a novel approach for the use of - geopp · reference stations and their removal from the original...

A Novel Approach for the Use of Informationfrom Reference Station Networks

Conforming to RTCM V2.3 and Future V3.0B.E. Zebhauser, H.-J. Euler, C.R. Keenan, Leica Geosystems AG, Switzerland

G. Wübbena, Geo++ GmbH, Germany

ABSTRACT

In applications requiring centimeter accuracy, singlebaseline positioning methods are being superseded bythose methods using information from permanent refer-ence station networks. The well known advantages af-forded by reference station arrays include improved mod-eling of the remaining tropospheric, ionospheric and orbitbiases. Methods and concepts show the improvements inperformance and reliability in some kind of closed systemapproaches. Standardization discussions underway withinRTCM target the interoperability between the referencestation systems and roving receivers from various manu-facturers. One obstacle in the discussion, and therefore inlater interoperability issues, is the creation and properdescription of the models used for deriving the biasesnoted above. This difficulty has to be mitigated and willvanish with time, but this interoperability is needed ur-gently. This paper details a different approach to utilizeand distribute the information from permanent referencestation networks in RTCM Version 3.0 compatible mes-sage types. Throughput calculations show their efficiency.The separation of different calculation tasks demands aneasy, powerful and broadcast sufficient standard for thetransfer and distribution of network information. Themethod proposed here does not restrict roving users to onespecific network positioning concept, yet provides theopportunity for applying simple interpolation methods.Such a concept may solve the current dilemma concerninginteroperability standards.

INTRODUCTION / MOTIVATION

Real-time messages for proper interoperability betweendifferent manufacturers’ equipment have been issued bythe RTCM Sub-Committee 104 (RTCM, 2001). All in-formation for precise positioning using baseline ap-proaches can be transmitted using Version 2 messageTypes 18-21 or in the future using Version 3 messageTypes 101 and 103 respectively.

Because the use of single reference stations has somedisadvantages in that the accuracy and reliability of inte-ger ambiguity resolution deteriorates with distance fromthe reference station, networks of reference stations arebeing developed to mitigate the distance-dependency ofRTK solutions. With such networks, a provider can gener-ate measurement corrections for receivers operatingwithin the network area and can supply this information tothe user in a standardised format. As the current kinematicand high-accuracy message types do not support the use ofdata from multiple reference stations, new standards musttherefore be considered to facilitate the valuable informa-tion afforded by reference station networks.

The standardisation of network information and process-ing models is also necessary to reduce the sizes of thenetwork RTK corrections, as well as the transmission ofsatellite-independent error information. A simplified ap-proach of transmitting data from reference station net-works to roving users is now presented in this paper in theform of a new message standard capable of supportingreference network operations. Its use should help over-come some of the problems encountered in current net-work RTK concepts.

For a short summary of the existing concepts “FKP” and“VRS” including a discussion of problems we refer toEULER et al. (2001).

Specialised models requiring detailed description anddiscussion are not used in this proposal.

FUTURE V3.0 EVOLUTION

The future Version 3 of RTCM DGNSS messages (RTCMV3.0 DRAFT, 2001) differ from those of Version 2.3 interms of increased flexibility and efficiency. This is pro-moted by a layer approach consisting of application, pres-entation, transport, data link and physical layer. In thispaper we focus only on the presentation layer that de-scribes the message formats and how to apply them. For athroughput analysis one has also to consider the transport

layer. With this higher efficiency was achieved by re-placing the 20% parity bits overhead per word with oneCyclic Redundancy Check at the end of the whole mes-sage. One message can reach 32767 bytes in length andalways fits the byte boundary. A short preamble andlength definition replaces the 2 words (60 bits) messageheader of Version 2.3. A short or long message type canbe chosen, which differ in the message overhead of 40against 48 bits in total. The message numbers start with101 in order to avoid any possible confusion with Version2, which utilizes message types 0-63. The followingscheme shows the message frame structure:

Fig. 1: RTCM Version 3 Frame Structure

The message types actually planned for RTCM V3.0 arelisted in the Appendix.

As we believe a RTK network message type has to bereleased as soon as possible, we propose it conforming toRTCM V3.0 as well as V2.3. This paper focuses attentionon V3.0, whereas the paper of KEENAN et al. (2002) doeson V2.3.

TRANSMISSION CONCEPT PROPOSAL:COMMON AMBIGUITY LEVEL AS THE KEYLINK

It is a well-known fact that the proper resolution of the so-called integer ambiguities is the key to high accuracypositioning for a single baseline. The power of networksolutions will be experienced with the proper integer am-biguity resolution between permanent reference stations.Following the resolution of integer ambiguities betweenreference stations and their removal from the originalobservations, a common integer ambiguity level can beestablished across the network.

While the future RTCM message Type 103 contains in-formation to process a single baseline, the proposed con-cept shall help to transmit information related to a refer-ence station network or part thereof. The future messageType 103 remains unchanged and is still part of the infor-mation transmission concept. Measurement information ofadditional reference stations is created by forming differ-ences of their Type 103 corrections with those of a “mas-ter” reference station. In essence, one will transmit to theuser the corrections of the master reference station viaType 103 messages and the additional smaller correctiondifferences between the master reference station and eachfurther (auxiliary) reference station in the network via the

new proposed message type. As a first proposal let’s call itType 125 message for Version 3.0. The presence of thesecorrection differences will allow each rover to directlyinterpolate spatially. Such an approach could afford cor-rections for a virtual reference station, or the reconstruc-tion of the original corrections in the sense of Types 103messages for each station relative to the master referencestation.

The master reference station coordinates have to be pro-vided to the rover using an RTCM Type 102 message,whereas the position information of the further (auxiliary)reference stations can be transmitted as coordinate differ-ences. This saves some bits due to the smaller numbers.Furthermore this rarely changing “static” information canbe split from the correction differences and form an addi-tional message portion, which can be broadcasted at avery low rate.

The proposed Type 125 message concept will supplymore information to rover users at the same transmissionrate. By providing a common integer ambiguity level, itsimultaneously allows the information to be directly usedfor spatial interpolation. This concept can be used withone-way communication and so the number of participantsis not limited as in a VRS concept realization. The deci-sion on the processing concept can be made at the rover.The roving receiver can decide whether to use the com-plete information of a single reference station or use partor the complete suite of transferred reference station datafor deriving its best solution.

Alternative means of transmitting this information aredeveloped in the further sections.

OBSERVATION EQUATIONS

The concept discussed in this paper uses correction differ-ences, i.e. between station single differences reduced byslope distances, receiver clock errors and ambiguities.These can be directly derived from the equation (2) ofEULER et al. (2001). Using ∆, one can denote the betweenstation single difference of a component. In our approachwe attempt to determine the components on the right handside of the resulting observation equation excluding thetropospheric and orbit (non-dispersive) as well as theionospheric (dispersive) parts, and subtract them from theleft hand side (the measurement). Supplemented withindices for station A (or B, C, … respectively), satellite jand frequency indicator (L1, L2 and in the future L5, forGalileo in a similar way) we get for the L1 carrier phase:

Φ

Φ

∆+∆

+∆

+

∆+∆+∆=

=∆⋅−∆⋅−∆−∆Φ

,121

21

,

1,1

,1,1,

)()(ˆ

)()(ˆ

~)(

εδ

δδ

f

tI

f

tI

rtTtT

Nfc

dtcst

jAB

jAB

BEjAB

jAB

jAB

jABAB

jAB

jAB

(1)

Preamble 8 bits

Length Parameter 1 bit

Data Message 0 … 127 bytesor 128 … 32767 bytes

CRC 24 bits

Length 7 or 15 bits

where1f frequency of L1

c speed of light in a vacuum t measurement epoch

)(1, tjAB∆Φ raw phase measurement in [m]

jABs~∆ geometric range single difference, in-

cludes antenna phase center variations and multipathing,which are already determined and applied by the networkprocessing software

Φ∆ ,1,ABdt total receiver clock errorjABN 1,∆ initial phase ambiguity in [cycles]

)(ˆ tT jA∆ modeled tropospheric refraction effect

)(tT jAδ∆ residual tropospheric refraction effect

BEjABr ,δ∆ broadcast orbit error induced distance

dependent effect on baseline AB

)(ˆ tI jA∆ modeled ionospheric refraction

)(tI jAδ∆ residual ionospheric refraction effect

Φ∆ ,1ε random measurement noise

and with ∆ denoting the between station single differenceof a component.For the other frequencies one gets equivalent equations.

This resulting residual shall be transmitted to the rovers,but as already mentioned above, problems arise in thestandardization of models. Note our approach proposes analternative and transmits the whole right-hand side ofequation (1), and therefore the problem is circumvented.

For the generation of a common integer ambiguity level,which is a key feature for this approach, we refer toEULER et al. (2001).

RTCM CORRECTION DIFFERENCES PROPOSAL

Before we describe the correction differences proposed fora message type in detail, we shall recap on the RTCMcorrections in the notation used here. The future RTCMV3.0 will include a message denoted as Type 103 corre-sponding to Type 20 in V2.3. The carrier phase correction(example given for L1) for a station A is defined to:

jA

jA

jA

jRTCMA dtcdtctts ΦΦ ⋅−⋅+Φ−=Φ ,1,1,1,,1, )()(δ (3.1)

The variable )(ts jA denotes the geometric range between

the position of the receiving antenna and the broadcastedsatellite position.

For a second station B, the correction can be written in thesame way, as well as for further reference stations:

jB

jB

jB

jRTCMB dtcdtctts ΦΦ ⋅−⋅+Φ−=Φ ,1,1,1,,1, )()(δ (3.2)

As discussed above a network algorithm will generate anambiguity leveled set of RTCM Type 103 messages forevery reference station. These messages could be directlytransmitted in parallel by every broadcast station, but dueto throughput issues, it is desirable to have more compactmeans than doing that. Therefore we are proposing only totransfer the differences for the auxiliary reference stations.

In the case, that station A denotes the master referencestation and the station B stands for one of the auxiliaryreference stations, one can form directly from the equa-tions above single differences always related to station A,following the equation (1).

As a result of the network processing either the singledifference or double difference ambiguities will be takeninto account. In the case of double difference ambiguities,it must be assured that they all relate to the same referencesatellite at each epoch, although a change of the referencesatellite may appear between epochs. As already stated,this bias remaining over all single differences can be re-duced by subtracting a constant (integer cycle) bias fromall correction differences relating to one reference stationpair. This yields a further reduction in the overall size ofthe numbers that have to be transmitted.Basically one has to transmit all information relating tothe master reference station. This will include corrections(message Type 103 in RTCM V3.0) and the master sta-tion’s coordinates and antenna information (message Type102 and 106). To avoid inconsistencies in the data proc-essing one should not mix up the two philosophies of“corrections” and “measurements”, e.g. by transmissionof master reference station information in measurementtype messages like Type 101. Relative to that master ref-erence station, the correction differences of the other(auxiliary) reference stations B,C,D,… are generated.These will be transmitted in the proposed message format.So the new message type contains all relevant data:

jAB

ABjAB

jAB

jAB

Nfc

dtctts

1,1

,1,1,1, )()(

∆⋅+

∆⋅+∆Φ−∆=∆Φ Φδ(4)

If one considers splitting the correction in two compo-nents, namely a dispersive and a non-dispersive part, onehas to form the geometry-free and the ionosphere-freelinear combination from L1 and L2, so that the two re-sulting correction differences will become:

jAB

jAB

dispLLjAB

ff

f

ff

f

2,21

22

22

1,21

22

2221,

1,

∆Φ−

−

∆Φ−

=∆Φ −

δ

δδ

(5)

jAB

jAB

dispnonLLjAB

ff

f

ff

f

2,22

21

22

1,22

21

2121,

1,

∆Φ−

−

∆Φ−

=∆Φ −−

δ

δδ

(6)

Both are expressed in meters and the dispersive part(equation 5) is related to the L1 frequency.

While the multiple frequency option (e.g. named asL1/L2/L5 option) is easier to handle, the dispersive/non-dispersive option has the advantage of being able to varythe rates of one part in comparison to the other, due to itssmoother behaviour.However there are two possible ways of using the disper-sive/non-dispersive option: firstly their direct interpolationsimilar to the FKP concept, or secondly in the reconstruc-tion of the message Type 103-like corrections. In the lattercase, when the dispersive and non-dispersive terms arederived by L1 and L2, one will miss information related toL5 in the future. To be able to reconstruct all three (L1, L2and L5) corrections as independent information, an addi-tional residual has to be provided. In the future, possiblyalready for the V3.0 release one has to include such amessage portion, that consists mainly of a value, whichcan be derived e.g. by

( )( )dispnonLLj

ABdispnonLLj

AB

dispLLjAB

dispLLjAB

f

f

−−−−

−−

∆Φ−∆Φ+

⋅∆Φ−∆Φ=

21,1,

51,1,

25

2121,

1,51,

1,AddRes

δδ

δδ(7)

wherein L1L2 and L1L5 indicate the frequency combina-tions the dispersive and non-dispersive parts are derivedfrom. This means, that firstly one has to compute the dis-persive and the non-dispersive part from L1 and L5 fol-lowing the equations (5) and (6); secondly one has to formthe differences to the equivalents computed from L1 andL2; thirdly the dispersive residual part has to be scaled toL5; and in the end one adds these two residuals to onescaled to L5. This value can then be directly applied to acorrection difference for L5 just derived from L1 and L2:

AddRes21,1,

25

2121,

1,21,

5,

+∆Φ+

⋅∆Φ=∆Φ

−−

−

dispnonLLjAB

dispLLjAB

LLjAB

f

f

δ

δδ(8)

A format proposal draft covering the dispersive/non-dispersive option can be found as Appendix. A verysimilar format proposal for the multi-frequency option canbe formulated, but has been neglected. The equivalentproposal for RTCM version 2.3 can be found in KEENANet al. (2002).

Note on the Used Ephemeris Set SynchronizationTo assure the use of the same orbits at the reference andthe rover sides, the Issue Of Data Ephemeris (IODE) willbe included into the future correction message Type 103.This insures a correct reconstruction of measurements

from corrections as well as from correction differences.An inclusion of the IODE into the correction differencesmessage (Proposed 125) of the auxiliary reference sta-tions, which are always related to one master referencestation, is not necessary, because the Type 103 messagesof this master reference station already contain this infor-mation and are always transmitted at the highest rate. Ifone wants to use the correction differences just for inter-polation as they are thought for, the IODE from the Type103 message must not be held available for these calcula-tions.

ASSESSING THE RANGES OF THE MESSAGECOMPONENTS

To get an idea of the ranges related to the dispersive andnon-dispersive effects, the impacts of ionospheric andtropospheric refraction as well as of orbit errors have to beconsidered. Note we have assumed that antenna phasecenter variations and multipathing have been mitigated toa negligible amount by the network software.

Satellite OrbitIn order to significantly reduce the orbit errors one coulduse IGS predicted orbits during data processing. Howeverin reality one has to use the model provided by the broad-cast message in a real-time approach. This less accurateorbit representation contains errors, which are satellite,distance and time dependent. The radial error generally isless than 10 m, equating to an error of 0.4 ppm (parts permillion) in baseline length. In extreme cases, 40 m ofradial error are possible (1.6 ppm). The change in satelliteorbits with time is very smooth unless a maneuver appearsor the broadcast representation shows a break. Assumingmaximum distances of reference station separations ofabout 300 km, then one gets certainly less than 0.5 m.

Ionospheric RefractionThe ionospheric effect is both distance and time depend-ent. The latter characteristic is treated in a further section.The error due to ionospheric refraction can reach someppm in baseline length, up to 15 ppm in mid-latitudes,whereas in equatorial zones some tens of ppm are possi-ble. SKONE (2001) detected gradients of 35 ppm in L1 atequatorial regions 1998 corresponding to 35/59 cm inL1/L2 range over 10 km distance. WANNINGER (1994)noted up to 80 ppm, e.g. in South Brazil in 1992. Suchlarge amplitudes vary slower (large scale travelling iono-spheric disturbances). Assuming maximal separation ofreference stations of about 300 km, this corresponds to±24 m maximum. With a correction difference resolutionof 0.5 mm, this needs a data slot of 17 bits if one considersa “dispersive” message slot. Problems in reference stationnetworks can result from irregularities of e.g. 20 – 50 kmrange and 0.5 m magnitude in L1 as detected by SKONE(2001) in the auroral region, because reference stationseparations longer than some tens of km do not cover suchspatial changes. However this is a particular problem ofthe network design not the message design.

Tropospheric RefractionWith large height differences in reference station networksand low satellite elevation angles, e.g. around 5°, themodeled single difference error contribution can exceedmore than 10 m. Theoretically one can think about speci-fying a very simple model in the RTCM standards toeliminate this part. However such a model has to consistof a standard atmosphere (using defined values for tem-perature, pressure and relative humidity, as well as of therelated gradients with height) and of a simple formula forthe tropospheric delay including the mapping function.The prime reason to create our pure observation correctionbased message is avoid the introduction of models!

For the whole non-dispersive amount one has to reserve atleast 15 bits. Considering a realistic maximum heightdifference of 2000 m over 300 km station distances, onegets maximal tropospheric refraction effects of less than±12 m. ±11 m is due to height and elevation differencecovered by a model and ±1 m due to non-modeled errors(i.e. time and location dependent variations). This is anoptimistic estimate so it is better to reserve 16 bits forsuch a message part equivalent to a range of ±24 m. Theother significant non-dispersive effects, i.e. the orbit er-rors, should not exceed more than one meter for baselinesless than 300 km. Consequently one can include this to-gether with the tropospheric effect to afford one “non-dispersive” or “geometric” part, and for this also only aslot size of 16 bits is required.

Range of the Correction DifferencesThe following table summarizes the maximum rangesnoted above for the ionospheric, tropospheric and satelliteerrors and also the number of bits required for 0.5 mmresolution. As already mentioned in the troposphere sec-tion the latter two components are often combined to ageometric (or: non-dispersive) group of systematic errors.

Effects Non-Dispersive DispersiveSpecific Orbit Troposphere IonosphereSitedependent

- < ±12 m(dh < 2000 m)

-

Distancedependent

< 1.6ppm

some few ppm < 80ppm

Sum(distance< 300 km)

max.±1 m

max.±12 m

max.±24 m

17 bits 17 bitsBitsRequired 18 bits

Table 1: Summary of Maximum Errors

Concluding the above, one has two possibilities withwhich to realize the approach. Firstly one can take intoaccount only one component comprising the whole cor-rection differences for all frequencies (L1, L2 and L5),that have to be broadcast at a high rate. Alternatively, oneconsiders the dispersive and non-dispersive componentswhich can allow the broadcast of e.g. the non-dispersive

component at a lower rate than the dispersive. In addition,the latter then has to be scaled by a frequency dependentfactor before it can be added to the non-dispersive partand subsequently used for correction of the L2 or L5 ob-servations. A discussion of the possibilities, that arisefrom observations on three frequencies, can be found inHAN and RIZOS (1999) or HATCH et al. (2000).

For the first case, one certainly has to expect less than±48 m. This results in 18 bits for a 0.5 mm carrier phaseresolution for each satellite and frequency (L1, L2, L5).For the second case one has to expect less than ±24 m forthe dispersive part (a data slot of 17 bits) and certainly lessthan ±24 m for the non-dispersive part (17 bits also). Intotal one gets 34 bits (plus two for indicating the kind ofcomponent) for each satellite. Consequently the secondcase has the advantage of saving around 200 bits for 12satellites considering the anticipated L1/L2/L5 scenario onGPS. As already mentioned the non-dispersive part can bebroadcast at a lower rate thereby also contributing to asignificant data throughput saving.

Range of the Additional Residual for a 3rd FrequencyIn the case of providing dispersive and non-dispersivecorrection differences, the necessary additional residualfor observables on a third frequency like L5 (GPS) con-sists of effects, that cannot be covered by linear combina-tions generated just from L1 and L2. These phenomenaare frequency dependent and will be summarized in thefollowing: The receiver noise and other unmitigated ef-fects should generally not exceed an amount of some fewmillimeters and even in extreme cases become less than 1or 2 centimeters. The main impact has to be expected fromhigher order ionospheric refraction, that is not covered bythe quadratic approach for the dispersive part. FollowingBASSIRI and HAJJ (1993) one can assess a maximum rangeof some few centimeters for this effect considering aminimum elevation of 5° and high ionospheric activity. Sothe whole additional residual should never exceed 6 cmand can be transmitted in an 8 bit data slot.

Resolution of the Coordinate DifferencesAs the coordinates of the master reference station are fullytransmitted via a message Type 102, the coordinate infor-mation of the auxiliary stations can be reduced to differ-ences relative to the master reference station.The resolution needed for the coordinate differences de-pends on whether the correction differences will be usedfor spatial interpolation purposes only or to reconstruct theundifferenced Type 101 like corrections from the pro-posed Type 125 messages.If the station coordinates will be used for the interpolationof correction differences, then the coordinates can beprovided with a lesser resolution of 2.5 meters. Thismeans that, in the limits of the specifications, the correc-tion differences in the proposed Type 125 message do notchange more than 1 mm from one interpolation grid pointto the other. These interpolation grid points are separatedin geocentric coordinate vector components dX, dY anddZ by 2.5 m.

If the rover will attempt to ‘reconstruct’ the measurementsof each auxiliary reference station, then the coordinates ofthe auxiliary reference stations should be transmitted to aresolution of 1 millimeter.

To cover all purposes we propose a further option: thehorizontal components were described by WGS84 latitudeand longitude differences with a 0.000025 degree resolu-tion and the height difference is given in meters with0.001 m resolution related to heights above WGS84 ellip-soid in a Reference Frame conforming to ITRF like that ofGPS. Such a representation is well suited to e.g. referencestation grids and helps to save some bits.

The definition of the antennas for all auxiliary referencestations has always to be consistent with the master refer-ence station. This can be easily solved by using the samemodel type antenna for all reference stations. In this ap-proach the phase center variations are totally eliminatedby applying absolute calibration values.

ASSESSING THE MESSAGE UPDATE RATES

General ConsiderationsWith the consideration of dispersive and non-dispersiveterms, different update rates can be specified. For exam-ple, the non-dispersive effects (tropospheric and orbiteffects) are assumed to change more slowly relative to thedispersive effects (ionospheric effects). A recent question-naire completed by RTCM members suggested the fol-lowing maximum tolerated latency values: orbits at120 seconds, troposphere at 30 seconds and the iono-sphere at 10 seconds.

Assessments for Update RatesAs already well investigated e.g. by WANNINGER (1999),rapid changes in corrections using data from referencestations in mid-latitudes can reach 1.5 ppm per minute forthe dispersive part and only 0.1 ppm per minute for thenon-dispersive part. This supports the proposal, that thesetwo components should be transferred at different rates.

Our own investigations have shown that both the disper-sive and non-dispersive components exhibit trends ofsome mm, sometimes with rapid changes of the short-termtrend including a change in its sign. In some extreme casesat low geomagnetic latitudes variations reaching 1 cm persecond have been found in the dispersive part. There arealso some high-frequency changes between epochs of1 Hz data that reach up to 4 cm in both dispersive andnon-dispersive parts, for baseline lengths of up to 300 km.

Following are some initial suggestions for the update ratesthat must be defined when using corrections within RTKnetworks, not just for the concept described here.

Probably the most important factor to be considered, whenassessing optimal update rates for all network information,is the implication that such rates will have on the Time-

To-First-Fix (TTFF). A roving user would demand that heshould receive all correction differences as soon as possi-ble after he begins surveying. This would mean that allrelevant network information (Types 102, 103 and 125)should be transmitted as often as possible so as to mini-mise the TTFF and provide rapid access to all networkinformation – an initial suggestion is at least every15 seconds. The dispersive terms should be transmitted atan equal or higher rate, as should the non-dispersiveterms. Ideally the first would be at least every 0.1 Hz.However it still has to be clarified if the correction differ-ences or at least a non-dispersive component can betransmitted at an even lower rate than the dispersiveterms.

THROUGHPUT ANALYSIS

Although we developed a format for the L1/L2/L5 option,only the dispersive/non-dispersive option is documentedin depth here and in the Appendix for brevity. Neverthe-less the related throughput assessments yield similar re-sults in the case of two frequencies.

We want to give a short outline of some results. For themaster reference station one has always to transmit the fullstation information present in the messages of Type 103,102 and 106; the latter two can be transmitted at a lowrate. But for each additional station, a larger proportion ofbits can be saved if using the proposed type 125 message.For the dispersive/non-dispersive case, this factor is about2 - depending on the difference in the rate of broadcastingthe dispersive and non-dispersive component as well as onthe number of frequencies - compared to a “conventional”102/103 realization. If not only L1 and L2 are taken intoaccount, but also L5 will be included in the future (3 fre-quency case), this factor will increase up to 3.

Fig. 2: Throughput of auxiliary reference stations for oneepoch tracking 10 satellites at 2 or 3 frequencies: com-parison of existing with future and proposed messages;scenarios of our proposal for V3.0 are plot with solid lines

Some different message type scenarios for the auxiliaryreference stations are presented in Figure 2. These are thetransmission of the reference station information as V2.3Type 20, 21 (leaving out Type 24 messages, which con-tain station information, that can be transmitted at a lowerrate) and as V3.0 Type 101 with and without Type 102(station information), and the proposed V2.3 Type 25message for both dispersive/non-dispersive as well as forjust the dispersive component. Furthermore different sce-narios using the proposed V3.0 message Type 125 withand without coordinate difference portion are shown:Dispersive part only, plus non-dispersive, or L1+L2+L5.The information regarding the master reference station isexcluded in this analysis. It has to be transmitted simulta-neously by Type 103/102/106 messages. The graph corre-sponds to 10 satellites using two (L1 and L2) or threefrequencies, i.e. considering that L5 (or Galileo) is avail-able. It shows the throughput in bits for one epoch setdependent on the number of auxiliary reference stations.

To aid in the discussion of how the reference station in-formation will be transmitted at realistic rates, let us con-sider what the RTCM 2.3 draft (RTCM, 2001) states:“However, the data update requirement of RTK is muchhigher than conventional differential GNSS, since it in-volves double-differencing of carrier measurements. Datamust be updated every 0.5 – 2 seconds. The data rate isdriven not by SA variations that are no longer relevant,but by the RTK technique, which requires measurement-by-measurement processing. As a consequence, the datalinks are more likely to utilize UHF/VHF, with transmis-sion rates of 4800-9600 baud.”Looking at the situation objectively, one has take intoaccount a lower than nominal throughput because trans-mission includes stop and parity bits. Considering thisoverhead of 20%, one gets for nominal rates of 4800 and9600 bps, real rates of 3840 and 7680 bps. With somespares for safety’s sake, one can expect about 3500 and7000 bps as the real data transfer rates. At present GPSsuppliers normally recommend radio modems with atransmission rate of 9600 bps or greater. In Europe, GSMmodems capable of 9600 bps are becoming more andmore the standard communication solution. Some newGSM modems also cover the new GPRS or HSCSD stan-dards, that allow data rates up to 50 kbaud and becomesmore and more competitors of early UMTS realizations.

Regarding the Future V3.0 Conforming ProposalThe proposed correction differences have already beenoptimized for the RTCM V2.3, so the respective Version 3proposal does not result in significantly better throughput.Just the general message structure change improves theperformance as well as the fact, that the master station’scorrections can also be transmitted by a new slim mes-sage, denoted as Type 103. This saves a lot of bits and soallows the data transmission for some more auxiliaryreference stations compared to RTCM V2.3.

Considering a nominal baud rate of 9600 bps (7000 bpseffectively), 10 satellites tracked simultaneously on 2 or 3

frequencies, results in a different number of auxiliaryreference stations, whose information can be transmittedin average in 1 second, using three different alternatives.The following table shows the results using the messagesa) Type 101 without extended data portion (XDP), be-cause 103 is not explicitly formulated yet, b) the proposedType 125 in the dispersive/non-dispersive option withoutcoordinate difference portion and c) this proposed Typealso in the dispersive/non-dispersive option, but transmit-ting the latter component in a significantly lower rate.

Number of Auxiliary Reference Stations Using …Message Type Format L1+L2 L1+L2+L5a) RTCM 101 w/o XDP 6 4b) dispersive/non-disp. 9…10 8...9c) dispersive at a high, non-disp. at a low rate (factor ≥5)

~ 14 ~ 13

Table 2: Number of auxiliary reference stations that canbe transmitted in average in 1 sec. at 9600 bps with 10satellites tracked at 2 or 3 frequencies, using differentmessage formats, excluding coordinates as well as theinformation of the master reference station

Note that the information of the master reference stationas well as the coordinate information of the auxiliarystations has been omitted from this theoretical calcula-tion. Including this information one would get a reductionof about 20…30%, depending on the scenario. But re-gardless of that these results advocate a new messageformat for the transmission of reference station networkinformation.

Broadcasting ConsiderationsThe need to transmit the information of more than e.g.five reference stations leads to the discussion, how oftenthe station information should be updated (especiallyregarding the high-frequency changes of the dispersiveeffects) and what latency can be accommodated by therover processing software. This may result in the distribu-tion of the information related to one specific measure-ment epoch over a specific number of seconds. The likelydemand of transmitting the data of as many stations aspossible favours the dispersive/non-dispersive optionbecause it allows the transmission of components at dif-fering rates. This will afford a great saving of bits and soallow the transmission of more stations’ data at the sametime.

CONCLUSIONS AND SUMMARY

A new message standard has been proposed to aid thesupport of reference network applications. Its use shouldovercome some of the problems seen in recent proposalsfor reference station network positioning concepts.

Our concept is based on transmitting messages of Type101/102/106 for a master reference station and a newlyproposed message type 125 for further (auxiliary) refer-

ence stations. The more additional reference stations aretransmitted using the proposed message type, the greaterthe savings are compared to using RTCM messages ofType 103 plus 102.

The main advantages of the proposed message formats areas follows:§ In general, there are less data bits to transmit§ The correction differences can be used for direct

interpolation, once adjusted to the common integerambiguity level

§ No detailed model specifications are needed (e.g. fortroposphere)

§ One-way communication is sufficient§ Kinematic applications can work without interrup-

tions and discontinuities unlike a VRS system, whichhas to change its VRS position during motion.

§ There are no restrictions on the number of users in areference station network

§ No dependency exists to a specific concept approach:the rover user gets the full reference station networkinformation and can independently use its own mod-els, interpolation and processing concepts

This approach can be used with update rates of up to 1 Hz,elevation masks down to 5° and reference station dis-tances up to 300 km.

With the implementation of L5, the dispersive/non-dispersive option will easily provide the most efficient useof throughput. Variations in the dispersive component willthen dictate the lower limit for the correction differenceupdates.

Further ConsiderationsIn addition to the issues raised in this paper, there arefurther ones that should be discussed. E.g. suppose alarger number of auxiliary reference stations. Their cor-rection differences shall be transmitted in sequence butrelated to the same epoch. This possibly takes longer thanthe update interval of the 103 messages of the masterreference station, that are updated at a higher rate relatedthen to more recent epochs. The potential for synchroni-zation problems has to be discussed.

Finally one can state, that the proposed concept can miti-gate or solve many of the problems of the existing con-cepts.

REFERENCES

Bassiri, Hajj (1993): Higher-order ionospheric effects onthe global positioning system observables and meansof modeling them. manuscripta geodaetica (1993)18:280-289

Euler H.-J., Goad C.C. (1991): On optimal filtering ofGPS dual frequency observations without using orbitinformation. Bulletin Géodésique (1991) 65: 130-143, Springer-Verlag Berlin.

Euler H.-J., Keenan C.R., Zebhauser B.E., WübbenaG. (2001): Study of a Simplified Approach in Util-izing Information from Permanent Reference StationArrays. Paper presented at ION GPS 2001, Salt LakeCity, Utah.

Han, S., Rizos, C. (1999): The impact of two additionalcivilian GPS frequencies on ambiguity resolutionstrategies. 55th National Meeting U.S. Institute ofNavigation, "Navigational Technology for the 21stCentury", Cambridge, Massachusetts, 28-30 June, pp.315-321.

Hatch, Jung, Enge, Pervan (2000): Civilian GPS: TheBenefits of Three Frequencies. GPS Solutions Volume3, Number 4, Spring 2000, pp. 1-9.

Keenan C.R., Zebhauser B.E., Euler H.-J., WübbenaG. (2002): Using the Information from ReferenceStation Networks: A Novel Approach Conformingto RTCM V2.3 and Future V3.0. Paper in prepara-tion for PLANS 2002, Palm Springs

RTCM (2001): RTCM Recommended Standards forDifferential GNSS (Global Navigation SatelliteSystems) Service, Version 2.3, RTCM Paper 136-2001-SC104-STD

RTCM V3.0 Draft (2001): RTCM Recommended Stan-dards for Differential GNSS (Global NavigationSatellite Systems) Services, Thirteenth DRAFTFUTURE Version 3.0. RTCM Paper 173-2001-SC104-265

Skone S. H. (2001): The impact of magnetic storms on GPSreceiver performance. Journal of Geodesy 75 (2001)9/10, 457-468

Townsend B, K. Van Dierendonck, J. Neumann, I.Petrovski, S. Kawaguchi, and H. Torimoto(2000): A Proposal for Standardized Network RTKMessages. Paper presented at ION GPS 2000, SaltLake City.

Vollath U., A. Buecherl, H. Landau, C. Pagels, B.Wagner (2000): Multi-Base RTK Positioning UsingVirtual Reference Stations. Paper presented at IONGPS 2000, Salt Lake City.

Wanninger (1994): Der Einfluß der Ionosphäre auf diePositionierung mit GPS. Wissenschaftliche Arbeitender Fachrichtung Vermessungswesen der UniversitätHannover, Nr. 201.

Wanninger, L. (1999): The Performance of Virtual Ref-erence Stations in Active Geodetic GPS-Networksunder Solar Maximum Conditions. Proceedings ofION GPS 99, Nashville, S. 1419-1427.

Wübbena. G, A. Bagge, G. Seeber, V. Böder, P.Hankemeier (1996): Reducing Distance DependentErrors for Real-Time Precise DGPS Applications byEstablishing Reference Station Networks. Paper pre-sented at ION 96, Kansas City.

Wübbena. G., S. Willgalis (2001): State Space Approachfor Precise Real Time Positioning in GPS ReferenceNetworks. Presented at International Symposium onKinematic Systems in Geodesy, Geomatics andNavigation, KIS-01, Banff, June 5-8, Canada.

APPENDIX: PROPOSAL FOR A RTK NETWORK RTCM V3.0 MESSAGE TYPE 125 FORMAT

Table 1. Contents of a Proposed Type 125 Header - RTK Network Correction Differences

PARAMETER NUMBEROF BITS

RESOLUTION UNITS RANGE

Message Number 12 1 Unitless “125”=0000 0111 1101

Epoch Time (TOW/TOD)(See Note 1)

23 0.1 Seconds GPS: 0 to 603,799.9 secondsGLONASS: 0 to 86,399.9 seconds

M = Multiple MessageIndicator

1 -- “0” – informs the receiver that thisis the last message of this typehaving this time tag“1” – informs the receiver thatanother message of this type withthe same time tag will follow

Number of Auxiliary Sta-tions Transmitted

4 1 Unitless 0 to 15

GS = Global SatelliteSystem Indicator(See Note 2)

2 -- Unitless “00” – Message for GPS satellites“10” – Message for GLONASSsatellites“01” – Message for GALILEOsatellites“11” – reserved for future systems

Network ID 5 1 Unitless 0 to 15Reserved 1 -- -- --Reserved 4 -- -- --Master Ref. Station ID(See Note 3)

12

S = 64 bits

1 Unitless 0 to 4095-> waiting for a 102 type message…

Note 1: Epoch Time (TOW) - For this data slot, if a resolution of 10 Hz seems sufficient for customers and suppliers (to bedecided) then 23 bits are required (from Type 101 message). Greater than 603,799.9 seconds will roll over to the followingGPS Week. Greater than 86,399.9 seconds will roll over to the following GLONASS Day.Note 2: Global Satellite System Indicator – self-explanatory really. GLONASS would use fewer bits within the Epoch Time asit is referenced to Time of GLONASS Day.Note 3: Master Reference Station ID – relating to that reference station designated as the reference for differencing purposes.Note: Total Number of Bits required for RTK Correction Differences Header is 59 bits

Compact Representation = 59 bits (6.750 bytes)Long Representation = 64 bits (7 bytes)

Table 2. Contents of the RTK Network Auxiliary Station Header Portion of a Proposed Type 125 message



Auxiliary Station ID 12 1 0 to 4095Reserved 2 -- -- --Coord. Type & ResolutionFlag (CTR Flag)(See Note 4)

2 -- Unitless 0 = no co-ordinate differences in-cluded (wait ..)1 = Cartesian ECEF Coord. Diff.,Resolution 2.5 m2 = Cartesian ECEF Coord. Diff.,Resolution 0.001 m3 = Ellipsoidal ECEF Coord. Diff.,Resolution 0.000025 deg and0.001 m

Reserved 1 -- -- --

# of GPS Satellite SignalsProcessed

7

S = 24 bits

-- Unitless 1 to 32

Note 4: Co-ordinate Type and Resolution Flag (CTR Flag) – this is required to define the resolutions of the Correction Dif-ferences according to the users applications and requirements.

0. No coordinate differences included (system must wait)1. Cartesian ECEF Coordinate Differences – resolution 2.5 metres2. Cartesian ECEF Coordinate Differences – resolution 0.001 metres3. Ellipsoidal ECEF Coordinate Differences – resolution 0.000025° and 0.001 metres

Having the CTR flag here complements the ‘Reference Station Motion Flag’ that is incorporated within Message Type 102 –RTK Reference Station Message. It would also afford the flexibility should a provider wish to transmit different levels of co-ordinate difference resolutions.Note: Total Number of Bits required for Auxiliary Station Correction Differences Header Portion is 21 bits


Table 3. Contents of the RTK Network Auxiliary Station Portion of a Proposed Type 125 message

CTR Flag Option 1: Cartesian ECEF Co-ordinate Differences, Low/Coarse Resolution 2.5 m



If CTR Flag = 1ECEF dXCo-ordinate

18 2.5 Metres ±327677.5 m

ECEF dYCo-ordinate

18 2.5 Metres ±327677.5 m

Reserved 2 -- -- --ECEF dZCo-ordinate

18

S = 56 bits

2.5 Metres ±327677.5 m

Note: Total Number of Bits required for CTR Flag Option 1: Low/Coarse Resolution 2.5 metres is 54 bitsCompact Representation = 54 bits (6.750 bytes)Long Representation = 56 bits (7 bytes)

CTR Flag Option 2: Cartesian ECEF Co-ordinate Differences, High/Fine Resolution 0.001 m



If CTR Flag = 2Reserved 2 -- -- --ECEF dXCo-ordinate

30 0.001 Metres ±536870.911 m

Reserved 2 -- -- --ECEF dYCo-ordinate

30 0.001 Metres ±536870.911 m

Reserved 2 -- -- --ECEF dZCo-ordinate

30

S = 96 bits

0.001 Metres ±536870.911 m

Note: Total Number of Bits required for CTR Flag Option 2: High/Fine Resolution 0.001 metres is 90 bitsCompact Representation = 90 bits (11.25 bytes)Long Representation = 96 bits (12 bytes)

CTR Flag Option 3: Ellipsoidal ECEF Co-ordinate Differences,Mixed Resolution 0.000025 deg in Latitude and Longitude, 0.001 m in height



If CTR Flag = 3WGS84 dLatCo-ordinate

18 0.000025 Degrees ±3.2767 degrees (corresponds to aresolution of always better than2.7 m even in high latitudes)

Reserved 5 -- -- --WGS84 dLonCo-ordinate

19 0.000025 Degrees ±6.5535 degrees (twice range ofdLat to accommodate high latitudeapplications: at 60° lat, the maxi-mum distance in longitude shortensdown to the same of dLat, i.e. about350 km; at 70° lat down to about240 km)

WGS84 dHgtCo-ordinate

22

S = 64 bits

0.001 Meters ±2097.151 m (Height aboveWGS84 Ellipsoid in a ReferenceFrame conforming to ITRF like thatof GPS )

Note 5: The WGS84 dLat, dLon, dHgt option can also be used to generate a grid of reference stations / correctionsNote 6: The WGS84 dHgt parameter is set to a higher (finer) resolution to accommodate the height sensitivity of atmosphericrefraction correction.Note 7: The WGS84 dLat, dLon, dHgt option uses the same datum definition (e.g. an ITRF realization) as the master referencestation but is always referenced to the surface of the WGS84 ellipsoid.Note 8: The Type 125 Header always contains the Master Reference Station ID as well as a Network IDNote: Total Number of Bits required for CTR Flag Option 3: Resolution 0.000025° and 0.001 metres is 59 bits


Table 4. Contents of the RTK Network Correction Differences Body Portion of a Proposed Type 125 Message- Dispersive/Non-dispersive Parameters -



Satellite ID 6 1 Unitless 0 to 63 (if GALILEO uses >32 SVIDs)

P/C = CA-Code / P-CodeIndicator

1 -- Unitless "0" – C/A-Code"1" – P-Code

Reserved 1 -- --Dispersive or Non-Dispersive Indicator(See Note 10)

1 -- Unitless "0" – Dispersive correction follows"1" – Non-dispersive correctionfollows

Ambiguity Fixed/FloatFlag (AFF Flag)(See Note 11)

1 - Unitless 0 = Fixed Carrier Phase Ambiguity1 = Float Carrier Phase Ambiguity

Reserved 5 -- --Carrier Phase CorrectionDifference(See Note 10)

17

S = 32 bits

0.0005 Meters ±32 metres

Note: Total Number of Bits required for Correction Differences Body Dispersive/Non-dispersiveCompact Representation = 26 bits (3.250 bytes)Long Representation = 32 bits (4 bytes)

Table 5. Contents of the RTK Network Correction Differences Body Portion of a Proposed Type 125 message- considering Multiple Frequencies (L1; L2; L5) -



Satellite ID 6 1 0 to 63 (if GALILEO uses >32 SVIDs)

P/C = CA-Code / P-CodeIndicator

1 -- "0" – C/A-Code"1" – P-Code

Reserved 1 -- -- --Frequency Indicator(See Note 10)

2 -- "00" – L1"01" – L2"10" – L5"11" – reserved

Reserved 3 -- -- --Ambiguity Fixed/FloatFlag (AFF Flag)(See Note 11)

1 - 0 = Fixed Carrier Phase Ambiguity1 = Float Carrier Phase Ambiguity

Carrier Phase CorrectionDifference(See Note 10)

18

S = 32 bits

0.0005 Meters ±64 metres

Note 9: is not used hereNote 10: Instead of Dispersive Or Non-Dispersive Indicator, one can modify it to Frequency Indicator and put the wholecorrection difference of L1, L2 or L5 into the field Carrier Phase Correction Difference, that has to become 18 bits wide then.The V3.0 resolution is improved to 0.0005 m requiring one additional bit compared to version 2.x. The carrier phase correc-tion differences are represented in L1 cycles for consistency. They are always related to the phase center, i.e. the AntennaReference Point (ARP) of the reference stations involved (analogous to message Type 24) corrected by the model type antennaPCVs (e.g. a “nullantenna”). In any case their antenna phase center behaviour has to be consistent with that of the masterreference station.Note 11: AFF Flag - AMBIGUITY FIXED/FLOAT FLAG Indicates whether the carrier phase ambiguity for that particularsatellite has been resolved completely by the network software to a fixed level, or at least partially to a float level.Note: Total Number of Bits required for Correction Differences Body Portion is 28 bits


General Note: The term “Compact Representation” denotes the amount of bits really necessary, if one would not follow thebyte boundary considered in the RTCM Version 3.0 as a rule, which is denoted here by the term ”Long Representation”

________________________________

ADDENDUM: Future RTCM V3.0 Type Messages Overview (as already planned in Draft 13 of October 2001)

Type 101 – GPS RTK ObservablesType 102 – RTK Reference Station ParametersType 103 – GPS RTK CorrectionsType 104 – GLONASS RTK ObservablesType 105 – GLONASS RTK CorrectionsType 106 – Antenna Identification ParametersType 107 – System Parameters

Type 111 – GPS Differential CorrectionsType 112 – GLONASS Differential CorrectionsType 113 – No Op FillerType 114 - Radiobeacon AlmanacType 116 - ASCII CharactersType 117 - GPS EphemerisTypes 4001-4095 - Proprietary messages

a novel approach for the use of - geopp · reference stations and their removal from the original...

Documents