Architecture for a Future C-Band/L-Band GNSS Mission Part 2: Signal Considerations and Related User Terminal Aspects

working papers Architecture for a Future C-band/L-band GNSS Mission Part 2: Signal Considerations and Related User Terminal Aspects

Jose-Angel Avila-Rodriguez This column continues an exploration of possible use of the C-band radio frequency for GNSS Jong-Hoon Won navigation. Part 2 focuses on C-band signal design in the context of non-interference with Stefan Wallner other services in nearby RF bands, as well as user equipment design and performance. Marco Anghileri Bernd Eissfeller he radio navigation satellite higher free space losses due to the limi- Berthold Lankl service (RNSS) portion of the tations on the higher signal frequency. Torben Schüler radio frequency (RF) spectrum An omnidirectional C-band antenna at University FAF Munich T is already overcrowded, and the 5 GHz will be 3.2 times smaller in the Oliver Balbach bands suitable for new uses are very linear dimension than an equivalent L1- IfEN GmbH limited. This is especially true for the band antenna. (The latter signal has a Andreas Schmitz-Peiffer E1/L1 band occupied today by GPS and 19-centimeter wavelength at 1.575 GHz Jean-Jacques Floch Galileo. compared to the wavelength of 6 centi- Lars Stopfkuchen In addition, Japan’s quasi-zenith sat- meters at 5.015 GHz.) Dirk Felbach ellite system (QZSS) and potentially also Because of this wavelength-driven EADS Astrium Compass and GLONASS will be trans- design factor, the area of the C-band mitting navigation signals in this fre- antenna will be 10 times smaller than Antonio Fernandez quency band. But E1/L1 is not the only that of a standard L-band antenna. As a Deimos Space case. Even those RF bands that are not result, a C-band antenna receives only Rolf Jorgensen being used yet will certainly be shared 1/10th the broadcast power of its L-band TICRA by many systems in the near future. counterpart. (For details of relevant Enrico Colzi Thus, the search for unused frequen- research, see the articles by M. Irsigler et ESA-ESTEC/Vega in Space cy resources will almost certainly con- alia and A. Schmitz-Peiffer et alia (2008) tinue during the next years. The World in the Additional Resources section near Radio Communications Conference the end of this article.) 2000 (WRC-2000) allocated the por- Another important factor is the Authors Note: It is highly tion of C-band between 5010 and 5030 increased signal attenuation of C-band remarked that this column MHz for RNSS space-to-Earth applica- signals due to foliage, heavy rain, or is based upon a C-band tions. The general idea was to provide indoors, as well as other negative envi- GNSS study being conducted access to a frequency band that is not ronmental effects on signal tracking. On within the European Space yet overloaded by other signal sources the other hand, C-band exhibits much Agency (ESA) GNSS Evolution and, consequently, not so susceptible to smaller ionospheric errors for standard Program. Please note that interfering signals as guided by Inter- single-frequency applications. The hope the views expressed in the national Telecommunications Union is that technological progress might bal- following reflect solely the (ITU) regulations. ance some of the disadvantages from a opinions of the authors and Navigation in C-band presents both long-term point of view, given that an do not represent those of ESA. advantages and disadvantages, the actual application of C-band for RNSS is most important drawback being the not foreseen before the year 2020.

52 InsideGNSS july/august 2009 www.insidegnss.com working papers

We began our discussion in the pre- SPR-C GMSK PRS-C GMSK BPSK(10) Data vious column (May/June 2009, Inside BOC(5,5) Data GNSS) with an explanation of the scope of the C-band project, service analysis, sat- f(MHz) ellite constellations, ground segment, sat- ellite transmit signal power requirement, SPR-C GMSK payload design, spacecraft accommoda- BPSK(10) Pilot PRS-C GMSK tion, and end-to-end performance. BOC(5,5) Pilot In this column we talk about the C- 4990 5000 5010 5030 band signal design driven to respect the given constraints of other C-band servic- FIGURE 1 C-band GMSK Signal Plan es, and the C-band user terminal equip- ment design and performance analysis in tures by providing additional robustness selected signal plan for the C-band. the context of expected applications. in degraded RF situations. Moreover, the Note that both the SPR-C and PRS- Additional discussion of the naviga- proliferation of GNSSs and lack of high C services provide a data and a pilot tion message structure design and the precision signals that work on a single channel. related added value concerning the tro- frequency have also been important posphere corrections (e.g., the combina- drivers in the C-band study. Compatibility of C-Band tion of navigation data and numerical In order to design C-band signals Signals weather data from meteorological satel- the top-level requirements for both ser- Compatibility is the fundamental aspect lites), together with critical user-terminal vices were analyzed and established in in the design of any navigational sig- technologies needed to prepare C-band terms of geometric dilution of precision nal. Indeed, this criterion was assigned for use in a future GNSS constellation, (GDOP), availability, and continuity risk higher priority than other character- have been added to this digital and on- among other factors, and so on. In addi- istics such as navigation performance. line version of the article. tion to this, the SPR-C requires authen- As briefly mentioned earlier, the signal tication capability to provide robustness plan in the C-band must be compatible C-Band Signals Considered in terms of anti-spoofing while the PRS- with: Based on a thorough trade-off analysis, C needs code-encryption capability to • radio-astronomy (RA) band between the Service with Precision and Robust- provide enhanced anti-spoofing perfor- 4990 and 5000 MHz ness (SPR-C) and the Public Regulated mance. Both service signals should be • microwave landing system (MLS) Service for C-band (PRS-C) have been spectrally decoupled from each other. between 5030 and 5150 MHz identified for a future Galileo signal plan The C-band signal plan was opti- • Galileo uplink receiver (ULR) in C-band. mized for maximum occupied band- between 5000 and 5010 MHz A quick look at this service definition width and spectral separation between We will first describe our assump- reveals the main motivation for both the two provided services. In conse- tions in calculating the potential for services: 1) the SPR-C was to maximize quence, the signals presented next must C-band interference and describe the the possible user communities under be interpreted as an envelope of solu- GMSK signal in greater detail before C-band, following the civil/public dual- tions in the sense that derived alterna- reporting the results of our compatibil- use concept of satellite navigation; 2) tive signals with lower chip-rate and ity analysis. the PRS-C was to provide selected users lower sub-carrier frequencies would Radio-Astronomy. RA compatibility with the access to this service in order to also fulfill the criteria for compatibility is assured according to International fulfill high security requirements (e.g., with nearby C-band services. These are Telecommunication Union (ITU) regu- anti-jamming and anti-spoofing). namely the radio-astronomy service lations if the power flux density (PFD) As discussed in the first part of this (RA), the microwave landing system of the C-band downlink signals is not series, the PRS-C consists of two small (MLS) service (MLS), and the Galileo higher than a threshold value that is a spot beams with approximately 1,500 up-link (UL) service. function of the number of simultaneous kilometers of radius. Moreover, these Figure 1 shows the spectrum of the satellites within the very narrow beam of two spot beams shall provide high selected signal plan for C-band RNSS an RA telescope. geographic flexibility to point at any signals relying on the Gaussian mini- In our analysis we assumed that a required area on earth. mum shift keying (GMSK) modulation. maximum number of 10 C-band satel- In addition, use of C-band shall aim This scheme was found to satisfactorily lites could be seen at any time by any at mitigating problem areas of current L- accomplish the stringent requirements on RA antenna on the ground and that all band signals. In fact, the C-band Service spectrum confinement to ensure compat- the signals coming from these satellites Plan was designed to address the vulner- ibility with adjoining C-band services. have the same power at the surface of the ability of L-band in critical infrastruc- Table 1 summarizes the parameters of the Earth. Given that the antenna beam of

www.insidegnss.com july/august 2009 InsideGNSS 53 working papers

Frequency Band C-band, 5010 to 5030 MHz downlink signals, including output Service SPR-C PRS-C multiplexer (OMUX) filtering, basically Channel Data Pilot Data Pilot depends on the SSC between the down- Signal type BPSK(10) BPSK(10) BOC(5,5) BOC(5,5) link signals and the uplink signals of the Galileo Uplink receiver as well as on the Modulation GMSK BTc = 0.3 GMSK BTc = 0.3 GMSK BTc = 0.3 GMSK BTc = 0.3 power of the downlink signals as seen by Symbol rate 50 sps N/A -- N/A the uplink receiver. Maximum code length 51,150 k*51,150 -- -- In addition to these considerations, TABLE 1. Parameters of considered C=band signals in order to compute the power of the downlink signals that leaks into the uplink receiver, we need to consider the antenna decoupling between them. Note c (t) (t) s(t) k ϕ that the interference scenario between f(t) g(t) VCO the downlink signals and the uplink receiver does not correspond to a far field case. From the antenna point of FIGURE 2 GMSK generation scheme view, both the uplink receiver antenna as well as the downlink transmission the RA receivers is very narrow, no more -124.5 dBW/m2 in any 150 kHz band. antenna are located very close to each than a few satellites are expected to be in It has turned out that the most strin- other and, thus, near-field approxima- sight at the same time in the worst case. gent constraint on out-of-band (OOB) tions have to be taken into account. Furthermore, in the PFD computation emissions actually comes from the RA Given this situation, the common we assumed atmospheric losses of 0.5 dB band, while the MLS band seems to be solution of using a combined trans- in signal power and included the whole less problematic. Accordingly, the com- mit/receiver antenna will not work and, 10-MHz RA bandwidth in measuring patibility with the services on the right instead, we must consider an antenna the combined C-band PFD at the sur- (upper) part of the C-band spectrum has coupling approach. In our simulations, face of the Earth. proven to be relatively easy to accom- a value of -110 dB was calculated based Moreover, we included both the plish while the left (lower) part raises on the antenna design, and distance spectrum of the SPR-C and the PRS-C serious concerns. and power flux density. Additionally, in our calculations, although the SPR-C Galileo Uplink Receiver. In order approximately 4.4 dB were assumed for has global coverage and the PRS-C spot to measure the maximum tolerable the antenna losses in terms of negative beams limit their operation to small received power that can come from the gain of the uplink receiver antenna. regional “footprints.” Again, as one can C-band downlink signals without affect- imagine, this is very much a worst-case ing the correct functioning of the uplink Gaussian MSK (GMSK) scenario because most of the time only the receiver in the satellite, the minimum GMSK is a special case of continuous

SPR-C will potentially affect compatibility C/N0 for data demodulation shall not be phase frequency-shift keying (CP-FSK) with the RA band, and the impact of PRS- lower than a certain threshold required that employs Gaussian filtered frequency C signals will be restricted to local areas. to achieve a specified performance in pulses to smooth the transitions from Furthermore, to ensure compatibil- terms of bit error rate (BER). one point to the next in the signal status ity with the RA band, we calculated the This criterion basically relies on the constellation while minimum shift key- aggregate power flux density (PFD) of computation of the spectral separation ing (MSK) is obtained directly from the the downlink C-band signals of all satel- coefficient (SSC) between the OOB rectangular shape of frequency pulses. lites under consideration within the RA emissions of the proposed downlink The CP-FSK signal can be modelled band such that the maximum PFD shall services and the C-band uplink. In fact, mathematically after modulating it onto not exceed the corresponding threshold the underlying idea is to measure the the RF carrier as follows: value. increase of the equivalent noise that the Microwave Landing System. In order uplink receiver will observe when the not to cause harmful interference to the downlink signals leak into the receiver where P is the power of the carrier, fc is

MLS operating above 5030 MHz, the as additional noise. the center frequency, ϕn(t) is the phase aggregate power flux-density produced To compute the quantitative avail- of the modulated carrier, and ϕo is the at the earth’s surface in the band 5030– able margin we first need to compute the constant phase offset. Figure 2 5150 MHz by all the space stations with- C/N0 of the uplink receiver in the presents a simplified model in any RNSS system (space-to-Earth) absence of interference from downlink to generate GMSK signals. operating in the band 5000–5030 MHz signals. The contribution of the equiva- In the case of GMSK, the phase does shall not exceed a threshold value of lent noise due to the interfering C-band not evolve linearly in the time domain.

54 InsideGNSS july/august 2009 www.insidegnss.com working papers

MSK and GMSK MSK and GMSK Frequency Pulses Absolute Phase Evolution 2 π/2 Q + (0, +1) 1.8 (0, + 11)) � (—1,0)(—1,0) 1(+ 1,1, 0) � (0,(0 +1) (0, + 11)) (—1,0)(—1,0) 4(0,4(0, + 1)1) (+1,0)( 1.6 � � 1.4 1.2 1 π/4 (-1, 0) I 0.8 (+1, 0) 0.6 0.4 0.2 0 (-1, 0)) � (0,(0, --11) 7(0,7(0, -1) � ((+1,+ 0) 0 (0, -1)) (-1,(-1, 00)) 2(+ 1,1, 0) (0,( -1) Tc T Tc Tc Tc T Tc Tc � (0, -1) � n c n c (n–1) 2 2 (n+1) 2 (n+2) 2 (n–1) 2 2 (n+1) 2 (n+2) 2 Time Time

FIGURE 3 Comparison between frequency pulses and phase pulses of MSK FIGURE 4 Phase evolution of GMSK and GMSK

The evolution of the frequency over time where that the transition from one point of the adopts the following expression: constellation to another one is not real- ized at a constant linear rate but instead follows a Gaussian distribution. That where represents the time dura- In the definition of σ above, the is, GMSK begins with a slow velocity at tion of complex source codes. We need product is defined as the -3 dB the starting phase constellation point, to distinguish the individual bits or bandwidth-symbol time (BT) product. speeds up, and then slows down again chips (to be transmitted in the I-phase The higher this value is, the cleaner will when approaching the final phase con- or respectively in the Q-phase) from be the eye diagram of the signal, but the stellation point. the complex symbols that the I- and Q- higher the OOB emissions will be. On By doing this, we can ensure that, chips constitute together. These chips the other hand, the lower the selected independent of the sampling point in result indeed from multiplexing the product is, the more power will the receiver, the probability of being signal. This means in other words that be concentrated close to the center of near the constellation point of interest

whenever we refer to Tc, this will actu- the band, which is actually the objective. (widest point in the eye diagram) will be ally represent the chip duration of an However, this comes at the cost of higher higher. This is clearly depicted inFigure individual PRN-code chip sent either inter-chip-interference (ICI). 4. The density of points indicates that the on the I-channel or on the Q-channel. A typical value in communication state of the signal adopts this value with Consequently the duration of a I- and applications is =0.3, which is a higher probability. Q-chip tuple corresponds to a period of a good compromise between spectral The C-band study considered differ- . efficiency and ISI. As an example, the ent values for the bandwidth-symbol time Note that the single frequency pulse mobile communication standard GSM (BT) factor, with the solution = p(t) is no longer rectangular but can be is based on GMSK with =0.3. It 0.3 actually being the most interesting expressed as the convolution of a rectan- must be noted that inter-symbol-inter- due to its good compromise between gular pulse p(t) with the Gaussian filter ference (ISI) is basically the same if ICI spectral confinement and the ISI in the impulse response g(t): applies to chips and ISI to symbols. time domain. Figure 5 shows the com- The frequency pulses as well as parison in terms of power spectral den- the derived phase pulses for MSK and sities (PSDs) between two GMSK signal GMSK with different factors are plans. In the figure, we denote them as with shown in Figure 3 for comparison. GMSK1 (with =0.3) and GMSK2 As we can see from this fig- (with = 0.25), respectively. ure, the frequency pulse of the As we can recognize, the difference GMSK with lasts over between both options is minimal from The Gaussian filter g(t) adopts the approximately three chips, resulting in the point of view of their spectrum. following form in the time domain controlled but non-desired ICI. However, in the time domain GMSK1 The phase diagram of GMSK (with = 0.3) is shown to be is similar to that of the MSK more favorable as outlined in Figure 3 modulation with the difference and Figure 5.

www.insidegnss.com july/august 2009 InsideGNSS 55 working papers

-60 GMSK1 SPR-C (I+Q) -700 MSK 1 the very stringent requirements for GMSKGMSK BT = 0.0.22 GMSK PRS-C (I+Q) GMSKGMSK BT = 0.0.33 GMSK2 SPR-C (I+Q) compatibility with the nearby services. -800 GMSKGMSGSK BBTT = 0.0.55 GMSK2 PRS-C (I+Q) This is particularly difficult for the case -90 of the uplink receiver, which is spec- trally located directly on the left of the -1000 assigned downlink band. The simplest -1100 way to ensure compatibility would be to -1200 directly filter the signals after genera- tion, using a steep raised cosine filter, ower Spectral Density (dBW/Hz) P -1300 for example. Unfortunately, even though the sig- -51.1551 15 -40.6940 69 -30.6930 69 -20.4620 46 -10.2310 23 0 10.2310 23 20.4620 46 30.6930 69 40.6940 69 51.1551 15 nal might have been generated with a Offset with respect to the carrier (MHz) constant envelope, the desired constant envelope properties are lost after filtering. FIGURE 5 Comparison of GMSK signals for = 0.3 and with = 0.25 Furthermore, non-linear effects would appear during the high power amplifi- Compatibility of GMSK demodulation can be assured. cation (HPA) unless pre-distortion filters Signals • Compatibility with MLS: as we already or a linearized HPA are employed. Next we will summarize the most said, this is the least stringent con- The final effect is a spectral regrowth important results on compatibility of strain in the band. Indeed, the of those side-lobes we had attenuated the GMSK signals: GMSK signal plan is well below the previously, consequently losing all the • Compatibility with the radio astronomy: MLS PFD level of -124.5 dBW/m2 benefits of this intermediate filtering. the aggregate PFD of the composite It is important to note that, given This is important to keep in mind dur- SPR-C and PRS-C signal presents a the different directivity of the SPR-C ing the design because no matter how value with which the RA constraints, and PRS-C antennas, the contribution ideal our signal might appear regarding as outlined in the previous page, are of each service to the PFD on the ground its spectrum, if we cannot guarantee that met satisfactory. will strongly depend on the final equiv- the envelope of the signals will remain • Compatibility with the uplink receiver: alent isotropic radiated power (EIRP). constant after the power amplification as we have seen in previous sections, This was all taken into account in the (PA), all the efforts invested in reducing this is mainly driven by the SSC calculations. the side-lobes will be in vain (assuming between the uplink and downlink the need to have a constant envelope to signals of Galileo in C-band. Accord- Payload Constraints be a main driver). ingly, we show in Table 2 the values of As we have seen in previous sections, the Indeed, this was the main driver the SSC for each service separately. main constraint of C-band for naviga- when the constant envelope continu- With the GMSK C-band signal plan, tion is the very small amount of band- ous phase modulation (CECPM) was

the effective C/N0 required for data width that is available, together with selected as ideal candidate to meet all the

Multipath Envelopes - BW=20 MHz - Early-Late - Chip spacing=0.1 Multipath Errors Environment 1 BW=20 MHz - Early-Late - Chip spacing=0.1 5 2 GMSK1 SPR-C GMSK1 SPR-C RC1 SPR-C 1.5 RC1 SPR-C PSK SPR-C PSK SPR-C 1 MSK SPR-C MSK SPR-C 0.5 0 0 -0.5 Ranging Error (m) Ranging Error (m) -1 -1.5 -5 -2 0 10 20 30 40 50 0 1 2 3 4 5 Multipath Delay (m) Multipath Delay (m) x 104

FIGURE 6 Multipath Envelopes of SPR-C signals FIGURE 7 Multipath Error of SPR-C signals according to the LMS wide- band model

56 InsideGNSS july/august 2009 www.insidegnss.com working papers

SSC [dB-Hz] between requirements. The GMSK modulation is on C-band compatibility, any potential Signal Waveform a constant envelope by definition, con- signal plan needs to first prove its com- downlink and uplink siderably simplifying all these aspects of patibility with all the services around it. SPR-C GMSK BPSK(10) -112.9304 the payload as a result. As discussed in Only when the highest priority of com- PRS-C GMSK BOC(5,5) -111.9697 further detail in Part 1, for the SPR-C a patibility is assured, can we concentrate TABLE 2. SSC between downlink and uplink travelling wave tube amplifier (TWTA) on the proposed signals’ performance. was assumed, while for the PRS-C the For the simulations, a receiver band- In the case of the GMSK with direct selected amplifier is based on solid state width of 20 MHz was assumed. In addi- sequence spread spectrum (DSSS), ISI amplifier (SSPA) technology. tion, we used an early-minus-late dis- for data demodulation is not a problem criminator with a spacing of 0.1 chips. because of the use of long correlation GMSK Performance A quick look at the auto-correlation with a very short sampling time interval. We present in Figures 6–9 the multipath function (ACF) also reveals that, indeed, Therefore, the eye-diagram for ISI in this performance of the proposed GMSK sig- MSK has the sharpest slope around zero case is unimportant. nals. In addition, other solutions consid- compared to the other assessed signals, Consequently, we proposed to intro- ered in the definition of the C-band signal which explains the superior perfor- duce a new terminology of “ICI” (for and service plan are also presented. We mance of MSK in the previous figures. inter-chip or code-interference). The present first the results based on a single However, MSK could not demonstrate name was already anticipated in previ- static multipath reflection with a signal to that it met the OOB emission require- ous discussion. Note that the proposed multipath ratio (SMR) of -6.5 dB. ments; thus, the best signal waveform signal is the GMSK DSSS, which is dif- Although SMR values of -3 dB are among the compatible ones is GMSK. ferent from the original GMSK with well established for L-band, the selection This coincides with the results shown direct sequence frequency hopping of an SMR value of -6.5 dB for C-band in Figure 10 and Figure 11. (DSFH) that is widely used in GSM. seems reasonable due to the higher ratio To analyze the performance of the Figure 12 illustrates the eye-diagram of diffuse reflections that can be expect- various signals in terms of their ICI effect summary of SPR-C signal for different ed for C-band compared to L-band. on timing recovery and navigation data cases of GMSK (BT=0.3 and 0.25) along Later, the statistical wideband channel bit demodulation, the eye-diagram (and/ three-chips intervals. Note that the eye- model for land mobile satellite systems or phase diagram) should be employed. diagram of PSK should have a rectangu- (LMS) is employed. We highly recommend that here the ICI lar shape without any interference during As we can recognize, the best signal or eye-diagram should be done at the the phase changing interval in absence of in terms of performance is MSK, fol- chip level in order to show the effect of noise, i.e., the eyes are fully opened. lowed by GMSK and SRC. MSK does not bandwidth-efficient methods (for exam- The GMSK BT=0.3 is slightly better fulfill the compatibility requirements ple, GMSK, SRC, and so forth) compared than the GMSK BT=0.25, because the in the band and, accordingly, GMSK to the phase shift keying (PSK). ICI of GMSK is inversely proportional is shown to be the best option also in In general the theory used in GMSK to the bandwidth-symbol time (BT); terms of performance. We need to keep was derived for GMSK modulation of the so, a smaller BT provides a better per- in mind that, due to the tight constrains data sequence, not the code sequence. formance in the sense of ICI, but we

Multipath Envelopes - BW=20 MHz - Early-Late - Chip spacing=0.1 Multipath Errors Environment 1 BW=20 MHz - Early-Late - Chip spacing=0.1 5 2 GMSK1 SPR-C GMSK1 SPR-C RC1 SPR-C 1.5 RC1 SPR-C PSK SPR-C PSK SPR-C 1 MSK SPR-C MSK SPR-C 0.5 0 0 -0.5 Ranging Error (m) Ranging Error (m) -1 -1.5 -5 -2 0 20 40 60 80 100 0 1 2 3 4 5 Multipath Delay (m) Multipath Delay (m) x 104

FIGURE 8 Multipath Envelopes of PRS-C signals FIGURE 9 Multipath Error of PRS-C signals according to the LMS wide- band model

www.insidegnss.com july/august 2009 InsideGNSS 57 working papers

Autocorrelation Functions Autocorrelation Functions GMSK1 SPR-C GMSK1 SPR-C 1 1 RC SPR-C 1 RC1 SPR-C PSK SPR-C PSK SPR-C MSK SPR-C 0.8 0.8 MSK SPR-C

0.6 0.6

0.4 0.4

Normalized Amplitude 0.2 Normalized Amplitude 0.2

0 0 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 -22 -1.51 5 -11 -0.50 5 0 0.50 5 1 1.51 5 2 Chip Offset (chip) Chip Offset (chip)

FIGURE 10 ACF of SPR-C signals FIGURE 11 ACF of PRS-C signals have to pay a price in spectral ineffi- integrity information for safety-ori- an omni-directional antenna for L- ciency. Nonetheless, the GMSK BT=0.3 ented applications such as aviation band. The array antenna system has still complies with the ITU regulations and maritime operations. an amplitude and phase adjustment described earlier. • SPR-C receivers would have various functional block in which inputs multi-frequency components (e.g., are output from an ADC bank in Overall User Terminal L-band) to provide precision posi- an RF chain. In order to control Architectures tioning by efficient estimating of this amplitude and phase adjust- The C-band signal was designed to make ionosphere errors, that is, dual-band ment block, a navigation processor use of data and pilot channels. Using a user terminals (UTs). generates a command to control the pilot channel will provide a longer coher- • SPR-C UTs could have two different digital beam-forming based on the ent integration, thereby producing less architectures depending on antenna receiver’s heading and pitch angle noisy range information. designs — single antenna and array information. For an extended discussion of C- antenna systems. An SPR-C single- The main elements identified for the band user terminal design concepts antenna system could use two omni- PRS-C UTs’ core structure include: and performance analysis undertaken directional antennas for C- and • A decryption code generator for as part of the European GNSS Evolu- L-band. (See Figure 13). The SPR-C security access. tion program, see the three articles by array antenna system could incor- • Also, PRS-C UTs could have two J. H. Won et alia listed in the Additional porate an array antenna with digital different architectures depending Resources section. beam forming control technology on antenna systems: single antenna The main identified core structure for C-band for more high-accuracy system and array antenna system: of SPR-C user terminals or receivers and robust precision services and 1) For a single antenna system, an includes the following charac- Eye-Diagram SPR-GMSK (BT=0.3) Eye-Diagram SPR-GMSK (BT=0.25) teristics: • A multi-bit ADC. This could 1 1 be used for higher accuracy - 0 - 0 and is essentially required to achieve a more robust -1 -1 signal processing result 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 when employing GMSK in a receiver. 1 1 • A data authentication mod- σ 0 σ 0 ule should be used for com- mercial services that are, of -1 -1 course, a core of the SPR-C 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 service. This data authenti- Time (chip) Time (chip) cation module provides the FIGURE 12 Eye diagram summary of SPR-C signal SPR-C user group with high-

58 InsideGNSS july/august 2009 www.insidegnss.com working papers

Analog C-Band Digital Signal Navigation/User RF Part Processing Part Application Part Analog C-Band Digital Signal Navigation/User Omni-directional RF Part Processing Part Application Part Antenna (C-Band) Omni-directional Antenna (C-Band) Decryption RF Code Generator RF ADC Front-end ADC Front-end Data Demodulation Correlator Navigation Correlator Processor (Dual Freq Proc) RF Data Front-end ADC RF ADC Authentication Front-end Navigation Omni-directional 2N Channels Omni-directional 2N Channels Processor Antenna (L-Band) Antenna (L-Band) (Dual Freq Proc) UART Analog L-Band Analog L-Band RF Part UART RF Part

FIGURE 13 Schematic view on SPR-C UT architecture (single antenna FIGURE 14 Schematic view on PRS-C UT architecture with single antenna system) system

omni-directional antenna should be also available to the radians/second; ϕ is the carrier phase in radians, τ is the code

C-band RF part for cost effective governmental user group, delay in seconds, ωd are the angular Doppler frequency in rad/ such as military handheld users (Figure 14); and 2) For an sec, d is the data bit (±1), respectively; is the shaped pulse array antenna system, one array antenna with digital beam of SPR-C PRN code in the I-channel multiplied by a subchip forming control technology for C-band should be avail- that already accounts for the data bit (i.e., a data channel),

able to C-band RF part for a level of additional anti-jam- CQ,SPR is the shaped pulse of SPR-C PRN code in the Q-channel ming margin to user terminals that require a higher anti- (i.e., pilot channel), is the shaped pulse of PRS-C PRN jamming capability. The array antenna system of PRS-C code in the I-channel multiplied by a subchip considering data

has the same control logic with the SPR-C array antenna bit (i.e, data channel ), CQ,SPR is the shaped pulse of PRS-C PRN system. code in the Q-channel multiplied by a subchip (i.e, pilot), and

• Two antenna-driven architectures: a cost-effective, single- Tc represents the code chip duration of each service signal in antenna C-band with omnidirectional design for govern- seconds. mental users, such as those needing military handhelds We should emphasize that the use of , that is a shaped- (Figure 16), and an array-antenna design with C-band pulse code multiplied by a navigation data bit, is mainly digital beam-forming control technology for an additional for GMSK signals due to its continuous phase modulation level of anti-jamming capability and using the same control property.

logic as the SPR-C array antenna system. Note that the additional half-chip (Tc/2) code delay in the Q-channel comes from the “offset” in offset QPSK (OQPSK) Signal-In-Space (SIS) Model to restrict an instant phase change within ±90 degrees, thereby All GNSS signals-in-space (SIS) in the C-band (i.e., SPR-C and reducing the spectral leakage of the intended signals as much as

PRS-C) can be modeled as: possible. If we omit the terms Tc/2 from the preceding equation — that is, with no delay between I and Q — the signal model becomes a generic balanced QPSK. Let us turn our attention now to the models of the received signals. These models will then be used in the receiver design. with Moreover, it must be noted that for the GMSK data channel, the received signal model does not need to have because the correlation between the data-modulated shaped pulse code ( ) and the original shaped pulse code (C) that should be locally generated in a receiver would be done in a noncoherent way; so, the resulting correlation function has the same shape, and the polarity represents navigation data bit. • SPR-GMSK-QPSK(10)

where sSPR(t) and sPRS(t) are the SPR-C and PRS-C signal, respec- tively; I(t) and Q(t) are the output signals coming from the I and Q branches, respectively; P is the signal power (assuming:

PI,SPR = PQ,SPR and PI,PRS = PQ,PRS ); ω0 is the angular frequency in

www.insidegnss.com july/august 2009 InsideGNSS 59 working papers

FIGURE 15 Block diagram of signal acquisition system for shaped-pulse OQPSK

• PRS-GMSK-QBOC(5,5) The table incorporates pessimistic, moderate, and optimis- tic assumptions on the increased processing capabilities with respect to the current available processing power, considering by factors of 10, 100, and 1000, respectively. Here we assumed that Moore’s rule — a doubling of processing power every 18 months. The evolution of the processing power compared to Signal Acquisition the currently available processing power is assumed to be “140” Figure 15 illustrates the proposed acquisition system block dia- because the maximum processing performance measured in gram for OQPSK DSSS. Note in the figure that the delays in the teraflops/second is expected to rise from around 0.5 to 73 tera- integrators are set to Tc/2 seconds, and the upper code branch flops/second on a single chip by the year 2020+. is delayed by Tc/2 seconds. We used a “Hot Start Time” requirement in order not to be The acquisition system consists of two BPSK acquisition affected by the design of navigation data structure. detectors that produce the sum of I2+Q2 in a signal branch for The cells with a ratio larger than “1” means that the expected the first SS code. This is added to the sum of I2+Q2 in the other signal processing power at 2020+ may not be sufficient to meet signal branch for the second SS code in order to drive the non- the acquisition time requirement. For example, with only a 100- coherent integration. fold increase in processing power from current levels (e.g., 25 The block diagram contains a selection logic that combines multi-correlators and four times the sequential processing per the I and Q branches or selects one of them (representing com- channel), the processing-power figure for the SPR-C signal is on bined data/pilot or pilot-only processing, respectively). This is the order of 10-1 and fulfills the requirement, while the PRS-C because the I and Q codes of QPSK are synchronized to each needs still more signal processing capabilities. other, and we need to search a given code-search range for I or Nevertheless, we need a processing power redundancy of a Q that is the same as in the case of BPSK signals. parallel operation of correlators for fast time to first fix (TTFF) The only price that we have to pay in this case is a doubling and as high sensitivity as possible. Even with today’s advanced of hardware resources (i.e., correlators), but we will have an technology, the future of the signal acquisition in C-band is in important three-decibel gain in terms of C/N0. some sense bright. Table 3 shows the required processing power ratios that In order to analyze the effect of GMSK on signal acquisition, could be supported by the processing power anticipated to be we subtract the acquisition function of the unfiltered quadra- available by 2020 and later. These power ratios are needed to ture phase skip keying (QPSK) and quadrature binary offset fulfill a cold-start mean acquisition time (MAT) requirement carrier (QBOC) cases from the intended modulation cases. Fig- ure 16 with a sufficient C/N0 (e.g., 40 dB-Hz) and a small false alarm– illustrates the difference between normalized autocor- penalty time coefficient. relations of GMSK and the unfiltered rectangular code case.

60 InsideGNSS july/august 2009 www.insidegnss.com working papers

Processing Power Nproc Acquisition Ratio Due to the high dynamic stress require- System Signal Time (Require- compared ment for the SPR-C, the 18-Hz noise band- 10x140 100x140 1000x140 ment) to L2-band width usually used for L1-band is shown to Low-end civil < 45 s GPS L1CA 4x10-4 4x10-5 4x10-6 - be not sufficient to accommodate dynamic users (cold start) ranges. The use of a narrow noise band- GPS L2C 1.017x10-1 1.017x10-2 1.017x10-3 1 width in conjunction with a temperature- GPS L2P(Y) 2.034x100 2.034x10-1 2.034x10-2 20 controlled crystal oscillator (TCXO) does High-end civil < 30 s not offer a solution because of the lack of users 0 -1 -2 (hot start) Galileo E6-B,C 1.070x10 1.070x10 1.070x10 10 margin for the signal tracking loop design. SPR-C QPSK(10) 8.420x100 8.420x10-1 8.420x10-2 80 Increasing the noise bandwidth (e.g., 40 Hz) to give more design margin to the C- High-end GPS L2M 4.557x101 4.557x100 4.557x10-1 1 military (or < 10 s (hot band would provide us with better accu- Galileo E6-A 4.661x101 4.661x100 4.661x10-1 1.02 governmental) start) racy, but we would then lose some amount 1 0 -1 users PRS-C BOC(5,5) 9.322x10 9.322x10 9.322x10 2.06 of available C/N0 in dB-Hz. Accordingly, the best way to improve TABLE 3. Required processing power ratio at 2020+ for acquisition of GNSS signals performance of both C/N and accuracy is 0 to use a rubidium oscillator as well as to In those areas of the figure where the correlation value of increase the transmit signal power. The use of a rubidium oscil-

GMSK is larger than that of the unfiltered rectangular code, lator will retain the C/N0 ratio while providing a 4.0-degree acquisition is not a matter of concern; but the opposite case, i.e., tracking loop accuracy, which translates into a ranging accu- when the correlation value of GMSK is smaller than that of unfil- racy of about 0.7 millimeter. tered rectangular code, could pose a problem if the correlation For PRS-C, we assumed a vibration specification of 10 g/s value is smaller than the predefined acquisition threshold. to test the vibration-induced phase noise jitter. The C-band The latter situation forces us to reduce the code bin size to cover this smaller region, or to carefully choose the predefined SPR-C threshold value, which is controlled by setting the detection 0.2

and false-alarm probabilities. However, the analysis here was alue based on the unfiltered QPSK or QBOC cases. Therefore, if we 0.15 consider a more realistic case, e.g., a band-limited QPSK or QBOC, the result might be more similar to that of GMSK. 0.1 Note that, for the proposed bandwidth-efficient modula- tion schemes, the only thing that needs to be modified in the 0.05 signal-processing functional blocks of today’s GPS/Galileo receivers (which are based on the PSK modulation scheme), is 0 the correlator functional block. That is, if we replace the binary

code PRN (or BOC) generator of a GPS/Galileo receiver with Difference Normalized Correlation V -0.05 a code generator that creates the shaped-pulse codes proposed -3 -2 -1 0 1 2 3 for SPR-C and PRS-C, we can easily implement a GMSK-based Code Delay Error (chips) navigation receiver. PRS-C 0.35

Signal Tracking alue 0.3 As with the signal acquisition scheme, a code- and carrier- 0.25 tracking system for OQPSK DSSS signals was proposed con- 0.2 sisting of two BPSK tracking blocks. The main differences from the current BPSK-type Galileo 0.15 receiver’s signal tracking are 1) the shaped pulse code generator, 0.1 2) QPSK-type receiver, and 3) a block with an algorithm that 0.05 combines the carrier and code to deal with a QPSK signal. 0 Figure 17 shows the whole non-coherent phase-lock-loop -0.05

(PLL) noise jitters of SPR-C and PRS-C signals for a high-end Difference Normalized Correlation V -0.1 user terminal (e.g., a vibration isolation–equipped user termi- -3 -2 -1 0 1 2 3

nal). C/N0 threshold to maintain PLL lock (i.e., the cross point Code Delay Error (chips) of noise jitter line with the yellow threshold line) and tracking loop accuracy at high enough C/N for noncoherent PLL can FIGURE 16 Difference of normalized autocorrelation functions: GMSK or 0 SRC minus PSK be read from the figure.

www.insidegnss.com july/august 2009 InsideGNSS 61 working papers

Non-Coherent,, VibForHighUT g DLL Thermal Noise Jitter 30

25 otal Noise 20 0.1

15

10 0.05

5 DLL Thermal Jitter (chips) PLL Jitter (1-sigma) [deg] T 0 20 25 30 35 40 45 50 0 10 15 20 25 30 35 40 45 50 C/N0 [dB-Hz] Non-Coherent, VibForHighUT CN0 [dB-Hz] 300 FIGURE 18 DLL thermal noise jitter for GMSK together with other solutions 255 that were also investigated (0.5 delay spacing, BDLL=40 Hz) otal Noise 200 In order to analyze the effect of the GMSK design on UT tracking loops, we tested the delay-lock-loop (DLL) thermal 155 noise jitter performance. Figure 18 depicts the DLL jitter for sev- eral proposed signals. A 20-MHz bandwidth of receiver filter- 100 ing was assumed for the case of filtered PSK, but not for other 5 the other signals, because the bandwidth-efficient modulation scheme itself already includes the filtering effect. PLL Jitter (1-sigma) [deg] T 0 Obviously, if we use band-limited signals, we lose a level of 20 25 30 35 40 45 50 accuracy compared to the unfiltered ideal PSK. The accuracy C/N [dB-Hz] 0 of signal tracking (or, equivalently, ranging accuracy) depends on the sharpness of the correlation functions as seen in the UT. FIGURE 17 Non-coherent PLL total noise jitter for SPR-C (upper) and PRS-C (lower) in the case of vibration isolation–equipped user terminals As we can observe, GMSK performs the best in terms of this criterion, followed by filtered PSK and then SRC. noise jitter, which is four times higher than that of L-band, Undoubtedly, an ideal unfiltered PSK approach would pro- would bring the C-band noise jitter line near to the threshold vide the best result but is not achievable in a real-world environ- value (probability-of-false-alarm limit) even with a wide noise ment due to the use of band pass filters at the end of a GNSS bandwidth and a high quality oscillator. satellite’s RF chain and at the front-end of any type of typical Since this would finally make the use of PRS-C impractical receivers. Therefore, we conclude from the results shown in in a real-world environment, mounting the oscillator of a PRS- Figure 21 that GMSK provides a better accuracy than other C receiver using well-designed vibration isolators is strongly methods. Moreover, we must also note that, at a sufficiently recommended. A narrow noise bandwidth approach is not high C/N0, all methods provide a similar level of accuracy. applicable to PRS-C because of its high dynamic requirement. Moreover, using a rubidium oscillator we can obtain a compa- Boundary Condition Table 4 rable performance as an L-band system in terms of C/N0 and shows an SPR-C receiver’s signal power budget under the tracking loop design margin. typical operating conditions. In the table, the worst and the A better oscillator also allows better C-band performance in best cases depend on satellite elevation angles. Key design terms of ranging accuracy, but not vibration-induced oscillator parameters such as noise floor, antenna gain, correlation loss, phase noise jitter. Without a well-designed vibration-isolation noise figure, implementation loss and so on, were assumed solution in the receiver, C-band UTs cannot accommodate a in a reasonable way. The required C/N0 at correlator (i.e., the vibration environment even with a wide noise bandwidth. There- C/N0 threshold needed to maintain PLL lock) was obtained fore, vibration-isolation equipment is an essential design factor. from a signal-tracking stability analysis by reading the cross- Before leaving this part of the discussion, we should point ing point of a PLL noise jitter line with the predefined thresh- out that, given equivalent phase-tracking accuracy as an L1 old value.

(or L2) receiver, C-band ranging accuracy would be 3.2 times The typical receiver operation C/N0 region, where the noise superior (or in the case of L2, four times higher) than that of jitter line is aligned almost horizontally and produces a reason- L-band because the wave length of C-band is one-third that able accuracy of tracking results, was obtained at unshadowed of L-band. environmental conditions where the satellite-receiver link loss

62 InsideGNSS july/august 2009 www.insidegnss.com working papers

Satellite at low elevation Satellite at moderate Satellite at zenith ellite-specific information will be kept (10 deg) – worst case elevation (40 deg) (90 deg) – best case separate from the rest of the data. Received signal power available to an In addition to this, there is all com- isotropic antenna (C dBW) -158.3 -156.3 -158.3 mon information for providing a PNT service, such as GNSS system time con- Typical patch antenna gain (G dBic) R -5 +0 +5 version parameters, coordinated univer- relative to isotropic antenna sal time (UTC) conversion parameters, Receiver signal power available to typi- -163.3 -156.3 -153.3 ionospheric correction parameters, sat- cal patch antenna (C dBW) ellite health information, and almanac. Noise floor (No dBW) -203 (nominal operation) The C-band navigation message

LNA noise figure (Nf dB) -2 (already included in No evaluation) should contain additional data accord-

Correlation loss (Lcorr dB) -1 ing to the requirements of the PRS-C Implementation loss (L dB) -1 and SPR-C services. Compared to the most traditional content, the following Precorrelation CNo (CNo dB-Hz) at cor- 37.7 44.7 47.7 relator for nominal operation data will be included: fast clock dif- ferential corrections, digital signature Typical receiver operation region 37.7 ~ 47.7 dB-Hz (10 dB variation) (dB-Hz) for navigation message authentication (NMA), new encryption black key, and Design margin (dB) 14.0±5 dB zenith hydrostatic and wet delay data for Threshold CN (dB-Hz) for maintaining 0 28.7 (TCXO@40Hz) tropospheric corrections. PLL lock We should note that, in accordance TABLE 4 Receiver signal power budget at typical receiver operation region (SPR-C data) with former decisions, no additional integrity information will be transmit- only encompasses the free space loss, The main distinctions between these ted for both C-band services. atmospheric loss, and receiver antenna message types are due to the different Message Structure. The overall mes- polarization loss. Excluded from this transmission rate, resulting primarily sage structure, common to the two calculation were other losses stem- from the ranging code design and from message types, consists of a continuous ming from user environment–depen- the different requirements of the two stream of frames such as the one illus- dent power degradation factors, such services in terms of the size of the area trated in Figure 19. as atmospheric attenuation in tropical served and of the fields where they will Each frame divides into three sub- region, tropospheric scintillation, foliage find application. frames, which are sent one after each attenuation, interference and so on. As we will show in the following other and present remarkable differences From the link budget analysis, we discussion, the structure of the C-band assumed the received signal power avail- message contains a certain level of flex- able to an isotropic antenna was -158.3 ibility. Moreover, should the need arise FIGURE 19 Structure of the C-band message to -156.3 dBW, depending on elevation to further differentiate the PRS-C/NAV frame angles. Assuming the range of a typical from the SPR-C/NAV, a variable content patch antenna gain relative to an isotro- subframe still leaves very large margins concerning their function and content. pic antenna, we calculated the receiver for further changes, especially in the As we will detail shortly, the error pro- signal power available on a typical patch arrangement of additional data or in tection techniques employed for each antenna to be -163.3 to -156.3 dBW. The the definition of specific, new message message type will also be different. But

threshold C/N0 to maintain PLL lock in types. first, let’s take a closer look at the three the case of data-only tracking channel Information to Be Transmitted. Iden- subframes. is 28.7 dB-Hz from the extensive stabil- tifying the minimum required amount FCN Subframe. The first part of each ity analysis on PLL. This fact means the of data to be transmitted was crucial for frame is called the frame counter (FCN);

C/N0 design margins are 14±5 dB for the keeping the data transmission rate as it contains a progressive index, counting SPR-C. low as possible, while still meeting the for the number of frames that have been TTFF requirements. transmitted since the last update of sub- C-band Navigation Message The send-time information will be frame 2, which essentially contains clock Considering that two services (SPR-C transmitted in the form of progressive and ephemeris data (CED). and PRS-C) should be provided in the counters related to special time intervals The FCN field is generated and trans- C-band, two corresponding variations described a little later, while data to com- mitted simultaneously from all the satel- of the C-band navigation message were pute the satellite position and clock error lites, and since the value of the counter proposed: PRS-C/NAV and SPR-C/ consists of the traditional ephemeris and will be the same, symbol combining over NAV. polynomial coefficients. This kind of -sat multiple paths will be possible. In this

www.insidegnss.com july/august 2009 InsideGNSS 63 working papers

additional pilot channel could of 25 grid points). For the PRS-C ser- be available for precise signal vice, sending two of these granules (each tracking. extends for 12 x 12 degrees) will be more VC Subframe. The last than sufficient for covering the served part of each frame is reserved area (1,500 kilometers in diameter). for variable content (VC). This Different PRS-C/NAV messages will be means that each frame could transmitted on two separated signals, contain different data in this one for each individual area. part of the navigation message. In case global coverage is required FIGURE 20 Total frame length of the C-band navigation The content of this subframe by the SPR-C, given a grid-resolution message can be sent in an arbitrary of 3 x 3 degrees, this information must sequence because the user can be transmitted for 7,200 grid points, way the demodulation threshold for the identify the following content looking at grouped into 288 granules. contained bits is decreased and the over- the message type identifier at the begin- However, because no maritime users all system gains in robustness. ning of each VC subframe. are expected to require such high preci- The information contained in this Figure 20 summarizes the amount sion tropospheric corrections, the num- part of the navigation message plays of data forming the individual frames ber of grid points can be significantly a very important role in determining of the C-band navigation messages. reduced (to around 100 granules) if the the pseudoranges to the satellites: the Value-Added Data. We turn now points over the oceans are excluded. receiver clock can be set just by reading to a more extensive discussion about Furthermore, because the Arctic and this portion of the navigation message. numerical weather and the encryption/ Antarctic regions are very cold, the contri- Therefore, having such timing infor- authentication data. bution of zenith wet and hydrostatic delays mation available, especially in difficult Indeed, this data represents one of are expected to be very small. Therefore, environments such as urban canyons the most interesting points in the over- the total number of granules to be trans- or even inside buildings, may represent all message design because of its impor- mitted could be reduced further. a key feature for both SPR-C and PRS- tance in terms of the required service Detailed Content of the Subframes. C services. For this reason, as will be performance and because of the prob- Figures 21–25 show the detailed content shown later, this part of the navigation lematic aspect of the large amount of bits of each subframe and of three possible message will be particularly protected to be transmitted (especially concerning message types for the VC. As indicated against transmission errors. the wet and hydrostatic delay data for earlier, data transmitted within the VC CED Subframe. The clock and the SPR-C). subframe can be arranged in any order ephemeris data (CED) subframe con- Clock Differential Corrections.Fast according to specific requirements. Also, tains the fundamental data that allows clock differential corrections are sent via if additional content needs to be trans- the user to compute the satellite position the C-band messages, in order to meet mitted, numerous message types can be and time, providing precise ephemeris the stringent accuracy requirements of further defined and transmitted. data as well as clock correction coeffi- the two C-band services. The three defined message types con- cients. These corrections are transmitted as sist of a common type that is intended Other parameters present here are variations to the polynomial coefficients to be sent regularly, a second type con- the week number and the interval coun- that are used to compute the satellite taining the parameters to perform the ter in the current week, according to the clock error (δaf0 and δaf1), together with navigation message authentication, and system time related to the epoch when an accuracy index and a five-bit field a third type allowing for the manage- the current frame has been transmitted representing the PRN number of the ment of the encryption key, with a new by the satellite. satellite to which the clock corrections “black” key being provided to users. The CED subframe presents valid- refer. Error protection. We will now briefly ity duration of exactly two hours, after Zenith Wet and Hydrostatic Delay. explain the proposed techniques used to which new data is made available from For achieving high positioning per- protect the C-band message from trans- the ground segment and transmitted to formance, accurate tropospheric cor- mission errors. the users. With this guarantee, receiv- rections should be carried out by the Forward Error Correction. As intro- ers can combine symbols over multiple receivers. This kind of corrections can be duced previously, the three subframes frames in an effort to minimize the computed based on two key parameters, should be separately encoded, because transmission errors. namely the zenith wet and the hydro- of their different bit-error-rate (BER) Furthermore, the incoming data can static delay. performance requirement and because be wiped-off by multiplying it with the The proposed model for the trans- of the defined message structure. previously stored symbols belonging mission of this tropospheric data Given the importance of the time to the same validity interval; thus, an requires 638 bits for each granule (block parameter contained in the FCN sub-

63a InsideGNSS july/august 2009 www.insidegnss.com working papers

frame, a strong high-redundancy block by-column before being transmitted. code is to be used to encode the 9 bits The receiver will then perform the into 52 symbols. inverse process before starting decod- For subframe 2 and 3 the use of low- ing the symbols. The advantage is that density parity check (LDPC) codes is burst errors, which result from fad- proposed: recent studies, mostly in the ing and shadowing, would be spread FIGURE 21 FCN subframe and number of required bits framework of the design of the new GPS throughout a large portion of the mes- L1C message, strongly recommend the sage, facilitating the correction operated use of codes for protecting the data. by the decoder. Starting from the need for increas- Our design choice for the matrix is ing the correction performance and to have 48 rows and 51 columns. The

approaching the Shannon capacity limit following steps show the working prin- FIGURE 22 CED on the maximum amount of error-free ciple of the interleaver and how it applies subframe and digital data that can be transmitted, our to the C-band navigation message: 1) number of required bits study found that these codes outperform subframe 2 and 3 are put together for many other competitors in terms of cor- the transmission, 2) recting capability. Their performance is they are then writ- very close to that of turbo codes, but ten into the matrix unlike the latter codes, LDPC codes interleaver row-by- have no intellectual property constraints row, 3) the sequence on them, a factor that may be of great to be transmitted is importance for their future use. prepared reading the Table 5 presents the chosen cod- matrix content by ing schemes for the CED and VC sub- columns, and 4) the FIGURE 23 Type 1 message of the VC subframe and number of required bits frames, together with some specific interleaved sequence parameters. is transmitted. Especially concerning the use of Frame Duration and TTFF. Each the 4/5 code, we should note the moti- complete frame is received within a vation that led to this choice did not given time: 25 seconds for PRS-C/NAV come merely from the performance and 50 seconds for SPR-C/NAV. of this code. Obviously, lower coding Having available the satellite ephem- rates could perform better, especially eris and clock corrections, as well as the

in more realistic channel models. How- send time information (FCN) transmit- FIGURE 24 Type 2 message of the VC ever, when we had to make our design ted within each frame, the first pseudo- subframe and number of required choice, we had to consider at the same range measurements can be carried out bits time the requirement of having a data within the required TTFF (60 seconds transmission rate that should be as low for the SPR-C and 30 seconds for the as possible, the big amount of data to PRS-C). be transmitted (weather model and fast These considerations are based on clock corrections), and the need to guar- the assumption that a direct acquisi- antee a frame repetition time that could tion, without referring to coarse sat- meet the TTFF requirements. ellite position computations from the Therefore, a solution showing good almanac, is possible. This is true for BER performance while not introducing most environments, while in the case FIGURE 25 Type 3 message of too much redundancy in the informa- of weak signals almanac data not older the VC subframe and number of required bits tion to be transmitted would have been than one week should be available to the best choice. In any case, quite a good the users (stored in the devices) for margin remains for including further meeting the aforementioned TTFF Subframe Coding Scheme Used Coding Rate deterioration effects and still meeting requirements. CED LDPC (584,1168) 1/2 the requirement of a BER of 10-5. The following time would be required VC LDPC (1024,1280) 4/5 Block Interleaving. Another concept to retrieve both zenith wet and hydro- TABLE 5. Coding schemes for CED and VC subframes that turns out to be crucial and should static delay data and perform an accurate be employed is block interleaving. The tropospheric correction. First, for the encoded navigation symbols are fed into a PRS-C, two granules for the interested matrix row-by-row and then read column- area are transmitted within two frames

www.insidegnss.com july/august 2009 InsideGNSS 63b working papers

at 100 sps. A waiting time of around 50 seconds is expected. Second, for the SPR-C, all weather information on a global scale is grouped into about 100 granules. Any user on the Earth receives the corrections for an area of about 9,000 kilometers diameter with- in eight consequent frames. According to the chosen transmission rate, a latency of about 8x50=400 seconds is expected for reaching the so-called objective perfor- mance. (This navigation solution includes accurate tropospheric corrections.) Data Transmission Rates. The data for the two C-band services should be trans- mitted at the following rates: the PRS-C FIGURE 26 Grid of zenith wet delays (surface) as seen on July 28, 2008, at 12:00 UTC; values in units service at 100 symbols per second (sps) of millimeters. and the SPR-C service at 50 sps. These numbers are not limited by the total zenith path delays over several The ZHD is the largest part (often data demodulation itself. In fact, some years (carried out in an ESA study in 75–90 percent of the total tropospheric margin still remains for increasing the the framework of the Galileo testbed delay) and can be precisely determined data transmission rates and still achieve V1) revealed an RMS in the range of if pressure measurements are available the required BER performance. about 5 centimeters in zenith direction at the height of the user antenna. Conse- The main motivations for keeping under ordinary conditions (which map quently, the quantities the data rates to these values are due to to approximately 50 centimeters at an p(h): total pressure at a common ref- the proposed code lengths, which allow elevation of five degrees), but residu- erence height h (e.g., sea level) a better acquisition performance (longer als will tend to easily reach 1.5 to 2.5 qP: pressure scale height for vertical integration time and, therefore, lower decimeters under “unusual” conditions reduction at the user height

C/N0 requirements) as well as a better (which map to 1.5 to 2.5 metres at 5 will be needed. cross-correlation performance. degrees elevation). Conversion from pressure to zenith The idea for improving the perfor- hydrostatic delay is accomplished with Tropo Delay Corrections mance of these models is to retrieve the well-known Saastamoinen model in the Nav Message tropospheric delay corrections from which is internationally accepted and GNSS positioning requires tropospheric numerical weather fields, resample these of high accuracy. Vertical reduction is delays to be mitigated using a suitable data on a grid suitable for the broadcast carried out using an exponential func- correction model. Broadcast of these message format, and thereby supply a tion that relates the pressure at reference corrections is proposed for the C-band more precise compensation of this error height h to any height of a receiver hUSER services. component. Initial studies on the use with help from the pressure scale height. Often, so-called “blind” models are of such numerical weather model data Several precise mapping functions are used such as the RTCA MOPS. These indicate an accuracy in the range of 1.5 available to project the zenith delay into models are also fore-seen for Galileo. centimeters in the mid-latitudes. a slant direction. Such models are a kind of climatologi- Representing Tropospheric Corrections. The ZWD component is difficult to cal database, that is, they employ look- A two-dimensional (2D) representa- model using surface measurements. For up tables to figure out typical values tion of the needed delay quantities and this reason, vertical profile integration of describing the atmosphere at a certain vertical modelling functions is used to wet refractivity will be carried out using location. The hydrostatic and the wet reduce the amount of data to be trans- the 3D numerical weather fields first. The delay — the two components we nor- ferred to the user and to simplify the zenith wet delay will be directly given on mally have to distinguish — can then be computations. the correction grid: approximated without any knowledge of Numerical weather fields are 3D files, ZWD(h): zenith wet delay at a com- in situ measurements. and we need to integrate the refractivity mon reference height h Of course, such models are suitable profile in order to obtain those values q : ZWD scale height, exponential to describe the typical expected delay required to compensate for the tropo- ZWD trend for vertical reduction. under ordinary conditions but will fail spheric delays in GNSS positioning. In as soon as unusual conditions are pres- contrast, all needed quantities will be Similar to the treatment of the ent. Comparisons with GPS-derived given on a 2D grid here. hydrostatic component, vertical reduc-

63c InsideGNSS july/august 2009 www.insidegnss.com working papers

Typical range Number of Quantity (global grid) bits ZWD 256 mm 8

qZWD 0.5 - 5.0 km 9 p 60 hPa 6

qP 6 - 17 km 10 TABLE 6. Expected data ranges and number of bits for a grid point

Table 6 illustrates expected data ranges and the resulting number of bit reserved for data trans-mission. Optimization. A number of optimi- zation issues need to be taken into con- sideration to reduce the data load to an FIGURE 27 Grid of pressure/zenith hydrostatic delay as seen on July 28, 2008, at 12:00 UTC; values in acceptable minimum. Polar regions, for millimeters. instance, have low temperatures and relatively small temperature variations. tion can be carried out us-ing an expo- Encapsulation into Broadcast Message. This means that the zenith wet delay will

nential trend function ZWD(hUSER) Transmitting a global 3 x 3-degree data be by far less variable than in the tropics.

= ZWD(h)exp(-h/qZWD), but the scale grid in the navigation message will not For this reason, the grid resolution could height will be significantly different be possible as a single data package be coarser in these regions. from that employed to reduce the pres- containing all grid points. Instead, the Similarly, ocean surfaces might sure data. global grid is divided into sub-grids not be of large interest for a precision The original output resolution of (granules). positioning service. Con-sequently, global numerical weather models is cur- Figure 28 portrays such a granule a land-sea-mask is foreseen and data rently in the range of 0.5 to 1.0 degrees. with 3 x 3-degree grid containing wet over oceans can either be skipped or This amount of data cannot be handled delay data. In addition, hydrostatic data submitted on a coarser grid. Different in a broadcast message data stream. will be put on a 6 x 6-degree grid (blue data representation schemes, such as Hence, these grids are resampled to a dots). Each granule has 5 x 5 = 25 grid spherical harmonic coefficients, could lower resolution supplying smoothed points. An efficient distribution scheme be employed (but may be incompatible values on a coarser grid. is to be employed for the individual sat- with the granule fragmentation con- The spatial (horizontal) resolution of ellites that will allow the efficient broad- cept). Finally, the data sections can be the tropospheric grids differs because the cast of the various granules relevant to transmitted as compressed messages spatial correlation of pressure is substan- the user at a particular location in a (not studied here in detail). tially larger than that for the zenith wet minimum of time. Integration of Rain-Rate Information. delay. This fact is illustrated inFigure 26 Data Bits. The data to be transmitted We have only dealt with the problem (ZWD) versus Figure 27 (ZHD). In the comprise some header informa- figures, the hydrostatic component has tion (e.g., the granule identifier) a much higher correlation length (par- and the data grid itself. The data ticularly visible over the oceans where can be reduced by typical average the effect of topographic height varia- values in order to minimize the tions cannot be seen in the selected color number of bits as efficiently as scale). possible. In fact, it will be possible The following grid reso- to employ the standard (“blind”) lution is suggested: Galileo tropospheric correction ZWD, qZWD: 3° x 3° model for the data reduction of p, qp: 6° x 6° pressure and zenith wet delay. These grid sizes are a trade-off The difference between the between the desired accuracy and the actual quantity and the blind data value to be transferred to the user. model is the value to be trans-

They are based on an error analysis that mitted. FIGURE 28 Grid of tropospheric correction data for one takes into consideration the accuracy This method is particularly granule (sub-grid) with pressure (i.e., hydrostatic requirements as specified for the vari- useful for the pressure data, as delay-related quantities) data available at blue dots in addition to wet delay-related quantities. ous C-band services. can be seen in Figure 29.

www.insidegnss.com july/august 2009 InsideGNSS 63d working papers

In principle the weighting networks could be implemented at RF, IF, or baseband. From an interferer cancel- lation point of view, an RF weighting network would be preferable, because it would only require increased linear- ity for the low noise amplifier (LNA) and the weighting network (all stages in front of and including the weighting net- work). However, this RF solution would be expensive and bulky. A more attractive approach would be to put the beam-forming network into FIGURE 29 Grid of pressure/zenith hydrostatic delay as seen on July 28, 2008, at 12:00 UTC; values in the digital domain (IF or baseband). In millimeters. this case the whole front-end chain has to fulfill increased linearity require- of tropospheric delay compensation so Beam-Forming Antenna. Compared to ments in order to overcome the fact that far. Other value-added data could also L-band, in order to provide the neces- a receiver is significantly disturbed by be interesting for a C-band GNSS posi- sary signal strength for user equipment intermodulation products falling into tioning service. on the ground, the RF power of the the receive band. This is particularly true for rain-rate C-band transmitter should be higher. Use of a phased array introduces information, because rainfall attentua- However, this can cause problems in the similar considerations as for a fixed- tion is substantially stronger in C-band satellite, particularly in terms of power beam array. The individual weighting than in L-band. For this reason, such supply and interference (especially to the networks must track the movements of pieces of information can be vital for Galileo uplink). the receiver and the satellites in order to users in tracing signal reception prob- One solution to reduce this transmit maintain the nulling of the interferer. lems and similar phenomena. power requirement would be to use a Many algorithms are available in the Incorporation of such information beam-forming antenna in user termi- literature to address such tasks. How- into the navigation message would be nals — a quite imaginable prospect for ever, the phase center of the various possible. Up to five data bits would need handheld receivers. A possible further beams will cause problems, especially to be made available. Rain-rate informa- advantage of the beam-forming tech- when nulling is performed. Producing tion should be transmitted on the fine nique would be improved resistance a null in the beam pattern will cause a grid (3 x 3 degrees), because rain is often to jammers. Fixed beams or phased phase jump between signals entering the a local phenomenon (convection cells) arrays with variable beams could be beam in directions right and left from and thus will likely require the highest used, depending on the needs of the the null. resolution grid available. services. Low Noise Amplifier. The LNA in a Using a phased array antenna could GNSS receiver front-end has two pri- User Terminal Critical be especially beneficial for PRS-C ser- mary performance aspects: the noise Technologies vices, where jamming is a hot issue. figure, which mainly determines the The C-band project also sought to iden- Nulling algorithms could also be incor- sensitivity of the receiver, and the lin- tify critical technologies that would be porated into the receiver design in order earity of the LNA, which determines the strength and effect of interferers whose The problem of minimizing TTFF and the receiver’s intermodulation products are entering processing power consumption is even more the receive band. complicated than the direct long-code acquisition of For example, two wireless local area the C-band signal. network (WLAN) signals at 5250 MHz and 5290 MHz generate third-order needed to design and build user equip- to reduce strong jamming signals. intermodulation products at 5010 MHz ment capable operating with C-band Because a user terminal needs to and 5430 MHz. The product at 5010MHz in a future GNSS constellation. In the access signals from multiple satellites, falls into the receive band. following sections, we will discuss the it would have to have a multi-target The only way to reduce these out of most important of these technologies, phased-array antenna. This means band interferers and their intermodula- their significance for C-band receiver that the receiver design would need to tion products would be a narrow RF fil- operation, their availability and state of include a dedicated weighting network ter in front of the LNA. Such filters, how- the art, and other relevant aspects. for each satellite. ever, tend to be either bulky and costly

63e InsideGNSS july/august 2009 www.insidegnss.com working papers

(waveguide filter) or lossy, which dete- clocks are expensive, large, (rubidium this difference. The power consump- riorates the noise figure of the LNA. So, clocks about 200–300 cubic centime- tion of such an integrated solution will use of an LNA with good linearity even ters) and consume much power (up to be significantly less than for a discrete at high input signal levels is desirable. 10 watts). With the reduced size, price, front-end. Of course, good linearity usually comes and power consumption expected in the Massive Parallel Correlator Technol- along with higher power consump- future, however, manufacturers will be ogy. The number of effective correlators tion. Currently, some discrete LNAs are able to incorporate atomic clocks into that can be integrated into a single chip already available for the C-band. their receivers. should be an essential critical technol- In the C-band the manufacturers An alternative to the temperature ogy to acquire a direct long-code acqui- offering C-band LNAs are more dedi- controlled crystal oscillator (TCXO) sition of the proposed C-band signals (or cated to microwave components. This could be the microcomputer controlled codes). Long codes need long integration can be expected to change in the next crystal oscillator (MCXO). The price times and high requirements on Doppler years due to the WLAN applications that of the latter is quite high today, but the accuracy. are driving C-Band technology. technology is based on a crystal oscilla- Signal acquisition time may be RF Mixer and Filter. The RF mixer is tor with some licensed electronics. reduced by massive parallelization of important with respect to the linearity The main idea for improvement in correlations, achieved by integrating a of the system. In our work, we identified this area is to design a dual-mode oscil- high number of hardware correlators several examples from various manu- lator that uses the quartz in its funda- onto a single chip, possibly in combi- facturers of passive double- or triple- mental mode and in its third overtone. nation with other methods. The corre- balanced mixers with high linearity for The frequency of the third overtone is sponding hardware requirements have discrete C-band mixers. higher than that of the fundamental to be taken into account (for example, Most of today’s C-band mixers are based on gallium arsenide (GaAs) tech- Signal acquisition time may be reduced by massive nology, but some of them use a comple- parallelization of correlations, integrating a high mentary metal oxide semiconductor number of hardware correlators onto a single chip. (CMOS) technology capable of operating up to 6 GHz. The linearity of these pas- mode by approximately a factor of three fast Fourier transform, circular convolu- sive mixers depends on the local oscil- — but not exactly. A small temperature- tion, and so forth). lator drive level. High linearity requires dependent difference exists that can be As is well understood in the GNSS high drive levels — up to 20 dBm. Again, estimated by the MCXO electronics, and community, the problem of acquiring this increases power consumption but the oscillator output frequency then cor- and tracking GNSS signals involves a at levels similar for both L-band and C- rected accordingly. two-dimensional search in Doppler band. As use of such a technology becomes and code delay. In modern all-in-view Technologies for RF filters in C-band more widespread, the price could be receivers with multiple correlators per range from waveguide cavity filters to expected to fall into a region comparable receiver channel, each channel can have ceramic filters. Cheap filters in C-band to TCXOs. different Doppler frequencies and code are already available due to WLAN Integrated Front-End. An integrated delays in order to reduce the acquisition applications. The performance of such front-end would be especially attractive time (TTFF). filters is not as high as with cavity filter for antenna array applications, where The problem of minimizing TTFF types, but reasonable for the purpose of more than one front-end section is nec- and the receiver’s processing power con- receiver design. essary. In the L-band highly integrated sumption is even more complicated than Clock Oscillator. Clock quality is an front-ends are available in a commercial the direct long-code acquisition of the important performance parameter for market, typically within 5x5 millimeters C-band signal. Therefore, receiver design high-quality receivers. A precise clock size and consume about 40 milliwatts should consider a fast acquisition engine reduces the noise of the oscillators and of power. Complete one-chip solutions that might be based on a massively par- allows for narrower loop filters in low including processing are also available. allel array of correlators. dynamic scenarios, thereby reducing the We can also expect a similar solution Combined C-/L-Band Tracking. First of measurement noise. The receiver clock for C-band. In the field of WLAN appli- all, in order to process L- and C-band may also be synchronized to the system cations, multiple C-band RF front-ends signals simultaneously, both signals must time after a longer loss of the signal, dra- are already integrated for the WLAN be transmitted from the same satellites matically reducing reacquisition times. MIMO standard 802.11n. The per- without any clock offset. The dominant A precise clock is also essential for formance of such available integrated error source of single-frequency users is integrating the signal over a long inte- solutions will be worse than an opti- due to the ionosphere, which is inversely gration time interval, as is needed for mized discrete solution, but progress in proportional to frequency squared.

low C/N0. Currently, precise atomic semiconductor technology will reduce A key point of combined signal pro-

www.insidegnss.com july/august 2009 InsideGNSS 63f working papers

cessing of L- and C-band is to combine trical-mechanical-system (SoC-MEMS) bandwidth can be significantly reduced a common part of ionospheric propaga- integration has been considered as the as the IMU Doppler-aiding removes tion error between L-band and C-band next milestone in the MEMS research most of user dynamics from the signal- efficiently. Fortunately, the ionospheric area. In the past, MEMS research has tracking loop. This improves the quality error of a C-band receiver is 3.2 times mainly focused on physics level and of the measurements and the anti-jam- smaller than that of the L1-band receiv- component-level designs. This has made ming properties of GNSS receivers. er. Ionospheric scintillation may reduce distinct MEMS-IMU sensors affordable The efficiency of this integration the accuracy of a signal-tracking loop for a lot of applications. method depends on the quality of the and cannot be compensated in a single- Nowadays, the research effort has Doppler estimates derived from the frequency receiver. The use of L- and gradually migrated from components IMU; therefore, there is an upper limit C-band frequencies together makes it design to systems design, thus extend- of coherent integration time. easier to correct ionospheric errors suf- ing the operational performance into The RF front-end part of GNSS ficiently. the SoC field for achieving better sys- receiver block might be expected to be A signal-tracking Kalman filter in a tem performance and more effective implemented in a separate single die for local signal channel simultaneously han- cost reduction. the same reason as with SoC MEMS: dles the combined code/carrier tracking Unfortunately, up to now it seems the decoupling problem of CMOS and of a single- or dual-band (or higher) sig- that integrating a MEMS structure and the analog part. For example, integrat- nal with appropriate ionospheric delay a digital processing part into a single die ing the high-frequency RF part together estimation. Integer ambiguities can be would be problematic due to decoupling with the digital part in a single die has included in the state vector of the sig- of CMOS and the MEMS structure. been difficult up till now. Using an SiP approach, all analog parts and the digi- The SoC-type micro-electrical-mechanical-system tal signal processing portion including (SoC-MEMS) integration has been considered as the microprocessors and memory (RAM next milestone in the MEMS research area. and ROM) can be integrated into a sin- gle package. nal-tracking loops and also calculated Therefore, only a system-in-a package Multi-Bit ADC. A sampling num- more easily by combining L- and C- (SiP) type product is now available, which ber higher than the usual one or band carrier phases (e.g., widelane or means two individual dies integrated two bits is desirable in C-band user narrowlane). into a single package (i.e., a chip). equipment due to the continuous After integer ambiguities resolved, Currently SoC MEMS is implement- wave form property of the proposed the state vector of signal tracking loop ed with field programmable gate arrays GMSK signals, the constraints in the contains only signal parameters —for (FPGAs) because of their flexibility and C/N0, and the high performance require- example, code delay, carrier phase, Dop- easy configurability that make it possible ments for the services. pler frequency, and so forth — that need to implement all components into a sin- However, the main driver for multi- to be estimated. gle system. This type of system provides bit sampling is not so much to reduce SoC-Type INS-Aided Tracking. Over the a platform that simplifies implemen- the quantization loss. Rather a higher years the integration of GNSS and iner- tation of any type of digital hardware sampling bit number can help cope with tial measurement unit (IMU) technolo- solution in a short development cycle. high dynamics in terms of the received gies has advanced from the system level As a result, it is really efficient for a low incoming signal strength. It can also to deep inside the software/hardware production volume. help avoid a saturation of the analog-to- level. Today, a one-chip solution in sin- The main SoC-MEMS architecture is digital converter (ADC) due to a jammer, gle-die GNSS/IMU, including all digital divided into two components: a sensor enabling an active processing of the jam- signal processing components and self- cube and the signal processing/commu- ming signal. Multi-bit conversion results alignment functionalities, is even pos- nication interface circuitry. The sensor in more processing power being needed sible even with digital technology. cube has the whole set of sensors and in the digital part of the receiver. Moreover, many optimistically envi- sends all captured information to the In addition to the items discussed sion the possible implementation of a FPGA, which includes a signal process- here, many other critical technologies low-cost, system-on-chip (SoC) level ing part in which the received data are — such as power consumption, form GNSS/IMU in the near future. Such stored for later treatment. factor, board, processors, and memory an integration would consist of a single A lot of architectures have been — were identified for the early develop- (or multiple) die on a single chip that proposed for GNSS/IMU integrations. ment of C-band navigation systems. includes all systems, such as a GNSS Among these the deeply coupled meth- receiver and IMU as well as navigation/ od is known as the most advanced tech- Conclusions control processor blocks. nique. The main advantage of this tech- A C-band signal plan was designed to However, the SoC-type micro-elec- nique is that the carrier tracking loop fulfil the high-level requirements for

63g InsideGNSS july/august 2009 www.insidegnss.com working papers

[5] Schmitz-Peiffer, A. (2009), and L. responsible for research both identified services, namely the SPR- Stopfkuchen, J. J. Floch, A. Fernandez, R. Jor- activities on GNSS C and PRS-C. The effort focused on sig- gensen, B. Eissfeller, J. Á. Ávila-Rodríguez, S. signals, including BOC, nal modulation schemes to comply with Wallner, J. H. Won, M. Anghileri, B. Lankl, T. BCS, a n d MBCS the stringent requirements on spectrum Schüler, O. Balbach, and E. Colzi, “Architecture for modulations. Avila- confinement set out to ensure compat- a Future C-band/L-band GNSS Mission - Part 1: Rodriguez is one of the ibility with other services, according to C-band Services, Space- and Ground Segment, CBOC inventors. He is ITU regulations, with the neighboring Overall Performance,” Inside GNSS magazine, involved in the Galileo program, where he supports bands (e.g., radio-astronomy, uplink May/June 2009. ESA, the European Commission, and the Galileo receiver, and MLS) as well as to protect [6] Won, J. H. (2008), and B. Eissfeller, B. Lankl, A. Supervisory Authority (GSA), through the Galileo the Galileo uplink receiver. Schmitz-Peiffer, and E. Colzi, “C-Band User Termi- Signal Task Force. He studied at the Technical As a result, GMSK (with BT=0.3) nal Concepts and Acquisition Performance Analy- Universities of Madrid, Spain, and Vienna, Austria, and has a M.S. in electrical engineering. modulated both on I and Q channels sis for European GNSS Evolution Programme,” Pro- Jong-Hoon Won received was selected. Based on an extensive sig- ceedings of ION GNSS 2008, Savannah, Georgia, USA his Ph.D. in control and nal performance analysis together with instrumentation engi- [7] Won, J. H. (2008a), and J. Á. Ávila-Rodríguez, user terminal aspects, this modulation neering at Ajou Universi- S. Wallner, B. Eissfeller, J.-J. Floch, A. Schmitz-Pei- scheme was further optimized for maxi- ty, Suwon, Korea. His ffer, and E. Colzi, “C-Band User Terminal Aspect for mum bandwidth occupation and spec- thesis was on signal pro- Bandwidth Efficient Modulation Schemes in Euro- tral separation between the two identi- cessing, simulation, and pean GNSS Evolution Programme,” International navigation algorithms for software-based GPS fied services. Detailed signal parameters Symposium on GPS/GNSS 2008, Tokyo, Japan such as chip rate, chip length, and so on receivers. Currently he is working at the Institute of [8] Won, J. H. (2008b), and B. Eissfeller, A. were designed to satisfy the requirement Geodesy and Navigation at the Federal Armed Schmitz-Peiffer, and E. Colzi, “C-Band User Termi- that C-band navigation services shall be Forces (FAF) University Munich, Germany. His cur- nal Tracking Loop Stability Analysis for European rent research activities are software GNSS receivers, competitive with current or planned L- GNSS Evolution Programme,” Proceedings of ION GNSS/INS coupling systems, and user terminals. band services. GNSS 2008, Savannah, Georgia, USA Stefan Wallner studied at [9] Won, J. H. (2008), and B. Eissfeller, A. the Technical University Acknowledgements Schmitz-Peiffer, and E. Colzi, “C-Band User Ter- of Munich and graduated Authors Note: It is highly remarked that minal RFI Effect Analysis for European GNSS Evo- with a Diploma in tech- this column is based upon a C-band lution Programme,” Proceedings of the Fifth ESA no-mathematics. He is GNSS study being conducted within the NAVITEC-2008, Noordwijk, The Netherlands now research associate European Space Agency (ESA) GNSS at the Institute of Geod- Evolution Program. Please note that the Authors esy and Navigation at the University of the Fed- views expressed in the following reflect “Working Papers” explore the technical and sci- eral Armed Forces Germany in Munich. His main solely the opinions of the authors and do entific themes that underpin GNSS programs and topics of interests can be denoted as the spread- ing codes, the signal structure of Galileo together not represent those of ESA. applications. This regular column is coordinated by Prof. Dr.-Ing. Günter Hein. with radio frequency compatibility of GNSS. Additional Resources Prof. Dr.-Ing. Hein is head Marco Anghileri is a of Galileo Operations and research associate and [1] Ávila-Rodríguez, J. A. (2008), and S. Wallner, Evolution for the Euro- Ph.D. candidate at the J. H. Won, B. Eissfeller, A. Schmitz-Peiffer, J.-J. pean Space Agency. He is Institute of Geodesy and Floch, E. Colzi and J.-L. Gerner, “Study on a Galileo a member of the Euro- Navigation at the Univer- Signal and Service Plan for C-band”, Proceedings pean Commission’s Gali- sity FAF Munich. He of ION GNSS 2008, Savannah, Georgia, USA leo Signal Task Force and studied at the Politecnico [2] Irsigler, M. (2004), and G. W. Hein, and A. organizer of the annual Munich Satellite Naviga- di Milano, Italy, and at the Technical University Schmitz-Peiffer, “Use of C-Band Frequencies for tion Summit. He has been a full professor and Munich, Germany and has an M.Sc. in telecom- Satellite Navigation: Benefits and Drawbacks,” GPS director of the Institute of Geodesy and Navigation munication engineering. His scientific research Solutions, Wiley Periodics Inc., Volume 8, Number at the University of the Federal Armed Forces work focuses on GNSS signal structure and on sig- 3, 2004 Munich (University FAF Munich) since 1983. In nal processing algorithms for GNSS receivers. Bernd Eissfeller is full [3] ITU Regulations, www.itu.int/pub/R-REG- 2002, he received the United States Institute of professor and director of RR/en Navigation Johannes Kepler Award for sustained and significant contributions to the development the Institute of Geodesy [4] Schmitz-Peiffer, A. (2008), and D. Felbach, of satellite navigation. Hein received his Dipl.-Ing and Navigation at the F. Soualle, R. King, S. Paus, A. Fernandez, R. Jor- and Dr.-Ing. degrees in geodesy from the Univer- University FAF Munich. gensen, B. Eissfeller, J. Á. Ávila-Rodríguez, S. sity of Darmstadt, Germany. Contact Prof. Dr.-Ing. He is responsible for Wallner, T. Pany, J. H. Won, M. Anghileri, B. Lankl, Hein at . teaching and research in and E. Colzi, “Assessment on the Use of C-Band José-Ángel Ávila-Rodríguez received his PhD in the field of navigation and signal processing. Until for GNSS within the European GNSS Evolution Pro- signal design at the Institute of Geodesy and the end of 1993 he worked in industry as a project gramme,” Proceedings of ION GNSS 2008, Savan- Navigation at the University of FAF Munich. He is manager on the development of GPS/INS naviga- nah, Georgia, USA. tion systems. From 1994 to 2000 Eissfeller was

www.insidegnss.com july/august 2009 InsideGNSS 63h working papers

head of the GNSS Laboratory at University FAF engineer at EADS Astrium. His work is mainly Munich. He is author of more than 215 scientific focused on Galileo signal design at the system and technical papers. level and evaluation of performances and robust- Berthold Lankl received ness of Galileo signals. his Dipl.-Ing. from the Lars Stopfkuchen received Technical University his Diploma at the Uni- Munich and his Dr.-Ing. versity of Cooperative from the University FAF Education Ravensburg, Munich. He has 20 years Germany in communica- experience in the devel- tions engineering. He opment of radio relay systems at several positions joined EADS Astrium in the industry. Since 2003 he is a professor of GmbH in 2003 as design engineer for FPGA and communications engineering at the University FAF ASIC developments and participated in GIOVE-B Munich. and other ESA projects. Currently he conducts sys- Torben Schüler received tem engineering tasks for electronics subsystems his diploma in geodesy and is responsible within the C-band study for the and cartography from navigation payload design. the University of Han- Dr. Dirk Felbach received nover in 1998. He later his Ph.D. in 2001 from joined the Institute of the Technical University Geodesy and Navigation Munich in Radio Fre- (IGN), University FAF Munich, as a research asso- quency Engineering. ciate, where he earned a doctorate and received Since then he is with the habilitation in geodesy and navigation. Schül- Astrium working as a er is currently head of the GNSS/INS laboratory at Systems Engineer and Project Manager on naviga- IGN. His major research work is focused on precise tion payload equipment for time and signal gen- GPS/Galileo positioning, including atmospheric eration. delay modeling. Antonio Fernández Oliver Balbach joined IFEN received his M.S. degree GmbH in 2006 where he is in aeronautical engi- involved in the Galileo neering from the Poly- testbed project GATE and technic University of in projects concerning the Madrid and an M.S. in development of GNSS physics from the UNED receivers. He received his University of Spain. He has been working in the Diploma in electrical engineering from the Technical field of GNSS since 1996. Fernández co-founded University München and has worked as a patent DEIMOS Space in 2001, where he is currently in examiner at the German Patent Office and as research charge of the GNSS technologies section. associate at the University FAF Munich IGN. Rolf Jorgensen received Andreas Schmitz-Peiffer his MScEE degree from received his Master’s and the Technical University Ph.D. in atmospheric of Denmark. An antenna physics at Kiel Universi- expert with long experi- ty, Germany. He has 19 ence, he has been years of experience in involved in the design system engineering and support for development of C-band and Ku-band project management in earth observation and satellite communication payloads for ESA and navigation. Since 2000 he has been working at INTELSAT and responsible for the antenna design EADS Astrium GmbH as project manager in the in numerous other ESA projects. Galileo program, where he is presently leading the Enrico Colzi, received his C-band study and an indoor navigation project. Ph.D. from the Technical Jean-Jacques Floch was University of Delft, The graduated with a Dip- Netherlands and his M.Sc lom-Ingenieur in elec- in telecommunication tronics and telecommu- engineering from the nication at the Institut University of Florence, Supérieur Electronique Italy. He was with the European Space Agency Numérique in France. He (ESA), the Italian Space Agency, and the Univer- has worked in the area of mobile communications sity of Florence. Currently he is working for Vega for several years. Since 2002 he has been working Group PLC as senior RF payload engineer in the RF in the field of navigation satellites as system Payload System Division of ESA/ESTEC. 63i InsideGNSS july/august 2009 www.insidegnss.com