Open Access
Add to BibSonomyAbstract
We demonstrated ultra-wide input power range receivers for noise-loaded WDM applications without using optical variable attenuators. Required OSNR at 6E-5 BER exhibited less than 1dB variation across receiver input power range from -22dBm to +4Bm.
©2008 Optical Society of America
1. Introduction
Avalanche photodiode (APD) receivers are widely used in today’s WDM transmission systems. In comparison to a pre-amplified receiver where an optical amplifier is used in conjunction with a low sensitivity p-i-n receiver, an APD receiver has the advantages of much lower cost, simpler implementation, and compactness. State-of-the-art APD receivers in amplified noise-loaded WDM transmission links can only accommodate very narrow range of input power levels, e.g. from -22dBm to -10dBm, limited by detector chip avalanche breakdowns and sensitivity as well as saturation effect in trans-impedance amplifiers [1]. To accommodate wider range of power levels in noise-loaded amplified links, variable optical attenuators are often used in front of receivers to adjust input power to within receiver operating range. Variable optical attenuators add additional complexity to the receiver in terms of packaging and control. In addition, variable optical attenuators also add to the cost of receivers. In this paper, we report ultra-wide input dynamic range optical receivers capable of operating up to +4dBm power level with sensitivity better than -27.5dBm at 10.7Gb/s. The required optical signal-to-noise ratio (OSNR) at 6E-5 BER level for both BB and over-fiber transmission exhibited less than 1dB variation across receiver input power range from - 22dBm to +4dBm. The wide dynamic range operation of the receivers was achieved by high sensitivity/high breakdown threshold APD chip, wide range linear TIA, as well as adaptive APD bias control.
2. Principle of operation
The operating dynamic range of APD receivers is limited by their signal-to-noise ratios as a function of input powers. The noise sources for APD receivers consist of shot noise, multiplication noise, and thermal noise. The multiplication noise dominates and it is a function of the cascaded impact ionization process itself, which relates to the statistical correlation between the random gain and avalanche buildup time in the presence of dead-space effect [2, 3]. At very low input power levels (<-26dBm), the multiplication factor is chosen to be around 10 where the signal-to-noise ratio (SNR) is optimized amid the trade-off between gain and excess noise factor. APD chips with thinner multiplication layers are desirable for low power operations because thinner multiplication layer translates into lower noise and larger bandwidth. At very high input power levels (>0dBm), the multiplication factor need to be small to minimize the gain as well as the excess noise factor where large amount of carriers collide within minimal dead-spaces. In fact, APD chip needs to have thicker multiplication layers to minimize collision noise and maximize avalanche breakdown damage threshold at high power operations. To enable APD operation at wide dynamic input range from very low power to very high power, trade-offs need to be made in the design of the multiplication layer in terms of layer thickness and electron-hole ionization ratios. Based on above mentioned considerations, our APD chip was designed with 280nm thick InP multiplication layer with double InGaAs absorption layers. Figure 1 shows the simulated M-factor and SNR for our APD chip as a function of input power. The simulation was based on the carrier multiplication model incorporating the effects of dead-space proposed by Saleh et al. [2]. We see that reasonably flat SNR can be achieved for our APD chips across wide input dynamic range with M factors optimized for each power level. The optimization of M-factor for each power level is the critical enabling factor in realizing APD’s wide dynamic range. The M-factor tuning as a function of APD input power can be realized by adaptive APD bias control and the details are given in the following sections. Nakanishi et al. [4] recently proposed the concept of M-factor switching in an effort to improve the dynamic range of APD burst receivers, but their concept and implementation was limited to discrete switching only. Our concept and implementation is based on adaptive control of M-factor where M-factor is adaptively controlled as a function of input power levels to the receiver.
Besides the APD chip capable of operating at wide dynamic input ranges, the trans-impedance amplifier (TIA) that amplifies the APD photocurrent signal also needs to have large dynamic range. Linear TIAs with minimal distortions are desirable in most noise loaded WDM receiver applications. The linearity range for a TIA scales inversely with TIA gain. Wide dynamic range TIA is therefore a trade-off between TIA sensitivity or TIA gain and saturation current. Advances in GaAs technology have made it possible to have linear TIAs with very good sensitivity as well as very high saturation current. For this study, we used a GaAs based TIA with differential trans-impedance gain of 750 Ohms and saturation current greater than 3mA average.
3. Experimental setup
The receiver under test was in a standard MSA-compliant ROSA package (multi-suppliers-agreement receiver optical sub-assembly). The experimental setup is shown in Fig. 2. We tested transmission performance of our wide dynamic range APD receiver with respect to a reference conventional 10.7Gb/s NRZ tunable wavelength transmitter with extinction ratio better than 12dB. The reference tunable transmitter consists of a tunable external cavity laser and a negative chirp Mach-Zehnder Lithium Niobate modulator (chirp parameter alpha=-0.7). A 100km spool of standard single mode fiber was used to provide ~1600ps/nm chromatic dispersion at 1545nm. An optical switch was used to switch transmitter path between back-to-back and over 100km fiber spool. An erbium doped fiber amplifier (EDFA) along with a variable optical attenuator (VOA) was used to adjust OSNR level and launch power into the receiver. The input OSNR into the receiver was measured using an optical spectrum analyzer with its resolution bandwidth (RBW) set at 0.2nm. The receiver output was fed into a BERT (bit-error-rate-tester), which consists of PPG (pulse pattern generator) and ED (error detector).
4. Results and discussions
APD bias voltage was adaptively adjusted using APD reverse current monitor as the feedback. For example, the APD bias voltage which determines the M-factor at any given input power level was adjusted adaptively through an analog feedback loop to track and maintain the reverse current level based on a pre-determined optimization curve derived from APD chip characteristics. The adaptive control allows APD to operate at the optimum multiplication factor for best signal-to-noise ratio performance. Adaptive APD bias control along with receiver decision threshold control was implemented using fast opamp analog electronics to suppress potential receiver signal errors triggered by network power transients in real-world systems. Figure 3 shows the receiver output eyes at different receiver input power levels with OSNR=30dB for back-to-back transmission. We observed very little degradation in the eye quality as the receiver input power varied from -22dBm to +4dBm. Above +4dBm, the receiver performance is limited by TIA saturation where onset of TIA limiting causes eye distortion. At +4dBm, the APD current was measured to be ~3.5mA and the corresponding multiplication factor for the bias level was estimated to be ~1. The receiver sensitivity was measured to be -27.5dBm at 10.7Gb/s. Figure 4 shows the back-to-back OSNR performance at different receiver power levels, -22dBm, -10dBm, +2dBm, and +4dBm. The data were taken with receiver decision threshold level optimized for each OSNR set point at each receiver input power level. The best OSNR performance was achieved around -10dBm and slightly degraded as input power deviates away from -10dBm. The OSNR variations at 6E-5 BER was ~0.7dB across receiver power range from -22dBm to +4dBm. The slight variation in OSNR performance is from non-ideal behaviors from both APD and TIA. Further work is being conducted to better optimize APD and TIA.
Figure 5 shows the over-fiber OSNR performance at different receiver power levels, -22dBm, -10dBm, +2dBm, and +4dBm. The data were again taken with receiver decision threshold level optimized for each OSNR set point at each receiver input level. In consistent with back-to-back OSNR performance, the best OSNR was achieved around -10dBm receiver input power level for the same reason as described above. Again, less than 0.6dB OSNR variation was achieved at 6E-5 BER level as receiver input level varies from -22dBm to +4dBm. Figure 6 shows the BER curve as a function of receiver decision threshold (also known as V-curve) at various receiver powers from -22dBm to +4dBm for BB transmissions. The openings of V-curves are indicators on TIA linearity across receiver input power range. We saw fairly consistent V-curve openings across the receiver input range. We did observe slight closing of V-curve at higher receiver input power levels at +4dBm, where TIA starts to deviate from linear behavior near the onset of saturation or limiting.
Fig. 2. Experimental setup. SW: switch, OSA: optical spectrum analyzer, ASE: amplified spontaneous emission, Tx: transmitter, Rx: receiver
Fig. 3. Receiver output eye at receiver input power levels of -27dBm, -10dBm, +3.8dBm (the input eye is NRZ signal at 30dB OSNR at 10.7Gb/s).
Fig. 6. BER as a function of receiver decision threshold level for BB at various receiver input powers
5. Summary
We have demonstrated a wide dynamic range APD receiver for amplified noise-loaded optical transmission systems. The enabling technologies include high sensitivity/high breakdown threshold APD chip, wide range linear TIA, and adaptive APD bias control. We showed that the required OSNR to achieve 6E-5 BER was less than 1dB variation for receiver input powers ranging from -22dBm to +4dBm for both BB and over-fiber transmissions. In addition, the wide dynamic range receiver exhibits excellent sensitivity better than -27.5dBm at 10.7Gb/s.
References and links
1. G. P. Agrawal, Fiber–optics communication systems, 2nd ed. (Academic, New York, 1995).
2. M. Saleh, M. M. Hayat, P. P. Sotirelis, A. L. Holms, J. C. Campbell, B. A. Saleh, and M. C. Teich, “Impact ionization and noise characteristics of thin III–V avalanche photodiodes,” IEEE Trans. Electron Devices , 48, 2722–2731 (2001). [CrossRef]
3. P. Sun, M. M. Hayat, B. A. Saleh, and M. C. Teich, “Statistical correlation of gain and buildup time in APDs and its effects on receiver performance,” J. Lightwave Technol. 24, 755–768 (2006). [CrossRef]
4. T. Nakanishi, Y. Fukada, K–I. Suzuki, N. Yoshimoto, M. Nakamura, K. Kato, K. Nishimura, Y. Ootomo, and M. Tsubokawa, “Wide dynamic range and high sensitivity APD burst receiver configuration based on M–switching technique for 10G–EPON system,” LEOS2007, paper MH2, October 2007.
