HOMING IN ON WIRELESS LOCATION
In the choice between GPS-based and network- based methods for locating cellular callers, GPS is viewed as coming up short. However, an enhanced GPS approach can produce fast, reliable fixes.
By Norman Krasner
When it comes to implementation issues, wireless handset location presents the wireless industry with some critical choices. Wireless technology is fundamentally different from that of wireline networks, so
implementing wireless device location goes beyond simple software upgrades. Fortunately, the technology exists to locate callers inexpensively and with precision, even in the most challenging wireless environments.
Whether wireless location determination is being used for 911 emergency services or value-added services such as driving directions, mobile yellow pages, or roadside assistance, the challenge of providing location data is twofold. First, the chosen technology must be able to locate callers where
wireless users place calls. Second, it must provide location data that is accurate enough to enable lucrative location-based service offerings. Technology that is unable to locate callers inside skyscrapers, shopping malls, parking garages, and in densely wooded areas will be impractical for location service offerings. Inferior technology will draw considerable ire from customers paying fees for location-based services.
Many solutions for implementing wireless location are available. Some vendors offer
network-based radio location systems that track the signals from wireless callers. Others offer handset-based solutions that rely on the US military’s Global Positioning System (GPS) satellites. But only one technology, enhanced GPS (EGPS), has demonstrated the ability to locate callers quickly and accurately in nearly every calling environment, including locations where signal blockage and the incidence of reflected signals run high.
EGPS
GPS has been viewed with skepticism in discussions about
wireless location. Traditional GPS has a poor track record for dealing with weak signals and multipath errors that occur when signals reach a site through two or more paths. As a result, traditional GPS solutions can’t locate callers in places where the GPS receiver doesn’t have a direct line of sight to the GPS satellites.
EGPS technology improves on conventional GPS performance by:
- Sharing processing functions between the digital signal processor (DSP) built into
most wireless handsets and a network server operating within a carrier’s network
- Using information available from the wireless network itself
- Processing only a snapshot of data in software, rather than processing it continuously in hardware
The network server, running NT-class software, is linked to a fixed reference receiver that constantly sends the system updates of all GPS satellite positions. When a wireless user requests a location-based service or dials
911, the handset queries the server and receives about 50 bytes of aiding data for all GPS satellites within its view. The handset takes a snapshot of the GPS signal, uses the aiding data to process it, and transmits location information to the server. The server computes longitude and latitude data and performs a series of calculations that significantly improve the accuracy of the location fix.
EGPS is superior to traditional GPS on two levels. First, the server allows improved sensitivity and
acquisition times and enhances system accuracy. Second, by utilizing a snapshot and the processing power of the DSP, these benefits are made practical to implement in a handset.
Highlights of EGPS technology are:
- High sensitivity.
EGPS can acquire and provide fixes in conditions with as much as 25-dB signal attenuation or blockage. Traditional GPS technology can have difficulty acquiring signals when the attenuation exceeds 5 to 10 dB. This increase in signal sensitivity allows EGPS
to operate in difficult environments such as building interiors, automobiles, dense foliage, and urban canyons.
- Low time-to-first-fix.
Traditional GPS receivers require from 30 sec to several minutes to acquire and track satellites, depending on the receiver’s environment and how much information it has previously gathered. In worst-case environments, however, EGPS provides a first fix in a few seconds. In open-sky situations, EGPS can acquire a first fix from a cold start in less
than 1 sec.
- Low power dissipation.
EGPS technology uses a snapshot of data — typically 0.1 to 1 sec, depending on the sensitivity required. After this information is obtained, the RF unit powers down. The entire location operation takes only a few seconds for a cold start in a heavily blocked signal environment, and is significantly faster if prior information (such as local oscillator offset) is known or signal strength is high. In applications that do not require continuous
high-rate positioning, the low duty cycle EGPS receiver dissipates a small fraction of the power of a portable wireless handset.
Conventional GPS processing
GPS functions can be broken into four primary areas:
- Determining the code phases (pseudoranges) for the various GPS satellites
- Determining the time-of-applicability for the pseudoranges
- Demodulating the satellite navigation message
- Computing the position of the receiving
antenna using the pseudoranges, timing, and navigation message data.
Most commercial GPS receivers perform all of these operations unaided. In these conventional receivers, the satellite navigation message and its inherent synchronization bits are extracted from the GPS signal after it has been acquired and tracked. The level of received signals from all satellites must be high — approximately -135 dBm — to allow a navigation solution.
Distributed system concept
In EGPS,
the server-aided system architecture distributes the four primary functions previously described between a GPS reference receiver, a location server, and a GPS-enabled wireless handset.
Figure 1
describes the basic system model.
The GPS reference receiver gathers navigation message and differential correction data for all of the satellites in view. To serve a larger area, such as the continental US, the GPS reference receiver could be replaced by feeds from a wide area
reference network (WARN) of distributed receivers.
The network server gets and stores data from the GPS reference receiver(s) and provides aiding data to mobile units. The server also performs navigation solutions (calculating longitude, latitude, and altitude) after receiving pseudorange measurements from the handset. In general, the location server is remote from the final application. An example of this application would be a service center providing display and operator services.
Aiding data is
sent to each handset on demand. The handset’s approximate location (the location of the basestation communicating with the handset, for example) is used to identify what aiding data the server should send. Although the message is small, the aiding data provides all the information the handset needs to compute pseudorange information from a snapshot of GPS data.
The server also performs a series of calculations that provides complex error correction. For instance, the server has access to a
terrain elevation database that provides accurate altitude aiding for ground-based applications. This allows EGPS to have improved accuracy in urban environments.
Terrain elevation provides an extra range measurement that improves overall reliability and accuracy. In independently audited tests of EGPS, altitude aiding resulted in a 30% accuracy improvement in the urban canyons typical of most downtown environments. The tests confirm that altitude aiding is important in weak signal environments because it
improves satellite geometry and provides a reliable extra measurement in the navigation solution.
Another departure from traditional GPS is the distributed system’s ability to mitigate multipath and reflected signal effects. This is done at the server by properly weighting pseudorange informationthrough the use of a sequential measurement optimization (SMO) technique. The server also eliminates cross-correlations from strong signals onto the pseudorandom noise (PRN) codes of weaker satellites, and
corrects for atmospheric delays.
Because it doesn’t need to decode the navigation message, an EGPS-enabled handset can track far weaker GPS signals than a conventional GPS receiver. Rapidly and accurately identifying these weak signals requires a powerful signal-processing element that can search over the large number of PRN codes, times-of-arrival, and off-set frequencies that result from Doppler errors and local oscillator frequency errors.
Conventional GPS correlation
Conventional GPS receivers use correlation methods to compute pseudorange. A classic hardware correlator-based receiver multiplies the received signal by a stored (or generated) replica of the appropriate PRN code and then integrates, or low-pass filters, the product to obtain a peak correlation signal. The initial determination of the presence of a correlation peak is called an acquisition.
Once a signal is acquired, the process enters the tracking mode in which the PRN code is removed (or despread). This
signal has a narrow bandwidth commensurate with the 50-bps navigation message modulated onto the GPS waveform. After the PRN code is despread, the navigation message may be reliably demodulated if received signal strength is above approximately -135 dBm for the duration of the message being received.
The conventional acquisition process is time-consuming, especially if received signals are weak. To improve acquisition time, most GPS receivers utilize multiple correlators, usually at least thirty-six for a
12-channel receiver. The use of multiple correlators allows a parallel search for correlation peaks as a function of time-of-arrival, PRN code, and frequency offset. In some cases, massively parallel (up to 240) correlators have been used to improve acquisition speed and sensitivity. However, it would take more than 8,000 hardware correlators to match the speed and sensitivity of the fast convolution processing technique available in EGPS systems. The performance of fast convolution processing will improve
even further as DSP technology advances.
The EGPS handset
Figure 2
is a handset view of an EGPS system. Note that the conventional tracking loops are replaced by snapshot memory and fast convolution processing. In a handset implementation, most of the digital hardware would be placed on the cell phone’s baseband digital chip.
When a wireless caller requests location-based services (such as 911 assistance or driving directions from a third-party
provider), the server sends aiding data, including Doppler predictions, from satellites within view to the handset’s approximate location. After a snapshot of GPS satellite RF data has been stored in the handset’s memory, the RF unit powers down and processes the data using a DSP. Pseudorange measurements are calculated and returned to the server, along with other relevant statistical information. This snapshot approach allows the handset to gather GPS data when it isn’t transmitting, eliminating
potential self-interference.
Each message sent between the handset and the location server is small (50 to 100 bytes). This represents a significant reduction in the required communications bandwidth when compared to the bandwidth needed to deliver differential corrections, almanac, ephemeris, and/ or satellite trajectory data to the handset.
In some EGPS systems, fast convolution processing techniques provide higher sensitivity and faster acquisition speed by performing a large number of fast
Fourier transform (FFT) operations together with special pre- and post-processing operations. It is expected, in a worst-case scenario, that a pseudorange calculation in a 66-MIPS DSP will require approximately 5 sec, with much faster performance in cases where higher-performance DSPs are used or when signal strength is high.
In the system pictured in
Figure 2
, received data is down-converted to a suitably low intermediate frequency (about 2 MHz), digitized, and stored in a
buffer memory. The data is then operated on using a programmable DSP integrated circuit. Unlike the continuously-tracking, hardware correlator-based receivers, the snapshot processing software used in EGPS is not subject to the fluctuating signal levels and changing nature of the signal environment.
Pseudorange computation
Each received GPS signal (C/A code) is constructed from a high-rate (1-MHz) repetitive PRN pattern of 1,023 symbols, commonly called chips. These chips resemble the waveform
shown in
Figure 3
. Further imposed on this pattern is low-rate data, transmitted from the satellite at 50 baud. All of this data is received at a very low signal-to-noise ratio (SNR), measured in a 2-MHz bandwidth. If the carrier frequency and all data rates are known with great precision, and no data is present, then the SNR could be greatly improved by adding successive frames.
For example, there are 1,000 C/A code epochs over a period of 1 sec. The first such epoch
could be coherently added to the next, the result added to the third, and so on. The result would be a signal with a duration of 1,023 chips. The phasing of this sequence could then be compared to a local reference sequence to determine the relative timing between the two, thus establishing the pseudorange. Doppler and local oscillator uncertainty complicate this process. The server reduces these uncertainties by providing Doppler estimates.
Pseudorange determination is carried out separately for each
satellite in view from the same set of data in the snapshot memory, since the GPS signals from different satellites generally have different Doppler frequencies and different PRN patterns.
The presence of 50-baud data, superimposed on the GPS signal, limits the coherent summation of C/A code epochs to a period of 20 ms. This means that at most, twenty 1-ms epochs may be coherently added before data sign inversions prevent further coherent summation (unless this data is provided by the server). Additional
processing gain may be achieved by adding (or squaring) the magnitudes of the coherently summed intervals, providing the sensitivity and accuracy shown in the curves of
Figure 4
and
Figure 5
.
Network-based solutions
A discussion of the technical aspects of wireless location determination would be incomplete without mention of network-based radio location technology. This technology is fundamentally different from a handset-based
GPS solution. Essentially, network-based systems turn each basestation in a carrier’s network into a radio tracking post. As a result, they require the installation of $15,000 to $75,000 worth of hardware and software at each cell site. Furthermore, because cell sites in existing networks are situated to optimize voice coverage area (instead of optimizing location) new receiver sites would be required in many places.
While GPS also relies on radio location technology, the perfect spacing of GPS
satellites optimizes the geometry for pinpointing callers, and the system is accessible anywhere a wireless call can be placed. This eliminates the need to spend millions (if not billions) of dollars for ground-based tracking gear that will never replicate the optimal triangulation characteristics of GPS.
In addition to enormous costs up front, perhaps the greatest drawback for network-based systems is their poor performance, particularly in dealing with multipath. All network-based technology tested to
date has performed inadequately in densely-crowded urban areas, where multipath is most problematic. Even if carriers could turn a blind eye to the staggering starting costs of network-based technology, the technology’s inability to deal with multipath eliminates it from serious consideration as the only basis for a robust location-based service offering.
EGPS, E911, and wireless location services
The force driving US wireless carriers to implement location technology is the Federal
Communications Commission (FCC) mandate, requiring wireless carriers to provide the location of wireless 911 callers to emergency service dispatchers by October 1, 2001, with an accuracy of better than 125 m at least 67% of the time. The performance of an EGPS system (predicted in
Figure 4
and
Figure 5
) has been verified experimentally in independently-audited field tests. The accuracy of the EGPS system varied from 3 m in open-sky situations to 100
m in environments where severe signal attenuation and multipath made any other location technology impossible. Thus, where all other technologies fail, EGPS provides substantially better location information than the 125 m required by the FCC mandate.
In overseas markets, the impetus behind wireless location is the provision of location-based services. Here in the US, EGPS is the logical technology choice. EGPS is the only technology proven to deliver accurate location data to users in harsh
environments. When users are paying for service, they will demand quality and coverage. EGPS is the only technology that has shown it can deliver both.
Norman Krasner is vice president of technology at SnapTrack, Inc. in San Jose, CA. He holds an MS and a PhD in electrical engineering from Stanford University, and a BS in electrical engineering from MIT. He can be reached
at
nkrasner@snaptrack.com
.
Return to
Table of Contents
Puzzled by a network processing design issue?
