Potential GNSS Back-up Improves to GPS-Level Accuracy
A new enhanced differential Loran system demonstrates 5-meter accuracy not achievable by the current DLoran system, and requires less expensive reference stations. A prototype tested in Rotterdam’s Europort area uses standard mobile telecom networks and the Internet to reduce correction data latency — a key source of error — by one to two orders of magnitude.
By Durk van Willigen, René Kellenbach, Cees Dekker, and Wim van Buuren
For maritime applications, Loran is considered as the most promising backup for GNSS for situations where the use of navigation satellite signals is denied. For this reason, the Dutch Pilots’ Corporation askedReelektronika to investigate whether differential Loran could meet the Dutch Pilots’ 5-meter accuracy requirement for a harbor navigation system. This proved to be an enormous challenge, as preliminary tests showed that even 10 meters was difficult to achieve with differential Loran (DLoran) as promoted by Trinity House, the UK lighthouse authority. This led to a thorough renewed investigation of all possible error sources of a complete differential Loran system. The outcome of this research is very promising, as a couple of major error sources could be isolated. This made the complete system better understandable, so adequate countermeasures could be taken.
The development of Loran-C started in the United States about fifty years ago. It is a terrestrial low-frequency (100 kHz) system organized as chains, each consisting of a master station with two or more secondary stations. Each station broadcasts in a strict time format series of 8 or 9 pulses of approximately 250 µs. The effective radiated power is in the range of 100 to 1,000 kW, depending on the required working range. These high powers are required by the high levels of atmospheric noise in the 100 kHz frequency band.
Figure 1 shows the test area of enhanced Differential Loran (eDLoran), using the Loran stations of Lessay (France), Sylt (Germany), and Anthorn (UK).
Radiating such high-power pulses requires large vertical transmitting antennae of about 200 meters height (Figure 2). These high power levels have long been seen as a drawback of Loran-C. However, the upcoming GNSS interference risks changed this apparent drawback into an advantage, as jamming such high field strengths is hardly achievable unnoticed. Loran-C is, unfortunately, less accurate than GNSS but it is nearly impossible to jam over large areas. This is one of the major reasons that Loran gets so much renewed interest by all who face risks in life-critical and environment-critical applications of radio navigation.
Standard Loran does not meet accuracy requirements for harbor entrance and approaches. The International Maritime Organization requires 10 meters (95 percent), which is at least 5 times more demanding than standard Loran can provide. So, differential techniques have been developed and implemented, which are comparable with differential GPS. Although the error sources of GPS and Loran are quite different, the major common error source in both systems is the lack of accurate propagation models.
Several years ago, the General Lighthouse Authorities (GLAs) of the UK and Ireland implemented Differential Loran (DLoran) in the test area around Harwich. DLoran is based on a Loran reference station in the area of interest which measures temporal deviations of the measured pseudoranges. These “errors” are then sent to the user receiver through the Eurofix Loran Data Channel. This technique strongly resembles that of differential GPS. Unfortunately, for a number of reasons it proved to be impossible to achieve absolute accuracies of better than 10 meters with DLoran.
This has led to a new research project to find a more accurate differential Loran technique. All possible error sources have been investigated again where possible, producing unexpected solutions regarding accuracy and cost.
The total position error of Loran depends on the accuracy in time of the high-power generated Loran pulses feeding the antenna, the stability of the physical phase center of the Loran transmitter antenna, stability of the tuning of the antenna circuit, the accuracy of the measured additional secondary phase factor stored in the Additional Secondary Factor (ASF)database, and the quality of the Loran receiver. ASF is the additional delay when Loran signals propagate over land with a varying conductivity. As the ASF data are not fixed but vary slightly over time, temporal de-correlation, differential techniques have been developed to counteract that effect. In standard DLoran systems, the differential corrections are sent to the user through the Eurofix data link. Particular error sources include:
Transmitter Timing Accuracy. A Loran transmitter sends about 100 pulses per second. Each station has three cesium clocks time-synchronized to Coordinated Universal Time (UTC) via a time-transfer network. A two-way satellite time-transfer system will make it simpler and more accurate.
Antenna Phase-Center Stability. Loran transmitter antennas are vertical towers approximately 200 meters high to provide vertical polarization. Its phase center, at the published position, does not move more than about 1 meter according to the station crew at Sylt.
This situation is very different for a wire antenna as installed at the station at Anthorn in Northern England. The top-loaded wire antenna is installed between two towers 200 meters tall and separated by 675 meters (Figure 3). In stormy weather, the antenna position is not stable and does not continuously coincide within 1 meter of the published position of the antenna.
ASF Data. The net travel time of the Loran signal from the transmitter to the receiver antenna is the sum of the propagation through the atmosphere (primary factor or PF), some extra delay due to traveling over seawater (secondary factor or SF), and finally ASF. The PF and SF are calculated from models, while the ASF must be measured. These calculations can only be accurate if the net travel time can be accurately determined and the distance between transmitter and receiver can be calculated with the help of GPS-RTK. The time stamps of the signal when leaving the antenna are not sufficiently accurate. The time stamps on the received signals are established by using a GPS-disciplined rubidium (Rb) clock. In conclusion, we cannot accurately measure and compute the absolute ASF values. All mentioned possible errors led to the use of differential techniques.
As it is not possible to measure ASF data to sufficient accuracy, time-stamp errors at the transmitter can be circumvented by applying differential techniques over a limited area of interest. The receiver at the reference site and the rover receiver experience the same transmitter timing error, which makes it a common error and harmless in differential Loran. It is more difficult to cope with the offset of the Rb clocks at the reference and the rover sites, which is, unfortunately, not common-mode. Differential clock errors of a moving rover receiver may easily rise to 20 ns, equivalent to 6 meters. This type of error limits the achievable accuracy of an ASF data base.
The measured/calculated ASF data are due to weather effects on propagation slightly moving with time. That is the reason to use a reference receiver to measure these temporal variations and send these as AFS corrections to the rover receiver via the 30 bps Eurofix data link. Unfortunately, this rather slow data link introduces significant data latency, which turned out to be the source of significant differential Loran errors.
In the UK, many tests have been conducted to measure these ASF shifts and use the Eurofix data link for sending correction data to the user receiver. DLoran data are sent as pseudorange corrections per station. A complete set of DLoran correction data takes about 90 seconds. The UK plans to send correction data from multiple reference stations via a single Eurofix channel. The resulting very large data latency will preclude accuracies any better than 10 meters. The main reason of this conclusion was found by further analysis of measurements of the position of the rover receiver. These positions are shown as a scatter plot in Figure 4.
The left-hand plot gives the position deviation of 2,500 independent measurements, where the scattering was thought to be caused by noise on the measurements. The middle plot is the result after being corrected by DLoran data with a 90-second data latency, which shows a somewhat modified form but still quite noisy plot. However, when the DLoran data were low-pass filtered, it appeared that often all separate measurements more or less formed lines, which would not happen with just atmospheric noise. Further, the scattering after filtering did not decrease much, which would happen if the disturbances were due to noise. See the right-hand plot in Figure 4.
This demonstrates that the source of the problem is the slow data rate through the Eurofix channel, in combination with the movements of the phase center of the transmitter antenna at Anthorn. Apparently, the solution had to be found in a much faster data link which could neither be offered by Eurofix (30 bps) nor by the U.S.-proposed OFDM technique with a data rate of approximately 1 kb/s. This unexpected result forced us to drastically change the concept of differential Loran to get much better accuracy results, as requested by the Rotterdam pilots.
Enhanced Differential Loran
The above mentioned difficulties with DLoran have led to a new concept of differential Loran which had to fulfil three important primary improvements. The first is a significant reduction in the latency of the data in the data channel; the second is that a large number of reference stations should be allowed to send correction data to the user without saturating the data channel. Finally, it is necessary to measure ASF data more accurately without being dependent on atomic clocks.
The simple conclusion was that Eurofix could not meet the first two improvements. As Eurofix is an invention of Delft University in the Netherlands, it was somewhat painful for the Dutch to admit that a much faster data link is absolutely needed to achieve a two-fold better differential Loran position accuracy. However, Eurofix is still the prime GNSS backup candidate for distributing accurate UTC over very large parts of Europe. Further, Eurofix has the capability to send short messages, which might be encrypted for secure communication purposes that might then form a terrestrial backup for Galileo PRS.
Finally, the third improvement to generate more accurate ASF data cannot be found on a pseudorange method but has to be established on position bases.
Instead of using the Eurofix channel, eDLoran uses the public Global System for Mobile (GSM) network to send the differential corrections to users. eDLoran receivers therefore contain a simple modem for connection to the GSM network. The eDLoran reference stations are also connected to the Internet, which may be implemented via a cabled access or also via a GSM modem.
Fortunately, today many GSM networks are robust in respect of GPS outages. The eDLoran concept is quite simple and is shown in Figure 5.
The eDLoran infrastructure is not connected with any Loran transmitter station and operates completely autonomously. An eDLoran reference station is connected to a central eDLoran server by its connection to the Internet.
The measured positions of these reference receivers are processed in the eDLoran server, which returns the resulting correction data to the user, also via the Internet. Data latency will be not more than 2 seconds. The rover receiver starts the entire process by sending the raw position to the server, which will then return the optimal ASF database for that particular area. Corrections can be calculated by using data from multiple reference stations. Reference stations for eDLoran are small and consume not more than 10 Watts. Two types of reference stations are under development. A portable simple battery-powered version, not larger than 2 meters, can operate for 8 hours. This version is meant to do interference analysis on selected candidate locations. For a permanent installation, a continuously operating solar-powered unit is also under development. See Figure 6.
It has been mentioned that measuring accurately the departure and arrival times of Loran pulses is difficult. It is however needed in order to work out the ASF data on the pseudorange measurement for each Loran station in view. Therefore, a DLoran ASF measurement receiver concept uses Rb clocks and is relatively large and power-hungry. With eDLoran, position offsets due to ASF effects are measured and an eDLoran reference server outputs position- instead of pseudorange-corrections. Measuring positions is much simpler and more accurate and can be done with standard miniature low-power eLoran receivers. No GPS-disciplined Rb clock is needed, and total costs are significantly lower. The gain in accuracy of this simpler ASF measurement receiver together with the very low data latency is one of the reasons that the resulting eDLoran position accuracy is now approximately 5 meters instead of 10 meters with DLoran.
We conducted real-life static and dynamic tests to demonstrate the capabilities of this new concept. The static tests were done in post-processing with logged data from Hook of Holland and at Reelektronika, about 40 kilometers to the east. Only standard eLoran receivers, mostly equipped with E-field antennae, were used, and no atomic clocks were applied. At Reelektronika ,we used two 2-meter mini-reference stations, while in Hook of Holland the Loran antenna was mounted on top of the 30-meter lighthouse. Dynamic tests were done on board of the MS Polaris, the new pilot-station vessel of the Dutch Pilots’ Corporation.
Static Tests. To give a realistic image of the resulting errors of eDLoran, the scatter plots in Figures 7 and 8 are depicted in the position domain. The radial errors are shown in the time domain where the horizontal axis gives the 5-second epochs. The left and the middle plot show the results of the rover and the reference receiver, respectively. The eDLoran plots on the right depict interesting results, as those variations in ASF are largely cancelled while the scattering is smaller than that of the measurements at the rover and the reference receiver, individually. The scattering at the two locations was apparently partly due to low-frequency disturbances, for example because of the moving phase center of the antenna at Anthorn, or instabilities in the time-control loops in the transmitters.
Figure 7 shows the situation where the rover and the reference receiver were separated by 11 kilometers, while Figure 8 depicts the results when the rover receiver was at Reelektronika, more than 40 kilometers from the reference site at Hook of Holland.
This effect of movement of the phase center of the transmitter antenna is further made visible by applying an alpha-tracker (α = 0.9) on the position data of both receivers, which have an update rate of 5 seconds. The lining-up of dots on some parts of the scatter plots in Figure 9 are believed to be due to swaying of the transmitter antenna. Due to the low-pass filtering, the disturbances now contain fewer high-frequency terms.
Investigating the radial error plots of Figure 9, it is remarkable that the large excursions at event 1880 largely cancelled. The disturbance at event 1880 might be caused by antenna movement at Anthorn, which is nearly perfectly cancelled by eDLoran.
Investigating the radial error plots of Figure 8 and 9, it is remarkable that the large excursions around epoch 1900 largely cancel, while this is not happening at epoch 250. There, some local interference might have been the cause. The disturbance at event 1900 might be caused by swaying of the Anthorn antenna which is then a common-mode error at both receivers and is therefore strongly reduced in the eDLoran plots.
Dynamic Tests. Dynamic testing on board the Polaris at sea (Figure 10) is somewhat more complex to do correctly. The eDLoran receiver was installed about 1 meter above the GPS-RTK reference receiver. In this way, the lever-arm problem of not installing the antennae of the two receivers at the same location was avoided. The next issue was measuring ASF position data, which should happen synchronously with the GPS measurements. Time synchronization can be achieved by using simple GPS receivers at both Loran receivers. Some months later, the eDLoran concept was tested by using the stored AFS data and using a Reelektronika eDLoran receiver as a portable pilot unit (PPU) which looks identical to the GPS-based units the Rotterdam pilots use, manufactured by AD Navigation in Norway.
The results have been demonstrated to the harbor authorities in real-time on the laptop of the pilots on which the GPS-RTK and the eDLoran position were simultaneously shown. See Figure 12, where the large gray ship model represents the position and heading derived from GPS-RTK. The width of the ship model is 10 meters. The red triangle gives the eDLoran position; it remains within the borders of the ship symbol. For further demonstration purposes, the logged GPS-RTK data could also be plotted on a Google Earth map (Figure 13). The track was widened to 10 meters, as the accuracy requirements are 5 meters on either side of the track. The raw eLoran track is also shown, as well as the final white eDLoran track of unfiltered raw eDLoran data, which stays well within the 5-meter boundaries.
During the sea trials, the eDLoran receiver was connected to the eDLoran server on land via a miniature GSM modem to the Internet. All differential data were read via this mobile link. The required data bandwidth is very low, approximately 150 bps per ship (client).
The outcome of this research opens some new and quite surprising possibilities for multiple applications:
- eDLoran offers the best possible eLoran accuracy, as it does not suffer from unstable transmitter antennas, sub-optimal timing control of the transmitter station, and differential data latency.
- There is no need to replace older Loran-C stations with eLoran transmitters; this potentially would save large amounts of money. Further savings may be obtained by containerizing the transmitter and operating the stations unmanned.
- Installing eDLoran reference stations is fast, simple, and very cost-effective.
- The Eurofix Loran Data Channel can be freed from a relatively large stream of DLoran data, which leaves the full data bandwidth available for UTC and short-message services over very large areas.
- As there is no data channel bandwidth limitation, multiple reference stations can be installed, offering increased reliability and making the system more robust to terrorism and lightning damage.
- Single or multiple eDLoran servers can be installed in a protected area. There is hardly a practical limit to the number of differential reference stations to serve.
- The server selects the most optimal differential data based on the raw position of the user (client) and the available reference stations.
- As there is no need for any Loran data channel, eDLoran can be installed in all locations where Loran or Chayka coverage and access to the Internet are available. Required data bandwidth is approximately 150 bps per user.
- Standard eLoran receivers used on controlled trajectories (for example, pilots and ferries) collect position data when accurate DGNSS is available. The collected GNSS and eLoran data can be uploaded to the server to further refine the ASF data base. It is basically a self-learning system.
- All eDLoran reference stations monitor the eLoran and GNSS positions to offer alarm services in case of GNSS jamming or spoofing.
Acknowledgments We are very grateful for the near-endless hospitality of the Rotterdam Pilots and especially the crew of the MS Polaris and the MS Markab. Without their help, we would not have obtained the eDLoran results presented here. During the days at sea, we learned how much experience and professionalism is needed to bring those extremely large vessels safely in the harbor of Rotterdam.
We thank Martin Rumens and Dave Kelleher of Babcock International Group for their valued comments and diagrams.
DURK VAN WILLIGEN is a retired professor of electronic systems for navigation at the Delft University of Technology. He is founder and president of Reelektronika B.V., and started the development of Eurofix in 1985. He received the Thurlow Navigation Award of the Institute of Navigation (U.S.) and the Gold Medal of the Royal Institute of Navigation (UK).
RENÉ KELLENBACH graduated from Delft University of Technology in electrical engineering. After joining Reelektronika as a systems engineer, he has been involved in designing hardware and software for radionavigation and radar systems.
CEES DEKKER graduated from the Delft University of Technology in electrical engineering. He worked previously at Philips Research Labs and now assists Reelektronika B.V. with the development of Loran systems and GPS-related projects, and information systems.
WIM VAN BUUREN is a licensed maritime pilot in Rotterdam who took the initiative to develop a backup positioning system for the approaches to Rotterdam. He has been involved in the design and development of the hardware and software of Portable Pilot Units on a national and European level since 2000.