Pilot study on the validation of the Software- Defined Radio Receiver for TWSTFT

Transcription

1 University of Colorado Boulder From the SelectedWorks of Jian Yao 2017 Pilot study on the validation of the Software- Defined Radio Receiver for TWSTFT Available at:

2 Pilot study on the validation of the Software-Defined Radio Receiver for TWSTFT BIPM contribution to the validation of the SDR for TWSTFT Zhiheng JIANG and Elisa Felicitas ARIAS Time Department Bureau International des Poids et Mesures Biographies Dr Zhiheng Jiang obtained his PhD at Paris Observatory. He worked at the Research Institute of Surveying and Mapping (Beijing, China) at the National Geographical Institute (France) and since 1998 he is working with the International Bureau of Weights and of Measures (BIPM) as a principal physicist. His main interests are in time metrology and gravimetry, and he is deeply involved in time transfer for the generation of the international time scale, Coordinated Universal Time (UTC). zjiang@bipm.org. Dr Felicitas Arias was born in Argentina. She received the master s degree in astronomy from the University of La Plata (Argentina) and the PhD in astrometry, celestial mechanics and geodesy from Paris Observatory. She was director of the Buenos Aires Naval Observatory and is Professor at the University of La Plata. Since 1999 she is the Director of the Time Department at the International Bureau of Weights and Measures (BIPM), where she is responsible for the maintenance of the Coordinated Universal Time (UTC). Her fields of activity are the space and time references. She is author of about 130 publications in scientific journals and proceedings. farias@bipm.org. ABSTRACT The Software-Defined Radio (SDR) receiver [1-5] has been developed at the Telecommunications Laboratory (TL, Chinese Taipei) for implementation in two-way satellite time and frequency transfer (TWSTFT/TW) Earth stations, aiming at improving the stability of TW time transfer, with particular impact on the diurnal signature present in almost all the links, which is the major uncertainty source in TW time links. The BIPM and the Consultative Committee for Time and Frequency (CCTF) Working Group on TWSTFT launched in February 2016 a pilot study for validating the SDR receiver in view of its implementation for use in UTC time links. Participants to this pilot study are the laboratories contributing to UTC which operate the SDR receiver and the BIPM Time Department. Goals of the pilot study are first, to validate the efficacy of the SDR receiver for improving the TWSTFT uncertainty and significantly reduce the diurnal effects; second to implement routine TW code measurement data through SDR receiver to improve the UTC time comparisons. In the frame of the pilot study, the SDR receiver has been installed and is operational by the end of 2016 at TL, NICT, KRIS, NTSC in Asia, PTB, OP and SU in Europe and NIST in the US. The project to install the SDR receiver in all the TW laboratories is ongoing. We present the results obtained at the BIPM Time Department in the preliminary steps of the analysis and validation of the SDR measurement data by using different methods, such as the time deviation (TDev), the triangle closures, the comparison with GPSPPP [6]. We analysed the data obtained from the pilot study and observed that in some TW links there is significant improvement, while in others no noise reduction is visible. This preliminary result needs more investigation, together with other open points as the calibration, the discontinuities observed after restarting the receiver etc. If the conclusion of this pilot project is positive, a proposal could be discussed at the CCTF to include TW time links with the SDR receiver in the computation of UTC. We study also the potential strategies to further improve the TWSTFT SDR; its combination with GNSS links [7] and the computation of SDR indirect links [8, 9]. Key words: UTC, TWSTFT, SDR, uncertainty, diurnal signature, GPSPPP

3 Remark: In the so far publications, e.g. [1-4], the term SDR has been defined as Software-Defined Receiver. The CCTF WG on TWSTFT has agreed to use SDR to stand for the technique Software-Defined Radio. Therefore the term Software-Defined Receiver in the earlier publications should be replaced by SDR Receiver. In this paper SDR stands for the technique or the related measurement data and results. Correspondingly, Satre stands for the modem currently in use in TW Earth stations or related data and results. 1. INTRODUCTION The measurement uncertainty (u A ) of the TWSTFT link may attain at most 0.2 ns. The dominant instability source in TW is the diurnal signals of which amplitudes are recorded to be ±1 ns or even bigger in some extreme cases, 0.5 ns on average. Efforts have been made in last decade to investigate the physical causes of the diurnal but no convincing conclusion has been obtained. Since 2009, the combination of the TW and GPS carrier phase (through GPSPPP solution, PPP for short) has been used for the UTC time links [7], profiting of the advantage of the long-term stability in TW and the short-term stability of the GPS carrier phase. In the later, diurnals do not exist. The combination technique is successful in both that the diurnal and the white noise are significantly reduced meanwhile the TW link calibration is kept. The TW community continues investigating on the origin of the diurnals. The latest development is the SDR receiver for TW [1-4]. TW with a SDR is a technique under validation, which can significantly reduce the diurnal signals observed in most TW links. The diurnal increases the TW uncertainty, and is one of its major sources. The goal of the pilot project, jointly organized by BIPM and the CCTF WG on TWSTFT and started in February 2016, is to validate the SDR receiver in view of its use in UTC generation. The validation requests that adapted equipment is installed in TW UTC participating stations (PS) and that they provide regularly data along the project. The role of the BIPM is to coordinate with the pilot study participants to collect the data and perform the analysis to assess the capacity of the SDR to reduce the diurnals and improve the short-term stability. After this first step, feedback will be given to the CCTF WG on TWSTFT. A second discussion, which might be beyond the scope of this paper, would take place at the CCTF, with a possible proposal for inclusion of data obtained with the SDR receiver for the computation of UTC. We hope that this study will have positive impact on the equipment manufacturers, motivated by this new competitive technique. Our conclusion is that in some cases, the SDR significantly reduces the diurnals in the TW links and improves the time stability by a factor of maximum 3 (from 100~120 ps to 30~40 ps). However, for some trans-continental long distance links, the gain is not as significant as for short continental links. Further investigation is required. The software currently used at the BIPM for the computation of UTC has been adapted to process both the Satre and the SDR data together with the monthly computation of Circular T. Figure 1.1 Satellite coverage of the laboratories operating the SDR receivers. In this paper we present the results of SDR receiver applied to (Figure 1.1): - the Asia-Asia links between the laboratories of KRIS, TL and NICT, where the satellite Eutelsat 172A is used; - the Asia-Europe links between PTB, NTSC and SU, where the satellite Intersputnik Ekspress - AM22 is used; - the Europe-Europe and Europe-USA links between PTB, OP and NIST, where the Telstar 11N is used.

4 We observe that the gains of the SDR depend a lot on the regions or the satellite coverage. In general, we concluded that gains on links between stations in the same region are more important than those on baselines linking different regions. Therefore, the analyses in this paper are grouped into different regions and between them: Asia-Asia, Asia-Europe, Europe-Europe and Europe- USA. The analysed data span between Oct and Dec We investigated the TWSTFT Satre and SDR, and compared them with the GPSPPP results. The indicators used to validate the SDR are the Time Deviation (σ x ), the standard deviation of the measurement noise σ, the triangle closures, the differences between the two independent techniques TW and PPP (or double clock difference DCD), the gain factor, that can be estimated and defined as the ratio of the standard deviations of Satre and SDR σ Satre /σ SDR, and/or as the ratio of the Time Deviations. 2. A QUICK LOOK AT THE SDR SETUP New developments of TWSTFT technique may drive considerable improvement of the stability at averaging times of one day and less, e.g., employing dual pseudo-random noise (DPN) codes and carrier-phase (CP) based TWSTFT. As well known, the precision of the current TW links involved in UTC computation is still limited due to instabilities of the signal arrival time that suffers from diurnal disturbances. The sources of diurnals are still unclear, and may be not only the variation of the physical propagation delay, but also a composition involving signal interferences and imperfection of TW equipment [5]. A software-defined radio (SDR) receiver was originally designed for implementing the DPN-based or CP-based TWSTFT. In the scope of the pilot project it is used to precisely measure the arrival time of code signal transmitted by the Satre modem. In fact, it has been observed that in some cases the SDR receiver has the capacity to considerably reduce the diurnal variations [1, 2]. Figure 2.1 shows the setup configuration of the SDR in a TWSTFT system equipped with a Satre modem. TWSTFT Satre TWSTFT SDR setup A/D converter: analogue-to-digital converter; AMP: amplifier for optimizing the power level of the intermediate frequency signal. Figure 2.1 Setup of the SDR in a TWSTFT Earth station. There is no change in the standard TW ground station except for a splitter which is added to drive the arriving signal to both SDR and Satre modem. The SDR consists of a reference frequency (RF) amplifier for optimizing the power level of the intermediate frequency (IF) signal, an analogue-to-digital (A/D) converter for sampling the reception signal, and a personal computer for collecting the samples and performing the measurement by the software. In the software, some digital signal processing (DSP) algorithms are applied to measure the time of arrival (TOA) by code phase and carrier phase. Thanks to DSP, the multiple channels can be easily realized to measure more than one TOA at the same time. Since the setup enables independent operation from the Satre modem to the SDR, two parallel TW measurements can be performed and recorded simultaneously.

5 Since 2015, three laboratories in the Asia region, TL in Chung-Li (Chinese Taipei), NICT in Tokyo (Japan) and KRISS in Daejeon (South Korea), have experimentally operated SDRs in their TW systems. The results of these laboratories observations gave the indication that the SDR could significantly reduce the diurnal signal and the measurement noise. Based on this promising result, the BIPM and the CCTF WG on TW launched jointly a pilot study to validate the SDR technique towards its use in the computation of UTC time links. From July to November 2016, the SDR facility has been installed and operated in NTSC, PTB, OP, NIST and SU, with the technical support of TL. All the other UTC TW laboratories, USNO, AOS, CH, IT, NIM, PL and SP expressed their aim to participate to the SDR pilot study. The installation of the SDR facility at their stations is progressing. 3. DATA SETS, ANALYSIS AND RESULTS The SDR data were measured and collected from TL, KRIS, NICT, NTSC, PTB, OP, NIST and SU between Oct and Dec The SDR raw data used in these analyses do not include any TW calibration parameter (RefDelay, CALR and ESDVAR). Because the effects under study are mainly due to diurnal and measurement noise, the analyses and their results will not be affected. The calibration is important for the UTC time transfer but not the key issue in this section. There is no particular technical difference between the calibration of a Satre link and a SDR link. The study is based on the analysis of the DCD (double clock differences) between: a. the TW links obtained by the Satre and the SDR data; b. links obtained by TW and GPS PPP, since they are independent techniques,. The following indicators are used to describe the relative quality of the SDR: - The Time Deviation (TDev or σ x ) of a link technique on different averaging times, SDR vs. Satre. - The standard deviation (σ) of the DCD of the TW (Satre or SDR) and the corresponding GPS PPP links. PPP is used as a reference since the GPS carrier phase has low measurement noise and is almost free of the diurnal. Moreover, TW and GPS are independent techniques. The DCD uncertainty contains both, those of SDR and PPP. So the uncertainty of SDR estimated using DCD is bigger than the real uncertainty of SDR only. - The triangle closure which theoretically equals zero. The non-zero closures represent the true errors resulting from adding the measurement noise and the diurnal. In metrology practice, we rarely know the true error. The triangle closure gives a window to look into it; - The gain factor. Taking the σ x as an example, the gain is defined as the ratio of the two quantities σ x-satre /σ x-sdr, here the σ x-satre and σ x-sdr stand for the TDev of Satre and SDR; - Consistency between results from SDR and from Circular T provides indication of the long-term stability of the SDR. 3.1 Comparison between the TW Satre and SDR time links By comparing the TW Satre and SDR links, we can obtain two important pieces of information relevant to the UTC time transfer: 1) The short and long term calibration stability of the SDR with respect to the Satre link can be determined through the analysis the variations of the DCD between Satre and SDR links, in general their mean values should be constant and the scatter (represented by the standard deviation ) should be as small as possible for different lengths of periods; 2) The gain in SDR with respect to the corresponding Satre link. To do this, we analyse the Time Deviations of the Satre and SDR links as well as their DCD; We analysed the links distributed within a region and between regions covered by different satellites as described in the introduction. The link OP-PTB is presented as a detailed example of the analysis performed on all links. Figure (left) plots both the Satre (black) and the SDR (blue) time links for the baseline OP-PTB during a 100-day period between MJD and (UTC ). The SDR link has been aligned to the Satre link using the first month of data only, with a correction of ±0.415 ns. We observe that the noise is significantly reduced in the SDR link due to mainly the reduction of the diurnal effects. This is confirmed by the TDev in Figure (right). The analysis of the DCD of the Satre and SDR links shows that the amplitude of the diurnals in Satre raw data is about ±1 ns. The mean value of the DCD is ± ns, comes mainly from the diurnal plus the measurement noise. Similar information is obtained when using the GPS PPP link as reference, as will be presented in the next section 3.2. The time stability as obtained from TDev in Table at 2 hour averaging

6 time (Satre data interval) is 307 ps for the Satre and 94 ps for SDR, representing an improvement in the stability with a gain factor 3.3. At all the averaging times from 2 hours to 32 hours, the maximum gain factor is 5.2 at 8 hours and the minimum is 2.1 at 32 hours, and 3.7 on average. This agrees with the conclusion of an earlier study [1]. As indicated above, the SDR link was aligned to the corresponding Satre link using the data of the first month. The DCD values in Figure correspond to an interval longer than three months, with the mean value ±0.412 ns. This suggests that, once the SDR has been calibrated, it can be as stable as the Satre link. Later in the paper, we will show a DCD comparison over more than one year to investigate the long-term stability of the SDR. The analysis of the inter-continental link OP-NIST results in a gain when the SDR is compared to the normal TW link. This is shown in the TDev values in Figure and Table 3.1.2, where the gain factor is significantly smaller than in the short baseline OP-PTB for all averaging times, representing on average 10% improvement in OP-NIST. However, it seems that the gain in short terms is larger than the longer terms. No significant reduction of the diurnal is observed for this particular baseline TW/SDR σ=0.415 ns 1.0 ns TW/Satre MJD Figure left: the Satre and SDR links on the baseline OP-PTB; right: the corresponding TDev of the two links. Table Gain factor in TDev of the SDR versus Satre link (OP-PTB) at different averaging times Average time/h Satre σ Xsat /ps SDR σ Xsdr /ps Gain=σ Xsat /σ Xsdr Mean 3.7

7 Figure The TDev of the Satre and the SDR links on the baseline OP-NIST. Table Gain factor in TDev of the SDR versus Satre links (OP-NIST) at different averaging times. Av. time/h Satre SDR σ Xsdr /ps Gain=σ Xsat /σ Xsdr σ Xsat /ps Mean 1.10 Similar conclusions can be obtained from figure and table for the long baseline PTB-NIST. The gain factor is in general smaller than the inner-european links with average value 1.03representing only 3% improvement. Table Gain factor in TDev of the SDR versus Satre link (PTB-NIST) at different averaging times. Average time/h Satre SDR σ Xsat /ps σ Xsdr /ps Gain=σ Xsat /σ Xsdr Mean 1.03 Figure The TDev of the Satre and the SDR links on the baseline PTB-NIST. Figure (left) plots both the Satre (black) and the SDR (blue) time links for the clock comparison of TL-NICT during February We observe that the noise is significantly reduced in the SDR link, in which the dominant diurnal effect has been greatly removed. This is also confirmed by the TDev in Figure (right) and the Table The differences of the Satre and SDR links or DCD show that the amplitude of the diurnals removed is about 0.2 to 0.3 ns. The standard deviation σ=0.231 ns. The time stability as obtained from TDev at one hour averaging time is 127 ps for the Satre link and 35 ps for the SDR link, representing an improvement in a factor 3.6. This agrees with the result of the short link OP-PTB (Figure and Table 3.1.1) Table shows the gain factors in the SDR with respect to the Satre for different averaging times. The most important gains are in the short-terms less than one hour. Here a factor of 3 to 4 can be reached. The improvement at 12 hours corresponds to the reduction of the diurnals. The gain is reduced from the short term (3.6) to the longer term. And beyond 1-day (1.2), the gain decreases gradually to zero. This consists to the above results. The mean value of gains is 2.2 for this Asia-Asia baseline.

8 ns σ=0.231 ns TW/SDR TW/Satre MJD h/2 h 3h 6h 12h day 3d wk Averaging Figure left: the Satre and SDR links on the baseline TL-NICT; right: the corresponding TDev of the two links. The diurnal is mostly removed and the stability is reduced by a factor of 3.6 at averaging time of one hour and 2.2 on average. σx / ps TW link TL-NICT diurnal TW/Satre TW/SDR Table Gain factor in TDev of Satre versus SDR on the link TL-NICT Average time/h Satre σ Xsat /ps SDR σ Xsdr /ps Gain=σ Xsat /σ Xsdr Mean 2.2 A quick look to the links PTB-SU (Figure 3.1.5) and SU-NTSC (Figure 3.1.6) indicates that both the diurnal and the short-term stability are improved in the link SU-NTSC, while for the other two links, PTB-SU only the short-term stability is improved but no clear effect is visible on the diurnals. Figure TDev of the link PTB-SU with Satre and SDR. Figure TDev of the link SU-NTSC with Satre and SDR.

9 3.2 Comparing the GPSPPP to the Satre and SDR TW results In the section 3.1, the Time Deviation puts in evidence a significant reduction of the noise in some links, including the diurnal noise, and an improved stability in the SDR solutions w.r.t. the Satre TW solution. This is in the case of the inner-asia and inner- Europe links. We will check if the SDR solution is better representing the physical clock behaviour. This can be done using GPS PPP on the same baselines. Being independent from TW, PPP provides a good reference to validate the improvement in the SDR. Its measurement uncertainty is small (u A =0.3 ns [11]) with excellent short-term stability and it is almost not affected by diurnal. For TW Satre or SDR, the one that is closer to GPS PPP will better represent the clock differences. Comparing TW to GPS PPP allows also investigating the stability of the SDR in view of the calibration, a key issue in the UTC time transfer. We emphasize that, SDR and GPS PPP are independent and therefore the standard deviation and the time deviation of the DCD are accurate indicators. As mentioned above, the uncertainty of SDR estimated by using DCD is the pessimist estimation. Baseline TL-NICT Figure left plot shows the TW Satre and the PPP links for TL and NICT during February 2016, where the standard deviation of the DCD is σ=0.212 ns. Figure 3.2.1right plot shows, for the same interval, the TW SDR link and the PPP link where the standard deviation of the DCD is σ=0.135 ns. These results suggest that the SDR link is closer to that of PPP hence represents the clock behaviour better than the Satre link. The gain factor is the ratio of the two standard deviations, and equals 1.6. The GPS PPP solution seems to present a jump of about 0.3 ns on MJD followed by a very slight drift until MJD This should not affect the conclusion that there is a gain of 60%. Figure low plot shows the TDev of the DCD. It is evident that the diurnal and the noise are significantly reduced with respect to Satre when PPP is the reference. On all the averaging times, the SDR agrees to PPP better than the Satre and represents trustfully the clock differences TW/SDR 2.0 TW/Satre ns =0.135 ns GPSPPP ns σ=0.212 ns GPSPPP MJD MJD σ x / ps Figure up left: PPP and Satre links on the baseline TL-NICT. The standard deviation of the DCD is ns; up right: PPP and SDR links. The standard deviation of the DCD is ns. The gain factor in SDR vs. Satre is 1.6; low: TDev of the DCD. At all the averaging times, the TDev of SDR agrees to PPP much better than that of Satre. h/2 h d/8 d/4 d/2 day 3d wk Averaging time

10 Baseline OP-PTB: Figure left shows the comparison between the PPP and the Satre links over the baseline OP-PTB for November The black crosses represent the Satre and the blue circles the PPP measurements. Figure right plot compares the PPP link to the SDR link for the same baseline and the same interval. The standard deviations of the corresponding DCD are ns and ns. The gain factor in SDR vs. the Satre is 1.9. Figure (low plot) shows the corresponding time deviations of the DCD of Satre and SDR links against PPP. On all the averaging times, the TDev of SDR agrees to PPP much better than that of the Satre. A closer look shows that the gains first increase with time and reaches the maximum on averaging time of 6-12 hours, dominated by the diurnal biases, and decrease to minimum about 3 days on averaging time. Table gives the gain factors in TDev in the DCD of Satre versus SDR links against the PPP link on different averaging times from 2 hours to 32 hours. The gain attains the maximum on 4, 8 and 16 hours indicating that the short-term instability and the diurnal are greatly improved. The average gain is 2.0. As underlined above, although the PPP is still more precise than the SDR thanks to the GPS carrier phase information, it is not errorless. So the gains obtained by taking the PPP as reference are less significant than that given in Table for the same baseline due to mainly the uncertainty in PPP measurements. Hence the gain in this case can be considered pessimist estimation compared to the average value 3.7 in the Table ns 2.5 TW/Satre GPSPPP σ=0.440 ns MJD -9.4 ns TW/SDR GPSPPP 0.5 σ=0.231 ns MJD σx / ps -9.6 Satre-PPP SDR-PPP Figure up left: PPP and Satre links on the baseline OP-PTB. The standard deviation of the DCD is ns; up right: PPP and SDR links. The standard deviation of the DCD is ns. The gain factor in SDR vs. Satre is 2.0; low: TDev of the two DCD. On all the averaging times, the TDev of SDR agrees to PPP much better than that of Satre. Table Gain factor in TDev in the DCD of Satre versus SDR links against the PPP link. Average time/h Satre σ Xsat-PPP /ps SDR σ Xsdr-PPP /ps Gain σ Xsat-PPP /σ Xsdr-PPP Mean h 3h 6h 12h 24h 72h * Averaging time

11 Baseline OP-NIST: The standard deviations of the DCD of the Satre-PPP and the SDR-PPP are correspondingly ns and ns, see Figures left and right plots. The gain factor is 1.2 or 20% of gain. Table gives the time deviation values at different averaging times. As seen, no significant gain is visible at any averaging time and the gain factor on average is The mean value above two indicators of the standard deviation and the TDev is 1.13 or 13% of gains. Again it is pessimist estimation. Table Gain factor in TDev in the DCD of Satre and SDR links w.r.t. the PPP link for OP-NIST. Average time/h Satre σ Xsat-PPP /ps SDR σ Xsdr-PPP /ps Gain=σ Xsat-PPP /σ Xsdr-PPP Mean 1.05 It is interesting to point out that the Europe-America links have the minimum gains in SDR in all the links so far considered in this study. The gain factors are of 2 or 3 in the inner-asia and the inner Europe links where the dominant gains come from the significant reduction of the diurnal biases. This does not happen in the Europe-America links. ns GPSPPP σ(dcd)=0.288 ns TW/Satre h 58 3h 6h 12h 24h 72h Averaging time /s MJD Figure left: PPP link and Satre links on the baseline OP-NIST; right: TDev of the DCD of Satre and SDR links to PPP. At all the averaging times, the TDev of SDR agrees to PPP better than that of Satre. σx / ps Satre-PPP SDR-PPP Comparison of the triangle closures given by the Satre and SDR measurements Figure shows the closure variation of the Satre and SDR links of the triangle OP NIST PTB during about 3 months from MJD to Different to a GPS triangle, the three TW links are independent and the non-zero closures represent the true measurement errors present in the diurnal, with amplitudes of about 2 ns, bigger than that of the links due to the uncertainty propagation. This implies that the TW link measurements and the related diurnals are independent. Therefore, the closure is an excellent indicator of the gains in SDR versus the Satre. On about the MJD 57720, the SDR seems been disturbed by unknown reason. However, it will not affect our analysis result and the conclusion. In the data here analysed there are three triangles in three regions: OP-NIST-PTB, NICT-KRIS-TL and PTB-NTSC-SU. The statistics of them and the gain factors are given in Table The gain factor obtained by the ratio of the standard deviations (σ)

12 on average is 4.5. It is interesting to notice that while the instability of the links (diurnal plus measurement noises) varies from 180 ps to 470 ps of the triangle closures, the gain factor in the triangle closures is almost the same: 4.3, 4.5 and 4.8. The triangle analysis suggests: 1) The diurnal is link dependent rather than site dependent; 2) The gain in SDR w.r.t. Satre link is almost constant. Figure Triangle closures involving the Satre (blue) and SDR (red) links over the triangle OP-NIST-PTB. Table Gain factor of the SDR vs. the Satre links given by the triangle analysis. Triangles σ(satre)/ps σ(sdr)/ps Gain factor= σ(satre)/σ(sdr) OP-NIST-PTB NICT-KRIS-TL PTB-NTSC-SU Mean Long-term stability of the SDR Long-term stability of SDR is no doubt one of the most important issues. Once the SDR receiver is calibrated, it should be kept stable. This can be expressed by the fact that the relation of the calibration of the SDR and the Satre links should be kept fixed or simply the DCD is constant. The calibration of TW links with a Satre modem is considered stable with respect to the combined calibration uncertainty of TW and GPSPPP, that is, (1.5²+1.7²)=2.3 ns.

13 Figure Comparison of the UTC(KRIS) UTC(TL) SDR link and Circular T for about 400 days of which there are two continuous 150-day segments. The standard deviation of the DCD (pink curve) is 1.09 ns. Figure illustrates the values of UTC(KRIS)-UTC(TL) measured by SDR and those obtained from the BIPM Circular T for about 400 days. There are two consecutive continuous periods, about 150 days each, between MJD and As seen, during these two intervals the results from the Circular T and the SDR are in good agreement except for the end of the first period and the beginning of the second. The DCD values range over -1.9 ns and 2.5 ns. The latter could be an outlier but all values are taken into account without rejections in the following statistics. On the segment about the MJD 57500, there are anomalies and interruption events where the DCDs are about 1-2 ns. The four events are illustrated in the figure and are repaired by the GPSPPP bridges. A constant of ±1.09 ns is applied to align the SDR to the Circular T. The standard deviation of the DCD is 1.09 ns which is the indicator of the fidelity of the SDR with respect to the true clock differences obtained through Circular T, noting that the 1.09 ns is the combined uncertainty of both the Circular T and the SDR. The value must be divided by the square root of 2. On all the other baselines here analysed, the two data sets of SDR and Circular T-derived are in good agreement, better than most of the DCD of GPS-TW Satre links in other regions, e.g. in Europe- Europe and Europe-America links. 4. FURTHER IMPROVEMENT IN THE UNCERTAINTY OF THE SDR Further improving the stability of the SDR is possible. This could be achieved by: 1) Combination of TW SDR and GPS Carrier phase through the PPP solution. This method has been used in the TW Satre and PPP combination for the UTC computation since Detailed discussion has been given in [7], see also the BIPM Circular T [11]; 2) The TW indirect link method is an effective method for the TW links under certain conditions. Some results are shown here below. Detailed discussion can be found in [8]. An advantage of this method is that no other time transfer techniques are involved; a disadvantage is that there is no certainty that all the TW links can be improved; Figure 4.1 shows the time deviations of the TW PTB-OP links computed with the Satre, SDR and the indirect-sdr between MJD and Table 4.1 gives the gain factors in TDev of the TW SDR direct link vs. the indirect links via NIST. The TDev are computed at 5 averaging times from 2 hours to 32 hours. The SDR direct link (blue line) has a significant improvement compared with the Satre link (red line). While the SDR indirect link (green line) further improves the SDR link. As shown in the Table 4.1, there are gains at all the averaging times. The gain attains maximum on 4 hours and decrease to near zero beyond one

14 day. The gain factor is 1.8 on average. The residual diurnal has mostly disappeared in the indirect SDR link. It has to be noted that improvement has also been demonstrated with indirect TW Satre links. Figure 4.1. TDev of the TW PTB-OP links computed with the Satre, SDR and the indirect-sdr methods between MJD and Table 4.1 Gain factors in TDev of the TW SDR direct link (OP-PTB) vs. the DSR indirect links via NIST. average time/h direct σ X-Dir /ps indirect σ X-Ind /ps Gain=σ X-Dir /σ X-Ind s Mean SUMMARY The uncertainty of the current TWSTFT is limited by the instabilities of the signal arrival time. The SDR implemented in the TWSTFT Earth stations allows the precise measurement of the arrival time of the coded signal and reduces the diurnal significantly in some of the cases here presented. SDR systems have been successfully installed at TL, NICT, KRISS, NTSC, PTB, OP, NIST, SU and IT. Experiments have being carried out since. Launched jointly by BIPM and the CCTF WG on TWSTFT, the pilot project aims at validating this technique by analyses involving stations equipped with SDR in Asia, Europe and USA. After validation at the results of the pilot project, consideration could be made by the CCTF on its use for UTC. This paper presents results of analysis performed on continental and intercontinental baselines using the current procedures for TW time link computation adapted to this particular case. The gain factor of the TWSTFT SDR receiver varies and seems to be correlated with the baseline length. In the cases where the gain is significant it is 2.0 on average in short-term stability and in the diurnal compared with the normal Satre modem solution. More investigation is necessary over links between Europe, Asia and the US before coming to a conclusion.

15 Results presented here are encouraging, but there is need to investigate the following issues: The offsets in the SDR vs. the normal TWSTFT link; The discontinuities and jumps in the SDR raw data; The calibration of a TW station equipped with SDR. It would be appropriate to: Extend the comparison to IPPP Extend the analysis to other links (data sets of inner-asia, inner-europe, inter Asia-Europe, inter Europe-USA) Agree on the convention for identifying the SDR TW equipment in the ITU data file. Finally, SDR receiver is a software modem that collects both the code and carrier phase data. In this study, we use only the code data. The carrier phase information could be used for more deep studies and various applications. ACKNOWLEDGEMENT We acknowledge the TWSTFT laboratories which provided the technical support and data, T. Gotoh, M. Fujieda, S. Yang, W. Wu, D. Piester, J. Achkar and V. Zhang. We appreciate very much the fruitful contributions of Yi-Jiun HUANG and Calvin S. Y. LIN who also provided some of the plots in this paper. DISCLAIMER Although equipment is identified for the sake of technical clarity, none of the BIPM or a Pilot project participant can or will endorse a commercial product. We further caution the readers that none of the described equipment s apparent strengths or weaknesses may be characteristic of items currently marketed. REFERENCES [1] Huang Y.J., Tseng W. H., Lin S.Y., Yang S. H. and Fujieda M. (2016) Introduction of Software-Defined Receivers in Two- Way Satellite Time and Frequency Transfer, Proc. EFTF2016 [2] Huang Y.-J., Tsao, H.-W. (2015) Design and Evaluation of an Open-Loop Receiver for TWSTFT Applications, IEEE Trans. IM, vol. 64, no. 5, pp , 2015 [3] Huang Y.-J., Fujieda M., Takiguchi H., Tseng W.-H., and H Tsao.-W., Stability improvement of an operational two-way satellite time and frequency transfer system, Metrologia, vol, 53, no. 2, pp , Mar [4] Huang Y.-J., Tseng W.-H., Lin S.-Y., Yang S.-h. and Fujieda M., Introduction of Software-Defined Receivers in Two-Way Satellite Time and Frequency Transfer, in Proc IEEE IFCS, New Orleans, LA, USA, 9-12 May 2016 [5] Hejc G. and Schaefer W., Tracking biases caused by imperfections in DLL receivers, in Proc. 41st Precise Time and Time Interval Syst. Appl. Meeting, 2009, pp [6] Yao J., Skakun I., Jiang Z., and Levine J. (2015) A Detailed Comparison of Two Continuous GPS Carrier-Phase Time Transfer Techniques, Metrologia, 52(5), [7] Jiang Z, Petit G (2009), Combination of TWSTFT and GNSS for accurate UTC time transfer, Metrologia 46, [8] JIANG Z, ZHANG V, PARKERT E, YAO J, HUANG Y J and LIN S Y (2017) Accurate TWSTFT time transfer with indirect links, Proc. PTTI2017 [9] Zhang V., Parker T., Zhang S. (2016) A Study on Reducing the Diurnal in the Europe-to-Europe TWSTFT Links, Proc. EFTF2016 [10] TWSTFT Calibration Guidelines for UTC Time Links V2016 ftp://tai.bipm.org/tfg/twstft-calibration/guidelines [11] BIPM Circular T 348, Dec. 2016, ftp://ftp2.bipm.org/pub/tai//circular-t/cirthtm/cirt.348.html