There are many commercial and military applications that require knowledge of precise location. Applications include finding children, lost pets and luggage, tracking search-and-rescue personnel or fire fighters, locating inventory in large warehouses or cargo containers, and rescuing military personnel in hostile territory.

This article describes integrated ultra-wideband transceivers, called localizers, for precise position location and low data-rate communication.

Localizers determine location by sharing range information within a network of units distributed in the environment. The range between pairs of localizers is determined by cooperatively exchanging ultra-wideband signals consisting of coded sequences of impulses.

As localizers are activated, each acquires as many contacts as possible. As local groups of nodes form into clusters, nodes in one cluster link with one or more nodes in other clusters, forming bridges between the clusters. Range information is constantly shared so that all localizers are aware of all other localizers in the network. Using precise timing techniques, the localizers are able to establish these ranges to an accuracy of about a centimeter.

The advantages of complementary metal oxide semiconductor (CMOS) ultra-wideband (UWB) localizer technology are:

• accurate position;
• low cost, low power units, potential for single-chip CMOS system;
• minimal interference with other communications systems due to UWB technology and low energy;
• with their wideband spread spectrum nature, the units are more tolerant to interference and multipath, and offer inherent privacy;
• UWB penetrates most substances, and can operate effectively in RF-hostile environments; and
• the network approach allows units to be great distances apart without requiring the power to directly reach that distance.

The block diagram of a localizer is shown in Figure 1. The technology used is based on the transmission of coded sequences of gaussian impulses and their reception and detection using correlation and pattern recognition. The code sequences are similar to those used in spread spectrum communications, but they modulate pairs of impulses, called doublets, which are generated by applying current steps through a current mode antenna.

For transmission, the basic task is sending coded sequences of gaussian impulses at precise times. For reception, the basic task is correlating the antenna signal against a known code sequence with precise alignment in time.

Since the transmission of a series of impulses would normally require current steps of ever-increasing amplitude, we devised a combination of a positive and a negative impulse to limit the peak current. A typical doublet, with an impulse spacing of 5 nsec, is shown in Figure 2a. A doublet is equivalent to a chip in spread spectrum terminology. An example of a doublet sequence is shown in Figure 2b.

To transmit a code sequence, we transmit a train of doublets. The impulses that form a doublet are baseband signals, without any carrier frequency, and have an ultra-wideband frequency content due to their short time duration.

For receive, the signal is picked up by a low impedance, current mode antenna and amplified by an RF amplifier before sending it to a correlator. In a typical spread spectrum receiver, the correlator slides the received signal past a reference code sequence. When the received signal and the reference code are in phase, a correlation peak emerges.

The localizers utilize a wideband analog signal with a spectrum of about 100 MHz to 1 GHz. The most efficient way to correlate against a signal with this bandwidth, and still keep the system low power and low cost, is to use the dual form of a correlator called a time-integrating correlator (TIC).

In this approach, the reference code is moved past the analog input signal as the product of the signal and code are summed in a set of analog integrators. A simple digital delay line for the local code sequence, which can be implemented with shift registers, replaces a hard-to-implement analog delay line for the received signal.

The output of each integrator represents a different phase alignment of the reference code with respect to the input signal. The operation of the TIC is shown in Figure 3.

In this example, the integration period is equal to the impulse separation, and the overlap between adjacent phases is equal to half the impulse separation. The discrete outputs from a TIC with multiple integrators correspond to samples of the continuous output of a sliding correlator. These samples can be processed subsequently in software to locate the correlation peak.

Unlike a sliding correlator that can detect the presence of a code sequence on the fly, a TIC needs an approximate time to position its correlation window to detect a peak. This knowledge comes from the acquisition process. Described here is a method to reduce the time required to achieve synchronization.

The localizer system makes use of CDMA to allow multiple localizers to communicate at the same time in the same area. Gaussian impulses are transmitted in a coded sequence, which are pseudo-random codes, similar to those used for direct-sequence spread spectrum systems such as GPS. Different sequences provide separate channels roughly equivalent to frequency bands.

Correlation is used to discriminate a particular code sequence from other signals, both non-sinusoidal and sine wave frequency-based. When a coded sequence of gaussian impulses is received by a localizer, it is compared to a locally generated reference sequence in a time-integrating correlator.

The output from the correlator is a function of the relative time-shift between the received signal and the reference signal. When the shift of the input signal matches that of the reference code, a correlation peak is produced. Consequently, the separation distance between localizers can be determined from the time domain output of each correlator.

Ideally, codes should be chosen so that the correlation of a code sequence against itself will have a single peak, making it easy to determine when the proper sequence has arrived. Maximal sequence codes and complementary codes approach this ideal.

The cross-correlation of one code sequence with a different sequence should not have correlation peaks for multiple localizers to operate at the same time. Fortunately, families of codes exist with tens of thousands of members that have both good autocorrelation and good cross-correlation properties.

Localizers, like other spread spectrum radios, have to find the precise alignment of the received signal with the reference sequence to detect a correlation peak. Without any prior knowledge, or an alternate back-channel, a spread spectrum receiver has to search through all possible times that a transmission may occur.

Using exhaustive search, the worst-case time required for synchronization grows as the square of the transmission repetition interval. Thus, a repetition interval 10 times as large requires 100 times the acquisition time. Also, the acquisition time is proportional to the fraction of each interval that the receiver can examine (such that 100 repetitions are required if 1 percent of the transmission interval can be examined each time). For low duty-cycle communication, this can lead to unacceptably large acquisition times.

Localizers need a rapid acquisition technique because the transmissions and receptions are spaced apart (episodic), and the correlation window spans a small fraction of the interval between transmissions. The latter limitation comes from what can be practically integrated on-chip. The reason localizers use episodic communications is to support very low power operation and precise position location in a multipath environment.

Extremely low power consumption for distributed unattended sensor networks ultimately comes from low duty-cycle episodic operation. This means reception, as well as transmission, is powered and active only during predetermined periods.

Accurate ranging for determining precise position location requires detecting the direct path signal sent between localizers. Other radios typically detect the strongest signal, which can be a multipath reflection. Localizers use episodic transmissions spaced further apart than the typical delay spread. Thus, the multipath from a previous transmission will have dissipated before the next transmission, and the direct path signal will not be obscured by multipath.

To better understand why exhaustive search is unacceptable as an acquisition technique, some details need explaining. Refer to Figure 4.

Localizer control software follows a repetitive timeline. The length of the timeline is referred to as an era. An era is composed of “N” packets, where N is a power of two (shown as 16 in Figure 4). Each packet consists of “K” epochs (shown as 32 in Figure 4). Each epoch is nominally one millisecond in length. One of the goals of the timeline approach is to minimize communications overlap between localizers within range of each other. To accomplish this, each epoch is further divided into 32 time slots 31.25 µs long.

The search process begins when one localizer randomly decides to begin transmission of a periodic signal. The localizer that performs the beacon process is referred to as the pinger. This node selects a minimally used packet, and transmits a single bit once every available epoch until it is detected by another node.

At the point of detection, the two nodes begin a protocol that culminates with the synchronization of the two nodes (acquisition). The goal of the searching node, or listener, is to find and acquire each pinger node as quickly as possible.

The PN (pseudo-noise) code sequences that we use for ranging and communication need to have certain autocorrelation and cross-correlation properties.

Specifically, the autocorrelation of a given code must have a single correlation peak with minimal sidelobes, and the cross-correlation with other members in a family can be no higher than the sidelobes of the autocorrelation.

Fortunately, Kasami codes have large families that match these requirements¹. However, these properties mean that a transmission can only be detected if it falls within the correlation window.

In the current generation of localizer chips, the correlation window is only 80 ns in length, and consists of 32 correlator bins. Thus, if the pinger transmitted once every epoch, or once every millisecond, then the listener would need to search as many as 16,667 times (includes 20 ns window overlap). Acquisition in this manner would require nearly 17 seconds (one millisecond per receive).

The key concept for the rapid acquisition process is finding a way to examine a larger fraction of the interval between transmissions with each reception. Imagine that there exists a PN code sequence that has autocorrelation sidelobes that can be detected with only partial overlap of the transmission and reception code sequences.

Then the lack of any detectable signal would mean that the reception time could be advanced with a step size much larger than the width of the correlation window. The other important property of this imagined PN code sequence is that its cross-correlation with the codes used for ranging and communication will be no worse than other code family members.

After extensive simulations and studies, we discovered a beacon code with the necessary properties. This beacon code is formed from 2^n/2 + 1 repetitions of the maximal sequence of length 2^n/2 - 1, which is one of the generator polynomials for a Kasami code of length 2ⁿ - 1.

Our application typically uses n = 10 to yield a code length of 1,023 doublets, or 33 repetitions of a 31-length maximum sequence code. The worst-case cross-correlation of this code with members of the small Kasami family is the same as the rest of the family members, which is approximately N^1/2.

Figure 5a illustrates the autocorrelation function of the beacon code, which is transmitted over approximately 10 µs, as would be seen by a series of localizer-receptions spanning 20 µs. Figure 5b illustrates an expanded view of the sidelobes, which occur at precisely-spaced intervals of 310 ns. Note that the amplitude of the correlation increases until the received code aligns with the expected code, which is a key component of the rapid acquisition process.

Use of this beacon code affords the following advantages:

• low cross-correlation, so other local receivers are minimally affected;
• significant cyclic autocorrelation sidelobes, for easy detection of the beacon; and
• a well-defined autocorrelation sidelobe structure that enables the rapid detection of the signal peak.

Since a sidelobe detection occurs every 310 ns (with increasing/decreasing amplitude) in a 20 µs window of time, it is only necessary to receive data for a contiguous 310 ns period once every 20 µs.

However, since the sidelobes are low at the edges, and may not be detectable through the noise, the search algorithm examines a 310 ns window once every 10 µs. Since the correlator window size is 80 ns, and some overlap is desired, this sidelobe search is accomplished by increasing the time (modulo 1° ms) in 60 ns increments for five consecutive windows. Refer to Figures 6a and 6b.

Each of these five windows is sampled over five contiguous epochs, which spans 5 ms. The time is then incremented by 10 µs, and the process is repeated. Thus, 100 windows, spaced 10 µs apart, are required as a maximum to detect one of the sidelobes of the beacon signal. Five receives are performed during each of the 100 windows, resulting in 500 receives by the Listener. Since each receive requires one epoch (1 ms), the maximum time for the listener to detect the pinger is 500 ms. This is faster by a factor of 33 over conventional search techniques. Figure 7 illustrates this rapid acquisition process.

Localizers can use ordinary crystal oscillators for two reasons.

• Localizers can exchange signals even if their clocks differ, as long as the differences do not accumulate over the short duration of an exchange. A good crystal oscillator (about 1 part per million accuracy and stability) is more than sufficient.
• In the process of exchanging signals, cooperating localizers can determine the amount by which their clocks differ, and can compensate for the different clock rates. This is possible because the short-term stability of a crystal oscillator can be much greater than its absolute accuracy.

For example, assume two localizers have determined the relative clock rate of each other's clocks. Each localizer can then measure range with a fractional error given by the fractional error in its clock during an exchange of signals. Within a network of localizers, only one localizer needs to have a very accurate clock for other localizers to absolutely calibrate their clocks. Independently knowing the distance between two localizers can also be used for calibrating the clocks of all communicating localizers.

Even with a perfectly accurate clock, a localizer will have unknown circuit delays that will affect the measured time-of-flight delay for a signal. As long as these delays are relatively stable, they can be measured and factored out of the range computation. This self-calibration technique requires a localizer to receive the same signal it sends. All the measured delay will then be due to circuit delays.

As previously illustrated, the fundamental problem in determining position is to obtain a measurement of the range between localizers in the system. The range between any two localizers is determined by measuring the round-trip transit time for a signal, dividing by 2, and multiplying by the speed of light. Figure 8 shows a typical ranging transaction.

Ranging requires sending an encoded signal from a first localizer to a second one, and then getting a different encoded signal back. Note that unlike radar, the return signal is not the echo, but a retransmission of a new code.

The process described in this article takes a systems approach to the design of localizers. Only the most time-critical and processing-intensive functions are performed in hardware, and most of the system complexity is handled in software. As much as possible, the sequential operation of the hardware is controlled by a general purpose processor.

This way, the details of operation can be worked out over time, and can be quickly changed if there are unanticipated problems. For transmission, the basic task is sending coded sequences of gaussian impulses at precise times. For reception, the basic task is correlating the antenna signal against a known code sequence with precise alignment in time.

The same type of pseudo-random noise (PRN) code sequence generator is used for both the transmitter and receiver sections. The design is a fully programmable 25-stage linear feedback shift register (LFSR). It is a universal code generator capable of generating maximal sequences, small Kasami codes, large Kasami codes and large Kasami-like codes, gold and gold-like codes, and more.

The receiver section includes the receiving antenna amplifier, the code sequence generator, and the time-integrating correlator (TIC) with 32 integrators. The receiving antenna amplifier and the code sequence generator feed the TIC with the amplified received signal and the reference code, respectively.

The same code sequence is fed to each integrator in the TIC, but delayed by a different amount for each “phase.” The analog-received signal is multiplied by the delayed digital code sequence and summed at the output of each integrator.

A phase corresponds to an integration window 5-ns wide, but the phases are spaced 2.5 ns apart, so their windows overlap.

Each integrator output is sampled by a sample-and-hold (S/H) circuit at the end of the code sequence. The outputs of the S/H circuits are multiplexed onto a common analog bus to be digitized by the analog to digital converter (ADC), which is read by the processor. Thus, after a reception event, the processor has 32 values that are equivalent to sampling the output of a sliding correlator during an 80 ns window.

The development of the ultra-wideband large current radiator (LCR) antenna by Dr. Henning Harmuth made it possible to radiate nanosecond-wide impulses with inexpensive complementary metal oxide semiconductor (CMOS) chips, using a physically small antenna.

The only restriction on its size is that it cannot be made larger than the equivalent width of the impulse being transmitted. The antennas on the fifth generation localizer prototype are evolved from the LCR, but use the same operating principles. Essentially, the current through a closed loop, and hence the radiation, can be controlled directly when the antenna is driven by a very low impedance source.

The antenna-driving output of the transmitter section looks like a standard H-bridge, as is commonly used to drive stepping motors. Each bit in the code sequence determines whether the current initially flows one way or the other through the bridge. A “surge” (derivative) of current through the antenna is created by connecting opposite sides of the antenna to VDD and Vss, thereby launching an impulse.

A reverse change of current, and an opposite polarity impulse, is generated by connecting both sides of the antenna to the same supply, VDD or Vss, which means there is always a closed path for current to flow. In other words, we generate a step change in current through the transmit antenna by causing a step change in voltage across the antenna.

Each of the four arms of the H-bridge has two programmable delay elements (one for each edge) to adjust making and breaking the switch connection in the respective output transistor. This can compensate for the different switching delays of the P-channel versus N-channel transistors, and for other circuit mismatches. Each arm is also divided into multiple (eight) sections that can be individually enabled for power control.

Localizers can use ordinary crystal oscillators because even if their clocks differ, localizers can exchange signals over the short duration of an exchange, and cooperating localizers can determine the amount by which their clocks differ, and can compensate for the different rates.

We achieved 5 ps timing resolution without using a 50 GHz clock. A stable crystal oscillator constantly clocks the coarse resolution part of a real time clock counter, and a PLL derived from this provides our 200 MHz clock. The 5 ns period of this oscillator is further divided by selecting a tap around the ring, and by using a programmable delay generator with 5 ps (or finer) resolution.

In large networks, it is inefficient if all nodes communicate with all other nodes since the number of links required multiplies with the number of nodes (possible (N*(N-1))/2 links) Therefore, to conserve links in a large network, the nodes, or localizers, aggregate into clusters of communicating nodes that have achieved a consensus clock.

Such a network of clustered nodes, presented as a concept of operation for a localizer-based personnel identification and location system for military or fire fighter applications, is presented in Figure 11.

These clusters have the following characteristics:

• Ad-hoc acquisition of new nodes and departure of current nodes (such as mobile ad-hoc network “manet”).
• Slot/packet alignment and multicast is possible within a cluster.
• Clusters can do beam-forming and/or focusing for increased range.
• Nearest-neighbor communication with alternative paths if nearest neighbor is absorbed or blocked.
• Projection of position determination using network linkages (called leap frogging) so that individual range errors are minimized in aggregate.

Inter-cluster communication is possible. In this scenario, packet scheduling must allow for clock drift between clusters. Also the timecode can be passed to synchronize events.

Multi-hop communication allows a network to span a distance much greater than the range of an individual localizer. Store and forward message passing introduces latency, but conserves power (it defeats the 1/Rⁿ received power reduction [n ≥ 3]) and reduces the probability of intercept.

Cooperative networking has other advantages. Data can be selectively shared over the network, and individual localizers can even be locked-out of the network. Also, range information can be hidden as a precaution against enemy targeting.

The fifth generation of CMOS UWB localizers have been developed with the following capabilities.

• Penetration

  Localizers are able to radiate impulse doublets with significant low frequency spectral content using physically small antennas.

• Low probability of detection

  High process gain (typically 30 dB) allows the instantaneous signal to be below the background noise level.

• Software-controlled parameters

  Parameters such as codes, timing, detection, and protocols are all software-mediated.

At the time of writing, localizers have demonstrated ad-hoc peer-to-peer networking of eight nodes, including rapid acquisition and flooding of range data. We achieved a maximum range of 30 meters, the range accuracy is several centimeters, and the node data rate is 1,000 bits/second. The prototypes have two full-custom CMOS chips (shown in Figure 12), an RF front end, an A/D converter, a TCXO, and a processor with RAM and EPROM.

The fifth generation prototype is a modular board stack, which includes new antennas and RF front end, and allows for incremental upgrades to the modules. With antennas and AAA batteries, the dimensions and weight for this prototype are 1.2 inches × 2.0 inches × 2.9 inches (actual size shown in the photograph on the above right) and 4.7 ounces, respectively.

1. Dilip V. Sarwate and Michael B. Pursley, “Crosscorrelation Properties of Pseudorandom and Related Sequences,” Proceedings of the IEEE, vol. 68, No. 5, pp.593-619, May 1980.

   M

Vincent Coli is the vice president of marketing for Æther Wire & Location, Inc. (www.aetherwire.com). He has 20 years of experience in the IC and electronic design automation tool industry doing applications engineering and marketing for user-customizable ICs and related design and prototyping software tools.

Coli has been associated with Æther Wire since 1993. Prior to Æther Wire, he has held management positions at Advanced Micro Devices Inc., VLSI Technology, and Aptix Corp. He has published several technical articles and holds three patents.

Coli received an M.B.A. and M.S. in electrical engineering from Santa Clara University and a B.S. degree in chemical engineering from Rensselaer Polytechnic Institute. He can be reached at vince@aetherwire.com.