A differential pulse coding modulation with noise shaping (NS-DPCM) is proposed to achieve better error vector magnitude (EVM) performance in a digital mobile fronthaul. Compared to previous DPCM based digital mobile fronthaul, a feedback loop combined with a finite-impulse-response (FIR) filter is added to the quantizer of DPCM encoder to operate as a quantization noise shaping technique block. The noise shaping technique increases the signal-to-quantization noise ratio by reshaping the spectrum of the quantization noise. Therefore, the noise power is at a lower level in subcarriers of OFDM signal where data is modulated and at a higher level in the subcarriers where no data is modulated. Different from the noise shaping of delta-sigma modulation, the proposed scheme utilizes the existence of unused subcarriers of OFDM signal, thus does not require oversampling at the transmitter and low pass filter at the receiver. In the experiment, the proposed NS-DPCM based mobile fronthaul transmission is demonstrated in a 25-Gb/s PAM-4 intensity modulation-direct detection optical link. Compared to the existing DPCM based mobile fronthaul, significant EVM performance improvement is achieved using the same number of quantization bits.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
As an attractive wireless network technology, the cloud radio access network (C-RAN) and mobile fronthaul (MFH) offer improved energy efficiency and cost-effectiveness for the 5G mobile network with densified small cells . Meanwhile, great challenges have also been incurred in MFH, such as demands of high capacity, low latency, and low cost . To resolve these issues, both analog [3–6] and digital [7–15] MFH technologies have been widely investigated. For analog MFH technology, mobile signals after baseband processing are firstly aggregated using frequency division multiplexing (FDM) scheme [3, 4], equalized time-division multiple access (TDMA)  or code division multiplexing (CDM) scheme  in the baseband units (BBUs) pool. Then, the aggregated signals are converted to analog signals by digital to analog converter (DAC) in BBU pool, and directly transmitted to remote radio headers (RRHs) via optical links. Compared to common-public-radio-interface (CPRI) technology , 15-digit quantization for I/Q components of the mobile signals is not required. Thus, analog MFH technology has the merit of high bandwidth efficiency. For instance, 59-Gb/s CPRI equivalent data rate could be supported by only 1.5-GHz bandwidth . However, analog MFH technology is seriously affected by the nonlinear impairments in the channel .
To circumvent the nonlinear distortion issues, many digital MFH architectures have been proposed. As a matured technique, CPRI based digital MFH digitized I/Q components of the mobile signals with 15 quantization bits’ pulse coding modulation (PCM), which provided excellent immunity to nonlinear distortions of the channel impairments . However, it suffers from extremely low bandwidth efficiency. Function split is an effective technique to greatly reduce the data traffic over digital MFH [16, 17]. However, some other issues such as incompatible with multiple-input multiple-output (MIMO), coordinated multi-point (CoMP) transmission, and 5G heterogeneous network are existed . On the other hand, advanced modulation formats such as 4-level pulse amplitude modulation (PAM-4), combined with quantization noise reduction techniques show remarkable benefits such as high bandwidth efficiency and perfect robustness against to the nonlinear distortions [7–14]. Among these techniques, differential pulse coding modulation (DPCM) effectively reduces the overall quantization noise of baseband orthogonal frequency division multiplexed (OFDM) signals [12, 13], and the quantization noise is evenly distributed in all subcarriers of the frequency domain. Since some subcarriers of the OFDM signals are not modulated with data, the quantization noise reduction in these subcarriers does not help to increase the signal-to-noise ratio (SNR) of the digitized signal. Thus, quantization noise shaping that change the frequency domain distribution of noise power is motivated. Delta sigma modulation based MFH [7–9] exploits a noise-shaping technique to push the quantization noise out of the data carrying subcarriers, but high order oversampling ratio is required. The noise shaping scheme without oversampling are also studied in the fields of audio requantization [18, 19], coding of color television signals  and direct-detection of discrete multi-tone . But no applications have been investigated in MFH, to the best of our knowledge.
In this paper, a DPCM with noise shaping (NS-DPCM) is firstly proposed to achieve error vector magnitude (EVM) performance enhancement of digital MFH. Compared to the previous DPCM , a feedback loop combined with a well-designed finite-impulse-response (FIR) filter is added in the quantizer of DPCM encoder to operate as a quantization noise shaping technique block. By reshaping the spectrum of quantization noise, the noise power is suppressed in the data carrying subcarriers of OFDM signals and enlarged in the unused subcarriers, which reduces the data distortion caused by quantization noise. Different from the noise shaping of delta-sigma modulation [7–9], the proposed scheme utilizes the existence of unused subcarriers of baseband OFDM signal, thus does not require oversampling at transmitter and low pass filter at the receiver. Digital MFH employing NS-DPCM is experimentally demonstrated over an intensity-modulation-direct detection (IM-DD) 25-Gb/s PAM-4 optical links. Significant EVM performance improvement can be obtained under the same number of quantization bits.
Figure 1(a) shows the architecture of digital MFH employing proposed NS-DPCM in PAM-4 optical links. In BBU pool, mobile signals after baseband processing are firstly encoded by the proposed NS-DPCM encoder module. The encoded digitized signals are aggregated through time domain multiplexing (TDM)  before transformed as PAM-4 signal and modulated to the optical domain. After transmission through fiber link, the received signal at RRH is firstly converted to the electronic domain and passing through a PAM-4 receiver. Then, the separated frames are obtained by TDM de-aggregation. After that, baseband signals are retrieved via conventional DPCM decoder and sent to the ratio frequency (RF) front end. The structure of the proposed NS-DPCM encoder is shown in Fig. 1(b). Compared to the previous DPCM with linear predictor , a feedback loop combined with a finite-impulse-response (FIR) filter is added in the quantizer of DPCM encoder to operate as a quantization noise shaping technique block. The input signal to the quantizer consists of two parts, one is the difference signal between the original signal and output of the linear predictor , the other is the filtered quantization noise . Then, the signal is quantized to get the output of the encoder as , and added to to update the input to predictor as . Meanwhile, the quantization error between the quantized value and the original value is measured as , and fed into the finite-impulse-response (FIR) filter whose discrete-time transfer function is denoted as . After that, the filtered quantization error is fed back and added into the next sample prior to its quantization. The FIR filter can be expressed as:Fig. 1(c). Thus, the Z transform of can be expressed as , where is the Z transform of . Then, we can calculate the Z transform of , denoted as :
To demonstrate the merits of changing the spectrum of quantization noise, we give some intuitive illustrations as shown in Fig. 2(a). For the baseband OFDM signal before encoding by DPCM or NS-DPCM, most subcarriers are used for data-carrying, which are located within in frequency domain. Here, denotes the bandwidth of data, denotes the sampling rate, and is satisfied. The other subcarriers are not used which are denoted as null subcarriers. When DPCM is employed, the overall quantization noise of OFDM signals is effectively reduced when comparing with PCM [12, 13], and the noise is evenly distributed in all subcarriers in the frequency domain. After OFDM demodulation, the data signals are affected by the evenly distributed quantization noise. NS-DPCM, on the other hand, reshapes the spectrum of the quantization noise in DPCM. Therefore, the noise power is at a lower level in data-carrying subcarriers of OFDM signal and at a correspondingly higher level in the null subcarriers. Therefore, under the same total noise power, the distortion of data signals affected by quantization noise will be smaller than that of DPCM after OFDM demodulation.
Different from the noise shaping of delta-sigma modulation [7–9], the proposed noise shaping in NS-DPCM utilizes the existence of unused subcarriers of baseband OFDM signal and well-designed FIR filter, thus does not require additional oversampling at the transmitter and low pass filter at the receiver. The design of is to choose the filter’s coefficients properly so that the output noise is as small as possible in the data-carrying subcarriers. Since the power spectrum of the quantization noise is constant under given quantization bits , increasing attenuation in the data-carrying subcarriers will unavoidably increase gain in unused subcarriers. We introduce a weighting function, whose Fourier transform is denoted as , to specify the noise power distribution at the different frequencies. In other words, is designed to minimize the total amount of weighted noise power as follows:Eq. (1) into Eq. (3), the objective function can be written as:Eq. (4) to time domain as:22] to solve the Eq. (6). Figure 2(b) shows the magnitude frequency responses of the introduced weighting function and the corresponding noise shaping function with 1-tap and 5-tap, respectively. The normalized frequency bandwidth of data is set as , which satisfies the configuration of typical baseband OFDM signals in MFH.
When applying carrier aggregation , NS-DPCM can also be employed to digitize the aggregated mobile signals instead of the baseband OFDM signals. In the practical setting of carrier aggregation, the unused band out of the aggregated signal band is always necessary because of a non-zero-width transition band in any anti-aliasing filter . Thus we can use the unused band to reduce the quantization noise, and the enhanced noise exists in the unused band instead of the guard band between component carriers.
3. Experimental setup
Figure 3 shows the experimental setup for the transmission of 25-Gb/s PAM-4 based digital MFH by employing proposed NS-DPCM. The previous DPCM scheme is also experimentally demonstrated for comparison. In the transmitter, OFDM is utilized to generate the baseband mobile signals with 30.72-MHz sampling rate, whose spectrum is shown in Fig. 3(a). Here, 2048-IFFT/FFT points are employed, the length of cyclic prefix (CP) 32, and the number of physical resource blocks (PRBs) is 100. The QAM orders vary among 4, 16, 64, 256, 1024, and 4096. The number of subcarriers is 1200 with 15.26-kHz subcarrier spacing. Then the in-phase and quadrature (I/Q) parts of OFDM signals are separately digitalized using DPCM/NS-DPCM encoder with 4-tap linear predictor. Each pair of digitized I/Q samples combined with 1/16 overhead compose one antenna carrier (AxC) container frame. After that, different AxC container frames are multiplexed in the time domain and encoded by 12.5-GBaud PAM-4 transmitter. The number of aggregated AxC containers is adaptively changed according to the quantization bits used in DPCM/NS-DPCM. For instance, with the number of quantization bits equal to 6, 25-Gb/s PAM-4 can support up to 50 (≈25-Gb/s / (30.72-MSa/s × 6-bit × 2 × (16/15) × (10/8))) AxC containers. Thus, 50 MFH links are aggregated for the experiment in this case. Reed-Solomon forward error correction coding (RS-FEC 528/514), just as used in CPRI , is employed to guarantee error-free transmission within the PAM-4 transmitter. A Keysight M8195A arbitrary waveform generator (AWG) is employed to generate the electronic PAM-4 modulated waveforms, which pass through a single-channel IM-DD system, including a 10GHz direct modulate laser (DML) operated at 1550nm, 20-km standard single-mode fiber (SSMF), a variable optical attenuator (VOA) and a commercial 10-GHz photo detector (PD). The detected electrical signals are sampled by a digital oscilloscope (OSC) and then processed by PAM-4 receiver including digital equalization  and FEC decoding. Then, TDM based de-aggregation is applied and the separated frames are passed to the DPCM decoder. Figure 3(b) shows the spectrum of the received digitized OFDM signals using NS-DPCM. Finally, OFDM demodulation and EVM performance evaluation of different schemes are conducted. The complementary cumulative distribution function (CCDF) of peak-to-average-power ratio (PAPR) for OFDM signals with various modulation formats are shown in Fig. 3(c).
4. Results and discussion
We firstly measure the quantization noise power spectrum of the received digitized OFDM signals using NS-DPCM with 5-tap noise shaping FIR filter and DPCM after 20-km SSMF transmission, respectively. The power spectrum of original OFDM signal is also calculated as reference. The QAM order of the OFDM signal is set to 64 and the received optical power (ROP) is −8dBm, where no transmission error incurred. As shown in Fig. 4, the quantization noise power in data-carrying subcarriers can be effectively reduced by using NS-DPCM. Specifically, our proposed scheme brings around 4.2-dB mean quantization noise power reduction in data-carrying subcarriers compared to the previous DPCM scheme, which could improve the EVM performance of recovered data.
Furthermore, we measure the EVM of each data-carrying subcarrier using DPCM and NS-DPCM with 1-tap, 3-tap, 5-tap, and 7-tap noise shaping FIR filter after 20-km SSMF transmission at ROP of −8dBm, respectively. As shown in Fig. 5, the EVMs of recovered data using DPCM are evenly distributed in all 1200 subcarriers, and the average EVM is 3.79%. By using 1-tap and 3-tap NS-DPCM, the average EVM is reduced to 3.09%, and 2.22%, respectively. However, they suffer from unevenly distributed noise power and EVM penalty at high-frequency subcarriers. To improve the performance of high-frequency subcarriers, pre-emphasis is required to increase the signal power of these subcarriers . However, pre-emphasis cannot completely eliminate the unequal noise floor  and will greatly increase the computational complexity due to the use of FFT/IFFT. When we increase the number of noise shaping FIR filter’s tap to 5, not only the average EVM is reduced to 2.19%, but also the quantization noise is much more evenly distributed across all subcarriers. Such improvements become less significant when the FIR filter’s tap increases to 7. Considering the cost of digital signal processing (DSP), 5-tap noise shaping FIR filter is chosen for NS-DPCM using in the rest of the experiments. Compared to CPRI, some additional processing latency due to the use of multi-tap FIR filters and feedback loops in DPCM/NS-DPCM is introduced. Considering the number of the filter’s tap is only 4 in linear predictor, the total increased latency can be estimated around 12 clock periods (4 clock periods due to delay unit, 1 clock period due to the multiplier, and 1 clock period due to the adder in both modulation and demodulation module). When the DSP clock rate is 30.72-MSa/s, the increased latency of DPCM is estimated as ~0.4 μs (~1/30.72-MSa/s × 12). By using NS-DPCM, another 5-tap FIR filter is introduced in the modulation module. Thus, the estimated increased latency of NS-DPCM is ~0.62 μs, which meets the low latency constraint of MFH.
Then, we compare the measured EVM at various quantization bits (QBs) by using DPCM and NS-DPCM after 20-km SSMF transmission at ROP = −8dBm as shown in Fig. 6. The modulation format is chosen according to the EVM thresholds specified by 3GPP (3.5% for 256QAM, 8% for 64QAM, 12.5% for 16QAM, 17.5% for QPSK) and DOCSIS 3.1 (1.68% for 1024QAM, 0.7% for 4096QAM). Besides this, a large performance margin is left for dynamic-range tolerance of RF signal delivery in a real system . Thus, for instance, 6-bit quantization is used for 256QAM. The same modulation format is employed for DPCM and NS-DPCM under the same quantization bit, which is depicted on the upper right of the dotted oval in Fig. 6. We can observe that the EVMs for both two schemes are significantly reduced with the increase of the number of QBs. Compared to DPCM, the EVMs at various quantization bits are all improved by using NS-DPCM. For instance, when QB = 6, EVM is reduced from 2.46% with DPCM to 1.4% with NS-DPCM.
The required number of QBs, supported number of antenna-carrier (AxC) containers and optical bandwidth efficiency improvement compared to CPRI using NS-DPCM based 25-Gb/s PAM-4 MFH for various modulation formats are summarized in Table 1. For instance, with the number of QBs equal to 6, 25-Gb/s PAM-4 can support up to 50 (≈25-Gb/s / (30.72-MSa/s × 6-bit × 2 × (16/15) × (10/8))) AxC channels of OFDM-256QAM signals. Compared to CPRI, the reduction of required QBs from 15-bit to 6-bit and PAM-4 modulation can improve the optical bandwidth efficiency by factors of 15/6, and 2, respectively. Therefore, the optical bandwidth efficiency can be improved by 5 times in total.
The measured (bit-error-rate) BER versus received optical power of the PAM-4 signals over optical back-to-back (B2B) and 20-km fiber transmission before FEC decoding are shown in Fig. 7. To obtain BERs below the RS-FEC 528/514 threshold (), the required received optical power is higher than −14.5 dBm at B2B. After 20-km SSMF transmission, less than 2 dBm degradation of received optical power requirement could be observed. The selected eye diagrams of the received PAM-4 signals are also shown in Figs. 7(b) and 7(c). With FEC decoding, error free can be guaranteed when BER is below the threshold of .
Under error-free condition, some representative constellations of recovered 5G-New Ratio (NR)-like wireless signals with 64QAM, 256QAM, and 1024QAM are shown in Fig. 8. As shown in Figs. 8(d)-8(f), by employing proposed NS-DPCM based digital MFH, 2.19%, 1.4%, and 0.55% of EVM values are obtained, respectively. Compared to the results using previous DPCM based digital MFH as shown in Figs. 8(a)-8(c), significant improvement up to 40% () can be observed, which shows the feasibility of the NS-DPCM based digital MFH solution for future 5G C-RAN.
A digital MFH solution employing NS-DPCM has been proposed for future 5G C-RAN. Compared to the previous DPCM based MFH, quantization noise power reduction in data-carrying subcarriers is obtained by utilizing the existence of unused subcarriers of baseband OFDM signal and well-designed noise shaping FIR filter, which enables EVM performance enhancement of the digital MFH. The proposed scheme has been verified in a single channel 25-Gb/s PAM-4 IM-DD optical links, and more than 40% EVM performance improvement could be achieved using the same number of quantization bits.
National Natural Science Foundation of China (NSFC) (61701359), 2016 key projects of Natural Science Foundation of Hubei Province (2016AAA012), 2017 Wuhan Basic Applied Research Project (2017010201010101, 2017010201010100), Natural Science Foundation of Hubei Province Grant (2016CFB032).
This work is supported by the National Natural Science Foundation of China (61701359), Key Projects of Natural Science Foundation of Hubei Province (2016AAA012), Wuhan Basic Applied Research Project (2017010201010101, 2017010201010100), Natural Science Foundation of Hubei Province Grant (2016CFB032).
References and links
1. NGMN, “Next Generation Mobile Networks 5G White Paper,” whitepaper v. 1.0 (2015).
2. A. Pizzinat, P. Chanclou, T. Diallo, and F. Saliou, “Things you should know about fronthaul,” inProceedings of European Conference on Optical Communications (2014), paper Tu.4.2.1.
3. X. Liu, N. Chand, F. Effenberger, L. Zhou, and H. Lin, “Demonstration of bandwidth-efficient mobile fronthaul enabling seamless aggregation of 36 E-UTRA-like wireless signals in a single 1.1-GHz wavelength channel,” inProceedings of Optical Fiber Communication Conference (2015), paper M2J.2. [CrossRef]
4. X. Liu, H. Zeng, N. Chand, and F. Effenberger, “Experimental demonstration of high-throughput low-latency mobile fronthaul supporting 48 20-MHz LTE signals with 59-Gb/s CPRI-equivalent rate and 2-µs processing latency,” inProceedings of European Conference on Optical Communications (2015), paper We.4.4.3. [CrossRef]
5. X. Liu, H. Zeng, N. Chand, and F. Effenberger, “CPRI-compatible efficient mobile fronthaul transmission via equalized TDMA achieving 256 Gb/s CPRI-equivalent data rate in a single 10-GHz-bandwidth IM-DD channel,” inProceedings of Optical Fiber Communication Conference (2016), paper M1H.3. [CrossRef]
6. H. Li, Q. Yang, M. Luo, R. Hu, P. Jiang, Y. Liu, X. Li, and S. Yu, “Demonstration of bandwidth efficient and low-complexity mobile fronthaul architecture via CDM-based digital channel aggregation.” inProceedings of Optical Fiber Communication Conference (2017), paper Th3A.5. [CrossRef]
7. J. Wang, Z. Yu, K. Ying, J. Zhang, F. Lu, M. Xu, and G. K. Chang, “Delta-sigma modulation for digital mobile fronthaul enabling carrier aggregation of 32 4G-LTE / 30 5G-FBMC signals in a single-λ 10-Gb/s IM-DD channel,” inProceedings of Optical Fiber Communication Conference (2016), paper M1H.2. [CrossRef]
8. J. Wang, Z. Yu, K. Ying, J. Zhang, F. Lu, M. Xu, L. Cheng, X. Ma, and G. K. Chang, “Digital Mobile Fronthaul Based on Delta–Sigma Modulation for 32 LTE Carrier Aggregation and FBMC Signals,” J. Opt. Commun. Netw. 9(2), A233–A244 (2017). [CrossRef]
9. H. Li, R. Hu, Q. Yang, M. Luo, Z. He, P. Jiang, Y. Liu, X. Li, and S. Yu, “Improving performance of mobile fronthaul architecture employing high order delta-sigma modulator with PAM-4 format,” Opt. Express 25(1), 1–9 (2017). [CrossRef] [PubMed]
10. S. Kim, H. Chung, and S. Kim, “Experimental demonstration of CPRI data compression based on partial bit sampling for mobile front-haul link in C-RAN,” inProceedings of Optical Fiber Communication Conference (2016), paper W1H.5. [CrossRef]
11. M. Xu, F. Lu, J. Wang, L. Cheng, D. Guidotti, and G. K. Chang, “Key Technologies for Next-Generation Digital RoF Mobile Fronthaul with Statistical Data Compression and Multiband Modulation,” J. Lightwave Technol. 35(17), 3671–3679 (2017). [CrossRef]
12. L. Zhang, X. Pang, O. Ozolins, A. Udalcovs, R. Schatz, U. Westergren, G. Jacobsen, S. Popov, S. Xiao, and J. Chen, “15-Gbaud PAM4 Digital Mobile Fronthaul with Enhanced Differential Pulse Coding Modulation supporting 122 LTE-A Channels with up to 4096QAM,” inProceedings of European Conference on Optical Communications(2017), paper W.2.B.6.
13. L. Zhang, X. Pang, O. Ozolins, A. Udalcovs, R. Schatz, U. Westergren, G. Jacobsen, S. Popov, L. Wosinska, S. Xiao, W. Hu, and J. Chen, “Digital mobile fronthaul employing differential pulse code modulation with suppressed quantization noise,” Opt. Express 25(25), 31921–31936 (2017). [CrossRef] [PubMed]
14. H. Xin, K. Zhang, H. He, W. Hu, and M. Zhang, “Fidelity enhancement in high-data-rate digital mobile fronthaul with sample bits interleaving and unequally-spaced PAM4,” Opt. Express 25(5), 5559–5570 (2017). [CrossRef] [PubMed]
15. C. P. R. I. Specification, V7.0, “Common Public Radio Interface (CPRI); Interface Specification,” (2015).
16. T. Pfeiffer, “Next generation mobile fronthaul and midhaul architectures [Invited],” J. Opt. Commun. Netw. 7(11), B38–B45 (2015). [CrossRef]
17. J. Terada, T. Shimada, T. Shimizu, and A. Otaka, “Optical network technologies for wireless communication network,” inProceedings of European Conference on Optical Communications(2016), paper Tu.1.F.2.
18. W. Verhelst and D. D. Koning, “Noise shaping filter design for minimally audible signal requantization,” IEEE Workshop on the Applications of Signal Processing to Audio and Acoustics,147–150 (2001). [CrossRef]
19. D. D. Koning and W. Verhelst, “On psychoacoustic noise shaping for audio requantization,” in Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing, 413–416 (2003).
20. B. Girod, H. Almer, L. Bengtsson, B. Christensson, and P. Weiss, “A Subjective Evaluation of Noise-Shaping Quantization for Adaptive Intra-/Interframe DPCM Coding of Color Television Signals,” IEEE Trans. Commun. 36(3), 332–346 (1988). [CrossRef]
21. W. A. Ling, “Shaping Quantization Noise and Clipping Distortion in Direct-Detection Discrete Multitone,” J. Lightwave Technol. 32(9), 1750–1758 (2014). [CrossRef]
22. S. M. Kay, Modern Spectral Estimation: Theory and Application (Prentice-Hall, 1988).
23. X. Li, S. Zhou, H. Ji, M. Luo, Q. Yang, L. Yi, R. Hu, C. Li, S. Fu, A. Alphones, W. Zhong, and C. Yu, “Transmission of 4×28-Gb/s PAM-4 over 160-km Single Mode Fiber using 10G-Class DML and Photodiode,” inProceedings of Optical Fiber Communication Conference(2016), paper W1A.5. [CrossRef]