Coverage Modeling for State of Vermont

The Vermont Center for Geographic Information (VCGI) is the State of Vermont’s designated agency to administer an ARRA grant from the National Telecommunications and Information Administration (NTIA) under the Broadband Technology Opportunities Program (BTOP).  The purpose of this NTIA program, called the State Broadband Data & Development (SBDD) Grant Program, is to map broadband Internet access in the State of Vermont, where broadband is defined as at least 768 kbps on the downlink and at least 200 kbps on the uplink.

The broadband wireless mapping effort is part of a larger State of Vermont broadband mapping project.   The purpose of this larger project is to map broadband Internet access in all 50 states so this information can be used by the public and by policy makers to increase broadband access.

Before the Vermont project, it was common for wireless broadband mapping efforts to rely solely on carrier-provided coverage maps to assess broadband access.  Virtually all other states approached the problem this way, but this approach was widely criticized, especially when no assumptions were provided with the maps.  For example, rarely did the carrier indicate what data rates could be expected within the coverage boundary nor did the carrier indicate whether the coverage boundary represented the mean threshold of service or some higher probability of service such as 95%.  In the carrier’s defense, its coverage maps are generated for the benefit of its subscribers and not for the NTIA.

VCGI wanted an independent assessment of wireless broadband data coverage and toward that end, the agency contracted with Pericle to collect transceiver data, model coverage and collect drive test measurements.  The principal objectives were an independent assessment of the carrier-provided coverage maps and actual measurement of wireless data throughput under real-world conditions.   Pericle was chosen because the firm had over 20 years experience in both modeling and drive testing for government and commercial radio users.

The Vermont Telephone Authority (VTA) partnered with VCGI to characterize wireless voice coverage in the State of Vermont.  Voice coverage was outside the scope of the NTIA-sponsored program, but all wireless carriers provide both data and voice services, so there were economies of scale if the projects were done jointly.

Pericle performed both coverage modeling and coverage measurements.  This case study addresses the coverage modeling task.

With assistance from Pericle, VCGI made formal requests to the four carriers for specific cell site information so this data could be entered into a database and used for computer coverage modeling.  All four carriers declined to provide this information but after executing non-disclosure agreements, the carriers did provide electronic versions of their coverage maps (as Esri Shapefiles).  No cell site information was provided by the carriers, but the State of Vermont’s Act 250 database and other information sources were used to determine the location and technical parameters of the carriers’ cell sites.  An extraordinary effort was made to collect the information that was finally used.

Most propagation models assume that the minimum loss is free space loss, given by 22 + 20log₁₀(d/λ) dB where d is the path distance and λ is the wavelength of the radio carrier.  Other losses are added to the free space loss to estimate the total path loss.  Free space loss is easy to compute, so the real problem is to predict the losses due to terrain and clutter. Let’s first address each of these losses and then examine some popular computer models used to predict these losses.

Terrain Loss and Digital Terrain Databases

Terrain loss is primarily diffraction loss and most models use principles of ray optics to estimate diffraction loss.   A popular method for sorting out the best way to treat multiple obstacles is the Epstein-Peterson method.  A computerized diffraction model is of little use without a digital terrain database.  There are several to choose from, some coarse and others fine.  Modern propagation studies should be done with the 30 meter database or its equivalent, if at all possible.  The 30 meter database is also referred to as the 1 second database because a distance of 30 meters is approximately equal to one second of latitude.

Clutter Loss and Clutter Databases

Clutter loss falls into two categories:  foliage and man-made.  Foliage loss is computed from a database of loss factors that are a function of both radio frequency and the type of foliage or it is included in a man-made clutter database.  Man-made clutter includes buildings, vehicles, bridges, etc.  Man-made clutter loss is usually calculated from a clutter database which applies a clutter category to individual tiles (cells) in the geographical area under study.  In the U.S., two land cover datasets are used to specify clutter type:  the USGS Land Use Land Clutter (LULC) database and the National Land Cover Dataset of 2006 (NLCD-06).  NLCD-2006 is available as grid data in which one of 21 land cover types is assigned to each 30 meter square cell.

Propagation Model Used in This Study

For the Vermont study, Pericle used the Anderson 2-D model which is specified in an industry standard, TIA-TSB-88.2-C.  The Anderson 2-D model predicts mean signal level using a combination of free space loss, terrain diffraction loss, and clutter loss.  Terrain diffraction loss is computed using the Epstein-Peterson method.  The NLCD-06 database is used for clutter losses with Table 17 in TIA-TSB-88.2-C mapping clutter category to clutter loss in dB for each of several frequency bands.  When employed by a particular software program, several parameters must be selected by the user to implement the Anderson 2-D model.

Software Tools

There are many software tools available to perform coverage studies and there is no single tool used by all wireless carriers.   For this study, Pericle employed the EDX Signal™ software program.  This program employs sophisticated algorithms and it allows the user to choose from several different propagation models with several user-selectable parameters associated with each model.  The model used for this study was Anderson 2-D with clutter loss.

Parameter Selection

The EDX Signal™ program allows the user to specify terrain databases, clutter databases and several physical and modeling parameters for each study performed.  For mobile wireless, the following databases and parameters were used:

• Terrain database = USGS 30 meter (1 second)
• Clutter database = NLCD-06
• Clutter loss factors = TSB-88
• Transmit EIRP = varies depending on site, see ACT-250 database or other sources
• Transmit antenna pattern = depends on site, see ACT-250 database or other sources
• Transmit antenna height = depends on site, see ACT-250 database or other sources
• Receive antenna pattern = omnidirectional
• Receive antenna gain = -3 dBi (includes 6 dB portable body loss)
• Receive antenna height = 1.8 meters (6 feet)
• Receiver sensitivity = depends on airlink standard, see Table 1

Note that only the downlink path (base station to subscriber) is modeled explicitly.  The carrier designs his network for downlink/uplink balance, so uplink performance is assumed to be roughly the same as downlink performance.

An important model parameter is the receiver sensitivity.  Receiver sensitivity is the minimum signal level (in the absence of interference) required by the subscriber receiver to achieve some performance criterion.

Service Threshold (dBm)  = -174 + 10log₁₀(ENBW) + NF + (C/N)_req

Where -174 dBm/Hz is the noise power density at room temperature, ENBW is the equivalent noise bandwidth in Hertz, NF is the receiver noise figure in dB, and (C/N)_req is the required carrier-to-noise ratio to achieve a channel performance criterion (e.g., BER < 1%).   The receiver noise figure varies by manufacturer and model, but a typical handset noise figure is 8 dB.    The (C/N)_req depends on the airlink standard and the manufacturer’s implementation.  Newer data standards perform very close to the theoretical Shannon limit or about 2 dB carrier-to-noise for the lowest data rate.  Channel bandwidths vary with airlink standard.  1XRTT and EV-DO channel bandwidths are 1.25 MHz, GSM and EDGE are 200 kHz, and UMTS/HSPA is 5 MHz.  Using 1XRTT as an example, we see from the equation above that a handset with a noise figure of 8 dB and a (C/N)_req of 5 dB has a receiver threshold of  -100 dBm.

Pericle collected receiver performance information from handset manufacturers, standards bodies and carriers to estimate the signal threshold for each type of service and for each airlink standard.  Table 1 lists the receiver signal thresholds used in computer modeling.  Only some carriers provided the service threshold used in their data maps (no carrier voice maps were provided), so the values of Table 1 may not match the carrier’s value in all cases.

Table 1 – Receiver Thresholds for Modeling
Carrier	Service Type	Airlink Standards	RX Threshold
AT&T Mobility	Data	EDGE, UMTS/HSPA	-101.5 dBm
	Voice	GSM, UMTS	-105.0 dBm
Nextel	Data	iDEN	-101.5 dBm
	Voice	iDEN	-101.5 dBm
Sprint	Data	1XRTT, EV-DO	-100.0 dBm
	Voice	1XRTT	-105.5 dBm
T-Mobile	Data	GSM/EDGE	-102.0 dBm
	Voice	GSM	-102.0 dBm
US Cellular	Data	1XRTT, EV-DO	-100.0 dBm
	Voice	1XRTT	-105.5 dBm
Verizon Wireless	Data	1XRTT, EV-DO	-100.0 dBm
	Voice	1XRTT	-105.5 dBm

 

In Table 1, the threshold for 3G data service is calculated for broadband data rates, not necessarily the lowest or highest rate offered.

Coverage maps were generated with EDX Signal™ and exported as Shapefiles for use in Esri Arcview.  For the four broadband data carriers who provided maps, both the carrier’s map and the Pericle map are plotted along with the location of known cell sites.  A sample coverage map for Verizon Wireless is shown in the sidebar.   Results were published for all six wireless carriers; the results for Verizon Wireless are summarized below:

• Verizon has the most complete site information of the cellular carriers in Vermont.  This information was obtained from ACT-250 (typically the most complete applications) and from observations during the drive test.  Serving cell locations reported by the Verizon network were not believed to be reliable and were not included.

• Verizon sites tend to be large, most often three sector and with 4 and sometimes 5 antennas per sector.  Verizon makes common use of tree towers (monopines) and these are often quite difficult to spot in Vermont.  Verizon typically occupies the top spot in most multi-carrier sites.

• A number of Verizon sites in Massachusetts and New Hampshire have been identified and included in the coverage map.  No Verizon sites in eastern New York were identified and this contributes to some of the gap between carrier and Pericle predicted coverage along the western Vermont border.

• Verizon makes good use of its licensed cellular and PCS band frequencies with more service provided in the cellular band in northern Vermont and the PCS band used more prominently in southern Vermont.  This is believed to be due to coordination with U.S. Cellular in these areas.

• Verizon provides good data coverage through large portions of Vermont.  76% of the data call attempts during the drive test were successful and 62% of these attempts were completed on infrastructure capable of broadband data speeds (EV-DO network infrastructure).  Of all the data calls attempted, 30% achieved downlink speeds exceeding the NTIA threshold of 768 kbps and 45% achieved uplink speeds exceeding the NTIA threshold of 200 kbps.  For data calls attempted within the Verizon provided data coverage area, 94% of these attempts were successful and 42% of downlink and 63% of uplink tests exceeded the relevant NTIA broadband data threshold.  By contrast, for data calls attempted within the Pericle estimated data coverage area, 92% were successful and 41% of downlink and 61% of uplink tests exceeded the relevant NTIA broadband data threshold.  At a state-wide level, the Verizon-provided and Pericle estimated coverage areas provide similar statistics.

• Verizon has good voice coverage over the majority of Vermont. 80% of the Verizon voice call attempts during the drive test were successful.  When compared versus predicted coverage, 70% of the voice calls were attempted within the Pericle estimated voice coverage area and 96% of these calls were successful.  This discrepancy is due to several sites within Vermont which did not have enough information to be mapped. Three sites along the western Vermont border, north of Orwell are good examples.  The site on Mount Robbins (southeast of Burlington) would help fill in the area between Burlington and Waterbury if it was mapped.  In northeast Vermont, missing sites (likely in New Hampshire) account for the discrepancy in Pericle versus carrier coverage estimation in this area.

For more information, see our report.