You are on page 1of 5

58 www.rfdesign.

com May 2002


M
i l l i meter -wav e (MMW), al s o s ometi mes
l oosel y referred to as mi crowave, data l i nks
ar e becomi ng an i ncr easi ngl y common means of
pr ovi di ng car r i er cl ass ser vi ce i n dense ur ban
areas. I f a l i ne-of-si ght exi sts between two poi nts
and r oof r i ghts and l i censi ng can be secur ed, a
MMW l i nk provi des a cost-effecti ve al ternati ve to
the sl ow and costl y pr ocess of l ayi ng fi ber . The
carri er has two key factors that the l i nk must sat-
i sfy i f the l i nk i s goi ng to provi de fi ber-l i ke servi ce:
avai l abi l i ty and bi t error rate (BER). I t i s the job
of the radi o and the l i nk desi gners to ensure that
these requi rements are met.
System parameters
A scan of Web si tes of some of the equi pment mak-
ers provi des some i nsi ght i nto the vari ous types of
specmanshi p. Some manufacturers provi de a sensi -
ti vi ty that i s based on a resi dual (after correcti on)
BER of 10
6
or 10
8
. For the data rates i nvol ved (100
Mb/s or greater), these ki nds of error rates resul t i n
uncor r ected er r or s occur r i ng onl y seconds apar t.
That i s hardl y carri er-grade servi ce. Thus, the fi rst
order of busi ness i s to determi ne the recei ver thresh-
ol d where the BER i s at an acceptabl e l evel . Thi s
val ue may not be avai l abl e from the manufacturer
and may have to be esti mated. Dependi ng on the
l evel of codi ng i nvol ved, an i ncr ease of 2 to 3 dB
shoul d move the system i nto the cl ear oper ati ng
regi on where the BER wi l l be 10 to 12 or better.
Once a usable threshold has been determined, the
next step i s to determi ne the maxi mum transmi t
power. Again, the user must be aware of what is actu-
al l y bei ng speci fi ed. The number of i nterest i s the
maximum value of the average transmit power at the
input to the antenna. The manufacturer may specify
the maximum peak envelope power emitted from the
transmi tter or del i vered to the antenna. Most sys-
tems using digital modulation and Nyquist filtering
will require anywhere from 4 to 10 dB of output back-
off (OBO) for linear operation
1
. Using the peak output
power results in overstating the transmit capabilities
of the radio, thereby overestimating the link perfor-
mance. I f the maxi mum average transmi t power i s
not given, the user may want to assume a reduction
from the peak power between 4 and 8 dB, to include
OBO and any fi l ter or wavegui de i nser ti on l oss
between the transmitter and the antenna.
The remai ni ng hardware parameter to address i s
th e an ten n a gai n . For th e 38 GHz ban d, th e
Federal Communi cati ons Commi ssi on (FCC) pro-
vi des a mi ni mum gai n of 38 dB, whi ch i s general l y
met wi th a 1-foot di ameter or l arger antenna
2
. For
the 28 GHz l ocal mul ti poi nt di str i buti on ser vi ce
(LMDS) band, there i s no mi ni mum gai n requi re-
ment onl y a maxi mum beamwi dth requi rement
so there i s a greater l i kel i hood of a data sheet
contai ni ng a marketi ng or typi cal gai n number. I n
addi ti on to removi ng any fl uff i n the antenna gai n
numbers, the l i nk desi gner may al so want to reduce
the antenna gai n by 1 dB to account for l ess-than-
perfect al i gnment.
Once a sati sfactory val ue for the antenna gai n i s
deter mi ned, the hop di stance and avai l abi l i ty
anal ysi s can be performed. The system gai n can be
defi ned as the maxi mum average transmi t power
mi nus the recei ver sensi ti vi ty expressed i n dB or
dBm. For dB, the formul a i s:
(1)
The l i nk gai n, whi ch may al so be referred to as sys-
tem gai n, i s defi ned as:
(2)
For the remai nder of thi s arti cl e, the term system
gain wi l l refer to Equati on 2.
G G P R L S T thresh ( ) max + 2 GRxAnt dB
G P R s T thresh max dB
Figure 1. Five-nines hop distance chart, 39 GHz, ITU rain model.
Tx/Rx
Performance
prediction for fixed
microwave data links
By J ohn S. Seybold
RF wireless links are becoming a popular
alternative to fiber. Getting solid numbers that
will permit estimation of link performance
with a high degree of certainty is a mandate.
60 www.rfdesign.com May 2002
Thi s system gai n deter mi nes how
the radi os, and thus the l i nk, wi l l per-
form duri ng a rai n fade. Dependi ng on
th e en v i r on men t th at th e l i n k i s
depl oyed i n, the desi gner may el ect to
reduce the system gai n by another dB
to pr ovi de an i nter fer ence mar gi n. A
common defi ni ti on of thr eshol d-to-
i nterference rati o uses a 1 dB degrada-
ti on of the system thr eshol d as the
poi nt where i nterference i s consi dered
a probl em.
Avai l abi l i ty i s the percentage of ti me
that the l i nk wi l l be oper ati onal . For
wi rel ess l i nks, thi s i s general l y consi d-
ered to be excl usi vel y due to rai n out-
ages and does not usual l y budget for
equi pment fai l ures. Thi s i s a shortcom-
i ng of the rai n avai l abi l i ty anal ysi s for
wi r el ess l i nks because wi r el ess l i nks
actual l y have mor e equi pment i n the
cri ti cal data path than a fi ber l i nk. For
the r emai nder of thi s wor k, the ter m
av ai l abi l i ty wi l l be tak en to appl y
excl usi vel y to l i nk outages due to rai n,
wi th the understandi ng that an addi -
ti onal anal ysi s may be r equi r ed to
account for avai l abi l i ty l i mi tati ons due
to equi pment fai l ures.
Note the concept of the two or three-
si gma desi gn as i t appl i es to r adi o
har dwar e. Whi l e thi s i s a common
method for system desi gn, the concept
must be addressed for the appl i cati on
to wi rel ess data l i nks. I f a system has a
thr ee-si gma system gai n, that means
that i t i s onl y expected to meet the
speci fi cati on 99.87% of the ti me. The
pr obabi l i ty of havi ng a system that
does not meet the speci fi cati on i s not
negl i gi bl e when compared to a four- or
fi ve-ni nes l i nk avai l abi l i ty.
Ther efor e, i t i s i mpor tant that the
key system parameters be known wi th
better than a thr ee-si gma cer tai nty.
Thi s i s best achi eved by careful accep-
tance testi ng ei ther duri ng producti on
or i mmedi atel y pri or to depl oyment. I f
the uncer tai nty i n the system gai n i s
not addressed, then the avai l abi l i ty of
the l i nk i s l i kel y to be l ess than the
desi gn tar get even i f the r ai n model s
are correctl y appl i ed.
I t shoul d be stressed that havi ng l ess
than the r equi r ed gai n r esul ts i n a
reducti on i n avai l abi l i ty. Thus, the l i nk
wi l l oper ate pr oper l y the major i ty of
the ti me, but i t wi l l fai l dur i ng r ai n
events that i t shoul d be abl e to handl e.
Rain models
Ther e ar e two popul ar r ai n model s
common l y u s ed for MMW RF
l i n k pl an n i n g: Th e I n ter n ati on al
Tel ecommu n i cati on s Un i on (I TU)
model
3
and the Crane model
4
. Each has
a cor r espondi ng set of empi r i cal r ai n
data wi th each model . Wh i l e each
model can be appl i ed usi ng the other
set of rai n data, that procedure i s not
di scussed herei n. Each set of rai n data
di vi des the gl obe i nto rai n regi ons that
characteri ze the rai n condi ti ons.
One recommendati on for use of the
I TU model uses the r ai n r ate, whi ch
occurs 0.01% of the ti me, or the 0.9999
rai n rate. Tabl e 1 shows the 0.9999 or
0.01% rai n rate data for use wi th the
I TU model . The rai n rate data for the
desi r ed r ai n r egi on ar e then used to
compute the path attenuati on usi ng the
expressi on:
(3)
Where: RR = the 0.9999 rai n rate for
the chosen r egi on, i n mm/hr and d i s
the hop di stance i n km.
Furthermore:
(4)
r
d
d

_
,

1
1
0
Atten a RR d r
b
0 01 . dB
Figure 2. Five-nines hop distance chart, 39 GHz, Crane rain model.
Figure 3. Four-nines hop distance chart, 39 GHz, Crane rain model.
62 www.rfdesign.com May 2002
Wi th :
(5)
Equati on 3 provi des an esti mate of
th e path atten u ati on th at i s on l y
exceeded 0.01% of the ti me. The I TU
model al so provi des an adjustment fac-
tor for avai l abi l i ty that i s appl i ed to the
path attenuati on i f avai l abi l i ti es other
than 0.9999 ar e desi r ed. The adjust-
ment factor i s appl i ed to the attenua-
ti on (expressed i n dB) that i s computed
by (3). The expr essi on for the adjust-
ment factor i s:
(6)
Where p i s the desi red outage proba-
bi l i ty expr essed as a per centage (p =
0.01 for 0.9999 av ai l abi l i ty ). Th e
adjustment factor i s val i dated for avai l -
abi l i ti es from 0.99 to 0.99999.
The a and b factor s ar e empi r i cal
val ues that have been tabul ated i n the
references and are the same val ues for
both Crane and I TU model s. The val -
ues are a functi on of the frequency and
pol ari zati on of i nterest. Parameter val -
ues for frequenci es that are not tabul at-
ed can be computed by i nter pol ati on,
usi ng a l ogari thmi c frequency scal e and
l ogar i thmi c scal e for a and a l i near
scal e for b.
The Cr ane model takes a di ffer ent
approach to model i ng rai n attenuati on.
Cr ane does not use an avai l abi l i ty
adjustment factor on the attenuati on,
but r ather uses the r ai n data for a
number of di ffer ent avai l abi l i ti es
some of whi ch ar e shown i n Tabl e 2
al ong wi th the rai n rates. The rai n rate
data for the desi r ed avai l abi l i ty ar e
then used i n Cranes empi ri cal model to
deter mi ne the path attenuati on as a
functi on of range. The Crane model i s
di vi ded i nto two segments, dependi ng
on the actual r ai n r ate i nvol ved. The
fi rst segment i s for di stances between 0
and (d) where:
(7)
I n thi s regi on, the rai n attenuati on i s:
(8)
Where:
(9)
For the other case, (RR)<d<22.5 km,
the attenuati on i s gi ven by:
(10)
Where:
z = b(0.026 0.03l nRR)
(11)
Hop charts and availability
Usi ng the precedi ng expressi ons for
ei ther of the model s, i t i s strai ghtfor-
ward to pl ot the rai n attenuati on for a
gi ven frequency, pol ari zati on and avai l -
abi l i ty as a functi on of hop di stance.
Once the frequency, pol ari zati on, avai l -
abi l i ty and rai n regi on are sel ected, the
a an d b coeffi ci en ts , r ai n r ate an d
Atten aRR
e
y
e e e
z
b
y RR
zd z RR RR

( )
+
( )

( )
( ) . . ( ) ln
1
0 83 0 17

1
]
1
1
1
1
1
y b
RR
RR


( )

1
]
1
1
0 83 0 17 . . ( ) ln

Atten
aRR e
y
d RR
b yd

( )
<
1
dB
where: 0< ( )
RR RR ( ) ( ) 3 8 0 6 . . ln km
Atten
Atten
p
p
0 01
0 12
0 546 0 043
.
.
. . log

+ ( ) ( )
d e
RR
0
0 015
35
.
km
A B C D E F G H J K L M N P
8 12 15 19 22 28 30 32 35 42 60 63 95 145
Table 1. ITU rain rate data (mm/hr) for 0.9999 availability.
Availability A B B1 B2 C D1 D2 D3 E F G H
0.99 0.2 1.2 0.8 1.4 1.8 2.2 3.0 4.6 7.0 0.6 8.4 12.4
0.999 2.5 5.7 4.5 6.8 7.7 10.3 15.1 22.4 36.2 5.3 31.3 66.5
0.9999 9.9 21.1 16.1 25.8 29.5 36.2 46.8 61.6 91.5 22.2 90.2 209.3
099995 13.8 29.2 22.3 35.7 41.4 49.2 62.1 78.7 112 31.9 118 283.4
0.99999 28.1 52.5 42.6 63.8 71.6 86.6 114.1 133.2 176 70.7 197 542.6
Table 2. Crane rain rate data in mm/hr vs. availability.
Figure 4. Five-nines hop distance chart, 28 GHz, Crane rain model.
64 www.rfdesign.com May 2002
adjustment factor (i f any) are fi xed. The
rai n attenuati on then becomes a func-
ti on of the di stance, d, onl y.
Fi gure 1 shows a pl ot of the expected
rai n fade for a 0.99999 rai n vs. hop di s-
tance for several di fferent rai n regi ons.
The col ored curves represent the fi ve-
ni nes rai n attenuati on vs. di stance for
the respecti ve rai n regi ons. The bl ack
system gai n curves represent the avai l -
abl e fade margi n for systems wi th di f-
ferent system gai ns.
The Fr i i s equati on gi ves the fr ee-
space l oss as a functi on of di stance.
(12)
For ter r estr i al l i nks at fr equenci es
bel ow about 55 GHz, gaseous absorp-
ti on can be i gnored unl ess the l i nk di s-
tance i s substanti al . Mul ti path fadi ng
i s usual l y mi ni mal because the l i nks
are stati onary and the antenna beams
are rel ati vel y narrow. Thus, the free-
space l oss provi des a good esti mate of
the path l oss for a poi nt-to-poi nt l i nk.
The pri mary source of outages i s pre-
ci pi tati on. The av ai l abl e r ai n fade
mar gi n for a gi ven l i nk can then be
computed as the system gai n mi nus
the free-space l oss.
By superi mposi ng a pl ot of the sys-
tem gai n mi nus the free-space l oss as
a functi on of range onto the rai n atten-
uati on curves, a hop di stance chart i s
pr oduced. Fi gur e 1 shows a hop di s-
tan ce ch ar t for a 39 GHz, v er ti cal
pol ar i zati on l i nk for thr ee di ffer ent
system gai ns. The poi nt where a sys-
tem gai n curve i ntersects a rai n atten-
uati on curve i s the maxi mum 0.99999
avai l abi l i ty l i nk di stance for that sys-
tem i n the sel ected r ai n r egi on. The
poi nt where they i ntersect i s the range
where the rai n fade margi n of the l i nk
i s equ al to th e 0.99999 r ai n -fade
depth. For exampl e, for a 170 dB sys-
tem i n r ai n r egi on A, the fi ve-ni nes
hop di stance i s 7.6 km.
Fi gur e 2 shows a si mi l ar hop di s-
tance char t usi ng the Cr ane model
and data. Because the rai n regi ons for
Crane and I TU are defi ned di fferentl y,
i t i s not possi bl e to make di rect com-
pari sons between the two model s. I t i s
possi bl e to pi ck a par ti cul ar ci ty or
geogr aphi c r egi on and compar e the
pr edi cted per for mance at that l oca-
ti on. For i nstance, Or l ando i s i n I TU
regi on N and Crane regi on E, so a
170 dB system has an expected fi ve-
ni nes hop di stance of 970 meters usi ng
I TU and 1000 meters usi ng Crane. For
val ues of system gai ns not shown, two
curves can be l i nearl y i nterpol ated.
The hop di stance char t pr ovi des a
gr aphi cal means of deter mi ni ng at
what hop di stance the rai n attenuati on
i s equal to the fade margi n. The i nter-
secti on can be determi ned usi ng i tera-
ti on on a computer. Because the atten-
uati on model mi nus the free-space l oss
i s a transcendental functi on, i t cannot
be sol ved di rectl y. Whi l e not as preci se,
the gr aphi cal sol uti on pr ovi des mor e
i nformati on than a numeri cal sol uti on.
For ex ampl e, i n th e dr i er r egi on s ,
where the rai n attenuati on curve has
l ess sl ope, i t can be seen that a smal l
change i n the system gai n can resul t i n
a substanti al change i n the pr edi cted
h op di s tan ce. Th i s becomes mor e
apparent at l ower avai l abi l i ti es such as
th e fou r -n i n es h op di s tan ce ch ar t
shown i n Fi gure 3.
Fi gu r e 4 s h ows a fi v e-n i n es h op
chart at 28 GHz. At the l ower frequen-
cy, the i mpact of r ai n attenuati on i s
l ess severe than at 39 GHz, so the cor-
respondi ng hop di stances are l onger. I t
i s al so noteworthy that the rai n atten-
uati on i s sl i ghtl y gr eater on hor i zon-
tal l y pol ar i zed si gnal s than on ver ti -
cal l y pol ar i zed si gnal s. The a and b
factor s i n the attenuati on model ar e
di ffer ent for hor i zontal and ver ti cal
pol ar i zati on. Thi s i s a consi der ati on
for dual pol ari zati on systems because
the performance i n rai n wi l l be l i mi ted
by the hori zontal l y pol ari zed si gnal .
A common questi on encountered by
r adi o vendor s i s: My hop di stance i s
l onger than the maxi mum fi ve-ni nes
hop di stance. What wi l l the actual
avai l abi l i ty be? Thi s i s an i mportant
questi on. Fr om an oper ati onal stand-
poi nt, 0.99997 i s not profoundl y di ffer-
ent from 0.99999 avai l abi l i ty. Usi ng the
I TU model (wi th the adjustment factor),
a fami l y of curves showi ng avai l abi l i ty
vs. hop di stance can be generated for a
gi ven system gai n. Fi gur e 5 shows a
sampl e avai l abi l i ty chart for a system
gai n of 170 dB. I t must be remembered
that thi s curve i s onl y val i d for the stat-
ed system gai n. Usi ng such a curve, the
avai l abi l i ty can be esti mated for any
hop di stance once the rai n regi on and
system gai n have been establ i shed.
Summary
Thi s ar ti cl e di scussed how to l ook
past typi cal mar keti ng speci fi cati ons
and get to the sol i d numbers that wi l l
permi t esti mati on of l i nk performance
wi th a hi gh degree of certai nty. Once a
set of sol i d r adi o speci fi cati ons ar e i n
hand, the domi nant l i mi tati on of l i nk
avai l abi l i ty i s the rai n fade. Two popul ar
methods for model i ng rai n fade and how
they rel ate to avai l abi l i ty were exam-
i ned. A hop di stance chart was generat-
ed and di scussed. Fi nal l y, a pl ot of
expected avai l abi l i ty vs. hop di stance for
a gi ven system gai n was presented.
L
d

_
,

20
4
log

dB
Figure 5. Availability vs. hop distance for vertical polarization, 39 GHz.
Continued on page 66
66 www.rfdesign.com May 2002
References
[1] Output Back-Off Requi r ements
for Root Rai sed-Cosi ne Fi l tered Di gi tal
Si gnal s , John S. Seybol d, Ph.D., RF
Desi gn, 2001
[2] Code of Federal Regul ati ons, Ti tl e
47, Vol ume 5, 101.517, Revi sed as of
October 1, 2000
[3] I TU Recommendati on PN837-1,
Rec838, Rec.I TU-R P.530-7
[ 4] El ectr omagn eti c Wav e
Pr opagati on Thr ough Rai n, Rober t
Cr ane, John Wi l ey & Sons, Febr uar y
1996.
About the author
John S. Seybol d recei ved hi s B.S.E.E. from the Uni versi ty of Wi sconsi n i n 1982,
hi s M.S.E.E. from Cal i forni a State Uni versi ty, Ful l erton, i n 1986 and hi s Ph.D.
from the Uni versi ty of Central Fl ori da i n 1995. Seybol d i s an associ ate professor
of el ectri cal engi neeri ng at Fl ori da I nsti tute of Technol ogy i n Mel bourne, FL,
wher e he al so ser ves as the associ ate di r ector of thei r Wi r el ess Center of
Excel l ence. Seybol d hel d a vari ety of posi ti ons i n the i ndustry pri or to joi ni ng the
facul ty at Fl ori da Tech. Most recentl y, he was a seni or RF systems anal yst at
Tri ton Network Systems, where he was responsi bl e for RF network pl anni ng, l i nk
outage predi cti on and frequency re-use anal ysi s. Seybol d has al so worked i n radar
systems, syntheti c aperture radar (SAR) and communi cati ons systems, i ncl udi ng
spread spectrum. He i s a l i censed professi onal engi neer i n the state of Fl ori da, a
seni or member of the I EEE and a member of Phi Kappa Phi and Tau Beta Pi . He
can be contacted at: jseybold@mpinet.net.

You might also like