Zhongxia Simon He, Sining An, Herbert Zirath
©SHUTTERSTOCK.COM/RA2 STUDIO
Due to government and other agency regulations, microwave bands, which range in frequency from 6 to 42 GHz, often have a small bandwidth of a few hundred megahertz, which lowers the data rate for communications. On the other hand, high data rate communications can be realized in millimeter-wave (mm-Wave) frequencies (30–300 GHz). Tens of gigabits per second of bandwidth are available in mm-Wave bands, such as the D-band, from 110 to 170 GHz, and H-band, from 170 to 260 GHz [1].
High data rate transmission experiments have been demonstrated at mm-Wave in many publications. The published results can be divided into three categories: transmission with shared local oscillator (LO), where a single LO source is fed to both transmitter and receiver; off-line demodulation, where the transmitter and receiver use different LOs and the demodulation is implemented off-line; and real-time transmission, where the transmitter and receiver use different LOs and the demodulation is implemented in real time. Communication data rate up to 120 Gb/s using a shared LO source has been demonstrated [2], [3], [4], [5], with a 120-Gb/s 16-quadrature amplitude modulation (QAM) signal transmission over the W-band [3] and a 90-Gb/s 32-QAM signal transmission over a 230-GHz carrier [5]. Additionally, a 100-Gb/s wireless transmission employing 16-QAM over 2.22 m is accomplished with the help of offline digital signal processing (DSP) [6]. A 96-Gb/s 8-phase-shift keying (PSK) signal has been used to demonstrate wireless data transmission at 240-GHz carrier frequency across a distance of 40 m [7]. At the D-band, data transmission up to 60 Gb/s has been demonstrated with 64-QAM [8]. These demonstrations show the transceiver chipset’s bandwidth potential. For real-time communications, the system either has a high modulation order signal with a low data rate or a low modulation order signal with a high data rate. A list of reported mm-Wave communication demonstrations with a single carrier, single polarization, and single pair of transceiver modules is shown in Figure 1 [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26]. Real-time and nonreal-time transmission experiments perform differently. There is a data rate gap between them. This is due to difficulties in realizing high data rate real-time modulation and demodulation. The modulator and the demodulator must have a large working bandwidth to accommodate the wideband signal in order to effectively utilize the wide-band spectrum in mm-Wave. Additionally, higher modulation order signals are required in order to further improve the data rate.
Figure 1. An overview of published mm-Wave communication demonstrators with a single carrier, single polarization, and single pair of transceiver modules.
For modulators, the challenges are having high output power, wide bandwidth, good linearity, and low cost. It is not straightforward to generate high output power at mm-Wave frequencies while maintaining good linearity and wide bandwidth. For demodulators, there are two methods: coherent demodulation and noncoherent demodulation. For noncoherent demodulation, the carrier synchronization is not needed. The demodulator structure can be quite simple. It is typically used with low modulation order signals like amplitude-shift keying (ASK), differential quadrature PSK (DQPSK), and on–off keying (OOK). The spectrum efficiency is constrained in this case. Coherent demodulation is required for signals with higher order modulation schemes, such as QAM signals or multicarrier modulated signals, like orthogonal frequency-division multiplexing (OFDM). In these cases, carrier synchronization is a key problem in coherent demodulation.
This article provides a review of several mm-Wave modulator and demodulator solutions. The organization of the article is as follows: the “Real-Time Modulator Topologies” section presents a review of various real-time modulator topologies; the “Real-Time Demodulators Without Pilot” section discusses blind synchronization and demodulator techniques, including coherent detection and noncoherent detection. In the “Pilot-Based Real-Time MODEMs” section, various pilot-assisted receiver topologies are reviewed. Finally, we present the “Conclusion and Outlook” section.
A typical wideband single-carrier modulation transmitter usually uses an up-conversion mixer to up-convert quadrature baseband signals (BB) generated by high-speed dual-channel digital-to-analog converters (DACs) to the desired frequency band. Alternatively, a single-channel DAC can be used to generate a modulated intermediate frequency signal (IF), which can then be up-converted by the mixer. It should be noted that when interfacing with an IF signal, the DAC must support 1.5 to 2 times the bandwidth compared to dual-channel BB interfacing. However, due to various limitations of mm-Wave transmitters, such as limited output power, high component costs, and limited DAC component bandwidth, several alternative transmitter topologies have been proposed. This section discusses the outphasing transmitter [22], delta–sigma modulation (SDM) transmitter [23], and frequency multiplier-based transmitter [24].
Power amplifiers (PAs) operating at mm-Wave frequencies have a lower saturation power and third-order intercept point (IP3) compared to those at lower frequencies. Attempts to increase input power can force the amplifier into nonlinear operation, resulting in amplitude (AM–AM distortion) and phase (AM–PM distortion) distortion of signals. This is especially problematic for high-order modulation signals with a high peak-to-average ratio. In Figure 2, a simulation of 16-QAM transmission constellation diagrams using a realistic PA model is shown. The upper left constellation in Figure 2 shows the signal at 12-dB back-off at input power, while the lower right constellation shows the effects of both AM–AM and AM–PM distortion at saturation. As seen in Figure 2, the outermost constellation points, which have the highest input power, experience the most severe amplitude and phase distortion after amplification, leading to a high error vector magnitude (EVM) value, which is a common measure of transmitter signal quality.
Figure 2. Constellations of 16-QAM signal affected by a PA [22].
Despite the nonlinear effects of PAs on amplitude and phase, with a fixed input power, the AM–AM and AM–PM distortions remain constant. Therefore, for signals with a constant envelope, there is no need to back off in output power. If a signal can be generated by combining two signals of constant amplitude, then the AM–AM and AM–PM distortion can be disregarded, and the entire power range of amplifiers can be utilized. The outphasing technique is one way to achieve this. Figure 3 provides an illustration of the outphasing technique. The basic principle of outphasing is to split the desired waveform S(t) into two complementary signals, ${S}_{1}{(t)}$ and ${S}_{2}{(}{t}{)}$. The transmitted signal can be represented as ${S}{(}{t}{)} = {A}{(}{t}{)}{e}^{{j}{\theta}{(}{t}{)}}$, where the transmitted information is modulated into amplitude A(t) and phase ${\theta}{(}{t}{)}$ of the carrier. Signal S(t) can be decomposed into two signals. By controlling the outphasing angle ${\phi}{(}{t}{)}$, the amplitude of decomposed signals can be set to any value in ${(}{A}{(}{t}{)} / {2},{\infty}{)}$. Two signals can be chosen as: \begin{align*}{S}_{1}{(}{t}{)} & = \frac{{A}_{\max}}{2}{e}^{{j}{[}{\theta}{(}{t}{)} + {\phi}{(}{t}{)}{]}} \tag{1} \\ {S}_{2}{(}{t}{)} & = \frac{{A}_{\max}}{2}{e}^{{j}{[}{\theta}{(}{t}{)}{-}{\phi}{(}{t}{)]}} \tag{2} \\ {A}_{\max} & = {\max}{\{}{A}{(}{t}{)}{\}} \tag{3} \end{align*}
Figure 3. Symbol-based outphasing technique.
where the amplitude of ${S}_{1}{(t)}$ and ${S}_{2}{(t)}$ is a constant value as half of the maximum amplitude of ${S}{(t)}$, and ${\phi}{(}{t}{)}$ is the outphasing angle between decomposed signals and the original signal S(t). And the outphasing angle ${\phi}{(}{t}{)}$ can be represent as: \[{\phi}{(}{t}{)} = {\arccos}{\left(\frac{A(t)}{{A}_{\max}}\right)}{.} \tag{4} \]
With this decomposition technique, an arbitrary waveform (i.e. QAM-modulated signal) can be decomposed into two signals of constant envelope. When these two signals get amplified by two identical PAs, the AM–AM and AM–PM distortions are identical. Combining two amplified signals together, an amplified original signal is generated without any nonlinear distortion. At frequencies below 30 GHz, DACs can produce ${S}_{1}{(t)}$ and ${S}_{2}{(t)}$ signals directly. At high frequency, such as E-band and D-band, it is difficult. In these cases, the baseband I and Q signals can be generated by a DAC and then up-convert them to mm-Wave using a commercial quadrature transmitter, as shown in Figure 4. The baseband signals can be represented as: \begin{align*}{S}_{1}{[}{n}{]} & = \frac{{A}_{\max}}{2}{e}^{{j}{[}{\theta}{[}{n}{]} + {\phi}{[}{n}{]}} \tag{5} \\ {S}_{2}{[}{n}{]} & = \frac{{A}_{\max}}{2}{e}^{{j}{[}{\theta}{[}{n}{]}{-}{\phi}{[}{n}{]}}{.} \tag{6} \end{align*}
Figure 4. System measurement setup [22].
Then, two RF signals are combined at mm-Wave.
A single 16-QAM signal transmission with a single transmitter is measured to serve as a reference. Then two identical 16-QAM signals are combined in-phase with two transmitters. Finally, the proposed symbol-based outphasing power combining is tested with the same setup. The test results are plotted in Figure 5. The outphasing solution can provide lower EVM for a given large output power. For a given EVM, the outphasing solution allows higher output power. For instance, for an EVM of 12% [bit error rate (BER) lower than 1e-4], the outphasing solution allows for more than 12-dBm output power, while a two identical 16-QAM–combined signal is limited to 10 dBm and the single 16-QAM to 8.4 dBm. Thus, 2-dB gain may be obtained with an outphasing solution as compared to the conventional power combining.
Figure 5. Performance comparison of outphasing combining with conventional power combining and a single transmitter [22].
Sigma–delta over fiber (SDoF) employs SDM to encode an up-converted RF signal to a high-speed digital bit stream. The quality of the RF signal is preserved in the digital signal thanks to oversampling and noise shaping. With SDoF, the QAM modulated signal can be generated without using high-speed DACs.
As shown in the Figure 6, the source signal is processed by an SDM that uses a comparator for the following operation: \[{g}_{\text{cmp}}{(}{t}{)} = \begin{cases}\begin{array}{c}{-}{1},{\text{ if }}{V}_{\text{in}}\,{>}\,{0} \\ + {1},{\text{ if }}{V}_{\text{in}}\,{<}\,{0} \end{array}. \end{cases} \tag{7} \]
Figure 6. The structure of the proposed SDoF, including the waveforms and their corresponding spectra [23]. EOC: electrical to optical convertors; OEC: electrical-optical convertors.
The SDM converts an arbitrary input waveform into a binary pulse stream whose in-band spectrum remains the same while quantization noise is presented at out-of-signal band. Such a stream can be generated by an field-programmable gate array (FPGA) with several tens of gigabits per second transceiver interface and suitable for transmission over fiber. The original source signal can be recovered by the filter (BPF) directly at IF frequency or upconverted to wanted band with a mixer.
Examples of 64-QAM signal generation at 2 GHz with 10 Gb/s SDM rate are presented in Figure 7. Different symbol rates of 10 Msym/s, 15 Msym/s, and 20 Msym/s are presented [23]. To enhance transmission capacity, an approach could be to merge multiple narrowband high-order modulation channels into a single wideband signal. To achieve this, it is crucial to minimize the hardware complexity and power consumption of each narrowband signal generation with high modulation. The SDM transmitter solution, referenced in this section, is suitable for generating multiple synchronized high-order modulation signals without the need for a DAC, thereby lowering hardware expenses and power consumption.
Figure 7. The received constellation diagram for different symbol rates when the receiver is formed by the FPGA [23].
The transmitter at mm-Wave frequencies often includes a mixer, frequency multiplier, and LO synthesizer. In some cases, signal modulation can be achieved by feeding a frequency multiplier with a distorted low-frequency modulated signal. Simplified transmitter structures utilizing these techniques are demonstrated in [27], [28], [29], [30], [31], [32], [33]. In [24], an E-band M-ary PSK (MPSK) modulator was proposed that utilized a simple predistortion process involving an unwrapped phase retarder. The experimental setup structure and frequency multiplier monolithic microwave integrated circuit (MMIC) used in this work can be seen in Figure 8. The simple phase retarder used as digital predistortion can be implemented for real-time practical use in application-specific integrated circuit or FPGA.
Figure 8. Schematic and layout of the frequency sextupler (a) and (b), along with the block diagram of the experimental setup (c) [24].
Consider a baseband modulated signal \[{s}{(}{n}{)} = {\mid{s}{(}{n}{)}\mid}{e}^{{-}{j}{\phi}_{s}{(}{n}{)}} \tag{8} \] where ${\phi}_{s}{(}{n}{)} = {\arg}{(}{s}{(}{n}{)}{)}$. The digital samples of s(n) will be converted into analog waveform that drives the frequency multiplier: \[{x}{(}{t}{)} = {\mid{s}{(}{t}{)}\mid}{\cos}{(}{2}{\pi}{f}_{\text{IF}}{t} + {\phi}_{s}{(}{t}{)} + {\phi}_{\text{PN}}{(}{t}{)}{)} \tag{9} \] where ${f}_{\text{IF}}$ is the transmission frequency and ${\phi}_{\text{PN}}{(t)}$ is the phase-noise term. The multiplier generated harmonic tone at wanted frequency is: \[{y}{(}{t}{)} = \mathop{\sum}\limits_{{k} = {1}}\limits^{K}{r}_{k}{\mid{s}{(}{t}{)}\mid}^{{6} + {2}{(}{k}{-}{1}{)}}{\cos}{(}{2}{\pi}{(}{6}{f}_{\text{IF}}{)}{t} + {6}{\phi}_{s}{(}{t}{)} + {6}{\phi}_{\text{PN}}{(}{t}{)}{)} \tag{10} \] where ${r}_{k}$ is the nonlinear coefficient of the k th harmonic, and it can be seen the process implies potential phase-noise deterioration.
The proposed phase retarder is a division of the unwrapped phase, i.e., the retardation of the phase revolution, of the complex baseband signal in (8) and can be described as replacing ${\phi}_{s}(n)$ with ${\phi}_{\text{DPD}}{(}{n}{)} = {(}{\phi}_{s}{(}{n}{)} + {2}{\pi}{c}{(}{n}{)}{)} / {6}$, where c(n) is an arbitrary integer that may change with the index n representing the unwrapping operation, giving: \[{s}_{\text{DPD}}{(}{n}{)} = {\mid{s}{(}{n}{)}\mid}{\exp}{(}{j}{\phi}_{\text{DPD}}{(}{n}{)}{)}{.} \tag{11} \]
The importance of the unwrapping operation has been described in [30] as to prevent the discontinuity of the phase of the predistorted signal. In our work, it is also noticed that such operation helps retain the transition between data symbols and eases the process of subsequent signal processing at the receiver. After DPD, the resulting RF waveform at the output of the X6 now becomes: \[{y}{(}{t}{)} = \mathop{\sum}\limits_{{k} = {1}}\limits^{K}{r}_{k}{\mid{s}{(}{t}{)}\mid}^{{6} + {2}{(}{k}{-}{1}{)}}{\cos}{(}{2}{\pi}{(}{6}{f}_{\text{IF}}{)}{t} + {\phi}_{s}{(}{t}{)} + {6}{\phi}_{\text{PN}}{(}{t}{)}{)}{.} \tag{12} \]
Comparing (12) with (10), it can be seen that the phase impairment is resolved. The remaining amplitude nonlinearity and impaired phase noise will be considered at the digital receiver. It may be important to note that this implementation is not exclusive to a modulation format, but since the amplitude distortion is ignored, a high-order multiple modulation format, e.g., 64-QAM, is unsuitable.
The power spectral density (PSD) of the received signal with and without the phase retarder and an ideal PSD is shown in Figure 9. It can be seen that by implementing the phase retarder, the effect of RF bandwidth expansion by the six-time multiplier has been suppressed. The experimental results of received symbol constellations with the proposed phase retarder process are shown in Figure 10. By constraining to MPSK symbols, phase-noise compensation is done blindly using a modification of a standard Viterbi–Viterbi (VV) algorithm. The postdistorted symbols are stripped of modulation using VV. A 4.8-Gb/s transmission is demonstrated with EVM = 10.7% and BER < 2.1E-4.
Figure 9. Comparison of the received PSD with and without the phase retarder and an ideal PSD. Receiver noise floor is without input signal [24].
Figure 10. Received symbol constellation diagrams: (a) before postdistortion, (b) after postdistortion, (c) after phase-noise compensation [24].
Table 1 summarizes several alternative modulator topologies that have been discussed in this section. These topologies address practical implementation issues in terahertz transmitters. Outphasing has been utilized to enhance linearity, delta–sigma eliminates the need for a DAC, and frequency multiplier-based techniques remove the necessity for a terahertz mixer.
Table 1. mm-Wave modulator topologies.
The recovery of the transmitted bit stream at the receiver end typically involves the use of DSP to account for any frequency and phase offsets. If the demodulator takes in the IF signals from the mm-Wave receiver, high-speed analog-to digital converters (ADCs) are needed to convert the signal into a digital format before DSP can be applied. However, for data rates approaching 100 Gb/s, the bandwidth of the IF signal becomes too broad (often exceeding 10 GHz) to be effectively digitized by commercially available ADCs. For example, the highest reported single-channel data rate in real time mm-Wave communication with DSP platforms is below 5.3 Gb/s. Consequently, alternative demodulation techniques are being investigated and are presented in this section. For different modulation orders, different synchronization structures can be used.
For low modulation order signals (i.e., OOK or ASK), real-time demodulation can be achieved through noncoherent detections. Direct detection can recover transmitted data for signals like OOK and ASK [12].
For the MPSK signal, both coherent and noncoherent topologies can be used. For noncoherent, real-time demodulation for binary PSK (BPSK) and QPSK signals can be achieved through differential coding and detection [13]. For coherent detection, phase and carrier recovery is necessary. Costas loop [17] and multiplication [18] are two commonly used methods for demodulation. However, when dealing with high-bandwidth IF signals, designing frequency multipliers can be particularly challenging. Additionally, for wideband signals, a Costas loop may not converge for higher-order MPSK signals. The complexity of the demodulator increases with an increase in the modulation order of the MPSK signal, rendering these two solutions limited to low-modulation order signals.
For high-modulation order MPSK and QAM signals, real-time coherent detection is accomplished by utilizing a recovered carrier signal. This can be achieved through a full analog phase-lock loop (PLL) or injection-lock voltage controlled oscillator (IL VCO) [14], [19]. The authors of [16] and [20] have proposed analog–digital hybrid carrier recovery solutions that only require low-speed ADCs. Pilot signals are used to assist carrier recovery in these two articles. After the transmitter and receiver are synchronized, the data need to be recovered from the baseband signal for analog and hybrid solutions. Power detectors or phase detectors can be utilized to recover data for low modulation order signals, such as OOK, BPSK, and QPSK. However, for higher modulation order signals like 16-QAM, the I and Q channel signal becomes a four-pulse amplitude modulation (PAM) signal after synchronization. Intel offers an FPGA that can operate a PAM signal up to 28.9 Gbaud [21], negating the need for a high-speed ADC in the system.
In this section, several demodulation methods for single carrier data transmission without any pilot signal are discussed. First, noncoherent detection methods are explained, where carrier recovery is not required, for example, OOK demodulation and differential BPSK and QPSK detection. Second, several hardware structures for carrier recovery are also discussed in this section.
OOK and PAM modulate information solely on the amplitude of the carrier, where a carrier envelope detector is used for demodulation and receiving. The main challenge for designing a high data rate OOK link lies in the development of wideband mm-Wave transceivers. A 42-Gb/s link at 300 GHz has been demonstrated at a 1-m distance using a unitraveling carrier photodiode photo-mixing transmitter and Schottky barrier diode as envelope detector [12]. OOK modulation occupies an ultrawide bandwidth, and typically equals twice the baud rate, which is impractical for wireless transmission where such wideband operation will not comply with any frequency allocation regulation.
Alternatively, such simple modulations found their application in polymer microwave fiber (PMF) transmission. When a mm-Wave signal is transmitted over PMF, the electromagnetic field is constrained within or near the PMF where frequency allocation regulations are not applied. In [25], we have reported a 40-Gb/s PMF data transmission using a 60- to 90-GHz band. An RF DAC designed in a 250-nm indium phosphide double heterojunction bipolar transistor process is used as a PAM modulator, and the link demonstrated a high-energy efficiency of 1.2 pJ/bit.
For the additive white Gaussian noise channel, QPSK modulation requires the least signal-to-noise ratio (SNR) to achieve the same level of BER in comparison with higher QAM modulations, and yet it has higher spectrum efficiency than OOK or BPSK. QPSK modulation also has less requirement for transmitter linearity; therefore, it is widely adopted for satellite communication and high-frequency mm-Wave links. QPSK transmission modulates the information onto the phase of the mm-Wave carrier, thus at the receiver side, carrier recovery is required for correct demodulation.
Alternatively, differential encoder and noncoherent detection can be used jointly. In a DQPSK system, the baseband signal only employs binary status +/−1, therefore no DAC is required at the modulator side. A binary stream up to 40 Gb/s can be directly generated by programming an high-speed FPGA. On the receiver side, differential detection can be implemented with a wideband exclusive OR gate as phase detector and a long transmission line as symbol delay, this solution eliminates the need for costly and power-hungry ADCs. A block diagram of an analog DQPSK demodulator is shown in Figure 11.
Figure 11. Analog DQPSK demodulator [35]. LPF: low pass filter.
This DQPSK concept was presented in [34] and [35] with 2.5 Gb/s and 5 Gb/s, respectively. The symbol delay is a physical transmission line that is designed for the operational baud rate; therefore, the system only operates at a fixed data rate. In [36], a data-rate adaptable DQPSK modem solution for mm-Wave wireless fronthaul networks is proposed. By software adjustment of the differential encoding rule, several common public radio interface (CPRI) data rates can be supported without modifying the receiver hardware. A field demonstration of CPRI transmission over an E-band link was performed on 19 December 2011 jointly with Ericsson, as Figure 12 shows [37]. In [38], a 40 Gb/s DQPSK transmission designed for D-band link operation was demonstrated.
Figure 12. Demonstration of CPRI over E-band link, Beijing, 19 December 2011 [37].
DQPSK modem designs still have several drawbacks: when demodulation error occurs, the error affects the preceding symbol; therefore, higher SNR is required to achieve the same BER in comparison with a coherent QPSK receiver; due to the physical delay element, the receiver only works at symbol rates that integer times of a base rate. Several hardware solutions for wideband signal carrier recovery are discussed in the next part of this section.
An analytical study for oscillator’s behavior under signal injection was performed in [39], which implies injection locking can be used for carrier recovery in narrowband modulation. Several works with data rate up to 320 Mbps were reported in [40] and [41]. We have demonstrated a 12-Gb/s QPSK carrier recovery and demonstration using a GaAs differential Colpitts VCO in [19]. The block diagram of injection locking-based carrier recovery is shown in Figure 13; the differential ports are used for modulated signal injection and recovered carrier extraction, respectively. At wideband modulation, the modulated signal’s power density is much lower than the carrier signal (due to LO to RF signal leakage); by adjusting VCO tuning voltage, the VCO can inject locked to the carrier tone and modulation is suppressed. In Figure 14, the measured demodulated signal BER and SNR are given. The injection locking receiver can operate at an arbitrary symbol rate and the implementation loss is around 1 dB compared with the transceiver synchronization case.
Figure 13. block diagram of IL VCO-based carrier recovery [19].
Figure 14. measured received SNR and BER [19]. Tx: transmitter; Rx: receiver.
By frequency multiplication, the phase modulation can be removed. For example, the fourth harmonic of a QPSK modulated signal appears as a sinusoidal tone where all modulated phases are folded into the same phase. This principle can be used for feedforward carrier recovery; however, at mm-Wave frequencies, the frequency multiplier is not widely available. In [42], a dynamic divider MMIC was developed for carrier recovery, as shown in Figure 15. The developed MMIC has −54-dBm sensitivity, which is promising for weak signal reception.
Figure 15. PLL-based carrier recovery concept for a E-band communication link [42].
To reduce the complexity directly at mm-Wave frequency, a down-conversion stage can be used to obtain an IF frequency. In [43], a multiplication-by-4 then divide-by-4 CR structure was proposed, as Figure 16 shows. The MMICs are made in an IHP 250-nm SiGe bipolar CMOS process and it demonstrated up to 5 Gb/s QPSK and 8 frequency shift keying demodulation.
Figure 16. Schematic of the proposed analog-oriented receiver [43].
We have reported a 15 Gb/s 8-PSK receiver using a comparator and a divider in [26]. In this work, the comparator is used as harmonic generation instead of a frequency multiplier. By adjusting the comparator threshold, the harmonic components generated from a modulated signal can be controlled. The block diagram of the proposed 8-PSK receiver is shown in the Figure 17.
Figure 17. Structure of the proposed 8-PSK demodulator [26]. VGA: variable gain amplifier; PS: phase shifter.
FPGAs with high-speed fiber optic transceivers are now available up to 56 Gb/s. The electrical-optical convertors (OEC) can be used as single bit comparator. Thanks to high production volume, 10-Gb/s OECs become affordable. A10-Gb/s OEC as comparator can be used for MPSK signal demodulation and carrier recovery, as we have demonstrated [44].
For MPSK signal, the data are only modulated on the phase of the IF signal. The IF signal can be represented by a stream of binary code without loss of information. An MPSK transmitter and receiver structure using 10 Gb/s OECs is shown as block diagram in Figure 18.
Figure 18. An MPSK system structure of mm-Wave point-to-point radio link [44].
The received IF signal after limiting amplifier (LA) will be converted as a square wave and digitalized by OEC into a binary code stream, as Figure 19 shows. Then the binary code stream is processed by an FPGA for MPSK demodulation. The OEC is converted is embedded with a clock-data recovery (CDR) unit, which automatically aligns a digitizer clock with received IF frequency and a special signal process is required for symbol time recovery.
Figure 19. Waveforms of a received PSK IF signal and a MPSK IF signal after an LA [44].
An E-band MPSK data transmission test was preformed using this method and test setup and the result is plotted in Figure 20. It can be seen with a fixed 10 Gb/s OEC, QPSK, 8-PSK, and 16-PSK data transmission was validated with BER < 10e-12.
Figure 20. mm-Wave point-to-point synchronization baseband link test setup and test result: (a) experimental setup; (b) received EVM performance with QPSK, 8-PSK, and 16-PSK modulations; (c) received BER performance with QPSK, 8-PSK, and 16-PSK modulations [44].
Apart from the blind synchronization techniques discussed earlier, which do not require additional information, the insertion of a pilot signal can significantly aid synchronization implementation. The pilot signal carries information that the receiver already knows. By analyzing the pilot signal, the carrier offset can be determined, and synchronization can be achieved. However, the use of a pilot signal incurs an energy cost either in the frequency or the time domain ƒ. This section discusses various solutions for pilot-based synchronization and receiver topologies.
Inserting a single-tone continuous wave (CW) signal in the frequency domain is a common technique for pilot insertion. The pilot tone can be injected, as depicted in Figure 21, at various frequency offsets from the transmitted carrier frequency, including in the middle of the carrier, outside the modulated signal band, and embedded inside the modulated signal band [45].
Figure 21. Choices of a pilot tone insertion: (a) at carrier frequency, (b) out-of-band, (c) in-band [45].
In the first scenario depicted in Figure 21(a), the pilot signal is placed in the middle of the transmitted signal. Given that carrier leakage occurs naturally at the transceiver front end modules, this is a frequent way for inserting pilots. Normally, a heterodyne receiver is needed in this case [46]. The received signal is down-converted to an IF frequency where the IF information can be obtained from the pilot tone. As shown in Figure 21(b), the pilot tone can also be placed outside the modulated signal band. In this case, a BPF may easily extract it. Carrier recovery and phase-noise mitigation can be done with the assistance of this pilot signal. The downside of this approach is the spectrum efficiency deterioration brought on by the additional bandwidth requirement. Another option, shown in Figure 21(c), is the pilot tone being embedded inside the modulated signal band. In this case, the pilot signal cannot be removed from the modulated signal by a BPF, so the pilot signal will introduce additional EVM. Its power needs to be as low as possible while strong enough to be detected by the receiver.
There are also three widely used pilot extraction methods, shown in Figure 22 [45]: the digital feedforward, analog feedforward, and analog feedback methods. For digital methods, the advanced DSP pilot extraction solution can be used. However, the sampling rate of the ADC is the bottleneck of the communication data rate.
Figure 22. Pilot extraction methods: (a) full digital feed-forward, (b) analog feed-forward, (c) analog feedback [45].
In [16], a 21-MHz CW pilot tone is inserted inside the baseband signal. On the receiver side, a low-cost ADC with a sampling rate of 100 MSps is used to extract the pilot signal. The system structure of the E-band link is shown in Figure 23(a), and the block diagram of its synchronous baseband receiver is shown in Figure 23(b). At the transmitter, a fractional (1/N) frequency of the carrier is superimposed within the baseband signal as a pilot tone. At the receiver, a DSP platform digitizes this pilot signal, analyzes the frequency offset, and then performs receiver LO adjustment using a DAC and a single sideband (SSB) mixer. The carrier recovery is realized in an analog–digital hybrid way with a feedback loop topology, as Figure 22(c) shows. A proof-of-concept demonstration at E-band achieves 9 Gb/s 64 QAM and 16-Gb/s QPSK transmissions.
Figure 23. (a) The E-band link structure and (b) the block diagram of the synchronous baseband receiver in [16].
Besides inserting the pilot signal in the spectrum, inserting the pilot signal in the time domain is also a very common solution [47], [48], [49]. The pilot sequence is placed between data frames periodically in the time domain, as Figure 24 shows [45]. With this method, the pilot signal takes an additional time slot. The advantage of using a sequence instead of a single-tone CW as a pilot signal is the pilot carries more information. The additional information can provide more functionality. For example, it can be used for channel estimation and equalization in a multiple input/multiple output communication system.
Figure 24. Pilot insertion in the time domain [45].
The sequence can also be put in a distinct frequency band, as in Figure 25(a), where the pilot occupies extra bandwidth. Alternatively, the pilot can also be inserted inside the baseband signal, as Figure 25(b) shows [45]. In this case, the pilot and the data overlap each other. Instead of taking additional bandwidth, it takes extra power from the transmitted signal. A special processing technique is needed to separate them. In [20], a pseudorandom noise-coded pilot signal is used. The pilot signal is superimposed within the baseband signal, as Figure 25(b) shows. Its system structure is similar to [16], as Figure 26 shows; the difference is that a digital matched filter is used to extract the pilot rather than a BPF. A proof-of-concept communication link at the E-band is demonstrated with 18-Gb/s 64-QAM and 24-Gb/s 16-QAM with a requirement of ADC sampling rate of 40 MSps. A comparison of recent publications on mm-Wave transmissions with synchronization is shown in Table 2. This work demonstrated an approach to recovery of high data rate signal with a much lower sampling rate ADC.
Figure 25. Choices of a sequence pilot insertion: (a) out-of-band, (b) in-band [45].
Figure 26. (a) The E-band link structure and (b) the block diagram of the synchronous baseband receiver in [20].
Table 2. Recently reported real time transmissions over mm-Wave bands [20].
In this article, several real-time high data rate single carrier transmitter and receiver structures are summarized. For the transmitter, outphasing power combining is used for enhancing transmitter linearity, SDM is used for DAC-less high-order QAM generation, and a multiplier-based transmitter is proposed for a mixerless simplified transmitter structure. On the receiver side, both coherent and noncoherent demodulator structures are summarized. A noncoherent demodulator does not require any ADCs, and coherent detection only requires low-speed ADCs for wideband signal demodulation. These solutions mitigated the problem of lacking high-speed ADCs in traditional software-defined radio receiver architecture.
[1] “Ericsson mobility report,” Ericsson, Stockholm, Sweden, Jun. 2018. [Online] . Available: https://www.ericsson.com/en/reports-and-papers/mobility-report
[2] S. Carpenter et al., “A D-band 48-Gbit/s 64-QAM/QPSK direct-conversion I/Q transceiver chipset,” IEEE Trans. Microw. Theory Techn., vol. 64, no. 4, pp. 1285–1296, Apr. 2016, doi: 10.1109/TMTT.2016.2533491.
[3] K. K. Tokgoz et al., “A 120Gb/s 16QAM CMOS millimeter-wave wireless transceiver,” in Proc. IEEE Int. Solid-State Circuits Conf. (ISSCC), San Francisco, CA, USA, 2018, pp. 168–170, doi: 10.1109/ISSCC.2018.8310237.
[4] P. Rodríguez-Vázquez, J. Grzyb, B. Heinemann, and U. R. Pfeiffer, “A 16-QAM 100-Gb/s 1-M wireless link with an EVM of 17% at 230 GHz in an SiGe technology,” IEEE Microw. Wireless Compon. Lett., vol. 29, no. 4, pp. 297–299, Apr. 2019, doi: 10.1109/LMWC.2019.2899487.
[5] P. Rodríguez-Vázquez, J. Grzyb, B. Heinemann, and U. R. Pfeiffer, “Performance evaluation of a 32-QAM 1-meter wireless link operating at 220–260 GHz with a data-rate of 90 Gbps,” in Proc. Asia-Pacific Microw. Conf. (APMC), Kyoto, Japan, 2018, pp. 723–725, doi: 10.23919/APMC.2018.8617171.
[6] H. Hamada et al., “300-GHz. 100-Gb/s InP-HEMT wireless transceiver using a 300-GHz fundamental mixer,” in Proc. IEEE/MTT-S Int. Microw. Symp. (IMS), Philadelphia, PA, USA, 2018, pp. 1480–1483, doi: 10.1109/MWSYM.2018.8439850.
[7] F. Boes et al., “Ultra-broadband MMIC-based wireless link at 240 GHz enabled by 64GS/s DAC,” in Proc. 39th Int. Conf. Infrared, Millimeter, THz Waves (IRMMW-THz), Tucson, AZ, USA, 2014, pp. 1–2, doi: 10.1109/IRMMW-THz.2014.6956202.
[8] I. Ando, M. Tanio, M. Ito, T. Kuwabara, T. Marumoto, and K. Kunihiro, “Wireless D-band communication up to 60 Gbit/s with 64QAM using GaAs HEMT technology,” in Proc. IEEE Radio Wireless Symp. (RWS), Austin, TX, USA, 2016, pp. 193–195, doi: 10.1109/RWS.2016.7444402.
[9] I. Kallfass et al., “All active MMIC-based wireless communication at 220 GHz,” IEEE Trans. Thz Sci. Technol., vol. 1, no. 2, pp. 477–487, Nov. 2011, doi: 10.1109/TTHZ.2011.2160021.
[10] V. Vassilev et al., “Spectrum efficient D-band communication link for real-time multi-gigabit wireless transmission,” in Proc. IEEE/MTT-S Int. Microw. Symp. (IMS), Philadelphia, PA, USA, 2018, pp. 1523–1526, doi: 10.1109/MWSYM.2018.8439258.
[11] Z. He et al., “A hardware efficient implementation of a digital baseband receiver for high-capacity millimeter-wave radios,” IEEE Trans. Microw. Theory Techn., vol. 63, no. 5, pp. 1683–1692, May 2015, doi: 10.1109/TMTT.2015.2417541.
[12] T. Nagatsuma, “Breakthroughs in photonics 2013: THz communications based on photonics,” IEEE Photon. J., vol. 6, no. 2, pp. 1–5, Apr. 2014, doi: 10.1109/JPHOT.2014.2309643.
[13] H. Takahashi, A. Hirata, J. Takeuchi, N. Kukutsu, T. Kosugi, and K. Murata, “120-GHz-band 20-Gbit/s transmitter and receiver MMICs using quadrature phase shift keying,” in Proc. 7th Eur. Microw. Integr. Circuit Conf., Amsterdam, The Netherlands, 2012, pp. 313–316.
[14] P. Harati, A. Dyskin, and I. Kallfass, “Analog carrier recovery for broadband wireless communication links,” in Proc. 48th Eur. Microw. Conf. (EuMC), Madrid, Spain, 2018, pp. 1401–1404, doi: 10.23919/EuMC.2018.8541539.
[15] C. Wang et al., “0.34-THz wireless link based on high-order modulation for future wireless local area network applications,” IEEE Trans. THz Sci. Technol., vol. 4, no. 1, pp. 75–85, Jan. 2014, doi: 10.1109/TTHZ.2013.2293119.
[16] S. An, Z. S. He, J. Chen, H. Han, J. An, and H. Zirath, “A synchronous baseband receiver for high-data-rate millimeter-wave communication systems,” IEEE Microw. Wireless Compon. Lett., vol. 29, no. 6, pp. 412–414, Jun. 2019, doi: 10.1109/LMWC.2019.2910661.
[17] S. Huang, Y. Yeh, H. Wang, P. Chen, and J. Lee, “W-band BPSK and QPSK transceivers with Costas-loop carrier recovery in 65-nm CMOS technology,” IEEE J. Solid-State Circuits, vol. 46, no. 12, pp. 3033–3046, Dec. 2011, doi: 10.1109/JSSC.2011.2166469.
[18] A. Ç. Ulusoy and H. Schumacher, “Multi-Gb/s analog synchronous QPSK demodulator with phase-noise suppression,” IEEE Trans. Microw. Theory Techn., vol. 60, no. 11, pp. 3591–3598, Nov. 2012, doi: 10.1109/TMTT.2012.2216894.
[19] Z. He, D. Kuylenstierna, S. Lai, and H. Zirath, “A 12 Gbps analog QPSK baseband receiver based on injection-locked VCO,” in Proc. IEEE MTT-S Int. Microw. Symp., Phoenix, AZ, USA, 2015, pp. 1–4, doi: 10.1109/MWSYM.2015.7166905.
[20] S. An, Z. S. He, J. Li, X. Bu, and H. Zirath, “Coded pilot assisted baseband receiver for high data rate millimeter-wave communications,” IEEE Trans. Microw. Theory Techn., vol. 68, no. 11, pp. 4719–4727, Nov. 2020, doi: 10.1109/TMTT.2020.2993789.
[21] “Stratix 10 TX device overview,” Intel, Aug. 2019.
[22] S. An, Z. S. He, J. Chen, Y. Li, and H. Zirath, “An 8 Gbps E-band QAM transmitter using symbol-based outphasing power combining technique,” in Proc. IEEE Int. Symp. Radio-Freq. Integr. Technol. (RFIT), 2017, pp. 150–152, doi: 10.1109/RFIT.2017.8048232.
[23] I. C. Sezgin, L. Aabel, S. Jacobsson, G. Durisi, Z. S. He, and C. Fager, “All-digital, radio-over-fiber, communication link architecture for time-division duplex distributed antenna systems,” J. Lightw. Technol., vol. 39, no. 9, pp. 2769–2779, May 2021, doi: 10.1109/JLT.2021.3057609.
[24] D. Nopchinda, Z. He, G. Granström, M. Gavell, and H. Zirath, “8-PSK upconverting transmitter using E-band frequency sextupler,” IEEE Microw. Wireless Compon. Lett., vol. 28, no. 2, pp. 177–179, Feb. 2018, doi: 10.1109/LMWC.2018.2792687.
[25] F. Strömbeck, Z. S. He, and H. Zirath, “A RF-DAC based 40 Gbps PAM modulator with 1.2 pJ/bit energy efficiency at millimeterwave band,” in Proc. IEEE/MTT-S Int. Microw. Symp. (IMS), 2018, pp. 931–933, doi: 10.1109/MWSYM.2018.8439488.
[26] Z. He, D. Nopchinda, T. Swahsn, and H. Zirath, “A 15-Gb/s 8-PSK demodulator with comparator-based carrier synchronization,” IEEE Trans. Microw. Theory Techn., vol. 63, no. 8, pp. 2630–2637, Aug. 2015, doi: 10.1109/TMTT.2015.2444851.
[27] P. Nazari, S. Jafarlou, and P. Heydari, “A CMOS two-element 170-GHz fundamental-frequency transmitter with direct RF-8PSK modulation,” IEEE J. Solid-State Circuits, vol. 55, no. 2, pp. 282–297, Feb. 2020, doi: 10.1109/JSSC.2019.2945263.
[28] S. Kang, S. V. Thyagarajan, and A. M. Niknejad, “A 240 GHz fully integrated wideband QPSK transmitter in 65 nm CMOS,” IEEE J. Solid-State Circuits, vol. 50, no. 10, pp. 2256–2267, Oct. 2015, doi: 10.1109/JSSC.2015.2467179.
[29] K. Takano, S. Hara, K. Katayama, S. Amakawa, T. Yoshida, and M. Fujishima, “Quintic mixer: A subharmonic up-conversion mixer for THz transmitter supporting complex digital modulation,” in Proc. IEEE MTT-S Int. Microw. Symp. Dig., May 2016, pp. 1–3, doi: 10.1109/MWSYM.2016.7540420.
[30] Y. Park and J. S. Kenney, “Adaptive digital predistortion linearization of frequency multipliers,” IEEE Trans. Microw. Theory Techn., vol. 51, no. 12, pp. 2516–2522, Jan. 2004, doi: 10.1109/TMTT.2003.819771.
[31] J. H. Choi, S. M. Kang, and K. H. Koo, “Digital predistortion of frequency multiplier for dual band wireless LAN transmitter,” in Proc. IEEE MTT-S Int. Microw. Symp. Dig., Jun. 2005, pp. 1–4, doi: 10.1109/MWSYM.2005.1517136.
[32] Y. Liu, G. Liu, and P. M. Asbeck, “High-order modulation transmission through frequency quadrupler using digital predistortion,” IEEE Trans. Microw. Theory Techn., vol. 64, no. 6, pp. 1896–1910, Jun. 2016, doi: 10.1109/TMTT.2016.2561276.
[33] A. Chung, A. M. Darwish, E. Viveiros, H. A. Hung, P. Mitran, and S. Boumaiza, “Analysis and compensation of nonidealities in frequency multiplier-based high-frequency vector signal generators,” IEEE Trans. Microw. Theory Techn., vol. 67, no. 6, pp. 2270–2283, Jun. 2019, doi: 10.1109/TMTT.2019.2909901.
[34] Z. He, J. Chen, Y. Li, and H. Zirath, “A novel FPGA-based 2.5Gbps D-QPSK modem for high capacity microwave radios,” in Proc. IEEE Int. Conf. Commun., 2010, pp. 1–4, doi: 10.1109/ICC.2010.5501842.
[35] Z. He, W. Wu, J. Chen, Y. Li, D. Stackenas, and H. Zirath, “An FPGA-based 5 Gbit/s D-QPSK modem for E-band point-to-point radios,” in Proc. 41st Eur. Microw. Conf., 2011, pp. 690–692, doi: 10.23919/EuMC.2011.6101754.
[36] J. Chen, Z. He, Y. Li, T. Swahn, and H. Zirath, “A data-rate adaptable modem solution for millimeter-wave wireless fronthaul networks,” in Proc. IEEE Int. Conf. Commun. Workshop (ICCW), 2015, pp. 1–6, doi: 10.1109/ICCW.2015.7247066.
[37] Z. Ghebretensaé et al., “Transmission solutions and architectures for heterogeneous networks built as C-RANs,” in Proc. Int. ICST Conf. Commun. Netw. China (CHINACOM), Kunming, China, Aug. 2012, pp. 748–752, doi: 10.1109/ChinaCom.2012.6417583.
[38] S. An, J. Chen, Z. He, S. Wang, and H. Zirath, “A 40 Gbps DQPSK modem for millimeter-wave communications,” in Proc. Asia-Pacific Microw. Conf. (APMC), 2015, pp. 1–3, doi: 10.1109/APMC.2015.7411747.
[39] B. Razavi, “A study of injection locking and pulling in oscillators,” IEEE J. Solid-State Circuits, vol. 39, no. 9, pp. 1415–1424, 2008, doi: 10.1109/JSSC.2004.831608.
[40] Q. Zhu and Y. Xu, “A 228 µW 750 MHz BPSK demodulator based on injection locking,” IEEE J. Solid-State Circuits, vol. 46, no. 2, pp. 416–423, Feb. 2011, doi: 10.1109/JSSC.2010.2090611.
[41] M. Tarar and Z. Chen, “A direct down-conversion receiver for coherent extraction of digital baseband signals using the injection locked oscillators,” in Proc. IEEE Radio Wireless Symp., Jan. 2008, pp. 57–60, doi: 10.1109/RWS.2008.4463427.
[42] T. Messinger, D. Müller, J. Antes, S. Wagner, A. Tessmann, and I. Kallfass, “Divide-by-8 phase detector MMIC for PLL-based carrier recovery in E-band communication,” in Proc. German Microw. Conf., 2015, pp. 237–240, doi: 10.1109/GEMIC.2015.7107797.
[43] A. Çag˘ri Ulusoy and H. Schumacher, “A system-on-package analog synchronous QPSK demodulator for ultra-high rate 60 GHz wireless communications,” in Proc. IEEE MTT-S Int. Microw. Symp., 2011, pp. 1–4, doi: 10.1109/MWSYM.2011.5972692.
[44] W. Zhang, S. An, J. An, X. Bu, and Z. S. He, “An MPSK millimeter-wave point-to-point link with radio over fiber synchronous baseband receiver,” J. Lightw. Technol., vol. 40, no. 2, pp. 481–489, Jan. 2022, doi: 10.1109/JLT.2021.3110374.
[45] S. An, “Millimeter-wave communication and radar sensing — Opportunities, challenges, and solutions,” Ph.D. dissertation, Dept. Microtechnol. Nanosci., Chalmers Univ. Technol., Gothenburg, Sweden, 2021.
[46] W.-R. Peng, T. Tsuritani, and I. Morita, “Simple carrier recovery approach for RF-pilot-assisted PDM-CO-OFDM systems,” J. Lightw. Technol., vol. 31, no. 15, pp. 2555–2564, Aug. 2013, doi: 10.1109/JLT.2013.2270551.
[47] A. Spalvieri and L. Barletta, “Pilot-aided carrier recovery in the presence of phase noise,” IEEE Trans. Commun., vol. 59, no. 7, pp. 1966–1974, Jul. 2011, doi: 10.1109/TCOMM.2011.051311.100047.
[48] N. Noels, H. Steendam, M. Moeneclaey, and H. Bruneel, “Carrier phase and frequency estimation for pilot-symbol assisted transmission: Bounds and algorithms,” IEEE Trans. Signal Process., vol. 53, no. 12, pp. 4578–4587, Dec. 2005, doi: 10.1109/TSP.2005.859318.
[49] F. Rice, “Carrier-phase and frequency-estimation bounds for transmissions with embedded reference symbols,” IEEE Trans. Commun., vol. 54, no. 2, pp. 221–225, Feb. 2006, doi: 10.1109/TCOMM.2005.863782.
[50] S. Li, M.-Y. Huang, D. Jung, T.-Y. Huang, and H. Wang, “A mm-wave current-mode inverse outphasing transmitter front-end: A circuit duality of conventional voltage-mode outphasing,” IEEE J. Solid-State Circuits, vol. 56, no. 6, pp. 1732–1744, Jun. 2021, doi: 10.1109/JSSC.2020.3038882.
[51] D. Nopchinda, “OFDM upconverting transmitter using a frequency multiplier,” in Proc. 52nd Eur. Microw. Conf. (EuMC), Milan, Italy, 2022, pp. 20–23, doi: 10.23919/EuMC54642.2022.9924491.
[52] I. Jaffri, A. B. Ayed, A. M. Darwish, P. Mitran, and S. Boumaiza, “Novel baseband equivalent model for digital predistortion of wideband frequency-multiplier-based millimeter wave sources,” IEEE Trans. Microw. Theory Techn., vol. 68, no. 9, pp. 3942–3957, Sep. 2020, doi: 10.1109/TMTT.2020.2992914.
[53] C. Wu, H. Li, J. V. Kerrebrouck, C. Meysmans, P. Demeester, and G. Torfs, “A bit-interleaved sigma-delta-over-fiber fronthaul network for frequency-synchronous distributed antenna systems,” Appl. Sci., vol. 11, no. 23, Dec. 2021, Art. no. 11471, doi: 10.3390/app112311471.
Digital Object Identifier 10.1109/MMM.2023.3277361