It cannot go on forever… We have to find a solution!

Durk van Willigen Delft University of Technology, em. Prof Dr Reelektronika BV, CEO, The Netherlands .

The discussions in the United States among LightSquared, the Federal Communications Commission (FCC), the Global Positioning System (GPS) industry, Department of Defence (DoD), Department of Transportation (DoT) and users about the division of spectrum in the L-band should not be seen as just an internal US affair. The strength of both the camps, Telecom providers and the GPS industry indicate that this battle may easily expand to other parts of the world. For many, this battle was a surprise, particularly for a peaceful world like navigation, where discussions are more gentlemen like and either focused on which GNSS systems is the best, or on the backup of GNSS which is so vulnerable and where society cannot function anymore without GNSS. All this changed abruptly when LightSquared published plans to install 40,000 transmitters in a band adjacent to L1 band used by GPS. This led to numerous protests in the GPS industry and among its users as what has been published in many magazines. It is for us to see whether it is telecom versus GPS, or is there a sensible cooperation in reach?

Many may wonder whether these debates are really important and why this exploded so unexpectedly. A short look into radio navigation history may help to understand the underlying discomfort on both sides. Radio navigation started as terrestrial based systems, and later some 30 years ago space-based systems took over.

The first worldwide radio navigation system was Omega, which basically used four carrier wave signals in the 10-14 kHz band. For identifi cation and wide-laning purposes, the four carriers were switched on and off in an accurately defi ned pattern. The radiated signals propagated quite easily in the layers between the earth surface and the ionosphere. The attenuation in those layers is moderate, so, the 10 kW ERP signals could be received worldwide. As the propagation models are complex and not accurately known, the attainable accuracy was in the range of some kilometres. As Very Low Frequency (VLF) signals can also be received under water, therefore, the navy’s interest at that time was understandable. However, the overwhelming introduction of GPS made Omega disappear rapidly. However, it is important to remember that the on-off modulation pattern in Omega required a total spectrum bandwidth of approximately 160 Hz.

Loran-C, operating at 100 kHz, showed a much better accuracy. Although the system was originally specifi ed to achieve absolute accuracy levels of better than one quarter mile i.e. 463 metres, in practice much better results were obtained. The main reason was a better understanding of the propagation phenomena, and the possibility to discriminate between groundwave and skywave signals. This was achieved by applying pulse-like amplitude modulation instead of using carrier waves only. As the skywave path was longer than that of the groundwave, the receiver could relatively easily select the groundwave for the position determination. This ground/sky wave discrimination works well up to distances of 1,000 km or more. The models of the groundwave propagation are much more accurately known than that of Omega which results in accuracy levels down to 50 metres. If differential Loran-C techniques are applied, absolute accuracies better than 10 metres are attainable. Loran-C, and its successor eLoran, is in use in many parts of the world like northwest Europe, Russia (called Chayka), South Korea, Japan, India and the Middle East. To overcome the high atmospheric noise levels in the 100 kHz band, high-power transmitters in the range of 100 kW to 2 MW were a necessity. For a long time these high energy levels were considered as an economic disadvantage, but today it is a blessing as jamming these strong signals over larger areas is very diffi cult. This makes eLoran a very capable and effi cient backup for today’s GNSS systems which are easily denied over rather large areas with simple low-power jammers. The Loran pulse modulation of the 100 kHz carrier implies a larger claim on the spectral bandwidth and amounts to 20 kHz.

DECCA is another LF system which also has been decommissioned after GPS became operational. This system used a number of carriers which were on-off modulated in well-defi ned rhythms’ as identifi cation of the station. The required total spectral bandwidth was about 120 Hz. Due to the carrier-wave type of signals, ground-waves cannot be separated from sky-waves which, depending on the ionosphere conditions, limited the working range down to a 100 km which made it a typical coastal navigation system. The accuracy was quite good and could be around 50 metres.

When the Russians launched the fi rst Sputnik satellite, the Americans listened to the broadcast signals and observed a rather strong Doppler shift of the received signals. These Doppler shifts made computation of the satellite’s orbit feasible. Then the US scientists developed the fi rst space-based navigation system called Transit, while the Tsiklon in the former USSR, worked in an opposite manner. The orbits of the satellites are now accurately known and by measuring the Doppler shifts over some time, the users’ position could be established. Accuracies were in the order of 10 – 50 metres, and the systems could be used worldwide. Transit broadcasts two carriers on 150 and 400 MHz, respectively. The Doppler peak-to-peak shift ranged from 4 kHz on 150 MHz to 10 kHz on 400 MHz. So, the total spectrum bandwidth was a mere 14 kHz.

The real major step in all aspects of radio navigation was done by GPS in the US and GLONASS in Russia. These systems apply CDMA code for ranging and data transfer and the signals are spread-spectrum modulated on the L1, L2, and L5 bands. Although the accuracy of these worldwide systems is an impressive 2 – 20 m, the spectrum requirements are also impressive being approximately 60 MHz. This bandwidth is primarily needed for ranges which is based on correlation of the received signals with the replica code of the selected satellite. The data transmissions need less than 1 kHz bandwidth.

The table 1 shows the enormous differences in spectrum needs for a number of well-known systems.

This leaves us with two burning questions whether these enormous spectrum claims by GNSS are really needed, and were at the start of designing GPS, alternative solutions considered that have more modest bandwidth claims without losing performance?

CDMA offers many advantages in respect of accurate measuring the time of arrival of the signals, rejecting multipath, simple decoding of the received signals, and also that all satellites can broadcast in the same part of the spectrum. For the military it was important that that these signals could not be used when the code was not known. Further, the applied spreadspectrum technique makes it diffi cult to detect the unknown received signals which are some 20 dB below the galactic noise level. The disadvantage of CDMA, however, is that range measurements require a rather large bandwidth to obtain sharp correlation responses, a prerequisite for accurate ranging.

It is interesting to see how Omega was designed. All eight stations broadcast sequentially four different carriers in the VLF band. On basis of the way the frequency steps were formatted, the receiver could identify which station was broadcasting on one of the four frequencies. Although the accuracy was rather poor, the basic concept was quite ingenious as it offered worldwide a simple carrier tracking ranging technique with a total bandwidth of just a mere 160 Hz.

Although the GNSS signal concept cannot be changed anymore, it is still interesting to see whether a pure carriertracking technique could also lead to an accurate GNSS navigation system with a large reduction in spectral bandwidth. For example, assume that a satellite sends on L1, L2 and L5 a carrier, each with two sub-carriers. This would result in 9 carriers in total. Further, we assume that only carrier tracking is used for range measurements in order to keep the bandwidth limited. The largest unambiguous range should be about 30,000 km for Medium Earth Orbit (MEO) satellites. This can be achieved by using two carriers separated by 10 Hz. The phase difference between the two carriers which started at the same time in the SV is a measure of the distance between de SV and the receiver. We assume that this range measurement can be done with an accuracy of better than 10 percent of unambiguous range, so a 3,000 km in this example. The next step is then to use two frequencies at 100 Hz apart, which would yield an unambiguous range of this 3,000 km and with an accuracy of 300 km. In practice, the precision on L-band frequencies is better than 10 percent, more around 1 percent of the wavelength. If we use the following set of nine carrier frequencies as mentioned in table 2, we would end up in carrier tracking without ambiguous range problems down to a precision at L1 of a few mm. By adding quadrature modulation on for example L1 and L2, orbital and time information of each SV can be received by the user.

Ionosphere data can be retrieved from the differences between de basic carriers on L1, L2 and L5. This relatively simple concept is modest in its spectrum needs. Although only carriers are used, Doppler effects will consume 30 kHz on L1, 24 kHz on L2 and 21 kHz on L5, totalling to 75 kHz. The data modulation is just a small fraction of the Doppler. Further, as all satellites would operate on a different frequency, no mutual interference would be experienced. So, for 100 SVs, the total spectrum needed is limited to 7.5 MHz. Two additional advantages of this concept are that although more satellites will consume more spectrum, this will not lead to an increased noise fl oor as is the case with the CDMA structure of GPS. Due to the applied carrier technique it would be more diffi cult to jam; the power density is much higher, and the receiver tracking bandwidth is rather small. Jamming over the entire spectral bandwidth would require signifi cantly much more power than with CDMA techniques.

However, the above given idea is just an exercise as modifying the current GNSS structures is out of the question. But the present confl ict between LightSquared and GPS would be easier to solve if we can have such a spectrum-effi cient system.

The basic issue between LightSquared and the GPS spectrum is the so called comfortzone, the separation between the GPS spectrum and that of LightSquared. GPS receivers cannot apply brick-stone type of bandpass fi lters in front of the LNA in the antenna without sacrifi cing its noise performance and still be able to show excessive attenuation for LightSquared signals. Although JAVAD had published that they have developed a useful solution, we have not yet seen the objective tests to confi rm their claim. The large bandwidth need of GPS is easier to understand by looking at the eye-pattern of GPS and digital telecom signals. GPS receivers need very steep slopes of the eyes in order to make sharp correlation peaks feasible, while telecom receivers just need to discriminate between ‘ones’ or ‘zeroes’. The latter allows bandpass fi lters which are about as narrow as 50 percent of the span between the fi rst two ‘nulls’. GPS receivers prefer fi lters that are 2 to 3 times wider than the distance between the two ‘nulls’. See Fig. 1. High-end GPS receivers can be upgraded according to JAVAD’s approach. Unfortunately, this method is nearly impossible to implement in the millions of GPS receivers used in smart phones and car navigation systems. Should all existing GPS C/A type receivers simply be depreciated and destroyed? And who will pay for this enormous logistically challenging modifi cation or retro-fi tting process? We may expect an interesting series of law suits.
An underlying issue is the commercial spectrum business which is so different for GPS and for telecom. GPS and all other GNSS systems have free access to use 60 MHz wide spectrum while the telecom providers usually have to pay for this. For example, the G3 30 MHz wide spectrum in the UK has been sold by auction for 1.350 billion Euros/year which is 22 Euro/year/ capita. An identical approach happened in the Netherlands where the Dutch pay about 9 Euro/year/capita. These auctions yield relatively easy incomes for governments as the tax mechanism is simple. If this should also be done with the use of the GNSS bands, how to collect these taxes, and which government would be allowed to collect this money? Should the GNSS providers be authorized for this, and would, for example, South American users want to pay for US, Russian, Chinese or European spectrum claims? Who owns the spectrum anyway? Can anybody sell something which he doesn’t own? This lack of balance between the GPS community and the telecom providers might disturb the relations between the two large and powerful communities.

Conclusions

• Pressure on the GNSS spectrum will not just continue but will even increase in many parts of the world.
• GPS spectral claims are primarily based on current technology and on its enormous installed user base.
• Spectrum is becoming scarce, so the price will rise but the telecom industry shows that this does not block the business profi tability.
• Balancing the ‘spectrum rent’ costs between GPS and telecom users might help to relax the discussions.

Solution

A solution shall be found as both, GPS and telecom networks are essential parts of today’s economies. GPS might shrink their spectrum needs and the GPS industry/providers/users could start paying for the spectrum. Both solutions, at present, are most likely just wishful thinking. So, the only remaining solution is to improve GPS receivers so that they can withstand the new powerful neighbours. However, this may cause a costly operation and logistically nearly impossible to realize in a couple of years. But, JAVAD made the fi rst step which is diffi cult to ignore! Who’s next?