An Introduction to Ultra Wideband Communication Systems

6.2. MC-UWB Receivers

In multicarrier systems, a single data stream is split into multiple parallel data streams of reduced rate, each of them transmitted on a separate frequency (sub-carrier). Each carrier is modulated at a low enough rate that ISI is not a problem. Subcarriers must be properly spaced so that they do not interfere. For a N sub-carrier system each subchannel is tolerant of N times as much dispersion as the original single carrier system. Multicarrier UWB (MC-UWB) communication systems may use orthogonal UWB pulse trains and multiple subchannels to achieve reliable high bit rate transmission and spectral efficiency [7]. Some of the advantages of MC-UWB are discussed in Section 5.4. MC-UWB systems, however, suffer from interchannel interference (ICI) in multipath fading channels because the received subcarriers arriving by different paths are no longer orthogonal. Hence, frequency synchronization is extremely important [7, 8]. Multicarrier systems also have the problem of the peak-to-average power ratio problem, which must be taken care of to maintain the proper linear operating region of the receiver.

The received MC-UWB signal is a multipath channel described in (6.15) over one transmission interval, and can be written as

Equation 6.135

where b_k is the symbol that is transmitted in a transmission interval over the k^th subcarrier, N is the number of subcarriers, and b is a constant that controls the transmitted power spectral density (PSD) and determines the energy per bit. The fundamental frequency is , where T_p denotes the duration of waveform p (t). To detect the information in subcarrier m, we examine the output of the filter matched to p (t) exp (j2pkf₀t) given by

Equation 6.136

In the previous equation

Equation 6.137

and w_m (t) is the filtered complex noise. If we sample the output of the matched filter at times t_i, i = 0, 1, . . ., L 1, we get the following form

Equation 6.138

The first part of this output is the multipath channel response to the m^th subchannel. The second part is the ICI term. For best performance, we would like to separate the multipath contribution due to each subcarrier and combine them optimally to form the sufficient statistic that will be used in the detection process.

The optimum receiver for a multicarrier system is a multichannel maximum likelihood (MCML) detector that finds the data vector that maximizes the likelihood function over all subchannels. The complexity of this receiver increases exponentially with the number of subchannels. The optimum receiver is not practical when a large number of carriers are used. The study of the MCML detector is used to provide an upper bound on the performance of any other form of receiver for multicarrier systems. Suboptimal, less complex receiver structures also need channel parameter information, so channel estimation is an important block in the receiver design. The performance of a single channel RAKE receiver with channel information and that of a noncoherent receiver that uses no channel information is described in [8]. The single channel RAKE receiver treats ICI as additive Gaussian noise. The noncoherent receiver, dubbed peak detector, can be used with any orthogonal modulation schemes, such as on-off keying, but is inapplicable to more complex modulation schemes, such as quadrature amplitude modulation (QAM). The performance of these two receivers compared to the optimal MCML shows that the single carrier peak detector suffers the maximum in terms of BER performance. The single carrier RAKE is close to optimal performance at small SNR, but deviates from optimal performance at high SNR [8]. In the next two sections, we describe two interesting multicarrier UWB receiver architectures: CI-UWB and FH-UWB.

6.2.1. Carrier Interferometry (CI) UWB Receiver

The UWB nature of the signals in the frequency domain leads to ultra fine multipath resolution in the time domain (down to path differentials on the order of one nanosecond). The resolvable multipath components, regarded as independently faded delayed versions of the original signal, enable a path diversity gain (for example using RAKE reception as described in Section 6.1.3). In the originally proposed UWB system [6062], data bits are modulated onto short pulses with a repetition period larger than the time delay spread of the multipath fading channel. The receiver implements a combining across the resolvable multipath components enabling high performance at low complexity. However, the data rate of such a system is low (relative to the total transmission bandwidth). In order to achieve higher data rates or to support multiple users, spread spectrum techniques have been proposed, as described in Section 6.1.3. When an attempt is made to improve the spectral efficiency (throughput) by co-locating multiple bits within the channel's delay spread, multipath effects lead to severe ISI, or MAI in the multiaccess case, and dramatic performance degradation with a small increase in the number of users. Therefore, such types of UWB may not be suitable for multiaccess and high throughput applications [6365].

In [66], the use of the multicarrier carrier interferometry (CI) waveform as a means to enable higher throughput/improved multiaccess in UWB (without significant performance degradation) is proposed. The CI pulse waveform in UWB corresponds to the superpositioning of N orthogonal carriers. The idea here is that with carefully chosen phase offsets (spreading codes) that ensure a periodic main lobe in the time domain (with side lobe activity at intermediate times), we could null out ISI, MAI, or any other forms of interference. At the receiver side, the received UWB CI signal is decomposed into carrier components and recombined to exploit frequency diversity and minimize ISI/MAI from other information-bearing pulses.

The CI-UWB transmitter is discussed in Section 5.4.1. Because the CI pulse shape can be decomposed into N frequency components, the receiver can exploit the benefits of frequency diversity. Assuming a slow frequency-selective fading channel (typical of UWB transmission), frequency selectivity exists over the entire bandwidth. However, careful design of the CI waveform, i.e., proper selection of and , ensures that each of the N individual carriers making up the CI pulse experience a flat fade. Note that T_f denotes the frame duration, while T_c corresponds to the duration of the main lobe of the CI pulse waveform, p (t). In time-based receivers, the frequency-selective fade is usually modeled as a tapped-delay line with L resolvable paths. In the frequency decomposable CI signal, the received signal is best characterized by the unique flat fade on each carrier component. Therefore, the received signal (assuming 0^th transmitted frames) can be shown as

Equation 6.139

where denotes user K's information bit in the m^th frame, A is the amplitude of the carrier, a_n is the fading gain, and f_n = 2pnDft_l is the phase offset of the n^th carrier, t_l is the delay of the l^th multipath, and n (t) is the AWGN noise. The receiver of Figure 6.48 is applied to detect bit b^k₀ where the angle q_k in the figure corresponds to , while is the time-hopping sequence assigned to user k. That is, the received signal is decomposed into its N subcarriers, and crosscarrier combining is then performed to exploit available frequency diversity and to minimize ISI. The m^th branch signal in Figure 6.50, rⁱ_m, is given by

Equation 6.140

The first term in the preceding equation represents the desired information, the second term represents the interchannel interference, and the third term represents the noise. When cross-carrier combining is applied (based on the MMSE criteria for minimizing ICI and noise while exploiting frequency diversity), the final decision variable, Rⁱ [63] can be shown to be

Equation 6.141

where

Equation 6.142

The frequency-based processing of CI pulses in UWB systems enables high throughputs while maintaining excellent BER performance. CI-TH-UWB gives better performance when compared to conventional RAKE processing [63]. The idea has been extended to CI-DS-UWB in [64].

BER performance plots for the comparison of time-based processing versus frequency-based processing for TH-UWB is presented in Figures 6.49 a and b [5], and for DS-UWB in Figures 6.50 a, b, and c [64].

6.2.2. Frequency Hopped (FH) UWB Receivers

In [70] a residue number system (RNS) assisted multistage frequency hopped spread spectrum multiaccess scheme is studied. We refer to this system as FH-UWB. This scheme is capable of efficiently dividing the huge number of users into a number of reduced size user groups; multi-user interference only affects the users within the same group. Because the number of users within the same group is only a small fraction of the total number of users supported by the FH-UWB system, advanced multi-user detection algorithms can be employed for achieving near single user performance at an acceptable complexity. Some of the advantages of the FH-UWB systems are discussed in Section 5.4.2.

Preliminary results [70] show that FH-UWB is capable of supporting an extremely large number of users while employing relatively simple quadratic receivers and multilevel frequency shift keying (MFSK). The readers are referred to [70] for a rigorous analysis and simulation results for BER performance. Figure 6.51a^[2] depicts the transmitter block diagram. Figure 6.51b illustrates the receiver architecture of the FH-UWB system employing MFSK modulation.

^[2] Figures 6.51a and 6.51b are modified from Figures 6.1 and 6.3 of [70].

A MC-UWB receiver architecture using sigma-delta modulators for the IEEE 802.15 standard has been proposed in [71]. In this design, the elementary pulse duration is chosen based on the peak power and pulse length limitations and desired side lobe behavior in the frequency domain. The interpulse separation and number of pulses is selected based on channel spread and average power limitation. This MC-UWB using N tone sigma-delta modulators simplifies implementation. OFDM-UWB, a variation of MC-UWB, is increasingly more popular than conventional MC-UWB due to simple transmitter and receiver implementations via IFFT/FFT algorithms, respectively and is discussed next.

6.2.3. OFDM-UWB Receivers

OFDM is a special case of multicarrier transmission that permits subcarriers to overlap in frequency without mutual interference, resulting in increased spectral efficiency. OFDM exploits signal processing technology to obtain a cost-effective means of implementation. Multiple users can be supported by allocating a group of subcarriers to each user. OFDM-UWB is a novel system that has been proposed as a physical layer for a high bit rate, short-range (10 m20 m) communication network in high performance computing clusters. Traditional UWB communications use PPM or PAM modulation schemes, different pulse generation methods, pulse rate and shape, and center frequency and bandwidth, as described in earlier sections. Earlier UWB systems were designed to be carrierless. In contrast, OFDM-UWB is a multicarrier UWB system that relies on splitting orthogonal subcarriers in a train of short pulses, sending them over the channel, and reassembling them at the receiver to get orthogonality and to recover each subcarrier separately [72, 73]. This new system offers more flexibility in shaping the transmitted spectrum because it has more degrees of freedom. OFDM-UWB provides more multipath resolution than single carrier UWB. A RAKE receiver can therefore achieve more multipath diversity gain and improve the overall performance. Unlike narrowband OFDM, a given tone in OFDM-UWB is transmitted only during parts of the transmission interval. Reliable communication results from integrating several pulses and high throughput from transmitting frequencies in parallel. OFDM-UWB is being proposed as the physical layer standard in 802.15.3a Wireless Personal Area Networks [74].

OFDM-UWB is a MC-UWB system that uses a frequency coded pulse train as a shaping signal. The frequency coded pulse train is defined as follows

Equation 6.143

where s (t) is an elementary pulse with unit energy and duration T_s < T, and p (t) has duration T_p = NT. Each pulse is modulated with a frequency , where c (n) is a permutation of the integers {0, 1, . . ., N 1}. It can be easily verified that

Equation 6.144

Because , the signals

Equation 6.145

are orthogonal for k = 0, 1, . . ., N 1. Further, if T_s = T, then the signals given by (6.145) are orthogonal for any k. Costas showed that if c (n) is a Costas sequence, then the p_k (t)'s remain near-orthogonal under different delays. Further, (6.145) can be efficiently implemented by taking samples of the signal p (t) which are IFFT samples and passing them through a DAC. The most challenging part of this structure is designing the DAC. These DACs typically operate around hundreds of MHz.

In the next section, we describe the receiver architecture for OFDM-UWB using sigma-delta quantizers which trade off the high operating rate of DACs for lower resolution.

N-Tone Sigma-Delta OFDM-UWB Receiver

A new method for generating and detecting the OFDM-UWB signal using a modified sigma-delta modulator is proposed in [75]. Unlike narrowband OFDM, the OFDM-UWB spectrum can have gaps between subcarriers. Another major difference between OFDM-UWB and narrowband OFDM is their spectral shapes. Also unlike narrowband OFDM, IFFT/FFT cannot be used directly to generate and receive OFDM-UWB signals because of the high bit rates involved. To solve these problems, a procedure to move the bulk of the processing load from the analog devices to the digital baseband section is described in [75]. Sigma-Delta A/D and D/A converters are a good choice for high bit rate wireless communications [76]. Traditional versions of the sigma-delta modulators cannot be used in OFDM-UWB transceivers because they would require prohibitively high sampling rates. A modified sigma-delta modulator to fit the characteristics of OFDM-UWB signals is proposed that enables most of the processing to be done digitally in both the transmitter and receiver. This approach enables the efficient use of IFFT, FFT to generate and demodulate OFDM-UWB signals and also gets rid of the high peak to average ratio (PAR) problem that occurs with OFDM systems. The modified sigma-delta modulator, called the N-tone sigma-delta modulator introduces N zeros at the frequencies in the quantization noise spectrum. These zeros match the locations of frequencies used by the OFDM system, and the quantization noise spectrum fills the gaps in the spectrum of the OFDM-UWB signal. In fact, this new structure can be used in other UWB systems anytime we have gaps in the spectrum of the transmitted signal. All these advantages come at the expense of a lower spectral efficiency unless one uses more complex multiband implementations. The N-tone sigma-delta quantizer structure is shown in Figure 6.52a. Generating UWB-OFDM signals using the sigma-delta modulator is shown in Figure 6.52b. The corresponding receiver structure is shown in Figure 6.52c [77].

6.2.4. Example on IEEE Proposed Standard for MC and OFDM-Based UWB Receivers

Details of the IEEE 802.15.3a preliminary standard describing the MAC and PHY layers of a wireless personal area network (WPAN) for MC-UWB and OFDM-UWB can be found in Appendix 10.A. We briefly describe the IEEE proposed standard for the PHY layer by Intel Corporation [59] for MC-UWB and OFDM-UWB to get some insight into the implementation details.

Intel's Proposal for MC-UWB

The frequency band is subdivided into 13 smaller bands of 550 MHz, and time frequency codes are used for multiple access. The proposed system occupies the 3.25 GHz to 10.6 GHz range. Coexistence with other narrowband systems is achieved by band dropping. Using pulsed transmission one band at a time yields low peak-to-average ratio (PAR). Reduced pulse repetition frequency reduces equalizer complexity. Pulses of duration 3 ns are used with QPSK modulation. The rectified cosine envelope yields a bandwidth of approximately 700 MHz, thus leading to the overlapping of bands. The time frequency coding scheme uses a single RF conversion chain and supports six coexisting piconetworks. Four symbols emitted per subband hop are used with interleaved M-ary Binary Orthogonal Keying (MBOK), which enables the use of the low complexity decision feedback equalizer. The Outer Reed-Solomon (221,255) code is used for error correction. The raw signaling rate for a symbol duration of 3 ns is 333 Mbps, and 2 bits per carrier (QPSK) is equivalent to 666 Mbps. The multiaccess performance in multipath shows that the proposed system can operate with interference 6 dB above the signal, is predicted to operate with four other piconets within a 5m radius, and improves in more dispersive channels. The link budget assumes a 7 dB noise figure with 108 Mbps at 10 meters with 9.3 dB margin. The link and implementation margin is combined. The estimated power consumption is 218 mW for the complete receiver. The 108 Mbps operation is in a 0.18 m CMOS per chip.

Intel's Proposal for OFDM-UWB

The frequency band is subdivided into smaller bands of 528 MHz. Three subbands occupy the 3.1 GHz to 4.6 GHz range. Symbols are pulsed on one subband at a time, which is a form of time interleaving. Higher frequencies are reserved for future enhancements, and the 5 GHz UNII band is avoided. OFDM uses parallel transmission using multiple carriers (tones) with 128 tones at 4.125 MHz, which gives a total bandwidth of 528 MHz. QPSK modulation is used at 2 bits per tone. 100 tones carry data, and 12 pilot tones are used for phase tracking and channel estimation. 10 tones are user-defined, and the remaining 6 tones including DC are null. The DC tone is null to simplify direct conversion receiver structures that have DC offset caused by on-chip LO leakage, and the use of the DC tone eliminates DC cancellation. Other null tones address world regulatory compliance, making spectral holes for radio astronomy, and so on. User-defined tones can be used for spectral or waveshaping. The symbol duration of 312.5 ns leads to a symbol rate of 32 MHz that includes 9.5 ns guard time for frequency hop. The raw signaling rate is 640 Mbps. The multipath performance of OFDM-UWB predicts that there is less than 0.1 dB loss of multipath energy, a 60.6 ns cyclic prefix adequate for expected channels, and there is no need for RAKE structures in the receiver. The link budget assumes a 7 dB noise figure with 110 Mbps at 10 meters with 8.1 dB margin. The link and implementation margin is combined. The estimated power consumption is 205 mW for the complete receiver, which has 3 mm² in analog and 3.8 mm² in digital. The 110 Mbps operation is in a 0.13 m CMOS chip.