a 10g linear burst-mode receiver supporting electronic dispersion compensation for extended-reach...
TRANSCRIPT
A 10G linear burst-mode receiver supporting
electronic dispersion compensation for
extended-reach optical links
Peter Ossieur,1,*
Nasir A. Quadir,1 Stefano Porto,
1 Marc Rensing,
1 Cleitus Antony,
1 Wei
Han,1 Peter O’Brien,
1 Y. Chang,
2 and Paul D. Townsend
1
1Photonic Systems Group, Tyndall National Institute, University College Cork, Ireland 2Vitesse Semiconductor Corporation, Transport Systems Engineering, USA
Abstract: We present a novel 10G linear burst-mode receiver (LBMRx).
Equipped with a PIN photodiode, a high sensitivity of −22.7dBm (bit-error
rate: 1.1x10−3
) was achieved when handling bursts with a dynamic range of
22.7dB (each −22.7dBm burst was preceded by a 0dBm burst). The
LBMRx requires a 150ns preamble for fast gain adjustment at the start of
each burst and can handle bursts separated by a guard time as short as
25.6ns. With electronic dispersion compensation, 3400ps/nm (200km)
chromatic dispersion can be tolerated at 2dB penalty in ASE-impaired links
using C-band electro-absorption modulators.
©2011 Optical Society of America
OCIS codes: (060.2360) Fiber optics links and subsystems; (060.4510) Optical
communications.
References and links
1. P. Ossieur, C. Antony, A. M. Clarke, A. Naughton, H. G. Krimmel, Y. Chang, C. Ford, A. Borghesani, D.
Moodie, A. Poustie, R. Wyatt, B. Harmon, I. Lealman, G. Maxwell, D. Rogers, D. W. Smith, D. Nesset, R. P.
Davey, and P. D. Townsend, “A 135km 8192-split carrier distributed DWDM-TDMA PON with 2x 32 x 10Gb/s
capacity,” J. Lightwave Technol. 29(4), 463–474 (2011).
2. D. Chiaroni, R. Urata, J. Gripp, J. E. Simsarian, G. Austin, S. Etienne, T. Segawa, Y. Pointurier, C. Simonneau,
Y. Suzaki, T. Nakahara, M. Thottan, A. Adamiecki, D. Neilson, J. C. Antona, S. Bigo, R. Takahashi, and V.
Radoaca, “Demonstration of the interconnection of two optical packet rings with a hybrid optoelectronic packet
router,” in Proceedings European Conference on Optical Communication, (2010), Postdeadline paper PD3.5.
3. P. Ossieur, T. De Ridder, J. Bauwelinck, C. Mélange, B. Baekelandt, X. Z. Qiu, J. Vandewege, G. Talli, C.
Antony, P. Townsend, and C. Ford, “A 10Gb/s burst-mode receiver with automatic reset generation and burst
detection for extended reach PONs,” in Proceedings Optical Fiber Conference (2009), Paper OWH3.
4. K. Azadet, E. F. Haratsch, H. Kim, F. Saibi, J. H. Saunders, M. Shaffer, L. Song, and Y. Meng-Lin,
“Equalization and FEC techniques for optical transceivers,” IEEE J. Solid-State Circuits 37(3), 317–327 (2002).
5. K. Hara, S. Kimura, H. Nakamura, N. Yoshimoto, and H. Hadama, “New AC-coupled burst-mode optical
receiver using transient-phenomena cancellation techniques for 10Gbit/s class high-speed TDM-PON systems,”
J. Lightwave Technol. 28(19), 2775–2782 (2010).
6. J. Nakagawa, M. Noda, N. Suzuki, S. Yoshima, K. Nakura, and M. Nogami, “Demonstration of 10G-EPON and
GE-PON coexisting system employing dual-rate burst-mode 3R receiver,” IEEE Photon. Technol. Lett. 22(24),
1841–1843 (2010).
7. P. Ossieur, N. A. Quadir, S. Porto, M. Rensing, C. Antony, W. Han, P. O. Brien, Y. Chang, and P. D.
Townsend, “A 10G linear BMRX supporting electronic dispersion compensation for extended-reach optical
links,” in Proceedings European Conference Optical Communication, (2011), Paper Th.13.B.
8. C. Knochenhauer, B. Sedighi, and F. Ellinger, “40Gbit/s transimpedance amplifier with high linearity range in
0.13µm SiGe BiCMOS,” Electron. Lett. 47, 605–606 (2011).
9. “Implementation agreement for integrated dual polarization intradyne coherent receivers,” OIF, IA OIF-DPC-
Rx-01.0, April 2010.
10. H. G. Krimmel, T. Pfeiffer, B. Deppisch, and L. Jentsch, “Hybrid electro-optical feedback gain-stabilized
EDFAs” in Proceedings European Conference on Optical Communication, pp. 693–694 (2009).
#157513 - $15.00 USD Received 2 Nov 2011; revised 17 Nov 2011; accepted 17 Nov 2011; published 1 Dec 2011(C) 2011 OSA 12 December 2011 / Vol. 19, No. 26 / OPTICS EXPRESS B604
1. Introduction
Today, optical networks are emerging that transport 10Gb/s non-return to zero modulated
(NRZ) bursts over extended-reach (>100km) fibre paths. Examples include hybrid
wavelength multiplexed, time division multiple access passive optical networks (DWDM-
TDMA PONs) [1] and optical burst-switched (OBS) DWDM metro networks [2]. The traffic
transported across such networks consists of bursts (or packets) whose amplitude can vary by
over 20dB from one burst to the next. Burst-mode receivers (BMRxs) are required to handle
such high dynamic range signals [3]. As these networks typically use C-band wavelengths,
dispersion compensation is needed to enable reaches beyond 100km. Compared to costly
optical dispersion compensation, electronic dispersion compensation (EDC) offers the
advantages of being adaptive, having no optical insertion losses and a small physical
footprint. EDC however requires linear receivers as opposed to limiting receivers [4]. While
continuous-mode linear receivers are commercially available, most existing BMRxs include
limiting stages [3, 5, 6], preventing their use in combination with EDC especially close to the
overload range of the receiver. Here, we report the realization of a 10Gb/s linear BMRx
(LBMRx) [7] and demonstrate that the device can enable EDC to extend the range of high
(20dB) dynamic range optical links.
Fig. 1. Linear burst-mode receiver block diagram.
2. Design and operation of the linear burst-mode receiver
Figure 1 shows a block diagram of the LBMRx. The anode of a 10GHz PIN photodiode is
connected to a high-speed (3dB bandwidth: 8.5GHz, input referred current noise: 726nARMS)
transimpedance amplifier (TIA) A1 whose gain can be continuously adjusted from 50Ω to
1.8kΩ. The cathode of the photodiode is connected to a second, low-speed (3dB bandwidth:
75MHz) transimpedance amplifier A2. Amplifier A2 has a linear gain of 500Ω over the entire
input dynamic range (−25dBm to 0dBm, corresponding to a peak input current ranging from
5.2µA till 1.6mA at 10dB extinction ratio and a photodiode responsivity of 0.9A/W), hence
its averaged output swing is proportional to the optical input power. This separation of the
high-speed signal path from the amplitude measurement block allows separate optimization
of the functions of each path. Next, peak detector PKD1 measures the amplitude of each
burst. This peak amplitude is provided to the gain adaptation block AGC1 which quickly
(within 25ns) adjusts the gain of the transimpedance amplifier A1 such that its output swing
equals a given reference. The gain adaptation block AGC1 also provides half the peak current
to a replica A’1 of the transimpedance amplifier A1, thus creating a reference for the
subsequent single-ended to differential conversion using amplifier A3 (gain: 6dB). Additional
gain is provided by post-amplifiers A4, whose gain can be continuously adjusted from 4dB to
21dB. Using measurements of the amplitude of the burst with the peak detecting block PKD2,
the gain of the post-amplifiers A4 is quickly (within 25ns) adjusted such that its output swing
#157513 - $15.00 USD Received 2 Nov 2011; revised 17 Nov 2011; accepted 17 Nov 2011; published 1 Dec 2011(C) 2011 OSA 12 December 2011 / Vol. 19, No. 26 / OPTICS EXPRESS B605
equals a given reference. Figure 2(a) shows the values to which the gain of TIA A1, post-
amplifier gain A4 and total LBMRx gain are adjusted as a function of the average input
optical power (photodiode responsitivity: 0.9A/W, extinction ratio: 10dB), as well as the
resulting output amplitude swing. Note how the LBMRx was designed in such a way that the
gain of post-amplifier A4 reduces first with increasing input power before the gain of TIA A1.
This approach concentrates the gain in the front-end of the LBMRx, thus minimizing
penalties due to post-amplifier noise and electrical crosstalk. Note that the full output swing
is not realized until the average optical input power exceeds −19dBm, which is acceptable as
long as the minimum signal swing exceeds the sensitivity of the subsequent EDC chip. With
input sensitivities of commercially available EDC chips as low as 20mV, even for a minimal
input power of −25dBm the output swing (100mV) of the LBMRx is sufficiently large.
Fig. 2. (a) Simulated gain and output swing vs. optical power, (b) simulated THD vs. optical power.
The LBMRx is designed to have minimum total harmonic distortion (THD) (across the
input dynamic range of −25dBm to 0dBm. The simulated THD (250MHz sinewave with 6dB
extinction ratio, 10 harmonics taken into account) is shown in Fig. 2(b) (PIN responsivity:
0.9A/W): it remains below 5% for an average optical power ranging from −25dBm to 0dBm.
This complies with industry-agreed specifications for linear optical receivers and measured
results of continuous-mode linear optical receivers [8,9]. The weak non-linear distortion
stems mainly from the non-linearity of the transistor that acts as a voltage controllable
feedback resistor in TIA A1, which is required to realize the controllable gain. Between
bursts, an external reset pulse (width: 10ns) is required to reset the peak detectors thus
preparing the LBMRx for a new burst. The unavoidable dc-offsets stemming from mismatch
between transistors and resistors are eliminated in a calibration step when the LBMRx is first
put into use with the ‘DC-offset compensation’ block.
3. Experimental setup and electronic dispersion compensation chip
Figure 3(a) shows the setup used to characterize the LBMRx. The outputs of two DFB lasers
(wavelengths: 1550nm, linewidth broadened using a 2kHz, 3% modulation depth tone for
suppression of stimulated Brillouin scattering) were non-return to zero (NRZ) modulated
with 10Gb/s data using electro-absorption modulators (EAMs). The EAM bias voltages were
fixed at 0.9V for all experiments. Next, a stream of alternating ‘loud’ and ‘soft’ bursts was
generated using variable optical attenuators and two semiconductor amplifiers (SOAs),
whose bias currents were switched on or off in alternating fashion, and hence acted either as
boosters or shutters. Each packet was 3.27µs long and consisted of a 150ns preamble with
sequences of 1s and 0s for settling of the LBMRx gain according to the strength of the burst,
followed by 231
-1 PRBS data (on which bit-error rates (BER) were measured). The extinction
ratio on the ‘soft’ channel was 9dB, on the ‘loud’ channel it was 7dB. The packets were
separated with 25.6ns guard bands. For the transmission experiments, two gain-clamped
EDFAs [10] were used to provide sufficient launch power for two spans of standard (ITU-T
G.652) single-mode fibre. The launch powers were always kept below + 12dBm for the
#157513 - $15.00 USD Received 2 Nov 2011; revised 17 Nov 2011; accepted 17 Nov 2011; published 1 Dec 2011(C) 2011 OSA 12 December 2011 / Vol. 19, No. 26 / OPTICS EXPRESS B606
‘loud’ burst. A final EDFA before the LBMRx was used as an optical pre-amplifier. The
variable optical attenuators before and after this EDFA are used to control the optical signal
to noise ratio (OSNR) and power of the signal provided to the LBMRx. The signal was
filtered using an optical filter (0.5nm) and coupled to the LBMRx.
Fig. 3. (a) Experimental setup for characterization of the LBMRx (insets: alternating stream of
loud and soft bursts; eye at transmitter output). (EAM: electro-absorption modulator, SOA:
semiconductor optical amplifier, GC-EDFA: gain-clamped erbium doped fibre amplifier,
SSMF: standard single-mode fibre, VOA: variable optical attenuator), (b) block diagram of
electronic dispersion compensation chip (VGA: variable gain amplifier, FFE: feedforward
equalizer, DFE: decision feedback equalizer, CRU: clock recovery unit).
The output of the LBMRx is ac-coupled with 560pF capacitors to the EDC chip. The
EDC chip (see Fig. 3(b)) consists of a 9-tap feedforward equalizer (FFE) and a 4-tap decision
feedback equalizer (DFE). A separate clock recovery unit (CRU) extracts a clock which is
used in the DFE to perform the final data decision. The FFE and DFE taps are adjusted with
an eye monitor and an internal microcontroller such that the eye opening of the received
signal is maximized. In the burst-mode experiments the microcontroller was disabled and tap
values derived when receiving a continuous signal (no burst-to-burst amplitude variations)
were used. Finally, an additional external clock-and-data recovery module was used to
provide a 10GHz recovered clock for the error detector.
4. Experimental results
The LBMRx was fabricated in a 0.25µm SiGe BiCMOS process; the die measures
2.4x2.1mm2 and uses 650mW with 2.5V/3.3V supplies. It was flip-chipped onto an AlN
substrate and wire-bonded to a 10G PIN photodiode (see Fig. 4(a)).
Fig. 4. (a) LBMRx mounted on ceramic substrate, (b) LBMRx input and output traces.
Figure 4(b) shows the response of the LBMRx to bursts with 15dB dynamic range. The
inset shows the clear open output eye of the −15dBm burst. Figure 5(a) shows the optical-to-
electrical transfer curve (S-parameter S21) for various gain settings measured using a vector
network analyzer suitably equipped with a Mach-Zehnder modulator. The gain ranges from
#157513 - $15.00 USD Received 2 Nov 2011; revised 17 Nov 2011; accepted 17 Nov 2011; published 1 Dec 2011(C) 2011 OSA 12 December 2011 / Vol. 19, No. 26 / OPTICS EXPRESS B607
85dBΩ down to 47dBΩ. Excellent stability of the frequency response over this gain range
can be observed with a 3dB bandwidth ranging from 6.8GHz to 8.8GHz and a maximum
peaking of less than 2dB. Next, we performed bit-error rate (BER) measurements. First, the
‘back-to-back’ (no fibre, no EDFAs) BER of the LBMRx (no EDC) is evaluated as a function
of the power on the photodiode, see Fig. 5(b). When all bursts have equal power (called the
‘static’ case), the sensitivity is −23.2dBm at a BER of 1.1x10−3
(the threshold for
RS(255,223) forward error correction, see e.g. IEEE 802.3av 10GEPON). Next, the BER
during the loud burst was measured to test the overload of the LBMRx. The shape of the
BER curve is attributed to the fact that for input powers above −5dBm, the bias voltage
across the photodiode started to drop which decreased the photodiode bandwidth and
increased its capacitance. This results in eye closure and deterministic bit errors, which will
be removed in a new design. The power coupled onto the photodiode was limited to 0dBm
due to setup limitations. Finally, we measured the sensitivity penalty due to a preceding loud
burst of 0dBm (‘dynamic’ case) and found it to be 0.5dB at a BER of 1.1x10−3
. Hence, a
dynamic range of at least 22.7dB can be supported, which exceeds typical requirements for
access or OBS networks.
Fig. 5. (a) Optical to electrical transfer (S21), (b) Bit-error rate vs input power.
Next, we measured the LBMRx performance for an ASE noise impaired signal by adding
an EDFA before the LBMRx. All optical signal-to-noise ratios (OSNRs) were measured in a
0.1nm reference bandwidth. First, the BER vs. OSNR for a given power onto the photodiode
was measured in the ‘static’ case. Unlike conventional optically pre-amplified receivers,
which operate far away from the thermal noise limited region, this is not necessarily possible
for optically pre-amplified BMRxs as these need to support large dynamic ranges, whereby
the upper limit of the dynamic range may be limited by the BMRx overload or the maximum
output power of the optical amplifier. Hence an OSNR characterization parameterized vs.
power (on the photodiode) is required: see Fig. 6(a). The error floor for the −20dBm case is
due to the thermal noise of the LBMRx. The BER curves in the case where the burst under
consideration is preceded by a burst at 0dBm signal power are also shown. The OSNR
penalty due to the preceding loud packet is shown in Fig. 6(a): 1.4dB penalty can be seen for
an input signal power of −20dBm, which falls to 0.4dB for an input power of −16dBm. The
OSNR penalty as a function of input power due to a preceding loud burst (0dBm) is shown in
Fig. 6(b): it becomes negligible for input signal powers above −12dBm, which is attributed to
the fact that any ‘memory’ from the loud burst is negligibly small once the input power of the
burst under consideration is larger than −12dBm. Next, the linearity of the LBMRx is
demonstrated by coupling it to an EDC chip and performing transmission tests with the fibre
spans and EDFAs. As no EDC chip was available whereby the taps could be adjusted on a
burst basis (which is necessary as the up to 20dB dynamic range in launched burst power
results in differing amounts of self-phase modulation, and hence impairment levels for ‘soft’
and ‘loud’ bursts), we considered the static case. The required OSNR for a BER of 1.1x10−3
#157513 - $15.00 USD Received 2 Nov 2011; revised 17 Nov 2011; accepted 17 Nov 2011; published 1 Dec 2011(C) 2011 OSA 12 December 2011 / Vol. 19, No. 26 / OPTICS EXPRESS B608
is shown in Fig. 7(a) versus reach for signal powers (incident on the LBMRx photodiode) of
−20dBm, −15dBm and 0dBm. LBMRx output eyes for the −20dBm case at 0km and 240km
are shown at their required OSNRs. The 1.5dB higher OSNR for −20dBm compared to the
−15dBm and 0dBm cases (at 0km) is attributed to the impact of thermal noise which is
significant when the LBMRx is operated close to its sensitivity limit. Assuming 2dB path
penalty at each input power and RS(223,255) FEC, at least 200km reach (3400ps/nm,
17ps/nm/km) is achieved across the input dynamic range of the LBMRx.
Fig. 6. (a) BER vs OSNR, and (b) OSNR penalty due to preceding loud burst.
Finally, the performance of the LBMRx with EDC is compared against a conventional
limiting BMRx and no EDC (note that using EDC with a limiting receiver does not make
sense as the limiting action erases all amplitude information needed to compensate the
intersymbol interference due to chromatic dispersion and self-phase modulation). As no
suitable conventional BMRx was available, the output of the LBMRx was provided to a 10G
limiting amplifier, thus converting the LBMRx into a conventional BMRx. The required
OSNR is shown in Fig. 7(b). No data could be obtained for the conventional BMRx at 0dBm,
and the error detector lost synchronization after 127km in the −20dBm case (as a BER less
than 10−3
could not be achieved), hence only data at −15dBm is shown. It can be seen how
after 210km (240km), the conventional BMRx has a 4dB (9dB) higher required OSNR
compared to the LBMRx+EDC. These results confirm the significant advantage of the
LBMRx over a conventional BMRx in extended-reach applications.
Fig. 7. (a) Required OSNR vs. reach (LBMRx + EDC), (b) Required OSNR vs. reach
(conventional BMRx).
5. Conclusion
We have demonstrated a novel 10Gb/s linear BMRx that can support a 22.7dB dynamic
range with 0.5dB penalty. It is shown how the linearity of the LBMRx can be used together
with electronic dispersion compensation to achieve significantly reduced transmission
penalties (up to 9dB at 240km) compared to conventional limiting BMRx’s.
#157513 - $15.00 USD Received 2 Nov 2011; revised 17 Nov 2011; accepted 17 Nov 2011; published 1 Dec 2011(C) 2011 OSA 12 December 2011 / Vol. 19, No. 26 / OPTICS EXPRESS B609