pUltrawideband (UWB) has been described by some as one of the most promising technologies of our times. Early UWB systems were developed mainly as a military surveillance tool because they could "see through" trees and beneath ground surfaces. Only recently, however, has UWB technology focused on consumer electronics communications. To fully appreciate the potential of UWB in these applications, it is essential for the designer to firmly grasp the unique characteristics of this form of wireless transmission. For this, we will look at UWB fundamentals and at some of the different types of modulation and coding schemes used by UWB systems. We will then discuss the basic building blocks of a UWB system optimized for streaming digital video and audio in consumer-oriented entertainment products.
>UWB definition
Alternately referred to as impulse, baseband or zero-carrier technology, ultra-wideband systems operate coherently across a wide range of frequency spectrum relative to the center frequency. The wide relative bandwidth is of key importance because it governs how immune the radio will be to multipath interference while simultaneously penetrating walls or other material. A UWB signal can be typified by a series of low-power derivative-of-Gaussian pulses.
Each pulse is extremely short in duration (10 to 1,000 picoseconds), typically much shorter than the interval corresponding to a single bit. This pulse shape is similar to the unintentional emissions generated by consumer electronics devices. The radiated signal is often less than a PC is allowed to radiate unintentionally.
Because of the short duration of the pulses, the frequency spectrum of a UWB signal can be many gigahertz wide, overlaying the bands used by existing narrowband systems. A Defense Advanced Research Projects Agency (Darpa) study panel, the group that coined the term ultrawideband in the 1990s, defined it as a system with a fractional bandwidth greater than 25 percent. Fractional or relative bandwidth is the ratio of signal bandwidth over center frequency. If B is the bandwidth, Fc is the center frequency, and Fh and Fl are the high- and low-frequency cutoffs (e.g., -10 dB from peak in the current FCC proposed rules), the fractional bandwidth is then defined as given in Equation 1.
As a result of UWB's distribution of energy, the spectral density is extremely low. Pending FCC resolution, a UWB radio will probably transmit less than 75 nanowatts of power per megahertz of frequency bandwidth. This would be equivalent to an aggregated power of 0.26 milliwatts, in contrast to 30 to 100 mW for 802.11b radios and 1 mW to 1 W for Bluetooth radios.
The interval between individual pulses can be uniform or variable, and there are a number of different methods that can be used for modulating the pulse train with data for communications (see coding-schemes section). One common characteristic, however is elimination of a carrier frequency. That is why UWB is also termed sometimes "zero carrier" radio. In other words, a UWB system can drive its antenna directly with a baseband signal.
A general UWB pulse train signal can be represented as a sum of pulses shifted in time, as shown in Equation 2:
Here, s(t) is the pulse train signal, p(t) the basic pulse shape (e.g. derivative-of-Gaussian), and an and tn are the amplitude and time offset, respectively, of each individual pulse.
UWB has many advantages that make it suitable for indoor wireless networks in general and, more specifically, for consumer-type applications. Basic physics gives UWB an inherent ability to maintain high speed through walls and in cluttered high-multipath environments. This advantage is not offset, however, by high complexity in its implementation. There are no fast Fourier transforms or inverse FFTs required, as in OFDM radios. There is no matrix inversion required to deconvolve the channel. The high-resolution UWB signal provides the channel impulse response directly. Also, the transmitter output is low enough to eliminate a power amplifier; nor does it have to be linear.
No harmful interference
The end result is a technology that provides potentially unlicensed operation, simplicity, very low transmit power, multipath and interference immunity, and the capability to deliver data rates in excess of 100 Mbits/second-all the while consuming very little battery power and relatively small amounts of silicon area, translating to low cost.
Thanks to its low spectral density, unlicensed UWB radio emissions do not add up to cause harmful interference to other radio systems operating in dedicated bands. In fact, normal propagation attenuation causes the signals to dissipate faster than they can add up. Furthermore, the power spectrum can be adjusted to reduce levels even lower in sensitive bands, such as global-positioning system or personal communications services receivers. That means UWB devices can be co-located with GPS and PCS equipment.
Moreover, ultrawideband's low power, wide spectrum and coded waveform make it difficult for eavesdroppers to detect. A further security benefit lies in a UWB device's ability to detect the distance to another ultrawideband device based on round-trip delay information. The high precision of this range information allows a device to reject communications with another device unless it is at or within an authorized range. The corollary is that ultrawideband receivers can use range information to reduce their level of interference even further (using transmit power control) and optimize network configuration and traffic flow.
Carefully implemented, UWB is also highly immune to the effects of multipath interference. This effect can be understood from a frequency-domain perspective by realizing that the signal bandwidth of a UWB signal is similar to the coherence bandwidth of the multipath channel. In other words (or from a time-domain perspective), UWB's strong resolution capability also improves the performance of the radio by allowing the different multipath components to be resolved. There are many possible paths that a UWB radio can use for communication, so UWB is considered very robust. It actually works better in a high-multipath channel than out in an unobstructed "line of sight" environment.
Shannon rules
From a communications perspective, perhaps the most important characteristic of UWB systems is their low signal-to-noise operation. The impact of this is clearly seen in Shannon's equation for channel capacity, shown in Equation 3, where C is the channel capacity, B the signal bandwidth, P0 the signal power spectral density, and kT the noise spectral density.
Most conventional radio systems are bandwidth-limited, trading power to improve the data rate. But as seen from Shannon's equation, the required transmit power goes up exponentially with the data rate. For instance, 802.11a systems use 64-QAM modulation to send 6 bits per symbol when transmitting at 54 Mbits/s, meaning a high signal-to-noise ratio is mandatory to resolve the small Euclidian distance between constellation points.
But in UWB systems, channel capacity scales linearly with bandwidth and nearly linearly with power. This fact allows UWB radios to scale with semiconductor technology and hit very high data rates with very low transmit power.
Modulation and encoding
The regulatory limits placed on the power spectral density (PSD) of the transmitted signal by the FCC affects the choice of modulation in two ways. First, the modulation technique needs to be power-efficient. In other words, the modulation needs to provide the best error performance for a given energy per bit. Second, the choice of a modulation scheme affects the structure of the PSD (i.e., possible spectral lines) and thus has the potential to impose additional constraints on the total transmit power allowed.
Two of the most popular modulation techniques that have been proposed for UWB are pulse-position modulation (PPM) and binary phase-shift keying (BPSK) (Figure 1). These are best compared in terms of modulation efficiency and power spectrum.
PPM encodes information by modifying the time interval between pulses. Its name derives from looking at the pulse position relative to a pulse that would have occurred if a fixed interval were used. This choice of modulation dates back to when avalanche devices were the only way to generate pulses with subnanosecond rise times. And time was the only parameter to easily modulate.
In the case of BPSK, the pulse is sent at zero or 180 degrees , right side up or upside down. Many communications textbooks show that in low-S/N environments, QPSK and BPSK are optimally efficient, requiring the least energy per bit for any given noise level.
Power spectral density
So, one form of a UWB signal is a simple pulse train. If we assume that pulses are uniformly spaced in time, we can rewrite Equation 2 as shown in Equation 4:
where T is the period or pulse-spacing interval.
The power spectral density of this signal, Φss(f), is the Fourier transform of the signal autocorrelation. If we assume that the pulse weights correspond to the data bits to be transmitted and the data is random, we find that the PSD is as given in Equation 5.
Here, &sigma2a and μa are the variance and the mean, respectively, of the an sequences, P(f) is the Fourier transform of p(t) and δ(f) is a unit impulse. This PSD has both a continuous portion and discrete spectral lines. Note too that for an unmodulated signal (&sigma2a=0 and μa=1), Equation 5 can be simplified into Equation 6.
As expected, the PSD consists of a comb of spectral lines with energy spikes that are separated by a distance 1/T. The spectrum envelope is defined by the basic pulse shape p(t).
For a modulated signal, it is worth noting that the magnitude of the spectral lines depends on the mean of the weights, microa. If we use a whitener on the data to make the transmitted data bits random and equally probable, and we encode the data bits using BPSK by using an ε{-1,+1}, then &sigma2a=1 and μa=0. In this (μa=0) case, Equation 5 becomes Equation 7, and it is clear that the discrete portion of the equation (i.e., the spectral lines) disappears.
This ability to eliminate spectral lines is a key feature of BPSK. It is crucial for UWB to minimize the presence of those spectral lines since they might interfere with conventional radio systems. Also, the presence of spectral lines can lead to reductions in total transmit power in order to meet regulatory PSD limits.
PPM encodes the data bits in the pulse stream by advancing or delaying individual pulses in time, relative to some reference. In this case, Equation 2 for the PPM UWB signal becomes Equation 8
where bn ε{-1,+1} is the data, T is the reference interval between pulses, and βT is the amount of pulse advance or delay in time, relative to the reference (unmodulated position) for a given data bit. The modulation here tends to smooth the spectrum but the spectrum still contains some spectral lines since the pulses are only delayed or advanced by a fractional part of the pulse width.
To reduce the level of the lines further, an additional dithering sequence can be added. Dithering is a pseudorandom process that jitters the "reference" position of the individual pulses according to a known random sequence. But dithering increases the level of complexity of a UWB system and adds to the synchronization schedule.
Modulation efficiency
The modulation efficiency of the two techniques described can be determined by the intersymbol distance as a function of the energy per bit, Eb. The symbol constellations for the PPM and BPSK systems would show an orthogonal signaling scheme and an antipodal signaling scheme, respectively. This indicates that BPSK has a much greater intersymbol distance than PPM for equal energy per bit. This leads to a 3-dB advantage in efficiency, meaning that PPM requires twice as much energy as biphase to achieve the same bit error rate. This difference is even more pronounced in a multipath environment since multipath appears as data modulation in a PPM system.
Another way to look at this advantage is to say that biphase modulation is also more spectrally efficient. It uses only about half the bandwidth of a PPM system for the same raw chipping rate. If one assumes that biphase modulation and PPM have the same amount of time to encode a chip, PPM must be able to fit two pulse positions into the same time window that contains a single biphase pulse. Therefore, the biphase pulse can have about half the bandwidth of a PPM pulse or, by reciprocity, twice the data rate for the same bandwidth.
Biphase presents several other key benefits. First, it exhibits a peak-to-average power ratio (PAR) of less than 8 dB. Thus, an implementation using biphase does not require any external snap-recovery or tunnel diodes or power-amplifier circuitry. Instead, it can be driven directly from a low-voltage high-speed CMOS IC.
Finally, for reasons of clocking, biphase modulation has reduced jitter requirements. In PPM, the clocking path must include elements to accurately control arbitrary time positions on a fast (pulse-to-pulse) basis. This control requires a series of wide-bandwidth circuits where jitter accumulates. But a biphase system needs only a stable, low-phase-noise clock as the pulses occur on a constant spacing. Synchronization circuits can be narrowband so that they do not add significant jitter. As a result, less power and real estate are needed to implement the required circuits.
Building blocks
As stated, UWB radio systems in general can be very simple. The transceiver portion can be implemented in standard digital CMOS and requires none of the expensive SAW filters common to conventional radio technologies.
Figure 2 shows a functional block diagram of a UWB biphase radio system. On the transmit side, the pulse-forming network (PFN) block generates a sequence of derivative-of-Gaussian pulses that are directly produced by gates in the CMOS IC that drives the antenna directly (no power amplifier required).
The receive chain is also simplified. The PFN block generates a sequence of Gaussian pulses that gets correlated with the incoming data sequence. This correlation signal is integrated over the duration of the bit period, sampled and latched to produce the output data. On startup, the control block searches for a phase match of the local clock with the transmitted clock by adjusting the phase and frequency of the clock. The architectural simplicity is conducive to a small die size, hence lower cost as well as low power consumption.
John McCorkle is a co-founder and CTO of XtremeSpectrum (Vienna, Va.). McCorkle has more than 25 years' experience in radar and communication signal-processing systems, including time as the U.S. Army Research Laboratories' UWB radar project director. He can be reached at john@xtremespectrum.com".
