Chapter 2 RADIO-OVER-FIBER SYSTEMS USING
2.5 Summary
Figure 2-10 The schematic diagram of PAPR.
2.5 Summary
Optical RF signal generations using an external MZM based on DSB and DSBCS modulation schemes are reliable techniques. The DSB modulation signal undergoes performance fading that restricts the fiber transmission distance due to fiber dispersion. Although DSBCS modulation has been demonstrated to be effective in the millimeter-wave range, it cannot generate vector modulation formats which are of utmost importance in wireless applications. The transmission distance of the 60-GHz wireless signal is dramatically restricted in a small range due to the high path losses. The main propagation phenomenon for 60-GHz signal consists LoS, first and second order reflections. Wireless transmitter and receiver were discussed in this chapter. The impairments of 60-GHz RoF systems are also investigated in this chapter. They include inter-symbol interference, fading, I/Q imbalance, and PAPR.
Chapter 3
DIGITAL MODULATION TECHNIQUES
3.1 Preface
In this chapter, we will give the brief overviews for digital signal techniques. The digital modulation formats including 1: single carrier, 2:
orthogonal frequency-division multiplexing (OFDM), 3: single-carrier frequency-domain-equalization (SC-FDE), 4: single-carrier frequency-division multiple access (SC-FDMA), 5: single-carrier frequency-division multiplexing (SC-FDM). The digital signal processing for system impairments including: 1.
Adaptive bit-loading algorithm, 2: I/Q imbalance compensation algorithm. This chapter also introduces multiple-input multiple-output (MIMO) technique in 3.4.
3.2 Digital Modulation Formats
3.2.1 Single Carrier
The single carrier modulation is a popular modulation format in digital communication systems. A single carrier system modulates a sequence of digital symbols into a high frequency sinusoidal signal called carrier. The modulated signal has three parameters: frequency, phase, and amplitude. Figure 3-1 shows two basic single carrier modulations with one bit per symbol. They are amplitude-shift keying (ASK) and phase-shift keying (PSK) [41]. The ASK signal with one bit per symbol is also called on-off keying (OOK). The PSK signal with one bit per symbol is also called binary PSK (BPSK). For the BPSK signal, the signal has 180 degree phase different between 1 and 0. The
BPSK signal can be expressed as
cos 2 , for 1 cos 2 cos 2 , for 0 (3-1) where A and are the amplitude and frequency of carrier signal, respectively. The BPSK signals can be graphically represented by a constellation diagram. The constellation diagrams only consider the amplitude and phase information of transmitted signal.
The BPSK signal contains only one bit per symbol which is very low spectral efficiency. Modulation format with high spectral efficiency is very important for wireless communication systems. Figure 3-2 shows the constellation diagrams for the PSK, quadrature PSK (QPSK), and 8PSK signals.
The QPSK and 8PSK signal have 2 and 3 bits per symbol, respectively. Based on amplitude and phase modulations, a variety of modulation schemes can be derived from their combinations. For example, by modulating both amplitude and phase of the carrier, we can obtain a scheme called quadrature amplitude modulation (QAM). Figure 3-3 shows the constellation diagrams for the 8QAM, 16QAM, and 32QAM signals. The transmitted s(t) can be express as
cos 2π sin 2π (3-2) where I(t) and Q(t) are the in-phase and quadrature component, respectively. At the receiver, the I(t) and Q(t) can be obtained by multiplying the received signal with cosine and sine signals, respectively.
While transmitting through the channel, the signal suffers distortions, interferences, and noises which may introduce error after demodulation.
Therefore, bit-error rate (BER) is the preferred measurement to verify system performance. When the real-time BER measurement is impossible in some of
digital receiver, the BER estimation from the demodulated constellation is very important. The BER of square QAM signal can be related to the signal to noise ratio (SNR):
BER ∙ 1
√ ∙ ∙
∙ (3-3) The SNR can be derived from error vector magnitude (EVM):
SNR = –20 log (EVM/100%) (3-4) The EVM is defined as the noise power from constellation over the signal power from constellation, as shown in Fig. 3-4. The importance of the above equation is that it relates EVM to BER through the SNR. These equations assume that the noise is Additive White Gaussian Noise (AWGN) [41, 49].
By the way, the channel also induces inter-symbol interference (ISI). The single carrier modulation signal could use feed-forward equalizer or feedback equalizer to improve system performance. The principles of these time domain equalizers are that the equalizers capture a sequence symbols and using adaptive signal processing to find the best weight coefficients for the reduction of impairment of ISI [50-52].
Figure 3-1 The waveform of bandpass modulation signals.
Figure 3-2 Signal constellation: (a) BPSK, (b) QPSK, (c) 8-PSK.
Figure 3-3 Signal constellation: (a) 8-QAM, (b) 16-QAM, (c) 32-QAM.
Figure 3-4 The principle diagram of error vector magnitude (EVM).
3.2.2 Orthogonal Frequency-Division Multiplexing
Recently, the multicarrier modulations have attracted significant interest because of many advantages [53-56]. One of advantages is that the symbol rate of multicarrier modulations is much lower than single carrier modulations.
Thus the effect of ISI will be reduced and the equalization in the demodulation will be easier than single carrier modulation system with high symbol rate. In order to increase data throughput within limited bandwidth, the frequency spacing between subcarriers is set as low as possible. The minimum frequency spacing between subcarriers is equal to the symbol rate, so that all subcarriers are orthogonal to each other and can be separated without using filter in the receiver. This modulation format is called OFDM.
The block diagram of the OFDM transmitter consists of serial-to-parallel conversion, modulation mapping, inverse fast Fourier transform (IFFT), cyclic prefix (CP) insertion, and digital-to-analog converter (DAC), as shown in Fig. 3-5 (a). The serial binary data sequence is firstly transfer to parallel sequence for each subcarrier. The binary data is mapped onto high order modulation symbols that can enhance spectral efficiency. The symbols are
transferred into time domain waveform through an IFFT. After the IFFT operator, the symbols at different subcarriers are assigned to suitable frequencies which are orthogonal each other. Then, a generated single inserts a waveform referred to as a CP. The CP that is a copy of the last part of the block could provide a guard time to prevent ISI.
The block diagram of the typical OFDM receiver is shown in Fig. 3-5 (b).
This demodulation process includes synchronization, fast Fourier transform (FFT), equalization, and symbol decoding. After timing synchronization by using training symbols, the FFT operator is used to transfer time domain signal to frequency domain. The channel information can be found from received training symbols. Due to the bandwidth of one subcarrier is fairly narrow for OFDM signals, the equalization can be implemented by just multiplying a value for each subcarrier. This equalization is called one-tap equalization or frequency domain equalization.
Figure 3-5 Block diagrams of OFDM transmitter (a) and receiver (b). (IFFT:
inverse fast Fourier transform, DAC: digital-to-analog converter, ADC:
analog-to-digital converter, FFT: fast Fourier transform)
3.2.3 Single-Carrier Frequency-Domain-Equalization
The drawback of OFDM signals is the high peak-to-average power ratio (PAPR). The OFDM signal is the sum of all the subcarriers. If many subcarriers are in phase for some input data, the signal would have high peak power. There are some issues for high PAPR signal. Because the system is not completely linear, the high peak power of a signal would be clipped by the system.
Therefore, the high PAPR signals are easily distorted and cause performance degradation. On the other hand, since the high PAPR signals have large power distribution, the signal would suffer more quantization noise and also induce performance degradation. Moreover, some of components limit the peak power of the signal. The system would have less power budget that would limit the transmission length. Due to these reasons, high PAPR become a big issue for the OFDM signal.
The equalizer is needed to compensate ISI that is introduced by the multipath propagation channel. The traditional single carrier signal use conventional time domain equalizers that are not practical for the broadband signal because of the long channel impulse response in the time domain.
Therefore, the FDE is more practical for the 60 GHZ systems with broadband channels. The transitional single carrier signal has low PAPR, but without using frequency domain equalization (FDE). If we can combine the advantage of OFDM and single carrier signal, this SC-FDE signal would have higher quality than the OFDM and traditional signal carrier signal. Therefore, the SC-FDE is used in this research [57, 58].
Figure 3-6 shows the basic idea of the time domain equalizer and the FDE.
The input signal and channel impulse response are x and h, respectively. The
received signal y is the convolution of the input signal x and channel response h in time domain. For the linear time invariant system, the convolution operation between the time-domain signal and channel response is equal to a multiplication operation in frequency domain.
However, the Fourier transform of the entire data sequence is not a practical solution. Thus, the signal is separated to several blocks which are equal to the defined FFT size. As the FFT is the implementation of a circular convolution, adding CP to the signal maintains the continuity of the signal.
Therefore, the principle of SC-FDE is almost the same with OFDM signal. The advantage of frequency domain equalizer is only a one-tap equalizer is needed.
When the broadband signal suffers a serious ISI, the time domain equalizer needs more taps to compensate for system performance which increase the computation requirement of the algorithm. On the other hand, the FDE only needs to increase the length of CP and FFT size. Although, larger FFT size also increase the computation requirement but increase the number of processing symbols. Different CP size only increase overhead of signal but does not increase processing complexity.
Figure 3-7 shows the block diagram of SC-FDE transmitter and receiver.
The transmitter is almost the same with traditional single carrier modulation.
The only difference is the SC-FDE needs additional overhead that is CP for FDE. At the receiver side, the CP is removed from the signal and then transfer to the frequency domain using FFT operator. The FDE is used to compensate for ISI. Due to the original signal is in time domain, the IFFT operator is used to transfer the frequency domain signal to time domain after equalization.
There are some similarities and differences between OFDM and SC-FDE.
The similarities are the FFT based implementations with CP and FDE. The difference are: the SC-FDE has FFT and IFFT at the receiver, whereas OFDM has the IFFT at the transmitter and the FFT at the receiver. SC-FDE has lower PAPR than OFDM. SC-FDE has lower sensitivity to the carrier frequency offset than OFDM. SC-FDE could not use adaptive bit and power loading, whereas OFDM could use adaptive bit and power loading. Because the SC-FDE has lower complexity at the transmitter, the SC-FDE is very suitable for the uplink of communication system.
By the way, equalizer coefficients play important roles in FDE to compensate for ISI. The OFDM signal uses train symbol to find equalizer coefficients. Therefore, the SC-FDE signal also needs train symbol. However, the SC-FDE signal has different power for different frequency subcarriers. Due to this reason, some subcarriers easily suffer noise interference and hard to find correct equalizer coefficients. One solution is that the SC-FDE systems use the OFDM train symbol. Nevertheless, the OFDM train symbol have high PAPR problem. Consequently, the Zadoff-Chu sequence is used to solve this problem.
A Zadoff-Chu sequence is a complex-valued sequence which has constant power in time domain and frequency domain [59]. Therefore, this sequence provides low PAPR train symbols in time domain and without power variation in frequency domain.
Figure 3-6 Basic idea for frequency domain equalization.
Figure 3-7 Block diagrams of SC-FDE transmitter and receiver.
3.2.4 Single-Carrier Frequency-Division Multiple Access
Before start with SC-FDMA modulation format, we introduce orthogonal frequency-division multiple access (OFDMA) modulation format first. The OFDMA modulation format is a multi-user version of OFDM modulation scheme [60, 61]. Different users have different channel. The OFDMA could assign suitable subcarriers for different users. However, the OFDMA has the same problem, which is high PAPR issue, with the OFDM signal. Signal with a high PAPR require high linearity components to avoid inter-modulation
distortion. Therefore, the SC-FDMA signal is proposed to provide better system performance for multi-user system. The OFDMA modulation format is a multi-carrier signal whereas SC-FDMA modulation format is a single carrier signal. Because of this reason, the SC-FDMA modulation format has lower PAPA than OFDMA modulation format.
Figure 3-8 shows the concept of SC-FDMA generation and demodulation system for uplink [48, 62]. Therefore, the system has several transmitters. The principle of SC-FDMA comes from SC-FDE system. Due to the IFFT is the inverse operator of a FFT, the block diagram of SC-FDE transmitter could be change as Figure 3-9. The original binary data sequence is sent into a vector signal modulator to map the original data onto a constellation. Then, a serial data sequence transfer to a parallel data sequence by using serial to parallel converter. In order to provide multi-user version, the FFT length need to be reduced. The FFT length of system depends on user bandwidth. The N-points modulation symbols Xn provide N-points output symbols in frequency domain by using N-points DFT operator. The SC-FDMA generator then maps several set of N-points output symbols to M orthogonal subcarriers. Then, an IFFT convert the frequency domain signal into a time domain signal. Then, CP, which is a copy of the end of the block and provides a guard time to prevent ISI, is inserted to the start of the generated time domain signal.
Two mapping methods will be introduced in this thesis. Figure 3-10 shows a distributed subcarrier mapping and a localized subcarrier mapping. In the distributed subcarrier mapping the N-points output symbols are distributed within the entire channel bandwidth that has M subcarriers. In the localized subcarrier mapping, the N-points output symbols are assigned to adjacent
frequency subcarriers. The SC-FDMA signal with distributed subcarrier mapping is named DFDMA. The LFDMA is the SC-FDMA signal with localized subcarrier mapping. The other M-N subcarriers, which are not occupied by data symbols, are assigned to be zero. If the occupied subcarriers are with equal frequency spacing in distributed mapping method, this mapping method is also named interleaved subcarrier mapping. The interleaved subcarrier mapping of SC-FDMA is interleaved FDMA (IFDMA) which is a special case of DFDMA [63, 64].
Figure 3-8 Block diagrams of SC-FDMA transmitter and receiver.
Serial to parallel conversion
Parallel
to serial DAC
Cyclic prefix insertion Symbols
FFT IFFT
I
Q
Figure 3-9 Block diagrams of SC-FDE transmitter.
Figure 3-10 The distributed and localized subcarrier mapping modes for one user of SC-FDMA signal.
3.2.5 Single-Carrier Frequency-Division Multiplexing
From previous description, different user uses different subcarriers.
However, this thesis focuses on 60-GHz RoF system with only one transmitter which could provide high data throughput. Therefore, this thesis proposed a novel modulation format that looks like SC-FDMA signal as shown in Fig.
3-11. The proposed modulation format is named single-carrier frequency-division multiplexing (SC-FDM). The concept of SC-FDM combines the signal processing for different users. The difference between
SC-FDM and SC-FDMA modulation formats are the serial to parallel converter, FFT, and IFFT operator. Since the proposed modulation format wants to occupy the full channel bandwidth, the serial to parallel converter needs to provide M symbols from M/N N-points FFT operators. Because the IFFT is a linear operator, the generator only needs one IDFT operator to convert the signal which is the combination of M symbols from frequency domain to time domain. Since the SC-FDMA receiver demodulate the signal from all users at the same time, the SC-FDM modulation scheme has the same receiver.
Figure 3-12 shows the conceptual diagrams of SC-FDM with interleaved subcarrier mapping and localized subcarrier mapping. The figures show that the signals consist of three groups. The SC-FDM signal with interleaved subcarrier mapping is named SC-IFDM. The SC-LFDM is the SC-FDM signal with localized subcarrier mapping. When there is only one group in the signal, the signal is the SC-FDE modulation format. When the number of groups equals to IFFT size, the signal is very similar to OFDM modulation format.
Compared with OFDM signal, the advantage of the proposed modulation format is the low PAPR. This thesis use numerical analysis to investigate the PAPR properties for SC-FDE and OFDM signals. Using Monte Carlo simulation, this thesis calculates the complementary cumulative distribution function (CCDF) of PAPR. The Pr(PAPR>PAPRo) is the probability that PAPR is higher than PAPRo [65, 66]. Figure 3-13 shows the CCDF curves of PAPR for SC-FDE, SC-IFDM, SC-LFDM, and OFDM. The SC-FDM transmitter uses square-root-raised-cosine filter with 0.05 roll-off factor. The results show that the SC-FDM signals have lower PAPR than OFDM signal.
Since different modulation formats could be assigned to different group,
adaptive bit-loading technology can be also utilized in SC-FDM signals. Table 3-1 shows the performance of different modulation formats. The property of SC-FDM signal that depend on the number of groups is between OFDM and single carrier signal. Compared with OFDM and single carrier, SC-FDM can adapt to different environments. If the system has serious uneven frequency response and good linearity, the system needs more groups. Otherwise, the system needs fewer groups.
Different mapping methods also induce different system performances.
Localized subcarrier mapping and interleave subcarrier mapping are used in this experiment. For the SC-LFDM signal, the modulation symbols for a single group are assigned to adjacent subcarriers. In the interleaved subcarrier mapping, the modulation symbols for a single group are separated evenly within the entire channel bandwidth. Because of the single group of SC-LFDM signal is centralized in frequency domain, the signal will get more benefit from adaptive bit-loading algorithm. For the SC-IFDM signal, a single group of signal is separated into the used bandwidth. Therefore, the SC-IFDM signal does not get much benefit after bit-loading algorithm. However, the advantage of the SC-IFDM signal is the lower PAPR compared with the SC-LFDM signal.
Figure 3-11 Block diagrams of proposed SC-FDM transmitter and receiver.
Figure 3-12 The distributed and localized subcarrier mapping modes for SC-FDM signal with 3 groups.
0 1 2 3 4 5 6 7 8 9 10 11 12 13
SC-IFDM with 2 Groups SC-IFDM with 8 Groups SC-LFDM with 2 Groups SC-LFDM with 8 Groups OFDM
Pr(PAPR>PAPR
0)
PAPR (dB)
Figure 3-13 Comparison of CCDF of PAPR for SC-FDE, SC-IFDM, SC-LFDM, and OFDM with 512 subcarriers.
Table 3-1 The properties of different modulation format.
Properties\Modulation Single Carrier
OFDM SC-FDE SC-FDM
PAPR Low High Low Middle
Equalizer Complex Simple Simple Simple
Transmitter Simple Middle Simple Complex
Receiver Complex Simple Middle Middle
Bit-loading No Yes No Yes
I/Q Imbalance Time Frequency Frequency Frequency 3.3 Digital Signal Processing for System Impairments
3.3.1 Adaptive Bit-loading Algorithm
Since the 60-GHz communication systems suffers from uneven channel
Since the 60-GHz communication systems suffers from uneven channel