The future radio access network (RAN) will continue to adopt a high-capacity transmission medium (e.g., optical fiber [1]) to connect central and remote units, and use servers in far edge and edge clouds to support growing end-users with centralized or distributed manners [2,3]. Furthermore, the new edge will proliferate moving geographically closer to the end-users to support lower latency, and access points densify as capacity demand drives toward more small cells and higher radio frequency bands which have an inherently shorter range [3]. With the exponentially increasing cell density and the development of the future RAN, complex and power-hungry remote units are not desirable. In light of this, the low-PHY functional split options might regain interest in the near-future RAN [4]. As a result, the low PHY functional split options with densified cell sites in the urban areas will demand ultrahigh transmission capacity. However, the present optical-fiber connected RAN infrastructure is only feasible in fiber-rich areas, and fiber resource is a scarce asset in urban areas where it might be impossible to deploy it. Therefore, the deployment of densified RAN with low civil engineering costs and high flexibility remains a significant barrier [5] for the RAN.

Terahertz (THz) band communication is envisioned as a key technology for the future 6G wireless communications, which could provide comparable capacity with optical fiber links [6]. The realization methodologies of THz communication can be categorized as electronic THz and photonic THz communication systems. The photonic THz communication systems inherit the benefits from broadband optoelectronic devices, showing advantages like high capacity, low distortion, and feasibility of convergence with optical fiber transmission links [7]. In dense urban areas, the photonic THz communication systems could play an important role as the wireless access relay and extension for the fiber-based RAN, and provide seamless “last-mile” connections to the end-users in the dense small cells [8]. Therefore, the investigation and demonstration of hybrid fiber–THz fronthaul with low-PHY functional split options would be of interest to the RAN operators [2]. The application scenario of hybrid fiber–THz fronthaul with the potential option-9 split (delta-sigma modulation, Δ-Σ [9]) is shown in Fig. 1, where the delta-sigma modulation is used to improve the system efficiency and reduce the cost at the receiver side.

 

Fig. 1. Application scenario of hybrid fiber–THz fronthaul with the potential option-9 split (delta-sigma modulation, Δ-Σ).

Download Full Size | PDF

In this Letter, we propose and experimentally demonstrate a hybrid fiber–THz mobile fronthaul system supporting high-order wireless signal transmission with the delta-sigma modulation. A uni-traveling-carrier photodiode (UTC-PD) operating in the frequency range of 280–380 GHz (NTT Electronics Corp., IOD-PMJ-13001) is employed in the setup as the THz emitter. In the terahertz signal reception, a commercial sub-harmonic terahertz mixer (Virginia Diodes, WR3.4MixAMC-I, 220–330 GHz) is used to downconvert the THz signals to the intermediate frequency (IF). The mixer supports terahertz signals ranging from 220 GHz to 330 GHz, and the IF output bandwidth is 40 GHz. Thus, the THz carrier frequency is selected as 300 GHz in the experiment, which also fits the potential spectrum regularization of the future 6G communications.

The setup of the proof-of-concept experiment is shown in Fig. 2. In the photonic THz frontend, the optical carrier from a C-band external cavity laser (ECL-1) is launched into a polarization controller (PC) to control the polarization state of the signal into an in-phase and quadrature-phase Mach–Zehnder modulator (IQM, 40-GHz bandwidth), where an auto-bias-control unit and baseband amplifiers are integrated with the module. The delta-sigma sequence is modulated to 10-Gbaud QPSK (quadrature phase-shift keying) and digital-to-analog converted by an arbitrary waveform generator (AWG, Keysight M8194A, 3-dB bandwidth: 50 GHz; 8-bit vertical resolution), the length of the sequence is set to 524,288 samples. The optical baseband signal is then amplified by an erbium-doped fiber amplifier (EDFA) and filtered by an optical bandpass filter (OBF) to suppress the out-of-band amplified spontaneous emission noise. After 10-km standard single-mode fiber (SSMF) transmission, the polarization state is aligned with an optical local oscillator (ECL-2) by a PC, and a polarizer (Pol.) is used to align the polarization. The frequency difference between ECL-1 and ECL-2 is 300 GHz. The coupled optical spectrum is shown in Fig. 2(a). The coupled optical signals are finally sent into a UTC-PD for photo-mixing generation of THz signals [10,11].

 

Fig. 2. Setup of the proof-of-concept experiment: (a) optical spectrum of the combined signal before UTC-PD; (b) electrical spectrum of the IF signal.

Download Full Size | PDF

The THz signals are transmitted through a transmitting horn antenna (25-dBi gain) and radiated into a 1-m wireless link with a pair of lenses to collimate the THz beam. At the receiver, the THz signals are received by a receiving horn antenna (25-dBi gain). A sub-harmonic THz mixer is used to receive the signal. The output IF signal [as shown in Fig. 2(b)] is then analog-to-digital converted by a broadband real-time digital sampling oscilloscope (DSO, 160 GSa/s, Keysight DSOZ594A, 3-dB bandwidth: 59 GHz) for further signal processing [12].

The transmission performance of the delta-sigma sequence mapped QPSK signal over the fiber–THz link is shown in Fig. 3. The propagation attenuation of the THz signal over the air is comparatively larger than the optical fiber, the delta-sigma modulation over QPSK format is used to enhance the fronthaul system sensitivity and power penalty, and the access reach could be increased. For 1-bit and 2-bit delta-sigma modulation, the NRZ and PAM4 formats can also be used with Schottky barrier diode based direct detection, where the receiver sensitivity and transmission performance will be influenced. The bit error rate (BER) drops with the increase of the photocurrent of the UTC-PD, and the QPSK constellation graphs at different photocurrents are shown in the insets of Fig. 3. The photocurrent is proportional to the input optical power of the UTC-PD, which is controlled by a variable optical attenuator (VOA) before the UTC-PD. When the photocurrent of the UTC-PD is larger than 2 mA, the system can achieve error-free transmission.

 

Fig. 3. Transmission performance of the photonic THz link over 10-km fiber and 1-m wireless distance.

Download Full Size | PDF

Before the transmission performance of the fiber–THz fronthaul system is further investigated, we first analyze the principles of quantization noise suppressed delta-sigma modulation. The delta-sigma modulation can transform the analog signal to the digital domain at the transmitter and use just a filter to restore the analog signal at the receiving side with no need for a digital-to-analog-converter (DAC). The delta-sigma modulation oversamples the analog signal to expand its Nyquist zone so that the quantization noise is distributed over a wider frequency range to reduce the in-band noise [13]. More quantization noise can be pushed out of the signal band by noise shaping techniques. Therefore, analog signals can be converted to on-off keying (OOK) or QPSK with only one or two quantization bits. On the receiving end, the signal can be easily retrieved without needing a DAC by filtering out the out-of-band quantization noise. To further reduce the quantization noise of the delta-sigma modulation, the proposed fiber–THz fronthaul system adopts a quantization noise suppressed scheme, as shown in Fig. 4. Its modulation prat consists of two 1-bit delta-sigma modulators, a filter, and an encoder, and the demodulation part consists of a filter and a decoder [14]. The modulation steps are briefly described as follows.

 

Fig. 4. Block diagram of the quantization noise suppressed delta-sigma modulation.

Download Full Size | PDF

Through the steps mentioned above, the quantization noise of the delta-sigma modulator is further reduced, and the signal-to-noise ratio is increased from 45 dB of the traditional 2-bit delta-sigma modulator to 67 dB, which enables the system to transmit orthogonal frequency division multiplexing (OFDM) modulation up to 16384 order QAM mapping.

To verify the performance of the fronthaul system, a 10-carrier aggregated QAM-OFDM analog signal matching the 5G-NR channel parameters is designed as the input signal. The signal spectrum is shown in Fig. 5(a), and the spectrum of the output digital after the first delta-sigma modulator is shown in Fig. 5(b). The signal bandwidths of the 10 channels are 40 MHz and 60 MHz, respectively. The total bandwidth is 500 MHz and the carrier frequency is set as 3.9 GHz to fit the parameter setting of 5G-NR.

 

Fig. 5. Spectrum before and after the delta-sigma modulation. Spectrum of the input 10-carrier aggregated 5G-NR signal located at 3.9 GHz (red), spectrum of the output digital after the first delta-sigma modulator (blue).

Download Full Size | PDF

To further evaluate the performance of the system, we conduct experiments under different oversampling rates and obtain the relationship between the error-vector-magnitude (EVM) and the oversampling rate (OSR), as shown in Fig. 6. The value is the EVM between the transmitted and received analog 5G-NR signals, after reconstruction of the analog signal from the delta-sigma modulated signal. The dotted lines in Fig. 6 are the EVM thresholds corresponding to different QAM modulation orders [15]. When the oversampling rate is less than 7, the system cannot support QPSK modulation transmission. When the oversampling rate is larger than 8, the system can support 5G-NR signal transmission up to 64-order QAM modulation. When the oversampling rate is 9, the system can support up to 1024-order QAM modulation. When the oversampling rate is 10, 11, and 12, the system can support up to 16384 order QAM-OFDM 5G-NR signal transmission. It is clearly observed that the system transmission performance improves as the oversampling rate increases. However, a higher oversampling rate is not necessarily better. The modulation bandwidth of the system is related to the oversampling rate and sampling frequency as follows:

 

Fig. 6. Relationship between the EVM performance and the OSR.

Download Full Size | PDF

When the sampling frequency is fixed, a higher oversampling rate means a smaller bandwidth. In addition, the bandwidth is related to the transmission data rate of the system as follows:

On this basis, we further investigate the sensitivity of the system to the external system noise. Figure 7 shows the relationship between the EVM and the photocurrent with an OSR of 10. A larger photocurrent means a higher signal-to-noise ratio in the fronthaul link transmission and relatively less noise. The dotted lines in the figure are the EVM thresholds corresponding to different QAM modulation orders. When the photocurrent is larger, the EVM is smaller, and the modulation quality is higher. When the photocurrent is 0.56 mA, the transmission quality of the system is poor and the high-fidelity transmission of the QPSK-OFDM 5G-NR signal cannot be supported. When the photocurrent reaches 1 mA, the system can support the high-fidelity transmission of 256QAM-OFDM 5G-NR signals. When the photocurrent reaches 1.5 mA, the system can support 4096QAM-OFDM 5G-NR signals. When the photocurrent is greater than 2 mA, the system can support 16384-order QAM modulation. This reflects the excellent characteristics of the fronthaul system in terms of high-fidelity 5G-NR transmission. Since the delta-sigma modulation is an oversampling technique, its demodulation result depends on multiple consecutive bits with the same status. Therefore, if there are a few random transmission errors, it will not strongly affect the overall recovery result. In Fig. 8, we show the received signal constellation of the system for different orders of QAM-OFDM 5G-NR signals under the oversampling rate of 10.

 

Fig. 7. Relationship between the EVM and the photocurrent when the OSR is 10.

Download Full Size | PDF

 

Fig. 8. Constellation graphs of different orders of QAM-OFDM 5G-NR signals after fronthaul system transmission when the photocurrent is 2 mA and the OSR is 10.

Download Full Size | PDF

To conclude, we have proposed a hybrid fiber–THz mobile fronthaul system with noise suppressed delta-sigma modulation. It could support carrier aggregation of up to 10 40-MHz and 60-MHz 5G-NR channels at the radio carrier frequency of 3.9 GHz. With the design of quantization noise suppressed delta-sigma modulation, the supported QAM order is up to 16384 with the error vector magnitude (EVM) below 0.5%. When the QAM order is 1024, error-free and high-fidelity transmission can be achieved. Therefore, the proposed fiber–THz fronthaul system is a promising candidate for the future high-capacity RAN.