TDOA technology for locating narrowband cellular signals: Cellphone location involves several practical and technical considerations. Time difference of arrival (TDOA) technology provides accuracy for locating analog cellphones in urban environments.

Immediately after the first mobile user headed off into the sunset with a cellphone, someone else wondered where he went. Thus, the idea for locating cellphones was born.

Many location techniques have been tried with varying degrees of success. Based on our work with operational cellular systems, time difference of arrival (TDOA) has proven to yield the best results in urban, suburban, and rural areas for analog advanced mobile phone serv-ice (AMPS) cellphones. The following information examines the issues involved in locating cellphones, and supporting evidence for TDOA as the technology of choice for analog telephones is described. The same arguments can be extended for digital cellular and PCS systems, although these systems have not been field-proven yet.

Positioning options Many technologies for locating cellphones are based on modifications to the cellphone itself. Global Positioning System (GPS) technology is the most commonly discussed option. For locating unmodified cellphones, three basic alternatives are available: measuring signal attenuation (Doppler shift), angle of arrival (also called direction of arrival, or DOA), and TDOA. We have focused on TDOA because of its high degree of accuracy and affinity for use in cellular systems.

Time difference of arrival GPS is a TDOA-based system, as are most of the systems proposed for the Location and Monitoring Service (LMS) band recently allocated by the FCC. TDOA systems use location receivers at multiple sites dispersed over a wide area; each of the sites has an accurate timing source. A signal transmitted from a mobile device propagates at 1,000 feet per microsecond to all of the antenna sites, where the signal reception is time-stamped. The differences in time stamps are combined to produce intersecting hyperbolic lines from which the location is estimated.

TDOA systems are subject to many of the same urban multipath problems as angle of arrival systems. Multipath distorts the shape of the signal and the group delay, making it difficult to accurately determine the point in the signal to be measured by all receivers.

For this reason, most TDOA systems historically have been wideband (.1MHz). TDOA technology for narrowband signals was first examined by Turin with acceptable results for many applications. We have been able to both duplicate and enhance the results for AMPS cellphones.

Choosing the channel To maximize the potential for successful location, we chose the reverse control channel for our TDOA system. Some of the reasons are as follows: * Although both channels are always modulated during transmission, the reverse control channel’s modulation characteristics are significantly better. The reverse control channel’s sidelobe peak (i.e., bandwidth) is 610kHz vs. the 66 kHz for the supervisory audio tone (SAT) on the voice channel. The frequency deviation (i.e., amplitude) is 68kHz for the control channel versus 62kHz for the voice channel.

* The control channel is always transmitted at constant power and at the system maximum (typically 600mW), while the voice channel is power-controlled during the transmission over a 27dB range (6mW to 3W). TDOA algorithms depend on adequate signal-to-noise ratio (SNR) at all receiving sites. Although the power modulation does not significantly affect the SNR at the cell site communicating with the cellphone, other location receiving sites can be negatively affected by the wide range in transmitting power.

* Because the reverse control channel is bursty by nature, collisions and cochannel interference statistically occur less frequently than on the reverse voice channels.

* The voice channel has no identifying information associated with it; therefore, a TDOA system must request the cellular phone/voice channel mapping for the specific instant in time that the system will attempt to locate the phone. The control channel, by contrast, always has the identifying information contained in the transmission.

* Cellphones transmit on the reverse control channel when not engaged in a voice conversation. Many applications for which a TDOA system would be deployed should not have to depend on a user placing a call.

The reverse control channel does have one drawback: it cannot be used to locate a cellphone that is engaged in a conversation. The cellular air interface standard is structured so the call setup messages use the control channel, but hand-off messages are sent in-band on the voice channel to conserve resources. Most end-user applications can be easily supported with a so-called “point of origination” location system, but in a few cases tracking during the length of a call would be valuable; therefore, we included use of the reverse voice channel as a secondary means of location.

Cramer-Rao bound Although the AMPS reverse control channel is narrowband (,30kHz), TDOA systems can still locate cellphones with an accuracy of several hundred feet. The Cramer-Rao bound defines the theoretical lower limit of signal measurement accuracy as determined solely by the characteristics of the signal. This bound does not factor in the real, physical effects of actual antennas, analog components and environmental anomalies. The Cramer-Rao bound of any transmission is defined solely in terms of a signal’s bandwidth, transmission time and received SNR at the two antennas of a TDOA baseline. (A baseline is a pair of geographically separate antennas used for the time difference calculation.)

System hardware and accuracy When we began examining existing cellular base station equipment TDOA suitability, we discovered that most equipment was not suited, nor intended, for high-accuracy time measurement. This was not surprising because the demodulation of voice channels for communications purposes requires timing stability only on the order of microseconds. Therefore, new equipment designs were required for TDOA systems. We have developed several generations of test systems, increasing stability and accuracy. Following are the important areas of system hardware that affect accuracy:

* The timing reference used at each of the receiving times must have a low drift rate and low phase noise. Drift rate affects how often the system requires calibration, and phase noise affects the system’s ability to discern multipath effects from other systemic errors. For a baseline connecting any two receiving sites, using a rubidium oscillator as a timing reference was found to contribute less than 20ns of timing error, root-mean-square (RMS).

* Each receiving site in a TDOA system requires a variety of intermediate frequencies for internal mixing and sampling. These frequencies must be derived from the timing reference to avoid introducing additional timing errors. The phase-locked loops used for generating these local oscillators must minimize the phase noise created within the loop. Phase noise introduced within the loop will later affect the cross-correlation processes used for TDOA and the ability of the system to identify components of multipath.

* The reverse control channel band spans 630kHz for a typical 21-channel AMPS system. Most receivers will have several stages of filtering prior to sampling. Depending on the filter response characteristics chosen, a filter can have a significant delay variation across the band, which may vary with time, temperature and other factors. To reduce the filter contribution to TDOA error to 10ns-30ns, the filter requires periodic calibration by both external and internal signal sources.

* The choice of receiver technology also affects the TDOA system contribution to error. TDOA relies on cross-correlation of baselines, in which each receiver at the end of the baseline independently receives the transmitted signal. Wideband digital receivers (.1MHz) offer better performance and matching, both within a single receiver system where diversity antennas are used and between receiver systems at different sites. The wideband digital receiver contribution to error was found to be less than 20ns RMS, not counting quantization errors.

Our testing with TDOA systems has yielded increasingly better performance with each generation. The current system hardware generation is expected to affect accuracy on any baseline by no more than 50ns, RMS. This places the system contribution to error on par with that of the signal (i.e., Cramer-Rao bound) itself.

The effect of multipath Multipath, also known as delay spread, refers to multiple copies of a transmitted signal received at an antenna. Multipath has been generally found to be the single largest contribution to error in a TDOA system. Multipath causes problems in normal cellular system operation because it is possible for a direct path signal to be momentarily canceled out by a reflected (multipath) signal; diversity antennas at cell sites alleviate this cancellation problem because of the statistical independence of multipath even within a single cell site.

The RMS delay spread in an urban environment will vary with the “signature” of each city (Turin and Lee), but it is generally about 1ms to 5ms, which would translate into an error of 1,000 feet to 5,000 feet if only the multipath at a single cell site were considered. Past considerations of TDOA systems have focused too much on this single delay number. This fails to take into account that multiple receiving sites are used to receive the transmitted signal in a location system, while only a single site is used in a cellular system.

The amount of multipath experienced by an antenna can be greatly influenced by the type of antenna (omnidirectional or sectorized), the horizontal and vertical beamwidth of the antenna, and the amount of downtilt. If the antenna has more gain in the direction of the cellphone (i.e., the direct path signal) and less gain in the direction of multipath, then the SNR is effectively boosted. The type of sectorized antennas used at most cell sites (808 to 908 beamwidth horizontal, 108 to 158 vertical) meet this criteria.

In addition to sectorized antennas, there are “smart” antennas that use multiple antenna elements (eight or more) per sector; these antennas use electronic beam-steering to maximize the gain in the direction of the strongest signal and minimize gain elsewhere. [See “Antenna technique boosts capacity and coverage, reduces interference,” p. 80.] If these antennas are installed primarily for voice quality purposes, and if the antennas’ electronic beam-steering can respond quickly enough during the transmission time of a reverse control channel (100ms), then our TDOA systems certainly can take advantage of these antennas.

In general, our test results suggest it is not economically feasible to install these types of antennas solely for location purposes. While the antennas are an enhancement vs. sectorized antennas for wide-angle multipath, the antennas do not provide much assistance for narrow-angle (,158 between the reflected signal and the direct signal) multipath. Statistical analysis shows that narrow-angle multipath makes up a significant portion of multipath. This is because more reflectors are located nearer to the cellphone’s transmitting antenna (at street level) than are located near to the receiving antenna atop a cell site.

Geometric dilution of precision Geometric dilution of precision (GDOP) refers to the geometry of a TDOA system’s receiving antennas relative to a transmitting telephone. The geometry can significantly affect location accuracy, equal to or greater than the effect of multipath. GDOP can be viewed as a multiplier that can either boost or degrade the performance of a location system. For example, if the TDOA process has measured every baseline in a location estimate with an accuracy of 200ns (200 feet), but the geometry of the antennas resulted in a relative GDOP of 3.0, then the final location estimate could have an error of as much as 600ns (or 600 feet).

Modeling shows that the receiving antennas used for a location estimate should be evenly distributed in a circle around the cellphone. Practically, this is not always the case, but a good TDOA design should reflect the best geometric orientation to increase the odds that every location estimate has good GDOP. This has required the development of new modeling software that parallels the propagation models first created for cellular systems.

Other effects of system design A TDOA system must be designed for locational accuracy using techniques similar to those used for implementing a cellular network for voice. However, the techniques are not identical, primarily because a cellular network is designed using the forward channel as the driving criteria, while a TDOA system uses the reverse channel as the driving criteria. In general, the design coverage should attempt to yield a 20dB-25dB SNR (before integration and processing gain) at more than three receiving sites for any cellular reverse control channel transmission from any point in the covered area.

Some of the design factors that influence the location accuracy of a TDOA system are: * number of receiving locations. * number of diversity antennas at each site. * average distance from the transmitting cellphone to each of the receiving sites. * average height of the receiving antennas. * average antenna power gain in the direction of the transmitting cellphone.

Results from testing TDOA systems Using narrowband systems from the early 1970s, Turin reported location error results from urban environments of about 870 feet RMS. Using TDOA systems with analog receivers and GPS as the timing reference in a semi-urban and suburban environment, we have reported results of 355 feet RMS. Using TDOA systems with analog receivers and rubidium oscillators as a timing reference in a dense urban environment (high-rise, mid-rise, and low-rise buildings) with significant multipath, we reported results of 200 feet RMS. These tests with analog receivers were to verify the performance of previous TDOA systems before proceeding with more costly and more advanced wideband development.

References Lee, W.C.Y., Mobile Cellular Telecommunications Systems, McGraw-Hill Book Company, New York, 1989. Sakagami, S., et al, “Vehicle Position Estimates by Multibeam Antennas in Multipath Environments,” IEEE Trans. on Vehicular Technology, Vol. 41, No. 1, February 1992. Smith, W.W., “Passive Location of Mobile Cellular Telephone Terminals,” IEEE 1991 Proc. on Security Technology, October 1991. Tanaka, T., et al, “Effects of Antenna Beam Horizontal Rotating and Beam Tilting on Delay Spread Reduction in Mobile Radio,” IEICE Trans. Communications, Vol. E76-B, No. 2, February 1993. Turin, G. L., et al, “Simulation of Urban Vehicle-Monitoring Systems,” IEEE Trans. on Vehicular Technology, Vol. VT-21, No. 1, February 1972.