ESTB PERFORMANCE AGAINST LOSSING SATELLITES (FIRST RESULTS)
Jaume Sanz (1), Xa ie P a s (1), Richa d Fa nwo h (2), San iago Soley (2)
(1) Resea ch g oup o As onomy and Geoma ics, Uni e si a Poli ècnica de Ca alunya (gAGE/UPC),
Ba celona, Spain.
(2) EUROCONTROL Expe imen al Cen e, B e igny, F ance.
Key wo ds: Augmen a ion Sys ems, EGNOS, ESTB
Abs ac
The EGNOS Sys em Tes Bed (ESTB) is he EGNOS p o o ype which has been b oadcas ing a Signal in Space
since Feb ua y 2000. Accu acy, in eg i y, con inui y and a ailabili y a e he main concep s in his Eu opean
componen o he Sa elli e Based Augmen a ion Sys em which is designed o suppo en ou e h ough p ecision
app oach ai c a na iga ion.
In his wo k, o he i s hal o he yea o 2002, we s udy he ESTB pe o mance in ixed si es in Ba celona, and
we analyze how his pe o mance is a ec ed when losing a single ESTB moni o ed sa elli e. The da a se in ol es
a 24 hou s weekly measu emen s collec ed in wo ixed si es (UPC1 and UPC2), om Janua y o Augus 2002. In
pa icula , he numbe o Loss o In eg i y e en s (LOI) and he deg ada ion o he Ve ical Posi ion E o
pe cen ile and he APV-II a ailabili y a e analyzed.
As he main esul s, wo sa elli es ha e been iden i ied which loss could ha e p oduced a la ge numbe o LOIs (up
o hund eds) in wo pa icula days be o e he ESTB upda e o Ap il 16 h 2002. The imp o emen in he ESTB
pe o mance a e such upda e is also epo ed.
1. In oduc ion
In he ecen yea s, Global Na iga ion Sa elli e Sys ems (GNSS), ha e expe ienced a conside able g ow h o new
applica ions and use s. Sa e c i ical applica ions, such as ci il ai adiona iga ion, equi e high le els o accu acy
and in eg i y (sys em's abili y o bound i s own accu acy) ha a e no me in s andalone GNSS posi ioning.
Sa elli e Based Augmen a ion Sys ems (SBAS) p o ide di e en ial GNSS co ec ions and in eg i y in o ma ion o
single equency use s. These co ec ions a e combined wi h ecei e measu emen s, and o he local in o ma ion,
o ob ain he inal na iga ion solu ion and in eg i y p o ec ion bounds.
Since Feb ua y 2000 he EGNOS Sys em Tes Bed (ESTB), a p o o ype o EGNOS, has become ope a ional. This
has esul ed in he ini ia ion o a ious ac i i ies using he ESTB in p epa a ion o he ime when he eal EGNOS
Signal-in-Space alida ion needs o begin. P o o ype da a collec ion ools ha e been de eloped, no ably he
Pegasus ool [1], o p ocessing da a collec ed wi h he ESTB and a ious da a collec ion e o s ha e al eady
aken place.
Based on he expe ience collec ed wi h he ESTB da a collec ion, Eu ocon ol has de eloped documen s
s anda dising he way da a should be collec ed and he way he esul s should be displayed.
Wi h hese ools, ecei e s and associa ed documen s a ailable, a need was iden i ied o pe o m da a collec ion on
a mo e egula basis o es ablish a good unde s anding o he pe o mance o an SBAS Signal-in-Space o e a
longe ime in e al o e an a ea wi h a good geog aphical dis ibu ion. Wi h his objec i e in mind, Eu ocon ol
es ablished a small ne wo k o da a collec ion si es composed o ou Eu opean uni e si ies, among hem UPC
Resea ch G oup o As onomy and Geoma ics (gAGE/UPC).
Since Janua y o 2002, gAGE/UPC, has been collec ing ESTB da a in Ba celona as a pa o he abo e men ioned
ESTB Da a Collec ion ne wo k o Eu ocon ol. The da a se s a e p ocessed using he Pegasus*Plus ool p o ided
by Eu ocon ol and, in he case o he UPC si es, a e also p ocessed wi h BRUS (Basic Resea ch U ili ies o
SBAS) [2], which is a so wa e ool de eloped by gAGE/UPC. One o he a ge s o his small wo king g oup is o
analyze he ESTB pe o mance o he di e en ope a ion modes and o iden i y and s udy po en ial anomalies in
bo h, he Signal-in-Space and use domains.
As a esul o his e o , a huge da a base is c ea ed allowing se e al s udies abou he ESTB pe o mance, such as
he p esen ed in his wo k, whe e he ESTB obus ness agains loosing a single moni o ed sa elli e is analyzed
du ing he i s six mon hs o 2002. This pe iod includes he ESTB upda e o Ap il 2002, a e which a clea
imp o emen o he ESTB esul s is obse ed.
Be o e s a ing wi h he main opic o his wo k, a gene al backg ound abou he ESTB Sys em is gi en in nex
sec ion 2.
2. ESTB gene al backg ound
2.1 The ESTB sys em
The ESTB is a comple e p o o ype o he EGNOS sys em. I implemen s he main unc ions ( anging, GPS
di e en ial co ec ions and in eg i y) wi h eal ime elemen s enabling di e en use expe imen s o es s. O
cou se, he ESTB unc ions, complexi y and pe o mance a e educed by an o de o magni ude wi h espec o he
EGNOS sys em, p incipally conce ning he a ailabili y, he obus ness and edundancy o acili ies.
Figu e 1. ESTB Sys em A chi ec u e
The ESTB sys em is illus a ed in igu e 1. I is composed o :
• Remo e In eg i y Moni o ing S a ions (RIMS) ga he ing da a o he pu pose o gene a ing co ec ions
and in eg i y messages deployed o e Eu ope a hos ing si es,
• Cen al P ocessing Facili y (CPF) o compu e on-line di e en ial co ec ions and in eg i y da a, based a
Höne oss in No way (NMA p emises),
• h ee EURIDIS Re e ence S a ions implemen ing a wide iangula obse a ion base o anging pu poses
wi h he s a ions loca ed in Aussaguel (F ance), Kou ou (F ench Guyana) and Ha ebeeshoeck (Sou h
A ica),
• a Mission Con ol Cen e (MCC) loca ed in Toulouse, F ance, (CNES p emises), o compu e anging
messages and p epa e na iga ion messages o ansmission h ough he AOR-E payload,
• Two Na iga ion Land Ea h S a ion (NLES) based on exis ing s a ions o he INMARSAT ne wo k, one
loca ed in Aussaguel (F ance) implemen ing he b oadcas link o he AOR-E Na iga ion payload, he
o he loca ed a Fucino (I aly) implemen ing he b oadcas link o he IOR Na iga ion Payload. Each
NLES is complian wi h equi emen s om he INMARSAT SDM
• a g ound ne wo k composed o sub-ne wo ks o ansmi da a om e e ence s a ions o cen al p ocessing
acili ies, o he mission con ol cen e and o he app op ia e geos a iona y uplink s a ions.
Th ough he GEO sa elli e, he ESTB p o ides messages con aining co ec ions o he use o posi ion and
in eg i y compu a ion.
2.2 The ESTB signal in Space
The ESTB sys em ansmi s i s di e en ial co ec ions and in eg i y in o ma ion o use s in a 250 bi messages.
The co ec ions o indi idual sa elli es mus be combined wi h ecei e measu emen s and o he local in o ma ion
o o m he na iga ion solu ion and he p o ec ion bounds. The main componen s o he ESTB message a e
desc ibed as ollows [3]:
Fas Co ec ions
They a e necessa y o co ec he as changing e o s – usually he sa elli e clock e o . The as co ec ions,
p o ided as ange co ec ion alues, a e applied di ec ly o he ange measu emen s.
Toge he wi h he as co ec ions, in eg i y indica o s in he o m o UDRE es ima es a e b oadcas . These UDRE
a e an uppe bound on he e o o he pseudo ange a e he applica ion o as co ec ions, including he
possibili y ha he use misses any messages. The UDRE a e used o compu e p o ec ion le els and also wa ning
lags indica ing ha an indi idual PRN should no be used in he posi ion solu ion.
Slow Co ec ions
Slow co ec ions a e used o emo e he slowly a ying e o s such as he sa elli e posi ion e o s caused by e o s
in he epheme is ansmissions and he deg ada ion o he sa elli e posi ion calcula ions wi h he ime.
The sa elli e clock e o is also co ec ed by he slow co ec ion.
Ionosphe ic Co ec ions
In con as o he calcula ion o he as and slow co ec ions (which a e e y s aigh o wa d), he de e mina ion
o he ionosphe ic e o co ec ions is mo e complex. The slan delay and he slan esidual ionosphe ic e o
caused by he ionosphe e a e calcula ed using se e al s eps.
The SBAS ionosphe ic co ec ion messages a e ela ed o an in e pola ion g id abo e he ea h's su ace a an
al i ude o he ionosphe e. Fo each sa elli e, he pie ce poin o he di ec line-o -sigh om he sa elli e o he
use 's posi ion h ough he ionosphe e is de e mined. This pie ce poin is hen used o de e mine he app op ia e
g id poin s o he in e pola ion.
The in e pola ion uses a weigh ing scheme o all ou neighbou ing g id poin s o de e mine an es ima ion o he
e ical delay o he pseudo ange a he ionosphe ic pie ce poin . This e ical delay is inally modi ied o ake he
ele a ion o a pa icula sa elli e in o accoun . Thus, a slan delay o he ionosphe ic e o model and an es ima ion
abou i s esidual e o a e calcula ed.
2.3 The pe o mance
The Minimum Ope a ional Pe o mance S anda ds (MOPS) [4] o SBAS speci y he compu a ion o he Ve ical
and Ho izon al P o ec ion Le els (VPL, HPL) o he di e en ially co ec ed na iga ion solu ion which mus be
me a a p obabili y o 99.99999%. Thus he ue e o mus no exceed he p o ec ion le el mo e han once in 107
seconds.
Figu e 2 shows he S an o d plo [4] epo ing he Ve ical Pe o mance o he na iga ion solu ion. The ho izon al
axis is he ue e o o he EGNOS na iga ion solu ion ega ding he ue posi ion. The e ical axis co esponds o
he p o ec ion le el compu ed o he na iga ion solu ion. Each bin speci ies he numbe o occu ences in a
loga i hmic colo scale.
The ho izon al and e ical lines a 20m indica es he Ve ical Ala m Limi (VAL) o he ope a ion mode APV-II
de ined by ICAO’s GNSS SARPS. The sys em is decla ed una ailable when VPL>VAL.
Poin s wi h e o g ea e han he p o ec ion le el (VPE>VPL) indica es Loss-O -In eg i y ailu es (LOI). I
includes he Misleading In o ma ion (MI) and he Haza dously Misleading In o ma ion (HMI) e en s.
Figu e 2. Ve ical Sys em Pe o mance
The No mal Ope a ion Region is he le uppe iangle de ined by he condi ions: VPL>VPE and VPL<VAL. The
a ailabili y equi emen s de ined by SARPS speci y a minimum o 99% o he sys em in his no mal ope a ion
egion o e e y phase o ligh . In igu e 2, such equi emen is ul illed 99.33% o he ime.
Simila me ics and plo s a e de ined o he Ho izon al pe o mance (see [5]).
3. ESTB pe o mance analysis agains losing sa elli es
The in eg i y equi emen s de ined in p e ious sec ion mus be ul illed whene e he se , o subse , o n>=4
sa elli es used in he na iga ion solu ion wi h alid di e en ial co ec ions a ailable o he na iga ion mode. And
he loss o a single o mul iple alid sa elli es due, o ins ance, o space-c a maneu e s, should no p oduce
LOIs. Indeed, he deg ada ion in he na iga ion solu ion accu acy, due o he loss o sa elli es, mus be ollowed by
an inc ease o he p o ec ion le els in such a way ha he in eg i y be gua an ee.
As we a e going o show in he nex sec ion, i was iden i ied wo days, be o e he ESTB upda ing o Ap il 16 h, in
which he losing o a single sa elli e could p oduce a la ge numbe o LOIs. F om he ea e , we a e going o call
hese sa elli es as "c i ical sa elli es" in he sense ha hei losses p oduce in eg i y ailu es.
3.1 Compu a ions
The ESTB pe o mance agains losing a single sa elli e has been analyzed du ing he i s hal o he yea 2002,
using he ESTB da a se s collec ed a he UPC1 and UPC2 si es, loca ed in Ba celona and Cas ellde els, Spain,
wi h 16 Km o baseline.
The da a se s in ol e 24h weekly measu emen s s a ing on 10UT Tuesday up o 10UT o F iday, om Janua y o
Augus o 2002. In o al, i has been analyzed 29 days, wi h abou 28 sa elli es in iew pe day, which suppose
mo e han 70. 106 p ocessed epochs o each si e.
The da a p ocessing algo i hm consis ed o compu ing he na iga ion solu ion a e excluding a single sa elli e in
iew and i e a ing his compu a ion o all o he sa elli es, and o all o he da a collec ing days. The algo i hm is
summa ized in he ollowing scheme:
o each $DAY
o each sa elli e in iew $PRN
compu e he na iga ion solu ion wi h BRUS
“excluding he sa elli e $PRN”
end o
end o
This algo i hm has been easily implemen ed in a e y s aigh o wa d sc ip using BRUS unde LINUX ope a i e
sys em.
3 Resul s
The ESTB pe o mance esul s ob ained om he p e ious compu a ions a e gi en as ollows, whe e in eg i y,
accu acy and a ailabili y a e analyzed o bo h si es UPC1 and UPC2, be o e and a e he ESTB upda e in Ap il
16 h.
3.1 In eg i y: LOIs when losing a single sa elli e
The esul s o UPC1 a e shown in igu e 3, whe e he LOI ailu es a e gi en o he di e en da a collec ing days.
As i is shown, in gene al, he loss o a single sa elli e does no p oduce a signi ican inc emen o he LOIs, excep
o he DOY 87 (Ma ch 28 h) and DOY 101 (Ap il 11 h). The in eg i y ailu es on bo h days a e ela ed wi h he
exclusion o sa elli es PRN20 and PRN01. The mos c i ical day was Ma ch 28 h, whe e mo e han 250 LOIs
appea s a e excluding he PRN20. No ice ha bo h days a e p e ious o he ESTB upda ing o Ap il 16 h, and no
signi ican inc easing o LOIs appea s when deselec ing a single sa elli e a e such ime.
Simila esul s as in UPC1 we e ob ained o UPC2 si e, bu wi h a numbe o LOIs qui e g ea e , i.e., wi h mo e
han 350 e en s when excluding PRN20 in Ma ch 28 h (see ig 4).
PRN01
PRN01
PRN01
PRN01
ESTB upda eESTB upda e
PRN20
PRN20
PRN20
PRN20
Figu e 3. LOI ailu es when losing a single sa elli e in UPC1 si e. The ho izon al axis is he Day o he Yea (DOY) anging on Janua y 10 h
(DOY=10) o Augus 28 h (DOY=240) o 2002. The e ical axis is he numbe o LOI ailu es ob ained wi hou excluding sa elli es (squa es)
o , a e excluding a single sa elli e (diamonds). The day co esponding o he ESTB upda e, Ap il 16 h (DOY= 106), is also indica ed by a
e ical line.

PRN01
PRN01
PRN20
PRN20
ESTB upda e
Figu e 4. Same plo as in p e ious igu e 3, bu o UPC2 si e
The numbe o LOIs in unc ion o he PRN sa elli e o all o he analyzed days in UPC1 is gi en in igu e 5.
This plo allows o easily iden i y he c i ical sa elli es, which exclusion p oduce la ge numbe o in eg i y ailu es.
As i is shown, and acco dingly, i has been commen ed in he p e ious igu es, only he sa elli es PRN01 and
PRN20 can be iden i ied as a c i ical ones.
DOY 087
DOY 101
DOY 087
DOY 101
Wi hou
excluding
sa elli es
Wi hou
excluding
sa elli es
Figu e 5. LOI ailu es when losing a single sa elli e in UPC1 si e o all o he da a collec ing days ( om Janua y 10 h o Augus 28 h). The
ho izon al axis is he PRN numbe o he excluded sa elli e in he compu a ion. The “0” s ands o he solu ion compu ed wi hou excluding
sa elli es. The e ical axis is he numbe o LOI ailu es e en s. The days in which he la ge LOIs appea s a e also indica ed in he igu e.
Simila plo has been ob ained o UPC2.
A compa ison be ween he LOIs when deselec ing a single sa elli e, be o e and a e he ESTB upda e o Ap il
16 h, is gi en o UPC1 in igu e 6. As i can be seen, he sys em shows a sho e numbe o in eg i y ailu es a e
such upda e, being clea ly mo e obus agains losing sa elli es.
Be o e Ap il 16 h 2002
A e Ap il 16 h 2002
Figu e 6. Same plo as in p e ious igu e 5, bu dis inguishing he esul s be o e he ESTB upda e (uppe ) and a e (lowe ).
3.2 Accu acy: 95 h Pe cen ile o he VPE
The 95 h pe cen ile o he Ve ical Posi ion E o is gi en in igu e 7 o he UPC1 si e. As i is shown, he accu acy
could be signi ican ly a ec ed when losing a single sa elli e, be o e he ESTB upda e, which e o could be
deg aded up o wo imes he o iginal alues. Mo eo e , in some cases he accu acy is imp o ed when deselec ing
a single sa elli e. Such imp o emen could exhibi a po en ial anomaly in he di e en ial co ec ions b oadcas ed
o such sa elli e, because when losing sa elli es, a deg ada ion o he accu acy should be expec ed due o he mo e
poo geome y.
A e he ESTB upda e, he 95 h e ical pe cen ile is educed up o wo imes ega ding he p e ious pe iods wi h
ypical alues o abou 3 me e s. And he loss o a single sa elli e has a small impac o e he 95 h pe cen ile,
excep o Ap il 25 h (DOY 115). The DOY 115 was an anomalous day showing VPEs clea ly g ea e han he
ypical ones a e Ap il 16 h. Simila beha io was ob ained in he o he ESTB Da a Collec ing si es o such a day.
ESTB upda e
Sho e when
excluding
sa e lli e s
Sho e when
excluding
sa e lli e s
Figu e 7. Accu acy: 95 h pe cen ile o Ve ical Posi ion E o o UPC1. The ho izon al axis is he Day o he Yea (DOY) anging on Janua y
10 h (DOY=10) o Augus 28 h (DOY=240) o 2002. The e ical axis is he 95 h pe cen ile in me e s. The solid line indica es he esul s wi hou
excluding sa elli es. The diamonds indica e he alues when excluding a single sa elli e.
The esul s o he UPC2 si e a e compa ed wi h hose o UPC1 (wi h 16 Km o baseline) in igu e 8, showing
simila pe o mances, as expec ed. No ice also he abno mal beha io o DOY 115 in bo h si es as such has been
p e iously men ioned.
ESTB upda e
UPC1 95 h Pe cen ile VPE
ESTB upda e
UPC2 95 h Pe cen ile VPE
DOY 115
DOY 115
DOY 115
DOY 115
DOY 115
DOY 115
Figu e 8. The same plo such as in p e ious igu e 7, bu compa ing he esul s o UPC1 ( op) and UPC2 (bo om).
A compa ison o he 95 h pe cen ile o UPC1 si e be o e and a e he ESTB upda e is gi en in igu e 8 in
unc ion o he PRN o deselec ed sa elli e. As i is shown, a e Ap il 16 h, all sa elli es ha e simila impac o e
he 95 h pe cen ile when hey a e deselec ed om he na iga ion solu ion (excluding he DOY 115). This is no he
case be o e such upda e, whe e he accu acy is clea ly deg aded when excluding some o he sa elli es PRN20,
PRN31, o PRN07, among o he s.
UPC1 be o e Ap il 16 h 2002
Remo ing DOY 115
Remo ing DOY 115
Remo ing DOY 115
Remo ing DOY 115
UPC1 a e Ap il 16 h 2002
Figu e 9. Accu acy: 95 h pe cen ile when losing a single sa elli e in UPC1 si e o all o he da a collec ing days om Janua y 10 h o Ap il 11 h
(le ) and om Ap il 18 h o Augus 28 h, bu excluding he DOY 115, ( igh ). The ho izon al axis is he PRN numbe o he excluded sa elli e in
he compu a ion. The “0” s ands o he solu ion compu ed wi hou excluding sa elli es. The e ical axis is he 95 h VPE pe cen ile in me e s.
3.3 APV-II A ailabili y
The esul s o he APV-II a ailabili y in unc ion o he day o he yea a e gi en in igu e 10. In gene al i appea s
a dec easing o he a ailabili y when deselec ing he sa elli es, which is p oduced by he inc easing o he
p o ec ion le els due o he wo s geome y. Bu , on he con a y han in p e ious cases, he a ailabili y esul s o
he pe iods be o e and a e he ESTB upda ing a e mo e simila . Ne e heless, a e he ESTB upda e, he APV-II
a ailabili y seem o be less sensible o losing a single sa elli e, and could dec ease up o 5%.
ESTB upda e
Figu e 10. APV-II A ailabili y o UPC1. The ho izon al axis is he Day o he Yea (DOY) anging on Janua y 10 h (DOY=10) o Augus 28 h
(DOY=240) o 2002. The e ical axis is he pe cen age o epochs inside he No mal Ope a ion Region. The solid line indica es he esul s
wi hou excluding sa elli es. The diamonds indica e he alues when excluding a single sa elli e.
The esul s o he UPC2 si e a e compa ed wi h hose o UPC1 (16 Km o baseline) gi en in igu e 11. The same
beha io is ob ained in bo h si es, as expec ed.
ESTB upda e
ESTB upda e
UPC1 APV-II A ailabili y
UPC2 APV-II A ailabili y
Figu e 11. The same plo such as in p e ious igu e 10, bu compa ing he esul s o UPC1 ( op) and UPC2 (bo om).