use is of concern rather than longer term (e.g. 1 to 100s instead of a 1 hour or 24h average). Figure 9 shows how our unit behaved.

Figure 9

Conclusions

• The published specification for the HP58503A is stated as (GPS locked operation) 1 part in 10¹² over 1 day averaging period. The performance experienced by both NPL and ourselves appears to support this claim.

• The evidence shows that GPS-DO's with high performance crystals are capable of meeting the requirements of applications where the frequency accuracy over several hours needs to be better than ±1x10^-11. Where "instantaneous" frequency over a much shorter timescale is required to be better than (perhaps) a part in 10¹⁰, a standalone (undisciplined) atomic frequency standard is still preferred. Our data suggests that the 58503A's uncertainty as a reference for such measurements may not exceed ±2x10^-11 (not actually specified).

• Daily accuracy of this GPS-DO is at least comparable with the performance experienced from oscillators disciplined using the MS. 60kHz radio transmission. For the purposes of monitoring a local (Cs) standard using MS., it is necessary for NPL and HP use a rolling 10-day average to achieve a part in 10¹² performance. Equivalent precision with GPS is achieved in a day.

We will continue to collect data on the 58503A in our Standards Laboratory. This will be used later to support a proposal to maintain traceability in our UKAS(NAMAS) accreditation.

Acknowledgments
The authors thank the Time & Frequency department at NPL for their support and the use of some of their data in this paper and colleagues at Hewlett-Packard Company, Santa Clara Division for their help and support in both the study and writing of this paper.