## Abstract

This paper describes the demonstration of an energy efficient orthogonal frequency division multiplexing passive optical network using the dynamic signal to noise ratio (SNR) management and adaptive modulation. Controlling the calculation precision and modulation format minimizes the energy consumption of digital signal processor while satisfying the requirements of bit error ratio. We show that the calculation precision and modulation format can be optimized according to the optical received power, and realize a 58.7% effective energy efficiency per bit in an FPGA-based receiver experimentally.

© 2014 Optical Society of America

## 1. Introduction

Next generation optical access networks (NGOAs) require not only high network capacity, but also various advantageous features such as flexibility, elastic bandwidth allocation, energy efficiency, and a long reach [1]. The orthogonal frequency division multiplexing passive optical network (OFDM-PON) has the potential to achieve high network capacity and flexibility [2]. It is also a promising technology for elastic bandwidth allocation. However, an OFDM-PON consumes a lot of power due to the complex digital signal processor (DSP) used for fast Fourier transform (FFT) and inverse FFT (IFFT) [3], even though the energy efficiency of optical networks has been recognized as an important factor for realizing sustainable NGOAs. On the other hand, a long-reach PON may increase the gap of transmission quality between near and distant subscribers. Today’s PON is designed based on the subscriber with the worst transmission quality. In such a situation, a near subscriber enjoys a transmission quality that is much higher than necessary. Therefore, the transmitters and receivers of NGOAs should be tuned adaptively depending on the transmission distance to improve network usage. Adaptive modulation is a promising technique for improving network utilization according to the transmission quality needed for each subscriber [4].

In this paper, we describe that the energy efficient intensity modulation-direct detection (IM-DD) OFDM-PON using a dynamic signal-to-noise ratio (SNR) management to adaptive modulation, and demonstrates the energy efficiency achieved with the dynamic SNR management experimentally. We have already proposed the dynamic SNR management, which controls the calculation precision of FFT and IFFT depending on the transmission distance to reduce the DSP power consumption [5]. This paper describes the combination of the dynamic SNR management and adaptive modulation for an energy efficient OFDM-PON. Power and energy consumption experiments confirm the feasibility of our technique for adaptive modulation in an IM-DD OFDM-PON.

## 2. Dynamic SNR management and adaptive modulation

The dynamic SNR management controls calculation precision, in other words, the number of bits used for the OFDM frame generation, to minimize the DSP power consumption while satisfying the bit error rate (BER) requirement. A previous report indicated that decreasing the calculation precision of the FFT and IFFT reduces the DSP power consumption [6], along with reduction in the resolution of a digital analog converter (D/A) and an analog digital converter (A/D) [7].

An OFDM-PON system (de-)generates OFDM frames by signal processing. This procedure includes quantization and rounding errors as a result of the limited calculation precision. Figure 1 shows the concept of the dynamic SNR management for an OFDM-PON system and the SNR degradation factors. The proposed technique controls the signal quality degradation of (1) the quantization and rounding errors of the IFFT in an optical line terminal (OLT), (2) the transmission loss in optical fiber and receiver noise, and (3) the quantization and rounding errors of the FFT in an optical network unit (ONU), for the OFDM downstream direction. Although there are several factors contributing to SNR degradation such as chromatic dispersion, we focus on the main ones, namely calculation errors and the transmission loss. The quantization and rounding errors depend on the number of bits used for the FFT and IFFT. The OLT and ONU calculate using the minimum number of bits, X_{OLT}-bit and X_{ONU}-bit, according to the transmission loss and receiver noise, to reduce power consumption.

The adaptive modulation improves both network utilization and the energy efficiency per bit of OFDM signal transmission. Modifying the constellation modulation from quadrature phase-shift keying (QPSK) to 16-quadrature amplitude modulation (16QAM) slows down the speed of the FFT and IFFT calculations by half to achieve the same transmission rate. As a result, DSP power consumption can be reduced without transmission rate degradation. The ONUs, which can select higher-order constellation modulation such as 16QAM, upgrade the constellation modulation to improve energy efficiency. The other ONUs use lower-order constellation modulation such as QPSK for long-distance transmission. In this paper, we clarify the energy efficiency of adaptive modulation by changing QPSK to 16QAM. Note that the other constellation modulations such as 32-quadrature amplitude modulation (32QAM) or higher-order constellation modulations could be applied while satisfying the BER requirements for very short-distance ONUs. The combination of the dynamic SNR management and adaptive modulation has the potential to greatly improve the energy efficiency per bit while satisfying the BER requirement.

In Fig. 1, the dynamic SNR management defines the calculation precision according to the transmission distance when using the same constellation modulation. If the transmission distance is short, the permissible FFT and IFFT calculation errors become large. Thus, an ONU (b) demodulates an OFDM signal with a small number of bits to reduce power consumption below that of (c) for downstream. In addition, upgrading the constellation modulation from QPSK to 16QAM improves the energy efficiency for a short-distance ONU, as shown in (a).

## 3. Experimental results of signal transmission

We evaluated the BER characteristics of the dynamic SNR management with adaptive modulation experimentally. Figure 2 shows the experimental settings. We evaluated the dynamic SNR management in a 10-Gbit/s IM-DD OFDM downstream environment. We chose an IM-DD, because it is suitable for an OFDM-PON due to its cost effectiveness and simple analog front-end configuration.

On the OLT side, a generated pseudo-random bit sequence (PRBS) maps the multi-level constellation modulation of QPSK or 16QAM with a gray code format for an OFDM frame. Then the subcarrier mapping function forms a Hermitian symmetric input vector of the IFFT block to generate a real-value OFDM signal for IM-DD transmission. The X_{OLT}-bit IFFT creates an OFDM signal, which contains the quantization and rounding errors of the IFFT calculation. The IFFT matrix size was set at 256 to transmit OFDM signal using 96 (<256/2) subcarriers. The OFDM transmitter generates an OFDM frame digitally by adding a cyclic prefix (CP) and a preamble. Then, an arbitrary waveform generator (AWG) behaves as a D/A with a sampling rate of 10 GSample/s at QPSK, and 5 GSample/s at 16QAM to achieve 10-Gbit/s with a Hermitian overhead. The OFDM frame is transmitted by using a 1556.23 nm narrow-linewidth DFB-LD, which spectral width is 69 kHz, and an LN intensity modulator.

On the ONU side, the APD-TIA receives an OFDM baseband signal via a 20 km SMF. The oscilloscope, which acts as an A/D, forms digital data. The data are oversampled at 50 GSample/s. The DSP performs frequency domain equalization and down-sampling, and then demodulates the received OFDM frame. The X_{ONU}-bit FFT demodulates the OFDM signal with quantization and rounding errors. We varied X_{OLT}-bit and X_{ONU}-bit from 8 to 16, and set both with the same calculation precision.

The left and right sides of Fig. 3 show the BER characteristics and constellations with a −18 dBm optical received power, respectively, obtained in a 20 km IM-DD OFDM transmission experiment. The results allow us to draw the following conclusions.

- (1) A calculation with a large number of bits improves the BER for the same optical received power. The constellation becomes satisfactory even for the same optical received power by increasing the number of bits used for the FFT and IFFT calculation.
- (2) 8 bit FFT and IFFT calculations with 16QAM could not achieve the forward error correction (FEC) limit, which is a BER of 3.8x10
^{−3}for hard decision FEC with a 7% overhead [8], because the dominant characteristics of this transmission are the quantization and rounding errors caused by FFT and IFFT calculation. - (3) In particular, the increase in the number of bits significantly enhances the BER in a domain with a small number of bits. On the other hand, increasing the number of bits improves the BER characteristics slightly in a domain with a large number of bits, e.g. more than 10 bits with QPSK and 12 bits with 16QAM.
- (4) Several bit number and optical received power combinations satisfy the BER requirement. For example, a ≥10 bit calculation with 16QAM at −17 dBm optical received power, a ≥12 bit calculation with 16QAM at −20 dBm, and a 16 bit calculation with 16QAM at −21 dBm meet the FEC limit. These transmission results suggest that the calculation precision can be reduced for an ONU with a high optical received power, in other words, an ONU located at a short distance from the OLT.
- (5) Decreasing the number of bits for FFT and IFFT calculation degrades the BER with both 16QAM and QPSK. The dynamic SNR management can be applied for some multi-level modulation.

The experimental results show that the calculation precision can be reduced for a user of high optical received power, in other words, for short distance transmission. For example, we can reduce the number of bits for FFT and IFFT calculation from 16 bits to 10 bits by using 4 dB optical power budgets, which is corresponding to the 16 km in a 0.25 dB/km transmission loss environment, while satisfying the FEC limit in 16QAM. The results confirmed that the dynamic SNR management and adaptive modulation functions successfully IM-DD OFDM transmission.

## 4. Experimental results for energy efficiency

We designed and implemented a field-programmable gate array (FPGA)-based OFDM transmitter and receiver to evaluate power consumption. Figures 4(a) and 4(b) shows the transmitter and receiver, respectively. We implemented each function on Xilinx Virtex-7 FPGAs, an A/D and a D/A. We designed that the FPGA, whose power consumption is monitored periodically, can be reconfigured the number of bits for calculation between 8 and 34 bits. Figure 5 shows the measured power consumption of the 10 Gbit/s transmitter and receiver. The size of look up tables (LUTs) used for logic and memory rises according to the number of bits increases. The power consumption also increases against the number of bits. The QPSK has more power consumption than 16QAM because the FPGA for QPSK runs at twice the speed of the FPGA for 16QAM. In addition, the transmitter and receiver with ≥26 bit consume a large power consumption due to the limitation of the FPGA layout.

Figure 6 shows the optimized energy consumption per bit of the FPGA of the receiver used to demodulate the OFDM frame against the transmission distance to meet the FEC limit with QPSK and 16QAM. We defined + 6 dBm optical output power, a 0.25 dB/km transmission loss and a split ratio of 1x64. The ONU located at long distance consumes a large amount of power, because of its high calculation precision. The experimental results indicated that the dynamic SNR management reduces the power consumption of the receiver located at a short-distance by 23.7% in QPSK and 17.5% in 16QAM compared with 16 bit calculation, which is needed for the worst transmission quality. Furthermore, upgrading the constellation modulation from QPSK to 16QAM significantly improves the energy consumption per bit. At less than 9 km, the energy efficiency of the ONU increases 58.7% when using both the dynamic SNR management and adaptive modulation.

The previous discussion focused on the energy reduction of a pair of OLT and ONU [9]. As following, we evaluate the energy reduction of OFDM-PON system which has multiple ONUs, by assuming several user distributions. Figure 7(a) shows the relationship between subscriber distributions and transmission distance between an OLT and ONUs. Four types configurations are modeled in this paper. It is indicated that the distance between OLT and ONUs fits Rayleigh distribution in [10]; therefore, we chose three types of Rayleigh distribution (σ = 3, 5, and 10) with peak subscriber densities of 4 km, 8 km, and 18 km, respectively. The constant distribution represents that the subscriber density does not depend on the transmission distance between OLT and ONUs. Figure 7(b) shows the relative energy consumption using dynamic SNR management and adaptive modulation for Fig. 7(a) distributions. A 16 bit FFT and IFFT calculation with QPSK is defined as the normalized energy consumption. The dynamic SNR management and its combination with adaptive modulation reduce energy consumption by 20.6% and 57.2%, respectively, in Rayleigh distribution (σ = 3) because a large number of short distance subscribers leads to a condition with minimum energy consumption.

## 5. Conclusions

We expanded the dynamic SNR management to adaptive modulation, and experimentally clarified the energy efficiency of the dynamic SNR management. The dynamic SNR management with adaptive modulation improves the energy efficiency by controlling the calculation precision of the FFT and IFFT and the multi-level constellation modulation. We showed that the dynamic SNR management with adaptive modulation improves the energy efficiency by 58.7% for short-distance ONUs. Furthermore, the energy consumption was reduced by 23.7% in QPSK and 17.5% in 16QAM, which is simply the effect of the dynamic SNR management. The proposed dynamic SNR management with adaptive modulation effectively reduces energy consumption for short distance subscribers, and realizes energy efficient IM-DD OFDM-PON.

## References and links

**1. **K. Kitayama, Y. Yoshida, and A. Maruta, “Green, elastic coherent IFDMA-PON for next-generation access network,” in *Proceedings of 14th International Conference on Transparent Optical Networks*, paper Tu.B3.1 (2012).

**2. **N. Cvijetic, “OFDM for Next-Generation Optical Access Networks,” J. Lightwave Technol. **30**(4), 384–398 (2012). [CrossRef]

**3. **K. Kanonakis and I. Tomkos, “Energy-Efficient OFDMA-PON Exploiting Modular OLT/ONU Digital Signal Processing,” in *Proceedings of OFC 2013*, paper OTh3A.4 (2013).

**4. **N. Iiyama, S.-Y. Kim, T. Shimada, S. Kimura, and N. Yoshimoto, “Co-existent Downstream Scheme between OOK and QAM Signals in an Optical Access Network using Software-defined Technology,” in *Proceedings of OFC 2012*, paper JTh2A.53 (2012).

**5. **H. Kimura, H. Nakamura, S. Kimura, and N. Yoshimoto, “Numerical Analysis of Dynamic SNR Management by Controlling DSP Calculation Precision for Energy-Efficient OFDM-PON,” IEEE Photon. Technol. Lett. **24**(23), 2132–2135 (2012). [CrossRef]

**6. **R. Bouziane, P. A. Milder, R. J. Koutsoyannis, Y. Benlachtar, J. C. Hoe, M. Glick, and R. I. Killey, “Dependence of Optical OFDM Transceiver ASIC Complexity on FFT Size,” in *Proceedings of OFC 2012*, paper JW2A.58 (2012).

**7. **H. Takahashi, A. A. Amin, I. Morita, and H. Tanaka, “Required Resolution of Digital-Analog-Converter for Optical OFDM,” in *Proceedings of OFC 2010*, paper JThA4, (2010).

**8. **ITU-T Recommendation G.975.1, Appendix I.9 (2004).

**9. **H. Kimura, K. Asaka, H. Nakamura, S. Kimura, and N. Yoshimoto, “First Demonstration of Energy Efficiency of Dynamic SNR Management for Adaptive Modulation in IM-DD OFDM-PON,” in *Proceedings of ECOC 2013*, paper We.4.F.5 (2013).

**10. **F. Vacondio, O. Bertran-Pardo, Y. Pointurier, J. Fickers, A. Ghazisaeidi, G. de Valicourt, J.-C. Antona, P. Chanclou, and S. Bigo, “Flexible TDMA access optical networks enabled by burst-mode software defined coherent transponders,” in *Proceedings of ECOC 2013*, paper We.1.F.2 (2013).