EXECUTIVE SUMMARY There are many factors to consider when engineering access and short-haul millimeterwave radio paths in relation to rain fading and traffic outages. Long-established reference models, sophisticated software and high-speed computational devices can help narrow the decision-making window for the microwave transmission engineer. However, much also depends on the end-use applications and overall objective for any particular microwave network. Microwave transmission engineers have to use their experience and professional judgment in determining how to best use these rain fading and outage tools for their specific networking projects. This white paper examines some of the classic models and pertinent mission questions for microwave radio network rain fading and outages. AUGUST 2011
TABLE OF CONTENTS EXECUTIVE SUMMARY... 1... 3 EFFECT OF SNOW AND ICE ON ANTENNAS AND RADOMES... 3 PREDICTING RAIN OUTAGES... 3 RAIN STATS: LITTLE CORRELATION TO ATTENUATION... 3 LESS SUSCEPTIBILITY OF LOWER FREQUENCIES TO RAINFALL ATTENUTATION... 4 DEPLOYMENT LESSONS LEARNED... 5 RAIN CELL PASSAGE... 5 RAIN AVAILABILITY OBJECTIVES... 7 RELIABILITY OBJECTIVES... 7 PATHS AS FAILSAFE AS POSSIBLE... 8 PERFORMANCE OBJECTIVES IN LONG-HAUL VS. SHORT-HAUL SYSTEMS... 8 NOTE... 8 APPENDIX... 9 CRANE S RAINFALL OUTAGE MODEL... 9 P.530 RAINFALL OUTAGE CALCULATION... 9 2 AVIAT NETWORKS AUGUST 2011
Rain attenuation - rain fading - at the higher microwave frequencies ( millimeterwave bands) has been under intense study and investigation for more than 60 years. Much is known about the qualitative aspects, but the problems faced by microwave transmission engineers who make quantitative estimates of the probability distribution of the rainfall attenuation for a given frequency band as a function of path length and geographic area remains a most interesting challenge, albeit now greatly assisted by computer rain models. To estimate this probability of rain (also wet snow - dry snow and ice crystals do not impede microwave propagation) outage distribution, instantaneous rainfall data is needed. The available rainfall data is usually in the form of a statistical description of the amount of rain that falls at a given measurement point over various time periods. Rain-induced attenuation along a given path at a given instant in time is a function of the integrated effect of the rainfall existing at all points along the path and is affected not only by the total amount of water in the path at that instant but also by its distribution along the path in volume and drop size. For heavy rain rates, the instantaneous distribution of volume and drop size along the path is highly variable and difficult to predict with any accuracy from the rainfall data generally available. EFFECT OF SNOW AND ICE ON ANTENNAS AND RADOMES Snow and glaze ice accumulation on antennas and spherical or pyramidal, i.e. other than planar, radomes that cause signal attenuation and beam deflection can have a significant, if not catastrophic, affect on microwave link performance and uptime not amenable to mathematical predictions and modeling. This effect of snow and ice accumulations on microwave antennas and radomes is not discussed in this paper as it is complex and subject to the widely varied geoclimatic conditions of each individual microwave site. Rainfall fading is independent of TDM or Ethernet transmission considerations. Rain fading causes longterm traffic outages that affect TDM and IP radios traffic equally; thus error objectives for TDM links could equally apply to IP. PREDICTING RAIN OUTAGES One of the earliest and most comprehensive attempts at developing a workable prediction method was carried out by Bell Laboratories in the 1950s and was described in Hathaway and Evans in Radio Attenuation at 11 GHz and Some Implications Affecting Radio-Relay Systems Engineering. 1 In their paper, Hathaway and Evans developed a method of predicting annual outages for microwave paths operating in the 11 GHz common carrier band, as a function of path length, fade margin and geographical area within the contiguous United States. This study has proved to be a worthwhile prediction tool, and when used with recognition of its limitations, is still one of the best references available for microwave engineers working within the United States. Additional studies have been conducted in Europe and Asia. The combined information has been reviewed and published by the ITU-R in several Recommendations, e.g. ITU-R Rec. P.837, P.838, and P.839. Modern mathematical models for predicting rain outage in microwave hops have evolved from these beginnings to the user-friendly Crane 2 oritu-r P.530 mathematiocal models, which are described later, now resident in such microwave path design programs as Aviat Networks Starlink and Pathloss. RAIN STATS: LITTLE CORRELATION TO ATTENUATION The total annual rainfall in an area has almost no relation to the rain attenuation for the area. Within the US, the Northwest states, for example, have the greatest number of rainy days annually, but very few thunderstorm events. But they are characterized by long periods of steady rain of relatively low intensity at any given time that even provides high availability (low rain outage) even in 38 GHz millimeterwave hops. 1 S.D. Hathaway and H.W. Evans, "Radio Attenuation at 11 GHz and Some Implications Affecting Relay System Engineering," BSTJ, January 1959. 2 Robert F. Crane, Prediction of Attenuation by Rain, IEEE Transactions on Communications, September 1980. 3 AVIAT NETWORKS AUGUST 2011
Other areas of the country with much lower annual rates experience types of rainfall such as those accompanying thunderstorms and frontal squalls that produce short duration rains of extreme intensity, and it is the incidence of rain storms of this type which determines the rain attenuation characteristics of an area. It is interesting to note that the first deployment of Bell s 11 GHz TJ radio links was in the Pacific Northwest. Outage: In TDM radio networks, an outage is defined as 1 error per 1,000 bits transmitted (10-3 bit-error rate) or a loss of frame synchronization. With the ARQ Even rain statistics for a day or an hour have little relationship to the excess path attenuation. A day with only a fraction of an inch of rainfall may have a path outage due to a short period of extremely high localized rain cell intensity, while another day with several inches of rainfall may experience little or no path attenuation because rain is spread over a long time period or the high intensity rain cell misses the hop. LESS SUSCEPTIBILITY OF LOWER FREQUENCIES TO RAINFALL ATTENUTATION The most common reason for the strong preference for lower frequencies for even short-haul routes is the susceptibility of frequencies above 10 GHz to rainfall attenuation. Although the effect is present to some degree at lower frequencies, it increases rapidly with frequency. For example, as seen below (Figure 1), rainfall intensity causing only a few db of attenuation at lower frequencies could be sufficient to cause a long-term path outage at 18 GHz. retransmission of lost packets in LANs, there is no consensus definition for outage in IP radio links. Figure 1. An increase in fade margin reduces the number of rain outage events and their durations only minimally frequency band, polarization and path length have a much greater effect on rain outage. Although fades caused by rainfall are occasionally observed at lower frequencies (10-20 db fades at 6 GHz have even been recorded in tropical regions), this type of fade generally causes outages only on paths above 10 GHz. Outages are usually caused by a blockage of the path by the passage of rain cells (thunderstorms, etc.), averaging 3-5 miles in diameter and perhaps 5-15 minutes in duration. Rain fades exhibit fairly slow, erratic level changes, with rapid path failure as the cell intercepts the path. The fades are not selective, i.e. all paths and frequencies in both directions of a hop are affected simultaneously. Because rain drops tend to flatten because of air resistance while falling, vertical polarization is significantly less susceptible to rainfall attenuation than horizontal in all millimeterwave bands. As seen in Figure 1 above, increased fade margin is typically of minimal help in rain fading; but margins as high as 45 to 60 db have been used in some highly vulnerable links for increased hop availability ( uptime ). Increasing the path s fade margin by shortening path lengths or increasing antenna sizes are the most readily available tools for reducing the per-hop outage in a given area, albeit less effective if tandem hops are deployed. Route diversity (ring-protection) or a lower frequency band is highly effective in reducing traffic. 4 AVIAT NETWORKS AUGUST 2011
Crane s mathematical models are used to calculate rain fading and path availability from either Crane or ITU rain rate tables. Most path planning tools such as Starlink use only the Crane rain model, while others also have the P.530 model which produces similar but not identical results. DEPLOYMENT LESSONS LEARNED The following are a few basic considerations in deploying millimeterwave microwave hops: Rain outage approximately doubles in each higher millimeterwave band, e.g. 18 to 23 GHz Or, increased rain outage at 23 GHz may result in its path length lowered 50% as compared to an 18 GHz path Rain outage is directly proportional to path length assuming a constant fade margin for each hop Rain outage increases X2 to X3 for H-pol millimeterwave hops as compared to V-pol H-pol hops require ~6 db more fade margin than V-pol hops for equal rain outage Rain outage in tandem-connected short hops are the same as for a single long hop if they have the same fade margin Rain outage durations are typically 5-15 min in duration each. A Pathloss, Starlink, etc., path calculation that shows small annual rain outage, e.g. 1 min/year, actually equates (in this example) to one 10 min outage in 10/1 = 10 years Rain cells typically travel east to west, thus causing fewer but longer duration outages to East-to-West millimeterwave hops than to North-to-South hops. Total rain outage to both is the same Traffic is 100% protected from subscriber disconnect caused by a long-term rain outage in a millimeterwave hop with optimally configured ring ( route diversity ) protection Two hops out of a ring repeater site will not both be affected exhibit simultaneous rain outage if they are separated in azimuth by at least 60-80. Multipath fading in optimally aligned millimeterwave hops does not occur during periods of heavy rainfall, so the entire path fade margin is available to combat rain attenuation fades Neither space diversity nor in-band frequency diversity provides any improvement against rain attenuation fade outage RAIN CELL PASSAGE Heavy rainfall, usually in cells accompanying thunderstorms and weather fronts, has a great impact on path availability above 10 GHz in some areas, but this outage time is always kept separate from multipath outage time because such long-term rain fade outages that interrupt or disconnect traffic are categorized as Availability % rather than short-term Performance (Path Reliability %) events discussed later that do not interrupt traffic. Long-term rain outages increase dramatically with frequency and path length. Extended 10-15 minute duration fades to over 50 db similar to that in Figure 1 have been recorded on a 3-mile 18 GHz path in Houston, for example. A similar outage was seen on a 6-mile 11 GHz link in Tampa. Lesson learned: Do not deploy millimeterwave hops of any length in Tampa, US which is ground zero for the incidence of high-intensity rain cells and thunderstorms over 100 a year. The predicted annual outage predicted by a Crane or ITU-R P.530 model may not occur for years and then accumulate over a single rainy season for a long-term average. Rain has long been recognized one of the principal causes of unwanted signal loss in millimeterwave microwave terrestrial and satellite radio paths through the lower atmosphere. Rain is not the only cause. Variations in water vapor along the path or occurrence of higher humidity visually seen as liquid water clouds or fog on the path will introduce a fixed loss (like waveguide feeder loss in path calculations as Atmospheric Absorption Loss that peaks in the 23 GHz band. But in millimeterwave bands, rain is often the sole cause of increased attenuation and fading. 5 AVIAT NETWORKS AUGUST 2011
% of Time Rain Rate Exceeded Crane s 1980 North America Rain Region, mm/hr A B C D1 D2 D3 E F 0.1 6.5 6.8 7.2 11.0 15.0 22.0 35.0 5.5 0.05 8.0 9.5 11.0 16.0 22.0 31.0 52.0 8.0 0.01 12 19.0 28.0 37.0 49.0 63.0 98.0 23.0 0.006 19.0 26.0 41.0 50.0 64.0 81.0 117.0 34.0 0.001 28.0 54.0 80.0 90.0 102.0 127.0 164.0 66.0 Figure 2: These rain rates (from Crane 1980) in mm/hr exceeded _% of the time, are long-term averages over a 10- year or longer period. Rain rates (and high frequency link outage) are greater or smaller over shorter periods. The simplified P.530 model, easily adaptable to programmable calculators, interpolates rain outages from the 0.01% rain rates in the Crane 1980 (above), Crane 1996, or ITU-R tables. Except for border regions, always select the rain region where your microwave links are going to be implemented. In hops crossing rainrate border areas, select the regional numbers based on how optimistic or conservative you want the final Early studies both theoretical and experimental resulted from recognizing the importance of rain in designing microwave paths with path availabilities ( uptimes ) over 99 percent. Estimates of rain outage in the 1950-1970s was typically a manual procedure requiring access to thick volumes of rain rate tables with statistics for major cities in the U.S. and around the world, but with personal computers emphasis has been on establishing predictive techniques for statistical estimation of attenuation probability distribution for a particular path. Robert F. Crane, in his Prediction of Attenuation by Rain 1980 paper developed a model (see Appendix) for determining attenuation on factors including path length, frequency, polarization and point rain rates shown in Canadian/US rain rate tables [Figure 2] and rain region map (Figure 3). Figure 3. Each region of North America has a different value that is used in the Crane 1980 and ITU-R P.530 mathematical models for predicting rain attenuation. Crane 1996 and ITU-R P.837 worldwide rain region maps and rain rate tables for North America and the world are also used in these models. 6 AVIAT NETWORKS AUGUST 2011
The Crane model tends to produce higher attenuation than the ITU model. But the uncertainty of either of these models or alternatively the short-term expectation of fade is quite large. Uncertainty stems from variations from year-to-year and location-to-location. RAIN AVAILABILITY OBJECTIVES Unlike multipath fading that cause short-term hits but no traffic disconnects that affect the hop s performance (discussed later), rain fades always cause long-term (defined as a >10 CSES/event) traffic outages of 3-20 min durations that disconnect subscribers if not ring-protected. A typical per-hop rain outage objective in a millimeterwave access or spur link is around 26min/year (99.995 percent availability or uptime ), which equates to perhaps two to three rain outage events each year. If this if unacceptable and traffic is not otherwise ring-protected, it may be necessary to assign the hop to vertical polarization, increase its fade margin (for a minimal improvement), or migrate the hop to a lower frequency band. RELIABILITY OBJECTIVES Rain outage always causes long-term traffic interruptions if not ring-protected, from which the hop s availability or uptime" is computed and is therefore not included in path s performance (reliability) calculations that consider only short-term (<10 CSES/event) multipath outages that do not significantly affect traffic (Figure 4). A microwave hop s error performance is predicted and measured only during available periods when the hop is up ( as near to 100% as is economically and technically possible, not down due to a long-term rain, duct, decoupling, defocusing, etc. power fade, equipment outage, or infrastructure (site, power, antenna feeder system) failure. In considering how to establish realistic performance (short-term outage, path reliability objectives, several things need to be kept in mind. A single overall design objective for not more than X minutes or seconds outage over some period, such as a year, is an oversimplification. The character of the particular kind of outage and its effect on the system should be taken into account and perhaps there should even be different objectives for different types of outages. Reliability % Outage Time % Reliability and Outage Time Year 3 Month Fade Period Outage Time per Month (Average) Day (Average) 0.0 100.0 8760 hrs 2233 hrs 720 hrs 24 hrs 50.0 50.0 4380 hrs 1116 hrs 360 hrs 12 hrs 80.0 20.0 1752 hrs 446 hrs 144 hrs 4.8 hrs 90.0 10.0 876 hrs 223 hrs 72 hrs 2.4 hrs 95.0 5.0 438 hrs 117 hrs 36 hrs 1.2 hrs 98.0 2.0 175 hrs 45 hrs 14 hrs 29 min 99.0 1.0 88 hrs 22 hrs 7 hrs 14 min 99.9 0.1 8.8 hrs 2.2 hrs 43 min 1.4 min 99.99 0.01 53 min 13 min 4.3 min 9 sec 99.999 0.001 5.3 min 1.3 min 269 sec 0.9 sec 99.9999 0.0001 32 sec 8 sec 3 sec 0.09 sec Figure 4. Rain outage always causes long-term traffic interruptions in a hop that is not ring-protected and, thus, is not included in path reliability calculations that consider only short-term (<10 CSES/event) multipath outages. For example, propagation outages due to multipath fading are usually short. An annual outage of 30min/yr due to multipath fading might represent 1,000 or more individual SES (severely errored second) outages, most averaging less than 1 second in duration each in properly configured hops. On the other hand, propagation outages totaling an hour per hop due to rain attenuation, on a path with a large fade margin, might consist of four or five individual outages averaging ten to fifteen minutes each. The effects of these two types of system outage would be quite different in nature, long-term rain outage being an availability or uptime (disconnects traffic) parameter, and short-term multipath fade outage an error performance or path reliability parameter that does not disconnect traffic. 7 AVIAT NETWORKS AUGUST 2011
A distinction should be made between TDM and Ethernet/IP communication links for which an outage of a few seconds or a few minutes is just a nuisance or an inconvenience, and public safety links carrying critical real-time voice and computer traffic for which such an outage might result in a danger to life, great economic loss or other catastrophic consequences. The suitability or unsuitability of a rain-affected band such as 18 or 23 GHz as well as ring and other protection schemes to mitigate outage could differ widely for these two situations. PATHS AS FAILSAFE AS POSSIBLE Even if the maximum possible reliability objectives are established and a path or a system is engineered to the full limit of the state of the art, the possibility of an outage can never be eliminated but can only be reduced to a very low probability. Thus, it is imperative to make any ultra-important services as failsafe as possible against a loss of the communications channel. Therefore, regardless of the degree of reliability, a system should be engineered so that if an outage does occur it can be tolerated or its effects at least kept within reasonable bounds. It seems that in some cases, perhaps many cases, a somewhat more relaxed attitude might be taken toward rain-induced outages than toward multipath outages or even equipment outages. In several respects, such rain outages seem to be somewhat benign in nature. If the fade margins are kept high and the paths are not stretched out too much, even in the less advantageous areas of the country, the number of outages per year should not be very large, and the length of individual outages on a hop should only rarely exceed some 2 to perhaps 20 minutes. Short (less than 2-second) microwave outages that do not cause trunk conditioning, common on even a typical longer non-diversity digital microwave connection with a high fade margin, will not drop any voice or data links. Such outages quickly clear with all connections remaining intact with little note taken of these transient events except for critical real-time, non-repeatable control or data blocks. Longer outages associated with low fade margins, rain, etc. disconnect all subscribers, cause IP packet loss, and block access to the digital link for at least 10 seconds per event. Such events are unacceptable to most users. These vulnerable links clearly require diversity protection. PERFORMANCE OBJECTIVES IN LONG-HAUL VS. SHORT-HAUL SYSTEMS For high reliability systems, usually involving long-haul systems with a great many hops in tandem, the per-hop objective in long-haul microwave routes may be as stringent as 99.9999% per hop (AT&T s one-way objective for 25-mile hops in a long-haul route), allowing only about 30 seconds of short-term multipath fade outage per year. Short haul systems, up to say 10 hops, may be assigned a per-hop design objective of about 99.9995% for about 160 sec/yr of one-way multipath fade outage per year. Spur legs or single hop systems may be designed for approximately 99.999% or about 5min (315 seconds) of short-term outage per year. Objectives of these kinds are typical of those used in for public service networks. Public safety, homeland security and electrical utilities may demand better performance, while for others even 99.9% or about 9 hours/year outage may be acceptable. NOTE It is important to note that graphs, formulas and computer methods for calculating short-term multipath outages are all for one-way outages as they are selective in frequency, i.e., do not occur at the same time in the two directions of a hop. To derive two-way multipath fade outage (not ever recommended since all ITU and North American performance objectives are only-way only), they would have to be doubled. Long-term outages due to rain or other non-selective power fades do not have to be doubled since they occur simultaneously in both directions of transmission. 8 AVIAT NETWORKS AUGUST 2011
APPENDIX CRANE S RAINFALL OUTAGE MODEL (A complex algorithm, thus only used in all Path Engineering Calculation Computer Programs) A = Rain attenuation (required fade margin) exceeded p% of the year A = α R P% β [(e λβd -1)/λ β - (b β e cβd /c β) + (b β e cβd /c β)]; for d<d<do A = α R P% β [(e λβd -1)/λ β]; for D<d Where R P% = Rain Rate exceeded p% of the time, mm/hr D = Path length in kilometers λ = [ln (be cd )]/d b = -0.17 2.3 R P% c = 0.026-0.03 ln R P% d = 3.8-0.6 ln R P% α = Multiplier coefficient, a function of frequency and polarization (Figure 5) β = Exponent coefficient, a function of frequency and polarization (Figure 5) P'=P[Do/D], where Do =22.5 km, for D>22.5 km P.530 RAINFALL OUTAGE CALCULATION (Begins for a 99.99% availability) Example: The hop s fade margin exceeded 0.01% of the time = 53 min/yr outage in an 18.7 GHz, 16 km (10 mi) V-pol path in Chicago (Crane rain region D 2) meeting a 99.99% rain availability objective: A 0.01% = α R 0.01% β D [1/(1 + D/d)], db Where R 0.01% = Rain rate exceeded <0.01% of the time, mm/hr D = Path length, km (mi x 1.6093) α = Multiplier coefficient, ƒ(frequency & polarization), from Figure 5 β = Exponent coefficient, ƒ(frequency & polarization), from Figure 5 d = 35 exp (-0.015R 0.01%) = 16.8 km (related only to rain region) A 0.01% = 0.058 (49) 1.08 16 [1/(1 + 16/16.8)] = 32 db (40 db if horizontally-polarized since the α and β increase) P R = % of time rain attenuation exceeds A (% outage time) = 10 x Where: x = 11.628 {-0.546 + [0.29812 + 0.172 log (0.12 A 0.01%/A)] 1/2 } A 0.01% = Fade Margin for 0.01% (53 min/yr) Outage (32 db) A = Path s actual Fade Margin from path calcs (41 db) P R = Rain Unavailability objective (0.005% for 99.995% availability) Path Availability = 100 - P R % = 99.995% Annual Rain Outage Time = P R % x 31.5x10 6 = 1600 sec/yr This much simplified, at least when compared to Crane s, ITU-R rain outage model is easily programmed into scientific calculators and computers. The regression coefficients, multiplier α and exponent β in both the above Crane and ITU-R P.530 rain models are from the following ITU-R Rec. P.838 table (Figure 5). 9 AVIAT NETWORKS AUGUST 2011
Frequency (GHz) Vertical Polarization α V Vertical Polarization β V Horizontal Polarization α H Horizontal Polarization β H 7.0 0.0027 1.312 0.0030 1.332 7.5 0.0033 1.311 0.0037 1.329 8.0 0.0040 1.310 0.0045 1.327 10.6 0.0109 1.243 0.0122 1.258 11.2 0.0132 1.224 0.0145 1.242 13.0 0.024 1.183 0.026 1.203 15.0 0.034 1.128 0.037 1.154 18.7 0.058 1.080 0.064 1.112 22.4 0.089 1.047 0.097 1.080 30.0 0.167 1.000 0.187 1.021 38.0 0.278 0.942 0.314 0.954 60.0 0.642 0.824 0.707 0.826 Figure 5. The ITU-R Rec.838 table comprises the regression coefficients, multiplier α and exponent β, in both the Crane model and the competing ITU-R P.530 rain model. WWW.AVIATNETWORKS.COM Aviat, Aviat Networks, and Aviat logo are trademarks or registered trademarks of Aviat Networks, Inc. Aviat Networks, Inc. (2011) All Rights Reserved. Data subject to change without notice. _w_rain_attenuation_fading_25aug11 10 AVIAT NETWORKS AUGUST 2011