Novel cell-AGC technique for Burst-Mode CMOS Preamplifier with Wide Dynamic Range and High Sensitivity for ATM-PON System S. Yamashita, S. Ide, K. Mori, A. Hayakawa*, N. Ueno**, and K. Tanaka Fujitsu Laboratories Ltd., Fujitsu VLSI Ltd.*, FUJITSU LIMITED** E-mail:
[email protected] Abstract We propose a novel cell-AGC technique for an ATMcell based burst optical receiver on a 156 Mbps subscriber system. The cell-AGC controls transimpedance gain according to the burst cell power and enables reception of burst signals with low extinction ratio. In addition, to realize high sensitivity, we developed an amplifier that is stable under changes in ambient conditions and deviations of transistor characteristics on IC. By adopting these techniques in a CMOS preamplifier-IC, the detectable power difference between burst cells was enlarged to more than 30 dB, and the minimum sensitivity was improved to less than -39.3 dBm. These performances show our new IC fully satisfy high sensitivity specification of an ATM-PON system, and incorporating this IC in the system makes it more flexible and economical.
1. Introduction Because of growth in IP traffic, demand for the construction of a broadband access network has been rapidly increasing. An asynchronous transfer mode passive optical network (ATM-PON) is one of the best means to meet this demand in regards to bandwidth and cost[1]. Figure 1 shows the configuration of an ATMPON system in which several Optical Network Units (ONUs) are connected to an Optical Line Terminal (OLT) through optical fiber and a 1:N star coupler. The downstream transmission is in continuous mode, and the upstream transmission is in cell-based Time Division Multiple Access (TDMA) burst mode. In 1998, ITU-T announced a global standard for ATM-PON system as recommendation G.983.1. The high sensitivity specification of Class C was defined to construct more flexible and economical networks. In this system, the transmission-loss differs for each path, so the burst cells at an OLT have a large power difference. In addition, an LD bias current can be applied on burst cells to lower ONU cost because ONUs are installed on the subscriber side. Therefore, the receiv-er in an OLT must realize high sensitivity, wide dynamic range between burst cells (high loud/soft ratio) and receive signals with an extinction ratio as low as 10 dB. Conventional preamplifiers[2]-[5] which control its transimpedance bit by bit have difficulty receiving large
signals with low extinction ratio because of the nature of the Log-Amp operation. In this work, we designed an automatic gain control (AGC) circuit to achieve (1) high loud/soft ratio (2) fast response from the first bit of the cell (3) ability to receive signals with low extinction ratio. Moreover, to realize high sensitivity, we adopt a novel configuration for an amplifier that has a stable transconductance and bandwidth. By adopting these techniques in a CMOS preamplifier-IC, the IC achieves a high loud/soft ratio of more than 30 dB, fast response from the first bit and ultra high sensitivity of less than -39.3 dBm. These results fully satisfy the high sensitivity specification of Class C, and a more flexible and economical ATM-PON system can be constructed. Upstream: Burst #1
Burst Cells
ONU#1 ONU#2
#1 #N
#2
#3 #2 #1
OLT #N
Optical star coupler
ONU#N
#1 #2 #3
#N #1
Downstream: Continuous
Figure 1. Configuration of ATM-PON system
2. Circuit Design 2.1. cell-AGC circuit Several preamplifier ICs with a wide dynamic range have been reported so far[2]-[5]. In these preamplifiers, a current-bypass diode, or current-bypass circuit is connected in parallel with a feedback resistor. When the output signal amplitude exceeds the turn-on voltage of the bypass circuit, input current flows through the circuit, leading to lower transimpedance. We name such types of controls bit-AGC, because they control transimpedance in every bit of a cell. If a large signal with a low extinction ratio is input to a bit-AGC preamplifier, the output waveform has a large bias, as shown in Figure 2. The amplitude of the output signal is thus reduced, and it becomes difficult to
output voltage
output signal
input current RESET input signal
Figure 2. Operation of bit-AGC Amp1
IN
Amp2 Amp3
OUT
Rf
Gain control circuit
Bottom level detector
G2
output voltage
discriminate between "0" and "1" properly. In the worst case, the output signal disappears. In order to solve the problem of output signal reduction caused by a large signal with a low extinction ratio, we propose a novel circuit to control transimpedance cell by cell according to the amplitude of the output signal. We name this type of control cell-AGC. As shown in Figure 3, the cell-AGC circuit consists of a bottom level detector (BLD), gain control circuit (GCC), reset circuit and FET connected in parallel to a feedback resistor. BLD quickly detects the bottom level of the Amp3 output signal, and a capacitor in BLD holds this level. Depending on this level, GCC generates a constant voltage during operation in the same cell, and determines Vgs of the FET connected to Rf. As the IC input current increases, Vgs of the FET also increases, resulting in lower transimpedance. Such control can also suppress duty deviation for signals with a wide dynamic range. At a change of cell, the reset signal is launched into BLD, causing the hold capacitor to charge, and the output of GCC returns to the initial state. In order to increase transmission efficiency, a fast response from the first bit is required. We accomplished this by fastening the GCC response. Figure 4 shows the response of cell-AGC for burst signals with a low extinction ratio. In case of large-signal cell, transimpedance is set as G1, and it is set as G2 for a small-signal. Since the output voltage is proportional to the input current, the bias of the "0" level is not as large as that in bit-AGC. This proves that this circuit discriminates between "0" and "1" properly. Therefore, burst cells can be received even with a low extinction ratio.
G1 G1
output signal G2
input current RESET input signal
Figure 4. Operation of cell-AGC
2.2. Amplifier Configuration In order to achieve high sensitivity, we adopted 0.25 µm high speed CMOS technology aiming at noise reduction. The input noise current density of the circuit can be written as[6]
ii 4kT 2 1 C 2ω 2 ) = 2qIg + + 4kT * gm * ( 2 + 3 RF ∆f gm 2 Rf 2
where Ig is the gate leakage current, RF is the feedback resistor, gm is the transconductance of the input FET, and C is the input capacitance. The first term is the shot noise of Ig, the second is the thermal noise of RF, and the third is the thermal noise of the input FET. We optimized Rf at 40 kΩ to suppress the second term as little as possible in the restriction of f-3dB=G/(2π∗Rf∗C), where G is the open loop gain of the IC and f-3dB is 3dB-bandwidth. We designed the open loop bandwidth to be 1.7 times wider than f-3dB to keep the phase-margin. Figure 5 shows the calculated total noise in our CMOS technology. As gm of the input FET increases, its thermal noise decreases initially. However, since input capacitance increases along with its transistor size, the thermal noise begins to increase gradually. As for this IC, considering the dissipation power, we optimized gm of the input FET to around 50 mS. In order to develop a stable amplifier, we must suppress changes of gm and keep the value close to the minimum point for the power supply and process deviation. To meet this requirement, we propose series FET load configuration. Figure 6 shows the configuration of Amp1-3. On the condition of Vin=Vout under negative feedback, Vgs of M1, M2 is equal to that of M3, M4. Consequently, each gm is shown as follows. g mM 1 = 2 µ N COX ×
mW1 × I1 L1
g mM 2 = 2µ N COX ×
mW1 × I1 kL1
Reset
Figure 3. Proposed circuit block
input noise current density [pA/√Hz]
1.2 1 0.8 0.6 0.4 0.2 0 0
0.1
0.2
0.3
0.4
0.5
gm of input FET [S]
Figure 5. Noise characteristics where µN is the electron mobility, Cox is the capacitance of gate oxide, I1 is the reference current, and k is the ratio of the gate width. Since the gain of this Amp is equal to the ratio of gm, the gain only depends on parameter k as follows. G=
the middle of the cell for input power levels of -36, -21, -6dBm. By cell-AGC, duty deviation is suppressed within 14%. Figure 11 shows the frequency response of the IC at +2.5 V and +25°C. The 3dB-down bandwidth is 145MHz and the equivalent input noise current density is 0.98 pA/√Hz. The minimum sensitivity level for an error rate of 1×10-10 using the 223-1 PN pattern is -39.8 dBm. Therefore, ultra-high sensitivity can also be achieved. Table 1 shows the measured characteristics of this IC under the condition of +2.5 V±5% voltage supply and ambient temperature in a range of -40 to +85°C. Figure 12 shows changes of input noise current density. The changes are within 0.14 pA/√Hz, which demonstrates that our proposed configuration of Amp is insensitive to voltage deviations and temperature changes. Those results fully satisfy G.983.1 Class C. So, we have succeeded in developing a burst-mode preamplifier-IC that can make ATM-PON system more flexible and economical.
g mM 1 = k g mM 2
The current that flows through M1 is shown as m*I1, so G and gm can be stabilized under power supply and process deviations. As G stabilizes, f-3dB can be kept stable under the relation of f-3dB=G/(2π∗Rf∗C). Therefore, f-3dB can be designed to be stable around the optimum bandwidth. Figure 7. Preamplifier IC I1 M4: W2/k, L1
M2: mW2/k, L1
Previous cell
Vout M3: W1, L1
Vin
Following cell
M1: mW1, L1 CH1
Figure 6. Configuration of Amp1-3
CH2
Reset
3. Measurement Results Figure 7 is a photograph of the preamplifier IC. Its size is 1.3 × 1.12 mm2. In our experiment, we used a PIN photodiode with a responsivity of 0.9 A/W and capacitance of 0.6 pF. To confirm our method, we measured the waveform for burst signal inputs. Figure 8 shows the response for 155.52 Mbps burst cell with its extinction ratio is 10 dB. The input optical power is -6 dBm for the previous cell and -30 dBm for the following cell. Transimpedance is controlled as the input power, and the "0" level bias is suppressed successfully. Next, to confirm the response speed of cell-AGC, we measured the waveform for a large burst cell with its power of -6 dBm. As shown in Figure 9, cell-AGC operates quickly from the first bit and duty deviation is not observed. This result demonstrates that BLD operates successfully from the first bit. Figure 10 shows eye diagrams measured in
CH1 V: 20mV/div CH2 V: 100mV/div H: 2ns /div
Figure 8. Waveform for burst signal input
CH1
CH2
Reset
CH1 V: 20mV/div CH2 V: 100mV/div H: 2ns /div
Figure 9. Burst response
Table 1. Peformance Technology 0.25µm CMOS Supply Voltage +2.5V ± 5% Ambient Temperature -40 - +85°C Bit Rate 155.52Mbps Transimpedance 90.7 - 92.3dBΩ 3dB-bandwidth 122 - 164MHz Input noise current density 0.91 - 1.05pA/√Hz Maximum duty deviation 14.0 % Minimum sensitivity < -39.3dBm Maximum overload > -6dBm Power dissipation 60 - 71.4mW
V: 2 mV/div H: 2 ns /div
optical input power: -36 dBm
V: 20 mV/div H: 2 ns /div
4. Summary optical input power: -21 dBm
V: 50 mV/div H: 2 ns /div
optical input power: -6 dBm
100
6
90
5
80
4
70
3
60
2
50
1 0 1000
40 10
100
input noise current density [pA/√Hz]
transimpedance [dBΩ]
Figure 10. Measured eye diagrams
frequency [MHz]
input noise current density [pA/√Hz]
Figure 11. Frequency response 1.2 1 0.8 0.6
+85°c +25°c - 40°c
0.4 0.2 0 2.35
2.4
2.45
2.5
2.55
2.6
2.65
supply voltage [V]
Figure 12. Change of input noise current density
We have proposed a novel cell-AGC technique for burst optical preamplifier in a 156Mbps ATM-PON system. By adopting this technique, we can realize (1) a high loud/soft ratio (2) fast response from the first bit (3) ability to receive signals with an extinction ratio as low as 10 dB. We also developed a stable amplifier that effectively suppresses the fluctuations caused by ambient conditions and process deviations to realize ultra high sensitivity. By adopting these techniques, we successfully developed a CMOS preamplifier-IC and realized a high loud/soft ratio of more than 30dB under extinction ratio of 10dB, quick response from the first bit, and ultra high sensitivity of less than -39.3 dBm under +2.5 V±5%, -40 to +85°C. Changes of input noise current density are within 0.14 pA/√Hz, which proves the stability of the amplifier. We confirmed that incorporating our new preamplifier-IC in an ATM-PON system can make it more flexible and economical.
5. Acknowledgement We would like to thank Dr. M.Yano for supporting our work.
6. References [1] K.Mori et al., “155.52Mb/s Optical Tranceiver Modules for ONU/OLT on ATM-PON systems”, Proc. ECOC, pp. 363366, Sep., 1997 [2] Y.Ota et al., “Burst-Mode Compatible Optical Receiver With A Large Dynamic Range”, IEEE J. Lightwave Tech., Vol. 8, No. 12, pp. 1897-1902, Dec., 1990 [3] D.Yamazaki et al., “156Mbit/s preamplifier IC with wide dynamic range for ATM-PON application”, Electronics Letters, Vol. 33, No. 15, pp. 1308-1309, July, 1997 [4] M.Nakamura et al., “An Instantaneous Response CMOS Optical Receiver IC with Wide Dynamic Range and Extremely High Sensitivity Using Feed-Forward Auto-Bias Adjustment”, IEEE J. Solid-State Circuits, Vol. 30, No. 9, pp. 991-997, Sep., 1995 [5] M.Nakamura et al., “A 156-Mb/s CMOS Optical Receiver for Burst-Mode Transmission”, IEEE J. Solid-State Circuits, Vol. 33, No. 8, pp. 1179-1187, Sep., 1998 [6] P.R.Gray and R.G.Meyer, Analysis and Design of Analogue Inetgrated Circuits, New York: Wiley, 1993
