You are on page 1of 9

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 56, NO.

4, APRIL 2008

1085

Cumulative Fading and Rainfall Distributions for a 2.1 km, 38 GHz, Vertically Polarized, Line-of-Sight Link
Neil Forknall, Roy Cole, and Dave Webb
AbstractBetween February 1, 1997 and January 31, 1998, a 2.1 km vertically polarized 38 GHz communications link was studied in the subtropical city of Brisbane, Australia. According to the current ITU fading prediction method the link would be expected to experience a rainfall intensity of approximately 50 mm/h for 52.6 min during the year, when averaged over a 60 s integration time. This should have caused approximately 20 dB of fading for 52.6 min [1][3]. However the link actually experienced 20 dB of fading for 150.5 min, with fading of 37 dB exceeded for 53.5 min during the year. A rainfall intensity of 84 mm/h was exceeded for 64 min at the receive end, and a rainfall intensity of 72 mm/h was exceeded for 52 min at the transmit end of the hop. Index Terms38 GHz eld trial, availability, backhaul transmission link, fade margin, ITU-R P.530-11 rain attenuation model, measurement results, radio access network, subtropical fading distribution, subtropical rainfall distribution.

I. INTRODUCTION

NETWORK operator uses a number of parameters to predict the quality of the service that will be provided to its customers. One of the parameters is the amount of time in a year that the service is available. Therefore it is important to be able to predict the amount of time, during a typical year, that a backhaul transmission link to a cellular base station will operate at a predened bit error rate. 38 GHz radio links experience rain-induced attenuation, so it is necessary to include a fade margin for this rain-induced fading in the power budget. For a given link, which is free from obstruction, the size of this fade margin depends on the amount of time the link should be available during an average year. II. 38 GHz LINK

The 38 GHz, 2.1 km vertically polarized link had an azimuthal angle, relative to North, of 213 degrees. The transmit power was 11 dBm. At each end a 30 cm parabolic antenna with a radome and collar was used. These antennas had a boresight gain of 39 dBi and a half power beamwidth of 2 degrees. The capacity of the radio was 8 Mbits/s, transmitted using continuous phase modulation within a 7 MHz bandwidth.

Connector losses on the link were approximately 0.5 dB. Relative to a clear sky receive level of 43 dBm a bit error rate of one in a million could be seen on this link after attenuation of 31 dB, while an error rate of one in a thousand could be seen after attenuation of 36 dB. After 41 dB of attenuation the link, which was free from radio interference, showed a bit error rate of eight in a thousand, but after 42 dB of attenuation the link completely failed. These observations were made by counting the parity errors on a service channel, over a 10 s period. A laptop was used to log changes in the received power level, the automatic gain control voltage, the modem temperature, and the 10 s average of the bit error rate. A rain gauge with a 20 cm aperture was installed on the roof at the transmit end, and two 20 cm rain gauges were installed on the roof at the receive end of the hop. These rain gauges, which were located within 20 m of the antennas, continuously monitored the rainfall. Wind speed and direction were also sampled once a minute at the receive end of the link. Ambient temperature, humidity, and historical rainfall records were provided by the Australian Bureau of Meteorology [4]. The transmit antenna was located 50 m above ground, which is approximately 20 m above sea level. The receive antenna was located 10 m above ground, which is approximately 10 m above sea level. Fig. 1 shows that the terrain prole below the path is at between points 0.3 and 1.9 km from the transmitter, with an average terrain height 5 m above sea level. A river crossed the terrain prole, but not at the point of reection, which was 1.7 km from the transmitter. The path was free from obstruction and mature trees helped to prevent ground reections from arriving at the receiving antennas. During the measurement period there was no evidence that the receiver suffered from multipath reception caused by ground reections. Fig. 2 shows the worst case of non rain related variations in the received signal, which occurred on December 3, 1997, when peak to peak variations of 3 dB were recorded. This activity coincided with signicant changes in relative humidity in the Brisbane area, which were recorded by the Australian Bureau of Meteorology [4]. III. SOURCES OF ERROR

Manuscript received May 29, 2006; revised December 27, 2006. D. Forknall and D. Webb are with Leeds Metropolitan University, Leeds LS1 3HE, U.K. (e-mail: peter.forknall@uclmail.net). R. Cole is with University College London, London WC1E 6BT, U.K. Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2008.919158

In the absence of rain-induced attenuation the received power varied between 41 dBm and 44 dBm. Over a 24 h duration the non-faded signal would usually vary between 42 dBm and 43 dBm or between 43 dBm and 44 dBm. During a typical day a higher ambient temperature was associated with the stronger received signal and the lower received signal strength

0018-926X/$25.00 2008 IEEE

1086

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 56, NO. 4, APRIL 2008

Fig. 1. Path prole of the measurement link, with the transmitter on the left.

Fig. 2. Largest non-rain related variations in power level during the eld trial.

Fig. 3. Daily variations in received power and receiver modem temperature.

was associated with the lower ambient temperature. The receiver modem temperature was typically between 12 and 24 degrees Celsius higher than the ambient temperature. The slight differences in the non-faded received power levels were related to the resolution of the system used to measure the received power levels, the temperature of the outdoor radios, and the at-

mospheric humidity. A received power level of 43 dBm was used to represent the non-fade state. Fig. 3 shows the modem temperature and fading activity on December 27, 1997, which included one rainfall event between 1500 and 1600. During rainfall the modem temperature dropped due to increased cloud cover blocking solar radiation and a re-

FORKNALL et al.: CUMULATIVE FADING AND RAINFALL DISTRIBUTIONS FOR A LINE-OF-SIGHT LINK

1087

Fig. 4. Measuring logged receive power at different ambient temperatures.

Fig. 5. Denition of maximum power during fading activity (the minimum fade).

duction in ambient temperature resulting from the rainfall. The rain also helped to mix the atmosphere. Scintillation may have inuenced the receive power levels, which typically vary by 1 dB in the absence of rainfall. However, the measurements were intended to quantify rainfall-induced fading, which varied over seconds, rather than scintillations which vary more rapidly. Therefore, the changes in received power level were recorded with a resolution of 1 dB, which is adequate for quantifying rainfall-induced fading but unsuitable for quantifying scintillations. During heavy rainfall, scintillation should not have signicantly inuenced the results at 38 GHz. No attempt was made to identify or remove any contribution from scintillation during heavy rainfall. Since variations in the modem temperature can inuence the logged receive power level the entire fade margin was swept in 1 dB steps, at 4 different ambient temperatures, and the difference between the measured and the logged receive power level was recorded. Fig. 4 shows the method used in the laboratory to measure the inuence of ambient temperature on the measured power level. The ambient temperatures at which these measurements were performed were 15, 20, 26, and 30 degrees Celsius. This resulted in an envelope, which dened the area of uncertainty at ambient temperatures between 15 and 30 degrees Celsius, in the absence of solar radiation. This paper uses the maximum power, corresponding to the minimum fade, when compiling fading statistics. Fig. 5 shows the maximum power, or minimum fade, range for ambient temperatures between 15 and 30 degrees Celsius. The rainfall was integrated over 60 s. During the year rainfall intensity of 60 mm/h, or more, was observed on 20 days

at the receive end of the hop and on 17 days at the transmit end of the hop. Fading greater than 15 dB was observed on 20 days. The heaviest 0.01%, or 52.6 min, of fading during the year was caused by seven storms, which occurred on May 30, 1997, September 24, 1997, October 7, 1997, November 30, 1997, December 27, 1997, plus January 28, 1998 and January 31, 1998. When examining data from the 20 days during the year that contained fades greater than 15 dB it was found that the minimum daily temperature was 11.3 degrees Celsius on May 30, 1997. Other days containing both fades greater than 15 dB and minimum daily ambient temperatures below 15 degrees Celsius included, June 21, 1997 with 12.5 degrees Celsius and September 24, 1997 with 11.5 degrees Celsius. From these 20 days the maximum ambient daily temperature exceeded 30 degrees Celsius on January 16, 1998 with 31.5 degrees Celsius and on January 28, 1998 with 31.8 degrees Celsius. The wind speed measured at the receive end could exceed 5 m/s during heavy rain-induced fading activity. This could have introduced air turbulence above the rain gauge, which might have reduced the collection of rain drops, especially the smaller drops. Any turbulence-induced measurement error would be related to the vertical wind speed around a rain gauge. It was noted that the wind speed was much higher at the transmit end of the link, which was also subject to sudden very strong gusts of wind. The wind speed at the transmit end was not measured, but the wind speed at a height of 50 m above the ground may well have inuenced rainfall collection at the transmitter by introducing turbulence around the building and above the rain gauge. Although the focus of this paper is on the storms that produce the heaviest 0.01% of annual rainfall it is still interesting

1088

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 56, NO. 4, APRIL 2008

Fig. 6. Rain collected between February 1996 and January 1999.

Fig. 7. Annual rainfall collected between 1860 & 1993, with a 10 year rolling average.

to compare the total rainfall at each end of the link. The cumulative rainfall collected at the transmit end and receive end of the link show that every month less rain was collected at the transmit end. This can be seen in Fig. 6, which also shows the cumulative monthly rainfall collected between January 1, 1996 and February 28, 1999 at Brisbane Airport, 10 km from the hop. The data from Brisbane Airport was provided by the Australian Bureau of Meteorology [4]. Between 0000 on February 1, 1997 and 1800 on January 31, 1998 at least 0.2 mm of rain fell at the receive end of the hop on 133 days, while at the transmit end 0.2 mm, or more, of rainfall was recorded on 104 days. During this measurement period a total of 800.4 mm of rain was collected at the transmit end, with a total of 979.6 mm of rain collected at the receive end. From April 1997 to July 1998, El Nino was present in the Equatorial Pacic. To place the inuence of El Nino into some perspective the total recorded rainfall measured at each end of the 38 GHz hop has been compared with historical rainfall data provided by the Australian Bureau of Meteorology [4]. Fig. 7 provides this comparison with the total annual rainfall collected over 134 years, at a site 2 km from the hop. A 10 year rolling average trend line has been added to the historical data, which shows that Brisbane was slightly dryer than an average year during the eld trial. Approximately 20% of the water collected in the rain gauge funnel spills over the 5 ml tipping bucket mechanism, but this

is accounted for during the calibration process, where each tip of the mechanism represents 0.2 mm of rain collected over the rain gauge area. The rain gauges were calibrated in December 1996 and February 1998 for rainfall rates up to 100 mm/h, integrated over one minute. A gauze prevented material collected in the funnel from reaching the bucket and inuencing the balance of the gauge. Silt did not collect in the bucket. The measurement error for a known volume of tap water was less than 3%. Between April 1, 1997 and January 31, 1998 a second identical rain gauge was installed at the receive end, within 20 m of both the radio receiver and the original rain gauge. The collected rain was almost identical for each storm and over the entire 10 month observation period. Fig. 8 shows the rainfall recorded by both rain gauges at the receive end of the link during a rainfall event on December 27, 1997. Fading recorded between May 2 and 6,1997 was lost due to data corruption, however no signicant rainfall events, of 60 mm/h or greater, were recorded at either end of the hop during this period. Between June 22 and August 6, 1997 the radio at the transmit end of the link was temporarily removed. During this period peak rainfall of 60 mm/h was recorded on July 7 at the receive end of the link. Between October 8 and November 18, 1997 the radio at the transmit end of the link was again temporarily removed. During this period peaks of 84 mm/h and 60 mm/h were recorded at the transmit end on November 16 and 17, 1997, respectively. At the

FORKNALL et al.: CUMULATIVE FADING AND RAINFALL DISTRIBUTIONS FOR A LINE-OF-SIGHT LINK

1089

Fig. 8. Rainfall collected during a storm by both rain gauges located at the receiver.

Fig. 9. 60 s rainfall intensity of 60 mm/h or more (four events have no fading data).

Fig. 10. Receive end cumulative rainfall distributions for each month, and the year.

receive end peaks of 60 mm/h were seen on November 17 and 18, 1997. Fig. 9 shows signicant rainfall events that occurred, and also shows when the transmit end outdoor radio unit had been temporarily removed from the link. During commissioning the antennas at each end of the hop were carefully aligned by searching for the highest automatic gain control voltage in the horizontal and vertical planes, during

normal atmospheric conditions. The antennas remained xed throughout the eld trial. When new, rain forms beads of water on the radome, causing additional attenuation which is very low. Once dirty, the rain can form a lm on the radome during heavy rainfall. Tests carried out during measurements revealed that if two spray guns were used, simultaneously at each end of a hop on a warm dry day,

1090

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 56, NO. 4, APRIL 2008

Fig. 11. Transmit end cumulative rainfall distributions for each month, and the year.

Fig. 12. Cumulative fade distributions for each month, and the year.

the additional attenuation depended on the age of the radomes. With 30 cm antennas at each end the attenuation caused by new radomes was so low that it could not be measured, but after one year the maximum attenuation was 3 dB. Therefore, the maximum attenuation caused by heavy rainfall would not exceed 3 dB on aged and dirty 30 cm radomes, even when the wind was simultaneously blowing rain onto the radomes at both ends of the hop. Although the wind speed and direction were sampled once a minute at the receive end of the hop it is very difcult to compensate for the local contribution the wind may have played in driving the rain onto the radomes at each end of the hop, at one moment in time. To quantify this error it is expected that the average total dirty radome attenuation for the hop would probably have grown from near zero to less than 1 dB at the end of the one year measurement period. Therefore no attempt was made to correct for wet radomes, and no attempt was made to reduce the radome attenuation by periodically washing or polishing the radomes. IV. THE CUMULATIVE RAINFALL DISTRIBUTIONS Fig. 10 shows that at the receive end the heaviest 64 min of rainfall exceeded an intensity of 84 mm/h, when integrated over

one minute. The heaviest (0.01%) 52.6 min of rainfall during the year is expected to have exceeded 91 mm/h. It can be seen from the graph that a large proportion of the heaviest rainfall fell in January 1998. Fig. 11 shows that at the transmit end of the link the heaviest 52 min of rainfall, integrated over one minute, exceeded an intensity of 72 mm/h. It should be noted that January 1998 had a signicant amount of heavy rainfall, relative to the other months. V. CUMULATIVE FADE DISTRIBUTION Between February 1, 1997 and January 31, 1998, fading of 15 dB was exceeded for 3 h and 53 min, and fading of 37 dB was exceeded for 53.5 min. Fig. 12 shows the monthly cumulative fading distributions. It can be seen that the fading in January 1998 was signicantly heavier than the fading in the other months. Fig. 13 shows the cumulative distributions for fading and the rainfall at each end of the hop, plus the cumulative distribution for the average rainfall collected at both ends of the hop. Fig. 14 shows the data recorded during a storm on January 28 1998, which was responsible for the heaviest rainfall and the

FORKNALL et al.: CUMULATIVE FADING AND RAINFALL DISTRIBUTIONS FOR A LINE-OF-SIGHT LINK

1091

Fig. 13. Cumulative fade distribution and cumulative rainfall distributions.

Fig. 14. Most signicant rainfall and fading event recorded during the eld trial.

TABLE I POWER LAW COEFFICIENTS FOR 38 GHz, FOUND IN ITU-R P.838-3 [3]

however Fig. 14 shows the one minute average of rainfall intensity at rain gauge points located close to the antennas, at each end of the link. VI. ITU RAIN ATTENUATION MODEL The International Telecommunications Union outlines a method for predicting the long-term unavailability caused by rainfall, in ITU-R P.530-11 [1]. This method uses long-term rainfall intensity data, where intensity is averaged over one minute, to nd the rainfall intensity that is exceeded for 0.01% of the time. This rainfall intensity value is then used to predict the long-term attenuation expected for 0.01% of the time, at a particular frequency, polarization and hop length. The rst step in applying this method is to obtain the rainfall intensity for the area of interest, integrated over one minute, which is exceeded for 0.01% of the time. This parameter is termed 0.01. If reliable long-term local data is not available an estimation of

Fig. 15. Rainfall intensity map for Australia, source ITU-R P.837-4.

most signicant fading observed during the eld trial. The rainfall responsible for the fading has a three dimensional structure,

1092

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 56, NO. 4, APRIL 2008

TABLE II COMPARISON OF MEASURED & PREDICTED ATTENUATION FOR DIFFERENT RAINFALL RATES

this value can be obtained from the global rainfall charts found in ITU-R P.837-4 [2]. Fig. 15 shows that Brisbane is located in a 40 mm/h to 50 mm/h region. is known the specic attenuation , in decibels Once per kilometer, caused by this rainfall can be predicted by using a power law approximation provided in ITU-R P.838-3 [3]

Table I provides the coefcients, and , found in ITU-R P.838-3 [3]. These values, which are frequency and polarization dependent, are considered by the ITU to be reliable. The specic attenuation obtained is valid for a uniform rain medium. To take account of the nonuniform nature of rain a reduction factor , is used, which takes the hop length and rainfall into consideration intensity at

Where

is the hop length in kilometers and

is

When is less than 100 mm/h, should be calculated using the real rainfall value, which is exceeded for 0.01% of the is greater than 100 mm/h the value 100 should time. When be used for , in the equation, as this limits the effect of the reduction factor with increasing rainfall intensity. exceeded for 0.01% of the time The path attenuation can therefore be calculated from the expression

of the time is integrated over a one-minute period. However, rainfall measurements with a one-minute integration time only started to be collected in this area in 1996. Long-term gures have been predicted from records with a longer integration period. Using the specic attenuation values based on ITU-R P.838-3 [3] for a vertically polarized 38 GHz signal, and a reduction factor for spatial averaging of rainfall on a 2.1 km hop, based on the method outlined in ITU-R P.530-11 [1], the link attenuation gure exceeded for 0.01% of the time can be estimated. The key parameters used in the ITU prediction method are shown in Table II. If the long-term average rainfall rate exceeded for 0.01% of the time was 91 mm/h the parameters for effective path length (1.7) and specic attenuation (18.2) on this link would be used to predict a long term average fading value of (1.7 18.2) 30.95 dB, exceeded for 0.01% of the time. This is approximately 6 dB less than the measured fading value of 37 dB, which was exceeded for 53.5 min during the one-year eld trial. Table II shows that according to the ITU method the long-term predicted attenuation for 0.01% of the time, based on 0.01% rainfall rates of 40, 50, 60, 72, 91, and 117 mm/h, would be expected to exceed approximately 17, 20, 23, 27, 31 and 37 dB, respectively. During the one-year observation period measured attenuation values of 17, 20, 23, 27, 31 and 37 dB were exceeded for 203.5, 150.5, 118.9, 92.8, 72.9 and 53.5 min. It must be borne in mind that the fading was not recorded over the whole year. If the fading had been recorded over the whole year the fading gures might have been slightly larger. The ITU recommendations are intended to provide a longterm average fading gure. The data from the eld trial shows what actually happened over a relatively short, one-year period. The large difference between the ITU fading prediction and the measured data emphasizes the need for more measurement data in this area. A more detailed description of this eld trial can be obtained from University College London [5]. ACKNOWLEDGMENT The authors would especially like to thank S. Dimitrios (formerly with Cable & Wireless Optus) for his help commissioning and decommissioning the link in December 1996 and January 1998, and for his diligent collection of the raw measurement data during the eld trial. Thanks are also extended to J. Henriksson (Nokia) for his valuable suggestions during the eld trial, and to X. Gui (Massey University) for his recent comments.

The prediction procedure outlined above is considered to be valid in all parts of the world. ITU-R P.530-11 [1] provides a method for predicting fading value has been for other percentages of time, once the obtained. VII. SUMMARY According to ITU-R P.837-4 [2] Brisbane is located in a 40 mm/h to 50 mm/h region, assuming rainfall intensity for 0.01%

FORKNALL et al.: CUMULATIVE FADING AND RAINFALL DISTRIBUTIONS FOR A LINE-OF-SIGHT LINK

1093

REFERENCES [1] Recommendation ITU-R P.530-11, 2005, Propagation Data and Prediction Methods Required for the Design of Terrestrial Line-of-Sight Systems. [2] Recommendation ITU-R P.837-4, 2003, Characteristics of Precipitation for Propagation Modelling. [3] Recommendation ITU-T P.838-3, 2005, Specic Attenuation Model for Rain for Use in Prediction Methods. [4] Queensland Climate Services Centre Australian Bureau of Meteorology (BOM), 295 Ann Street. Brisbane, Queensland, Australia. [5] N. Forknall, Fixed millimetric and microwave network planning in a cellular environment, M.Phil. thesis, University College London, London, U.K., Dec. 1999.
Neil Forknall received the M.Sc. degree in telecommunications and data networks from Leeds Metropolitan University, Leeds, U.K., in 1995 and the M.Phil. degree in electronic and electrical engineering from the University of London (UCL), London, U.K., in 2000. He is currently conducting UHF radio propagation research at Leeds Metropolitan University. He was previously employed by Nokia for 18 years, where he worked with wireless systems. Mr. Forknall is a registered Chartered Engineer in the U.K. He is a member of the Institution of Engineering and Technology (IET), London, U.K. and the Institute of Physics, London, U.K.

Roy Cole received the physics degree from University College London (UCL), London, U.K., and the Ph.D. degree in electron beam spectroscopy from Birkbeck College London, in 1954 and 1969, respectively. In 1954, he joined EMI Electronics, Ltd., where he was involved with the design and development of microwave ferrite devices and components and radar systems. In 1962, he joined the Department of Physics, North London Polytechnic, London, as a Lecturer. In 1965, he joined the Department of Electronic and Electrical Engineering, University College London, where he is currently Emeritus Professor of Electrical Engineering. His research interests include elds of millimeter-wave propagation, where he has worked at 36, 55, and 100 GHz, atmospheric propagation at infrared frequencies, atmospheric measurement, and the use of millimeter wavelengths for mobile radio.

Dave Webb received the Physics Degree and the D.Phil. degree in space physics from the University of York, York, U.K., in 1975. Between 1975 and 1979, he worked as a Postdoctoral Research Fellow at Bell Laboratories and at the University of York, and was a Senior Scientic Ofcer at the Ministry of Defence, London, U.K. In 1979, he joined Leeds Polytechnic (now Leeds Metropolitan University), Leeds, U.K., where he is a Professor of engineering and Director of the Praxis Centre. His research interests include mobile technology and disaster relief. Prof. Webb is a Fellow of the Royal Astronomical Society.

You might also like