Dual frequency absolute calibration of GPS receiver for time transfer A. de Latour, G. Cibiel, J. Dantepal, J.-F. Dutrey, M. Brunet, L. Ries, J.-L. Issler CNES 18, avenue Edouard Belin, 31401 Toulouse Cedex, France - 33 (0)5 61 27 31 31 [email protected], [email protected]

ABSTRACT This paper suggests a receiver time delay calibration method using a GPS simulator. The method has been used to get the receiver time delay of an Ashtech Z12T receiver. The method enables to compute an absolute time delay in the L1- and L2-frequencybands within few nanoseconds of truth. The measured absolute time delay depends mainly on the GPS simulator calibration. This method has several advantages : the radiofrequency environment is controlled and the calibration does not depend on the UTC-GPS time offset.

1. INTRODUCTION Future Galileo receivers (Rx) and GPS Rx dedicated to the time transfer involve to be able to make accurate and repeatable calibrations of timing Rx. The most widely used approach to calibration of Rx is co-locating the Rx, which will be calibrated with a reference Rx. This calibration approach, differential calibration, has been established as a method for measuring relative electrical delays between two Rx [PET 00]. To accomplish this, one Rx is designated as the reference Rx and is in constant circulation among timing laboratories. A relative calibration is performed between this Rx and the laboratory Rx. Therefore, this method is cumbersome and the stability of the calibration cannot be determined. A better solution to this calibration problem would be a repeatable method to measure the electrical delays for each Rx individually. Work at the Naval Research Laboratory (NRL) has shown that the delays of an Ashtech Z-12T Rx can be measured relative to a GPS signal simulator [WHI 01], [LAR 03]. Our method is based on this one. The difference between the two methods comes from the different ways used to calibrate the GPS simulator: the NRL’s method used the zero-crossing method the CNES’s method used correlation functions The CNES’s method improves the simulator calibration, the time delay is indeed estimated over

more than one transition contrary to the zero-crossing method. This paper explains the method used to calibrate GPS Rx for time transfer. In addition, the results issued from the Ashtech Z12T Rx calibration are shown. The calibration is performed in two main steps: 1st step: the GPS signal simulator calibration in the L1-and L2-frequency bands 2nd step: the GPS Rx calibration The Ashtech Z-12T Rx has been chosen for this experiment because it has a repeatable internal reference. This internal reference is determine by the difference in the time of arrival at the Rx of the input 20 MHz clock signal (the 20 MHz is generated a 10 MHz frequency reference multiply by two) and the input one pulse per second (1 PPS) signal. Because the 20 MHz clock and 1 PPS input to the Rx are both coherent and externally generated, the calibration delay remains the same when the receiver power is cycled. During the presented experiments, Rx and GPS signal simulator have located in a thermally controlled area, where the temperature is equal to 20.5 degrees Celsius with a maximum deviation of 1 degree.

2. GPS SIMULATOR CALIBRATION 2.1 General strategy The internal delay of the simulator corresponds to the delay between the beginning of the PRN C/A or P ranging code and the internal one pulse per second synchronized with the GPS time. The simulator calibration is performed using only the PRN C/A ranging code for the following reasons: the chips in the C/A code and the P code have the same rise- or fall-time, because they share the same simulator bandwidth

the generation of the reference P code is complex, because the P code period is one week contrary to the C/A code, whose period is equal to one millisecond. The simulator delay is estimated using the correlation function, for each GPS frequency bands, between the following signals: the reference C/A code, synchronised with the simulator PPS and sampled at 10 GHz the C/A code generated by the GPS simulator at the radio frequency output and sampled at 10 GHz The 10 MHz reference frequency is generated by an active hydrogen MASER clock (EFOS 16 from Observatory of Neuchâtel).

The PPS signal is transmitted with a 50-Ohms RF cable, which induced a delay of 32.131 ns with a standard deviation of 0.007 ns. The RF GPS signal goes through different elements: an 50 Ohms RF cable with an isolator : mean group delay in the L1 frequency band: 31.578 ns standard deviation group delay in the L1 frequency band: 0.105 ns mean group delay in the L2 frequency band: 31.633 ns standard deviation group delay in the L2 frequency band: 0.105 ns an amplifier: Mini-Circuits ZFL – 2500 VH with the following properties : frequency range: 10 MHz to 2.5 GHz gain: 20 +/- 1.5 dB maximum input power: 10 dBm mean group delay : 0.550 ns standard deviation group delay : 0.010 ns the amplifier is fed by a DC voltage source (Agilent E 3630 A), which generates 23 Volts DC The RF GPS signal has a delay respectively of 32.128 ns and 32.183 ns with 0.105 ns of standard deviation, at the input of the digital scope, in the L1and L2-frequency-bands. 2.2 Dual frequency calibration method The calibration method enables to evaluate the GPS internal delay in the L1- and L2-frequency bands. The calibration is performed in four steps: 1st step: running the internal calibration of the GPS simulator 2nd step: creating a scenario for the GPS simulator

Figure 1 : Synopsis of the GPS simulator calibration

The GPS simulator is a Spirent STR 4760 (serial number 2041), which is calibrated by Spirent every year. To get the RF signal, we use the MON/CAL OUT 1 port, which allows the GPS signal to be monitored at a level approximately 60 dB higher than the one on the front panel. This output port enables to have a GPS signal level, which is well above the thermal noise floor. That’s one reason why this port has been used for the simulator calibration. The second reason is that this port is used by Spirent for the annual calibration. The PPS signal has a 1s period with a high time of 100 ms. The PPS signal is synchronised with 1s epochs in simulated GPS time.

3rd step: storing the PPS and the RF signals with the digital scope 4th step: processing the stored data to get the estimated delay Internal calibration The internal calibration is performed to have a better stability concerning the estimated simulator delay. After running the internal calibration, the GPS simulator has its optimum system performance. The calibration utility aligns the code and carrier phases of each channel within the simulator unit.

GPS simulator scenario To be able to calibrate the GPS simulator, the Single Channel Utility has been used. This utility allows one channel of the satellite simulator to be used as a simple GPS signal source. The GPS signal characteristics may be interactively controlled. These main simulator characteristics are the following: the PRN code, the navigation data message, the signal level, the Doppler. Storing the PPS and GPS RF signals The PPS signal and the RF signal have been stored using the digital scope Tektronix TDS 694 C. Each signal is stored in a file with a Mathcad format. In each file the sampling frequency is equal to 10 GHz and there are 120000 samples.

Processing the stored data The method used to estimate the simulator delay looks like a standard GPS signal acquisition. Two parameters have to be found: the code offset versus the PPS generated by the simulator the residual carrier frequency offset These two parameters are evaluated simultaneously. The number of residual carrier frequency bins that must be observed is equal to 21 with a step of 4 kHz. The number of code offset bins that are observed is equal to 120000 with a step of 100 ps.

The Figures 2 and 3 give a first approximation of the simulator delay, which seems to be equal to 100 ns (C/A code in the L1 frequency band). Looking at the Figure 3, we can say that the zero-crossing method cannot be used without pre-processing the GPS signal because of the noise level.

Figure 4: zoom on the correlation function near the maximum value (L1)

For each frequency offset, the following operations are performed:

Figure 2 : GPS signal (C/A code in the L1 frequency band) with the PPS signal generated by the simulator

the RF GPS signal is extended from 120000 to 131072 samples using the zero-padding method (131072 = 2.1017). The zero-padding improves the correlation function. the RF GPS signal is multiplied by an oscillator, whose frequency is equal to the L1-central frequency (1575.42 MHz) plus the residual carrier frequency offset the base band GPS signal is filtered with a Nyquist filter, whose parameters are the following: roll-off factor: 0.5 Nyquist frequency: 40 MHz

Figure 3 : zoom on the GPS signal (C/A code in the L1 frequency band)

the base band GPS signal is correlated with the reference signal, which is synchronised with the PPS signal generated by the simulator

if the absolute maximum value of the correlation function is not bigger than 90% of the roughly estimated maximum amplitude value of the input signal, then the whole process has to be made one more time adding 30 degrees to the local oscillator phase.

The correlation function enables to know the following values: the estimated time offset 231.5 ns (C/A code in the L1 frequency band) 234.5 ns (C/A code in the L2 frequency band) the estimated residual frequency offset: 8 kHz (C/A code in the L1 frequency band) 16 kHz (C/A code in the L2 frequency band)

second is the time, where the signal exceeds the 1 Volt level. This level has been used in the experiment (4). Considering these two remarks, we can respectively estimate the GPS simulator delay in the L1- and L2frequency bands: 100.005 ns and 108.475 ns. 2.3 Results Before storing the signals, an internal calibration is done each time, to get a more reliable delay estimation of the simulator calibration. Date

L1 correlation delay [ns]

L1 delay [ns]

L2 correlation delay [ns]

L2 delay [ns]

17/02/2005

239,9

105,105

243,3

108,45

18/02/2005

239,9

105,105

243,5

108,65

08/03/2005

239,8

105,005

243,4

108,55

09/03/2005

239,6

104,805

243,1

108,25

Average

-

105,005

-

108,475

Table 1: GPS simulator calibration results

These results have been reached adding a phase offset to the local oscillator phase.

L1 standard deviation [ns]

L2 standard deviation [ns]

PPS cable

0,1 ns

0,1 ns

RF cable

0,105 ns

0,105 ns 0,148 ns

GPS simulator*

0,122 ns

Interchannel biais**

0,333 ns

0,333 ns

Tick-to-phase

0,100 ns

0,100 ns

Correlation accuracy

0,100 ns

0,100 ns

Results

0,408

0,417

Table 2: Measurement errors for the simulator calibration * the measured value is below the SPIRENT measured value from 1.668 ns ** this value is given by SPIRENT.

3. GPS RECEIVER FOR TIME TRANSFER Figure 5 : base-band signal using the optimal residual frequency offset (1575.44 MHz) in the L1 frequency band

CALIBRATION

3.1 Calibration method Once the simulator internal delay has been computed, the calibration of the GNSS receiver can start. The calibration of the GPS receiver is based on the pseudoranges-measurements. The GPS internal delay is calculated thanks to the following equation:

R X R SR − − SD = R X D c c

Figure 6 : correlation function using the optimal residual frequency offset (1575.44 MHz) in the L1 frequency band

The stored signal length is roughly equal to 12 code chips and the pulse of the PPS signal is located in the middle. To improve the correlation function, the reference signal is shifted by 6 code-chips. The GPS

(1)

with, RXR: Pseudorange computed by the GPS receiver SR: Pseudorange simulated by the GPS simulator SD: GPS simulator internal delay RXD: GPS receiver internal delay c: light celerity (299,792,458 ms-1) The GPS simulator internal delay has been calculated during the simulator calibration. The pseudoranges simulated by the GPS simulator are stored in a file

with an ASCII format. The pseudoranges computed by the GPS receiver are stored in a file with a Rinex format. The comparison between the two sets of measurement enables the estimation of the GPS receiver internal delay. As mentioned already, the Ashtech Z12T Rx, involved in the experiment, are not the standard Z12 Rx. Indeed, these specific receivers were designed to perform repeatable measurements. These measurements depend on the so called “Tick to Phase” (TtP) value (time difference between the 1 PPS rising tick and the next rising zero of the 20 MHz). As a consequence the estimated GPS receiver internal delay is related to this TtP value. The whole experiment has been performed with a unique TtP value. 3.2 Calibration configuration The calibration of the GPS receiver is achieved with the following configuration. To calibrate the GPS receiver, the output level must be the standard GPS level. For this reason, attenuators are added in the RF link to compensate the 60dBamplification of the simulator (see 2.1 paragraph). The attenuation value used was of approximately 60 dB. The generated PPS signals have a 1s period with a high time of 100 ms. The PPS signals are synchronised with 1s epochs in simulated GPS time.

The effect of the RF link is as follow : L1 frequency band mean group delay: 5.99 ns standard deviation group delay: 0.55 ns L2 frequency band mean group delay: 5.42 ns standard deviation group delay: 0.5 ns 3.3 GPS Simulator scenario The GPS simulator can generate simultaneously a radiofrequency signal issued from several channels. Each channel is calibrated separately by the simulator, using the tool called Channel Alignment. The simulated scenario is a specific scenario with: a receiver is located at a fixed-position (0° N, 0°E) no ionosphere and no troposphere two transmitted pseudorandom codes in the L1 frequency band: the C/A code and the P code one transmitted pseudorandom code in the L2 frequency band: the P code the GPS constellation is the default GPS constellation from Spirent, with 24 healthy space vehicles 3.4 Calibration results The calibration has been performed on two Ashtech Z12 T Rx, one from the BIPM and the other from the CNES. Pseudorange difference The Table 3 is focused on the pseudorange difference, which corresponds to the difference between the simulated and the received pseudorange (see equation 1). Mean at L1 [m]

Std Dev at L1 [m]

Mean at L2 [m]

Std Dev at L2 [m]

CNES

96,763

0,165

96,557

0,682

BIPM

92,353

0,205

97,443

0,671

Figure 7 : Synopsis of the GPS receiver Z-12T calibration

The RF link between the simulator and the receiver has got the following elements: a DC block two attenuators : 60 dB a 50 Ohms RF cable

Table 3 : Pseudorange difference

The Table 4 corresponds to the delay of this pseudorange difference. Note the impressive difference between the deviation at L2 frequency band and the deviation at L1 frequency band. This point must be cleared for the next experiments.

Mean at L1 [ns]

Std Dev at L1 [ns]

Mean at L2 [ns]

Std Dev at L2 [ns]

CNES

322,767

0,55

322,079

2,275

BIPM

308,056

0,684

325,035

2,238

Table 4 : Delay difference

Tick-to-phase calculation The important aspect of the Ashtech Z-12T Rx is the repeatability of the internal reference. The 20 MHz zero crossing that immediately follows the rising tick of the 1 PPS signal marks the internal reference (zero) time for the Rx. Whether this internal reference can be determined by opening the receiver case and inspecting the jumper on the circuit board : its position is clearly marked as rising or falling. The 1 PPS input does not have a zero rise time. Therefore, the rising 1 PPS signal is measured at the point it achieved 1 Volt. A TtP of 10.1 ns has been measured. The accuracy of determining the TtP is assumed to be 0.1 ns. Rx calibration results The receiver delays are computed using the equation (1). (RxR-SR)/c at L1 [ns]

- SD at L1 [ns]

- RF cable delay at L1 [ns]

= RxD at L1 [ns]

CNES

322,767

-105,005

-5,99

211,772

BIPM

308,056

-105,005

-5,99

197,061

Table 5 : Receiver delay at L1 (RxR-SR)/c at L2 [ns]

- SD at L2 [ns]

- RF cable delay at L2 [ns]

= RxD at L2 [ns]

CNES

322,079

-108,075

-5,42

208,584

BIPM

325,035

-108,075

-5,42

211,54

The measured receiver delays must be computed with several TtP value in order to compare our calibration with the NRL’s calibration. However, the isolator and DC-bloc must be replaced in the next experiments. Moreover it would be interesting to measure the interchannel biais of the GPS simulator in order to reduce the error budget.

ACKNOWLEDGEMENT We would like to thank Joe White and Ron Beard from NRL for their advices during the experiment.

REFERENCES [ASC 98] Ascarrunz F.G., Parker T.E., Jeffers S.R., “Pseudo-random code correlator timing errors due to multiple reflections in transmission lines”, in Proceedings 30th Annual Precise Time & Time Interval (PTTI) Meeting, pp. 433-438, 1998 [LAR 03] Larson K. M., “Development of a Carrier Phase Time and Frequency Transfer System”, PhD dissertation, University of Colorado, 2003 [PET 00] Petit G., Jiang Z., Ulrich P., and Taris F., “Differential calibration of Ashtech Z-12T receiver for accurate time comparisons”, in Proceedings of 14th EFTF, pp. 40-44, 2000

Table 6 : Receiver delay at L2 L1 standard deviation BIPM-receiver [ns]

L2 standard deviation BIPM-receiver [ns]

L1 standard deviation CNES-receiver [ns]

L2 standard deviation CNES-receiver [ns] 0,5

RF cable

0,55

0,5

0,55

Tick-to-phase

0,1

0,1

0,1

0,1

SD

0,408

0,417

0,408

0,417

(RxR-SR)/c

0,684

2,238

0,55

2,275

result

0,973

2,333

0,884

2,368

Table 7: Measurement errors for the receiver calibration

These results can not be directly compared with the NRL’s result [WHI 00] [LAR 03], because the measures have not been performed at the same TtP.

4. CONCLUSIONS This experiment has succeeded in performing an absolute GPS receiver calibration. The GPS simulator has been calibrated thanks to a new method proposed by CNES. The new method called “acquisition like process” has improved the simulator calibration accuracy as expected.

[WHI 01] White J., Beard R., Petit G., and Powers E., “Dual frequency absolute calibration of a geodetic GPS receiver for time transfer”, in Proceedings of 15th EFTF, pp. 167-172, 2001