Cẩm nang dữ liệu không dây P12 pps

For roughly equivalent geography, BSWD’s technical choicespredispose it toward maximum subscriber potential at the cost of some penetration.CDPD is not closely comparable to either ARDIS

BUSINESS 102: OTHER THINGS ARE

IMPORTANT TOO

Copyright © 1999 John Wiley & Sons, Inc.

ISBNs: 0-471-31651-2 (Hardback); 0-471-22458-8 (Electronic)

Each of these decisions can cloud the certainty of coverage comparisons ARDISconfigures minimal infrastructure for maximum in-building penetration at the cost ofsubscriber capacity For roughly equivalent geography, BSWD’s technical choicespredispose it toward maximum subscriber potential at the cost of some penetration.CDPD is not closely comparable to either ARDIS or BSWD because, in afull-blown implementation, it would deploy many more base stations, each with adedicated channel, thereby achieving both high capacity and good penetration Butcarriers are hesitant about the cost of a CDPD deployment of that scope BellAtlantic Mobile has stated1: “CDPD will be offered in markets where it is expected

to be commercially viable, although equipment will not be installed at all cell sites

in those markets.”

This uncertainty does not mean that coverage is simply unknowable withoutcomprehensive field tests A great deal can be accomplished, without the special

171

The Wireless Data Handbook, Fourth Edition James F DeRose

Copyright © 1999 John Wiley & Sons, Inc.

ISBNs: 0-471-31651-2 (Hardback); 0-471-22458-8 (Electronic)

knowledge of a radio engineer, by analysis of site location licenses, judicious use ofvendor coverage maps, and ZIP code predictors Not all carriers make these toolsavailable, and there are often quality variations in those that do exist The base stationlicense and vendor coverage map techniques are most effective when judgingrelatively wide geography such as an entire county At their weakest these simplemethods permit the user to confine field testing to likely trouble spots At their bestthey are often powerful enough to be excellent predictors of coverage, eliminatingmuch tiresome, labor-intensive work.

Ultimately, some field tests are required to prove “contract coverage” in specificareas For a prospective user these tests need not be laden with instrumentation Farsimpler techniques can give the thoughtful reviewer valuable insight into coverageand retransmission performance that tends to vary with message length

Since 1994 JFD Associates has conducted coverage and building penetration tests

at roughly six-month intervals on ARDIS, BSWD, and, later, CDPD There are alsopublic domain summary reports from users such as Pitney-Bowes andSchindler/Millar elevator, as well as carrier reports from BSWD Illustrative extractsfrom these sources are used in this chapter

12.2.1 ARDIS

ARDIS employs each channel in a given area in single-frequency reuse (SFR) mode

As the user moves from one single channel cell to another, the device remains tuned

to the same frequency Higher level logic determines which base station will workwith the device Its 90% coverage areas are deliberately designed not to overlap, asillustrated in Figure 12-1

Since ARDIS employs SFR, the area of reduced coverage probability between twoadjacent cells can be exploited Extending beyond the solid 90% line are areas ofever-decreasing coverage The 70% curve is portrayed as a dotted line

Where two of these 70% zones overlap, the effective coverage is 90% since the usercan be handled by either of two base stations The explanation is straightforward:When in an overlap zone there is a 30% probability that the user cannot be heard bycell A; the same 30% probability extends to cell B But the probability that the userwill not be heard by either cell A or B is 0.302 = 9% The probability that the user will

be heard by at least one base station is thus 91%

Note that there are scattered areas in which the probability of success is very high(shown here as triple-coverage zones) The proper siting of base stations can causethese areas to become quite rich, thus improving building penetration As anillustration, the “typical user transmission in Chicago is heard by 8 to 10 basestations.”2 Clearly not all these base stations hear with anything like a 70%probability, but there is always some finite chance that a message that would be missed

by BSWD or CDPD will be picked up by ARDIS

When ARDIS adds a new channel, a new base station is typically installed at thesame physical site Stamford, Connecticut, has such an overlay When the devicemodem is powered on, it “listens” first for an RD-LAP channel If found, it beginsoperation there Lacking RD-LAP, it will switch to MDC4800 As the device movesaround, or as channel loads/conditions vary, it is quite possible to receive sequentialpackets from the same message on different channels, some RD-LAP, some MDC.These layers of channels are currently highest in metropolitan New York with a total

of eight 25-kHz-wide channels

12.2.2 BSWD

BSWD has a cellularlike base station siting plan, though on a smaller scale.Omnidirectional cells are grouped in clusters; no two cells in the same cluster employthe same frequencies The initial goal was to provide a 90% or greater probability ofstreet-level coverage With the drive toward two-way paging, the goal has evolved to

“coverage comparable to one-way paging, nationwide.”3

BSWD’s base station density does not approach that of cellular As an example, inMarch 1998 BSWD had only 26 base stations in the entire state of Connecticut Thereare other “helper” sites along the Massachusetts, New York, and Rhode Islandborders Since Connecticut is very small, about 5000 square miles, the average cellradius for just the 26 Connecticut base stations is ~7.8 miles This is not the same

Figure 12-1 ARDIS base station siting (representative).

12.2 KEY COVERAGE PHILOSOPHIES 173

design criteria as, say, metropolitan voice cellular, which sometimes operates with1/2-mile-radius cells and microcells.

Naturally the sites are not uniformly distributed BSWD has 12 of the 26 locatedjust in Fairfield County to improve Interactive Pager (I@P) coverage for the GreaterNew York metropolitan area Fairfield County is ~435 square miles, so the averagecell radius drops to ~3.4 miles in this densely populated sector

BSWD has stated that it is “able to reuse (each channel) easily four times.”4 Thisrequires a broad geographic spread employing perhaps 30 base stations While thereuse figure has been disputed, in this analysis BSWD’s claim is accepted as correct.Thus, say, Los Angeles County has ~4 channels per cell and 7 cells per cluster The90% coverage lines of each cell in the cluster overlap to ensure smooth hand-off Anoverly simplified representation of BSWD’s base station siting is shown in Figure12-2; the principal hand-off areas from the central base station are designated H

12.2.3 ARDIS Versus BSWD: Representative 90% Coverage Contours

The resulting 90% area contours are depicted in Figure 12-3 The key point is that forapproximately the same geographic area coverage contours are always somewhatdifferent There will be locations in which BSWD will operate and ARDIS will not,and vice-versa Note that in this example BSWD requires more cells to achieveroughly the same 90% area coverage

Figure 12-2 BSWD base station siting (representative).

12.2.4 Improving Building Penetration with More Base Stations

BSWD is clearly on a path to enrich its infrastructure for two-way paging buildingpenetration One must be wary about carrier base station counts; they seldomcorrespond to geographic locations ARDIS counts each new frequency deployed atthe same physical location as an additional base station BSWD counts twotransceivers at the same location as two base stations Nevertheless, BSWD’s August

1998 claim of 1900 installed base stations5 probably brings it to a parity position withARDIS in unique site locations—and BSWD is growing its infrastructure morerapidly than ARDIS Further, BSWD is focused on ~160 fewer cities, each of whichcosts ARDIS a base station In selected areas, BSWD likely has more actual basestations than ARDIS

An interesting example is Fairfield County, Connecticut BSWD has 12 physicallocations; ARDIS has only 9 As noted in Section 12.2.3, BSWD probably requires

Figure 12-3 Comparative areas of 90% coverage.

12.2 KEY COVERAGE PHILOSOPHIES 175

more base stations to achieve parity in an equivalent area Field tests at 48 street-levellocations in Fairfield County during January 1998 show that ARDIS and BSWD haveequivalent, and very good, street-level coverage.

Additional tests revealed that ARDIS had ~6% coverage edge in totally enclosedbuildings Continued enhancements in BSWD infrastructure would tend to offset anybuilding penetration shortfall Very large additions might permit it to outperformARDIS and gain more capacity as well But one must be mindful of the rules of planegeometry If BSWD drove its infrastructure to 18 locations, doubling the number ofARDIS sites, the average cell diameter would only fall from 3.4 to 2.8 miles This is

a significant reduction—and would surely be pleasantly noted by BSWD users—butthis major investment would not result in microcells

12.2.5 CDPD

CDPD, using the same physical locations (and, e.g., antennas, T1 lines) as voicecellular, follows the voice cochannel reuse rules In metropolitan areas where CDPDbuild-out is most complete, this means a trisectored layout Hand-offs occur in thezones marked H, as shown in Figure 12-4

The hand-off threshold varies somewhat by carrier; the right user device is key In earlyBAM dedicated channel implementations, using the PCSI PAL phone, it was possible to

be in the central hand-off area and be unable to transmit or receive Putting the phone in

“diag” mode, one could watch the hunt: channel 721, 555, 729, 721, 555, —unable

to decide which channel was the best! Now BAM has time and dBm selection

Figure 12-4 CDPD trisectored base station siting (representative).

thresholds Every 90 seconds the device “listens” for alternatives If the other channelsthat can be heard are not 8 dBm better than the current channel, no switching occurs.

12.2.6.1.1 ARDIS Pre-ARDIS, the KDT-800 device used in IBM’s Field Servicesystem had a transmit power of 4 watts Much of this energy was wasted because thehand-held “brick” had internal antennas (this was a dual-diversity receive unit) As theinfrastructure was enriched, succeeding internal antenna devices such as the KDT-840had their output power reduced to 3 watts This level remains the upper limit forvehicular devices employing, say, Motorola’s mobile radio modem (MRM) line ofmodems

With the coming of frequency-agile, hand-held devices a design trade-off wasmade The power-hungry synthesizers, which replaced simple crystal cut oscillators,drew so much current at the 3-watt transmission level as to make battery lifeunacceptably short Regrettably for some users who treated their devices carelessly,the antenna would have to be external The trade-off was a reduction in output power

to 1.5 watts and longer battery life This is the usual output power level for ARDIShand held devices

The Interactive Pager (I@P) returned to an internal antenna located inside itsflip-top lid The transmit power level was also eased down to 1 watt Performance isgood with the lid open It has been reported that IBM Field Service people whonormally work with the lid closed encounter reduced coverage as compared to the oldKDT-800

12.2.6.1.2 BSWD The Ericsson infrastructure employed in the original RAMdeployment had poor receive sensitivity, which limited the optimum radius of the basestation to ~5 miles (–115 dBm laboratory capability).6 In 1997, to prepare for thecoming of the I@P, BSWD began to retrofit all existing base stations to improve thereceive sensitivity to –121 dBm This pushed their effective radius to ~7 miles.Because of these early restrictions, the original external modems for portable users had

a transmit power of 2 watts AND an external antenna The version of the I@P builtfor BSWD also has 2 watts transmit power

The higher transmit power requirement forced Ericsson, then subsequent vendors,into excellent battery-saving techniques They are among the best available today.During one 14-day test I transmitted 157 messages each on both an ARDIS and aBSWD I@P The ARDIS unit was an early version, and there was actually a bit ofextra traffic on it, but that does not justify the very different results I was forced toreplace ARDIS I@P batteries three times during the 14 days; I never changed theBSWD I@P batteries (Note: In some heavy usage CDPD tests external modembatteries must be changed every 2–3 hours)

12.2 KEY COVERAGE PHILOSOPHIES 177

All BSWD I@P tests were performed with lids open A “lids-closed” test mighthave revealed shortcomings, but it is a moot point since, in August 1998, BSWDannounced the availability of the “no-lid” RIM I@P950.

12.2.6.1.3 CDPD After early consideration of 256 discrete power levels,CDPD settled on the same range as its voice cellular counterpart Except forexpensive, multipurpose vehicular units, CDPD modems generally operate with amaximum transmission power of 6 watts, identical to hand-held voice phones.This is generally satisfactory in richly endowed urban areas with small-diametercells In the period 1995–1996, before some urban areas were optimized forportable phones, CDPD performance was very poor The lack of power is stillevident along many interstates that are really designed for 3-watt vehiculardevices Very large, very troublesome coverage holes often appear in thissituation

Figure 12-5 ARDIS RF coverage contours: Naples, Florida.

12.2.6.2 External Antennas Another technique to extend the coverage area ofpacket modems is to use external antennas without an increase in transmit power Theuse of an external mag mount antenna in my office improves CDPD signal strength

by about 20 dBm Naturally, these improvements are terrain dependent In the flatgeography of Florida the effects can be quite pronounced Figure 12-5 shows theimprovement in the 90% ARDIS coverage area when a mag mount antenna isemployed with a PM100D class modem The 90% coverage line extends an additional

5 miles from the base station center Note, also, that coverage does not suddenlyvanish after the 90% contour curve is reached Some success will be had at evengreater distances

This technique is often employed with over-the-road trucks having multiprotocolmodems that feature satellite connections Extending the terrestrial coveragemarkedly lowers the higher cost satellite bills

repeaters—are used to extend coverage into areas without base stations Thistechnique is used in Germany’s DataTAC system, for example The use of repeaters

in the United States tends to be limited to frequencies below 450 MHz If selectedcarriers did gain FCC approval for repeaters in the 800/900-MHz bands, coveragemight well be extended into rural areas

12.3 ESTIMATING COVERAGE WITHOUT FIELD TESTS

For years most carriers filed a radio station license with the FCC Reference copies areavailable to all U.S citizens for a nominal fee The copious information on theselicenses includes a number of facts that can help a coverage analysis, including thebase station:

1 Address (street, city, county, state)

2 Latitude/longitude

3 Output power (watts)

4 Effective radiated power (watts)

5 Antenna height (meters)

Recently, BSWD license facts have become more difficult to check BSWDadministratively canceled its licenses as a part of the 900-MHz auction process by aletter to the FCC on September 26, 1996.7 However, even purged records are stillpublic information Much of the data can still be recovered A listing of base stationsites for ARDIS and BSWD in Connecticut is contained in Appendix I With position

12.3 ESTIMATING COVERAGE WITHOUT FIELD TESTS 179

information in hand, it is quite easy to plot base stations in a particular area of interest

to see how the “paper coverage” compares

Figure 12-6 is a plot of ARDIS coverage in part of New Haven County using a7-mile radius for the 90% coverage “reach” of the base station This is probably anoverstatement of the radius since each ARDIS base station simply does not have to

Figure 12-6 Representative ARDIS base station locations: New Haven county, CT.

cover that much territory with 90% power However, it is easier to represent thanplotting 70 and 90% curves.

Figure 12-7 is the identical layout for BSWD, also using a 7-mile radius—thestated capability of a retrofitted base station

Figure 12-7 Representative BSWD base station locations: New Haven county, CT.

12.3 ESTIMATING COVERAGE WITHOUT FIELD TESTS 181

A cursory look at these plots indicates that ARDIS will likely have betterin-building coverage in Hamden/New Haven Indeed, this seems to be true Inside-by-side testing during January 1998, BSWD hit 6 for 6 on street and 3 for 3 inbuildings with large windows But when severe conditions were attempted—inside aclosed stairwell, an elevator, the ancient stone City Hall, a Yale University building,

or the back room of a restaurant—BSWD failed all five while ARDIS was completelysuccessful

Most carriers provide coverage maps that are easily accessible on the Web Thequality of these maps ranges widely: some useless, some misleading, a few that areuseful There is still value in examining them, even when the greatest payoff is inunderstanding what is not mentioned A representative sampling of each class ofcoverage map follows as a guide for more determined study

seem to reveal something but, upon closer study, provide very little real insight intorealistic coverage expectations The maps are generally blobs or clouds of a ratherlarge geographic area No ability is given to the user to zoom in on specific cities Nomap scale is present for orientation, and geographic landmarks are generally limited

to interstate highways The coverage blobs are not identified as either contours forhigher powered mobile devices or the more limited range portable or hand-held units.Examples of this breed include:

1 Ameritech (Figure 12-8): There are only four maps in the entire repertoire Thesouthwestern Ohio area, spilling over into Indiana and Kentucky, embraces

~20,000 square miles One can guess at the main coverage by tracing theInterstates on a road atlas But there also odd splatter patterns aroundWashington Court House south to Greenfield and near the Hocking State Forest(I think; no towns or natural landmarks are on the map)

2 GTE Mobile (Figure 12-9): These maps have a scale and cover more limitedareas, typically ~6500 square miles, which is about 30% more land area thanthe entire state of Connecticut But since there is little detail, one is unsureabout, say, the San Mateo county edges in the San Francisco/San Jose map AreSaratoga and Los Gatos in or out? The map does not say

3 Bell Atlantic Mobile (Figure 12-10): BAM’s maps are especially disappointingsince they initially began with considerable detail Individual base stations wereidentifiable by the separated circles or arcs Areas without any coverage wereclear In a retrograde action BAM threw out all the detail and replaced it withcrude area maps like those shown for the New York City metropolitan area, anarea of 7600 square miles

12.3.2.2 Useful But with Careless Errors The second class contains scaled,sequential maps of ever-increasing detail, often to less than 50 square miles Not onlyinterstate highways but also secondary state and county roads often appear Theprimary example of this category is ARDIS One can progress from the 50 UnitedStates and its territories down to towns of less than 5000 people An example is shown

in Figure 12-11

This particular base station is a low-elevation “gap filler.” During September 1998field tests the signal strength was –74 dBm inside the windowless barroom of the225-year-old Griswold Inn in downtown Essex Two miles away, at the intersection

of Route 9 and I-95, it was –104 dBm Three miles away, west along I-95, the RSSIwas –111 dBm Contact was lost at the 4-mile mark

There are problems with the ARDIS coverage maps First, they do not statewhat to them is obvious: ARDIS contractually guarantees 90% probability ofsuccess for hand-held devices operating within its coverage contours That is whatthe maps depict, but the legend says nothing about this crucial piece ofinformation

Second, the map maker (American Digital Cartography) sometimes makesridiculous errors The expansion view for New London shows the path of the

Unionville Center

Fairfield

Jeffersonville Kettering

Gahanna Kirfersville

Lancaster Columbus

Cincinnati

Dayton

75 71

Figure 12-8 Ameritech: Southwestern Ohio CDPD coverage (8-26-98).

12.3 ESTIMATING COVERAGE WITHOUT FIELD TESTS 183

Morgan Hill Campbell

Sunnyvale Pacifico

Miles

30

Hayward Fremont San Bruno

San Ramon Berkeley

San Rafael Novato Petaluma

Vallojo Napa Fairfield

Pleasant Hill ArtisonWalnut Creek

Danville

Livermore Pleasanton

San Mateo

Palo Alto

San Jose San Francisco

Hampstead West Babylon Freeport

Hampton Bays Pine Valley Wading River

495

87/287

95

0 7.5 Miles 15

Figure 12-10 BAM: Greater New York City metropolitan area coverage.