AVAILABILITY OF A LINE-OF-SIGHT MICROWAVE LINK

Transcription

1 Appendix 1 AVAILABILITY OF A LINE-OF-SIGHT MICROWAVE LINK A1.1 INTRODUCTION Ž. The IEEE Ref. 1 defines availability as the long-term average fraction of time that a system is in service satisfactorily performing its intended function. Availability, in our context, is usually expressed as a percentage or decimal. It defines the time a system, link, or terminal is meeting its operational requirements. Equipment availability is expressed by the familiar equation A s MTBFrŽ MTBF q MTTR. Ž A1.1. where MTBF is mean time between failures and MTTR is mean time to repair. Both are measured in hours. As one can see, equation Ž A1.1. only treats equipment failure and its repair time; it does not reflect outage due to fading. By restating the equation, we can cover the general case: A s uptimerž uptime q downtime. Ž A1.2. Example. If a system has an uptime of 10,000 hours and a downtime of 10 hours, then A s 10,000rŽ 10,000 q 10. s or 99.9% In our discussion we will be working with una ailability, which we can call a state of nonservice that occurs due to failure, other outage or degraded BER Radio System Design for Telecommunications, Third Edition By Roger L. Freeman Copyright 2007 John Wiley & Sons, Inc. 815

2 816 AVAILABILITY OF A LINE-OF-SIGHT MICROWAVE LINK performance of at least 10 seconds duration. Our notation of unavailability is U, and U s 1 y A Ž A1.3. or U s 1 y uptimerž uptime q downtime. Applying this to the above example, we find U s 1 y s or 0.1% Ž. With radiolinks, we will deal with one-way e.g., west-to-east and two-way availabilities. If the unavailability objective of a two-way channel is 0.02%, and the outage probabilities in the two directions are independent, the objective for a one-way channel is 0.01% or about 105 minutesryear for a two-way system, and 53 minutesryear for an equivalent one-way system. The importance of a ailability cannot be understated. It tells us how well we can depend on the system, link, or terminal to do its job. In this appendix we discuss availability from several aspects and point up some weaknesses in traditional arguments on the subject. A1.2 CONTRIBUTORS TO UNAVAILABILITY Ž. CCIR Rep Ref. 2 lists five major contributors to outage on radiorelay systems: 1. Equipment a. Failurerdegradation of radio equipment such as modulators and demodulators. b. Failure of auxiliary equipment, such as switchover equipment. c. Failure of primary power Ž equipment.. d. Failure of antenna or feeder. 2. Propagation a. Deep fading causing noise to exceed a certain limit. This may be due to ducting and usually lasts for a fairly long time. b. Excessive precipitation attenuation that is caused mainly by heavy rainfall and, in some cases, heavy snowfall. Generally, the effect lasts a fairly long time. c. Fading causing short interruptions Ž dispersive fading. Žincludes ISI degradation and outage. 3. Interference: noise in excess of a certain limit caused by interference sources that may exist within or outside the system.

3 CALCULATION OF AVAILABILITY OF LOS RADIOLINKS IN TANDEM Support facilities: collapse of towers or buildings in disastrous circumstances. 5. Human error: this includes maintenance downtimeroutages. The items listed above are in order of greatest to least contributors to unavailability. Equipment failurerdegradation is certainly the greatest contributor. A1.3 AVAILABILITY REQUIREMENTS The AT & T unavailability objective is 0.01% for a one-way channel over a 4000-mi route Ž Ref. 3.. The equivalent availability is 99.99%. In Canada a tentative objective of 99.97% is used for a 1000-mi one-way radio system. This corresponds to 99.95% availability on a 2500-km base. The current United Kingdom availability objective for bidirectional transmission is % per 100 km, which corresponds to 99.84% for a 2500-km circuit Ž Ref. CCIR Rep Ž Ref. 2.. A1.4 CALCULATION OF AVAILABILITY OF LOS RADIOLINKS IN TANDEM In this section we derive per-hop availability given a system availability consisting of n hops in tandem with independent outages on different links. Often such a system availability will derive from the ITU-R hypothetical reference circuit, which is 2500 km long. Panter Ž Ref. 4. assumes that such a 2500-km circuit consists of 54 hops each 30 mi Ž 48 km. long. We describe the procedure by an example. If a one-way circuit requires an availability of 99.95% over a 2500-km reference circuit, what is the required per-hop availability? First calculate the system unavailability, which, in this case, is 1 y s Divide this unavailability by 54 or r54 s This then is the unavailability for one hop. Its availability is 1 y s %. ATT uses a 4000-mi reference circuit with a required availability of 99.99% or an unavailability of or 1 10 y4. If we assume as above that each hop is 30 mi long, then there are 133 hops in the reference circuit. We now divide 1 10 y4 by 133 and the resulting unavailability per hop is %. The equivalent availability per hop is then %. A1.4.1 Discussion of Partition of Unavailability At first glance we could apportion half the outage Ž unavailability. to equipment failure and half to propagation outage. Thus, in the case of ATT, the

4 818 AVAILABILITY OF A LINE-OF-SIGHT MICROWAVE LINK unavailability per hop would be %r2 or % for equipment and % for propagation outages. White of GTE-Lenkurt argues against this approach Ž Ref. 5.. Let us consider a 1-year or 8760-h interval. A year has 525,600 min or 31,536,000 s. What is the annual expected outage when the unavailability is %? 8760 h s h 525,600 min s min 31,536,000 s s s If we assign half of this number to equipment outage and half to propagation outage, we then have 11.8 s of outage per year for each. The next step is to apply the conventional formula for availability wequa- tion Ž A1.1.x and calculate MTBF in hours: MTBF % s MTBF q MTTR However, we first must assign a reasonable value for MTTR or repair time. Consider that most LOS radiolink sites are unattended, and when a failure occurs, a technician must be sent to that site. Hershe must be alerted, gather up tool kit and required parts, and travel to that site, possibly mi away, and, of course, time must be allowed to carry out the repair. Values for MTTR in the literature for this application are from 2 to 10 hours. We use the worst case, then MTBF % s MTBF q 10 MTBF s h or yr If we allotted half the outage to equipment failure and half to propagation, we must double the MTBF, requiring a MTBF of about 301 years! The argument then follows that propagation outage, say 32 sryr, might consist of many events Ž short fades. in 1 year, whereas with equipment reliability we are dealing with one event every 301 years. It would follow, then, that we treat these two types of outages separately and independently for they are truly not summable. It is like summing 6 apples and 4 oranges resulting in 10 lemons. What is driving us to these large values of MTBF is the large values for MTTR. On sophisticated military radio terminals Ž non-los radiolink., MTTR runs at about h. It is assumed that the technician is on site and on duty.

5 IMPROVING AVAILABILITY 819 A1.4.2 Propagation Availability We then treat propagation availability separately, but it too requires some special considerations of reasonableness. Again let us turn to an example. We use the Canadian values Ž Section A1.3. of 99.95% for a 2500-km reference circuit and the equivalent unavailability is Assume, again, 54 hops in tandem for the 2500-km reference circuit. The unavailability per hop is then r54 s or an availability of %. This, of course, assumes a very worst-case fading where all hops fade simultaneously but independently. This is unrealistic. Panter Ž Ref. 4. reports a more reasonable worst case where we would allow only one-third of the hops to fade at once, 54r3 s 18; thus we can say that the unavailability due to propagation Ž multipath fading. is 0.005r18 s or an availability of would be required per hop. This would be nearly in keeping with ITU-R F paragraph 1.2 or Ž 50r of any month Ž not year. where the availability required is %rmonth to a reference level of 47,500 pwp. We essentially derive the same value for a 2500-km reference circuit where L s 2500; then the unavailability is 0.001r54 or % availability per hop, following ITU-R F to the letter. Again the assumption is made that all hops in tandem are simultaneously subject to fading. Following Panter s reasoning, the divisor would be 18 rather than 54 or a required availability per hop of % to the reference level of 47,500 pwp. A1.5 IMPROVING AVAILABILITY Examining equation Ž A1.1., we can improve availability by increasing MTBF Ž improving reliability. and decreasing MTTR Ž mean time to repair.. MTBF can be increased by using Hi-rel Ž high-reliability. components, particularly for those components that have a history of numerous failures. Another approach is to use redundancy, such as hot-standby operation. If the necessary automatic changeover circuits are employed, redundancy Ži.e., have two identical equipments with automatic changeover on failure. squares the MTBF of the combination. For example, if a transmitter has a MTBF of 10,000 hours and we equip with a second, identical, transmitter with changeover, the resulting MTBF is 10,000 10,000 hours or 10 8 hours. The difficult and ambiguous part of the availability equation is MTTR. Unattended repeater Ž or drop and insert. sites require that a technician travel to that site when there is a failure or degradation. How long does it take to get there? Did hershe bring the right part, card, or subassembly with himrher? If not, is the card, part, or subassembly available in the storeroom? If not, it must be ordered from the equipment manufacturer, and suppose the manufacturer does not have the part.

6 820 AVAILABILITY OF A LINE-OF-SIGHT MICROWAVE LINK It is not feasible from an economic sense to have every single card, part, and subassembly available as a spare part in the storeroom. Most operations only stock high-failure items. Some do not stock any and depend on fast turnaround from the manufacturer. For this latter approach MTTR can extend over 24 hours. How long does it take the technician to find the failed part? This can be improved by using BITE Ž built-in test equipment., and if this information can be remoted to the servicing terminal via a service channel, all the better. BITE identifies on a gorno-go basis Ž binary. if a card, circuit, or subassembly is working. It may be working, but how well? These BITE circuits can be set for a certain threshold before kicking in an alarm. Thus they can alarm at a certain degradation point. CCIR Rep Ž Ref. 2. uses 10 hours for a MTTR value. The U.S. military often uses 20 minutes. Of course, the U.S. military assumes a technician on duty and that the spare part is available. Nearly all military electronic equipment have excellent BITE, so troubleshooting can be easy and sure. Note how availability drastically improves with a 20-minute Ž h. MTTR. Every item mentioned above costs money the price goes up. It boils down to how much we are willing to pay for good availability values. A1.6 APPLICATION TO OTHER RADIO MEDIA The same approach to availability and its calculation and improvements can be applied directly to over-the-horizon links such as troposcatter and diffraction. It can also be applied to satellite communications links. For cellularrpcs, HF, and meteor burst systems, modifications would have to be made for propagation effects. Certainly for HF and meteor burst systems, propagation takes on a much more important role, probably a dominant role. REFERENCES 1. The New IEEE Dictionary of Electrical and Electronic Terms, 7th ed., IEEE Std 100, IEEE, New York, A ailability and Reliability of Radio-Relay Systems, CCIR Rep , Annex to Vol. IX, XVIIth Plenary Assembly, Dusseldorf, Transmission Systems for Communications, 5th ed., Bell Telephone Laboratories, Holmdel, NJ, P. F. Panter, Communication Systems Design for Line-of-Sight Microwa e and Troposcatter Systems, McGraw-Hill, New York, R. F. White, Reliability in Microwa e Systems Prediction and Practice, GTE- Lenkurt, San Carlos, CA, 1970.