All-Optical Wavelength Conversion With Extinction Ratio ...

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 46, NO. 6, JUNE 2010 937 

All-Optical Wavelength Conversion With Extinction 

Ratio Improvement of 100 Gb/s RZ-Signals in 

Ultralong Bulk Semiconductor Optical Amplifiers 

Patrick Runge, Christian-Alexander Bunge, Member, IEEE, and Klaus Petermann, Fellow, IEEE 

Abstract—All-optical wavelength conversion in ultralong semiconductor 

optical amplifier (UL-SOA) is presented. Opposite to 

other SOA based single path wavelength converters the UL-SOA’s 

property as an extinction ratio (ER) regenerator can be used additionally 

to the wavelength conversion. When converting to lower 

wavelengths, cross-gain modulation as well as four-wave mixing 

(FWM) can be used as wavelength conversion mechanism. For the 

conversion to higher wavelengths only FWM has the ability for an 

ER improvement. 

Furthermore, the high-speed capability for the wavelength conversion 

with ER improvement in bulk UL-SOAs is analysed for 100 

Gb/s RZ-signals. The degradation due to bit pattern effects is investigated 

for RZ-50% PRBS of different lengths. No significant 

influence could be observed for PRBS orders up to 25. 

Index Terms—All-optical wavelength conversion, Bogatov effect, 

extinction ratio improvement, ultralong semiconductor optical amplifiers. 

I. INTRODUCTION 

WITH the advent of 100 Gigabit Ethernet and ever 

increasing demand for capacity, transparent networks 

and with them all-optical high-speed signal processing will be 

needed. In recent networks wavelength division multiplexing 

(WDM) is used to increase the transferred data rates. For 

this reason, all-optical high-speed wavelength conversion will 

be needed in order to route data signals to different WDM 

channels or to resolve blocking of coincident packets for 

the same output port [1]. There are several possibilities how 

all-optical wavelength conversion (AOWC) can be done but 

only semiconductor optical amplifier (SOA)-based solutions 

provide the possibility for monolithic integration. In general, 

there are three mechanisms in SOAs that can be used to convert 

the signal: four-wave mixing (FWM) [2], [3], cross-gain 

modulation (XGM) [4] and cross-phase modulation (XPM). A 

disadvantage when using FWM and XGM mechanism is the 

decreased extinction ratio (ER) of the converted output signal, 

compared to the input data signals. Since the XPM results from 

Manuscript received January 09, 2009; revised October 26, 2009 and January 

01, 2010. Current version published March 10, 2010. This work was supported 

by the Deutsche Forschungsgemeinschaft (DFG). 

The authors are with the Fachgebiet Hochfrequenztechnik, Technische Universität 

Berlin, D-10587 Berlin, Germany (e-mail: runge@hft.ee.tu-berlin.de; 

christian-alexander.bunge@tu-berlin.de; petermann@tu-berlin.de). 

Color versions of one or more of the figures in this paper are available online 

at http://ieeexplore.ieee.org. 

Digital Object Identifier 10.1109/JQE.2010.2041430 

the index change caused by the gain change, the XPM is small 

and can only be used for AOWC when using phase controlled 

switches. For example, Mach-Zehnder interferometer (MZI) 

are such phase controlled switches [5], [6]. In contrast to single 

path solutions, these MZIs are inherently narrowband and very 

sensitive to data rates or wavelengths. For this reason there have 

been several approaches to combine single path regeneration 

concepts and single path AOWC [7], [8]. But all these schemes 

are speed limited to the carrier lifetime (approximately 20 Gb/s) 

and cannot be used for high-speed data signals. On the other 

hand, there are also possibilities for high-speed single path 

AOWC [9], [10] but these solutions decrease the quality of the 

signal. Nevertheless, the demand for high-speed AOWC with 

signal regeneration has been reported in [11]. 

In [12], a promising novel single path solution based on ultralong 

semiconductor optical amplifier (UL-SOA) was presented. 

The scheme has the potential for 2R regeneration and since the 

regenerator mechanism is based on the fast intraband effects, 

it should be suitable up to several hundred Gb/s. A first proof 

of concept for the high-speed potential was presented with an 

80 GHz sine modulated signal [13]. In [14] it was shown that 

a Bogatov-like effect caused by the fast intraband effects in the 

UL-SOA’s saturated section is the reason for the ER improvement. 

In this paper, a combination of the AOWC due to XGM or 

FWM and the ER regeneration in UL-SOAs is presented. Moreover, 

so far it has not been investigated if the regenerative effect 

in UL-SOAs is truly useable for high-speed data signals. For 

this reason full numerical simulations with 100 Gb/s binary RZ 

PRBS signals are presented in order to investigate if the slow 

interband effects influence the data signal and bit pattern effects 

occur. The paper will be structured as follows: First, the basic 

properties of UL-SOAs are given in order to provide an understanding 

for the results presented in later sections. In Section III, 

the simulation setup for the investigations of Section IV and 

Section V is described. In Section IV it is investigated why no 

bit pattern effects for high-speed data signals occur and the explanation 

of the mechanisms of AOWC with ER improvement 

are described in Section V. The problems regarding later 2R regeneration 

are discussed in Section VI. 

II. PROPERTIES OF UL-SOAS 

The purpose of UL-SOAs is to benefit from the semiconductor’s 

fast nonlinear intraband effects like carrier heating and 

spectral hole burning, while slow interband effects should be 

suppressed as far as possible, in order to avoid bit pattern effects. 

0018-9197/$26.00 © 2010 IEEE 

Authorized licensed use limited to: Technische Universitaet Berlin. Downloaded on March 29,2010 at 05:14:58 EDT from IEEE Xplore. Restrictions apply.

938 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 46, NO. 6, JUNE 2010 

Fig. 2. Conceptual setup of the wavelength converter including the ER improvement; 

is either , +1 or 0 1 depending which 

kind of wavelength conversion mechanism shall be used. 

III. SIMULATION SETUP 

Fig. 1. Probe signal’s amplification asymmetry due to the dynamical part of the 

gain suppressions in the UL-SOA’s saturated section (1 = 0 ; 

jE j P ). 

Unlike in SOAs, the main part of UL-SOAs is saturated by 

the amplified signals and the amplified spontaneous emission 

(ASE) noise after a certain length. For this reason, UL-SOAs 

can be regarded as being divided into an amplifying section 

and a saturated section. For bulk SOAs with typical dimensions, 

the transition between these sections takes place after approximately 

1 mm of propagation. The amplifying section has the 

same properties as a short SOA and the saturated section can be 

regarded as another device with different properties. Because of 

the high optical power after the amplifying section, the carrier 

density in the saturated section is fixed to the net transparency 

level. If the signal levels change much faster than the carrier lifetime, 

the carrier density in the amplifying section is only dependent 

on the average optical power. If the signal remains at one 

of the states longer than the carrier lifetime, the length and the 

gain of the amplifying section changes. As a result bit pattern effects 

occur. For high data rates with fast changing signal states, 

the performance is not limited by the slow interband effects. 

Only the fast nonlinear intraband effects influence the signals in 

the saturated section. Their response time is about 1000 times 

shorter compared to the slow interband effects, but their impact 

is also weaker. For this reason, the desired operation point for 

signal processing is when the carrier density inside the UL-SOA 

is shifted as far as possible towards the front of the device in 

order to efficiently use the length of the device and to avoid the 

influence of back-propagating ASE. 

Moreover, the UL-SOA shows a Bogatov-like effect over several 

nm due to the increased efficiency of the fast intraband effects. 

In [15] Bogatov has demonstrated that the beating of two 

input signals creates dynamic gain and index gratings in nonlinear 

semiconductor media. Because of the interaction of these 

dynamic gratings, the weaker signal’s amplification is dependent 

on the stronger signal’s wavelength detuning. Due to the 

-factor causing the index pulsation, the probe signal is stronger 

amplified on the longer wavelength side of the pump signal 

(Fig. 1). In [15] the pulsations are due to the slow interband 

effects, while in [14], [16] a similar effect based on the fast intraband 

effects was presented. Due to the fast intraband effects, 

this Bogatov-like effect is suitable for high-speed operations. 

The conceptual setup of the AOWC with ER improvement 

is shown in Fig. 2. Due to the two input signals, a parametric 

amplification of the data signal in dependence of the CW signal 

can be obtained, resulting in a nonlinear transfer characteristic 

over 

. With the help of the nonlinear 

gain characteristic, the ER can be improved. To achieve 

a maximum ER improvement the UL-SOA’s operating point is 

set at a steep slope of the nonlinear transfer characteristic. 

The investigated device is a 4 mm-long bulk SOA. The geometrical 

structure has been optimised in order to increase the 

efficiency of the nonlinear effects (decreased coupling losses 

and increased mode confinement [20]). The values are given in 

Table I. The driving current is 300 mA/mm. 

In later applications the presented scheme might be used 

at a switching node in long-haul transmission systems. After 

being transmitted over several spans, a degenerated channel 

approaches the node. First the whole channel will be optically 

amplified with an EDFA before the WDM channel of interest 

will be converted to the new wavelength. For this reason, we 

assume the following parameters for the input data signal after 

an EDFA: , . The power for 

the additionally injected CW is: 

. The signals’ 

wavelength are 1560 nm and 1564 nm, respectively. The 

wavelength alignment depends on which kind of wavelength 

conversion mechanism shall be used. The data signals being 

investigated are 100 Gb/s OOK-RZ-50% modulated PRBSs 

of orders up to 25. The bandpass filter after the UL-SOA 

is a second-order Gaussian filter with a 3 dB-bandwidth of 

275 GHz. 

Due to limited computation capacity it was not possible to 

simulate complete PRBSs longer than . For this reason, 

critical bit sequences were taken to estimate the impact of pattern 

effects on the regeneration scheme. As already mentioned 

in Section II, pattern effects occur if the data signal remains long 

at one of the states. Therefore, the critical bit sequences contain 

various combinations of the following bit sequences: constant 

signal states, one inverted bit within a constant signal state and 

alternating bit sequences. These combinations of bit sequences 

are expected to create worst-case bit pattern effects. A detailed 

description of the critical bit sequence is given in the Appendix . 

The simulations have been done with the same model as in 

[21], [22] where a very good match between measurements and 

simulation results has been demonstrated. 


RUNGE et al.: ALL-OPTICAL WAVELENGTH CONVERSION WITH EXTINCTION RATIO IMPROVEMENT 939 

TABLE I 

PARAMETERS OF THE BULK SOA USED IN THE SIMULATIONS; THE MATERIAL 

PARAMETERS HAVE BEEN TAKEN FROM LITERATURE [13], [17]–[19] 

Fig. 3. Eye diagram for a PRBS with the length of 2 0 1 after the bandpass 

filter at without additional CW at the input; the signal is totally distorted 

due to bit pattern effects, saturation effects and SPM. 

IV. PATTERN EFFECTS FOR HIGH-SPEED SIGNALS 

In general the speed of bulk SOAs for all-optical signal processing 

is limited to approximately 20 Gb/s due to the carrier 

lifetime. Bulk UL-SOAs in contrast should be usable for signal 

processing of data signals with transmission rates fairly above 

the carrier lifetime. For SOAs under usual driving conditions, 

bit pattern effects are expected for 100 Gb/s so that the existence 

of bit pattern effects needs to be further investigated. 

Fig. 3 shows the eye diagram of a 100 Gb/s data signal before 

and after the UL-SOA that has been exclusively injected into 

the UL-SOA. As expected, the eye diagram at the UL-SOA’s 

output is totally distorted due to multiple effects: bit pattern effects, 

gain saturation, self-phase modulation (SPM) and ASE. 

Since 100 Gb/s are not fairly above the carrier lifetime, it is not 

clear how big the contribution to the distortions due to bit pattern 

effects are. For this reason, the standard deviation of the signal 

levels at the maximum eye opening in dependence on the equivalent 

PRBS order is evaluated (Fig. 4). With increasing equivalent 

PRBS order, the standard deviation of the signal states increases 

resulting in a decreasing signal quality. The tremendous 

increase of the standard deviation’s “1”-level between the first 

and fourth equivalent PRBS order can be ascribed to the growing 

overshoot on top of the eye opening. Moreover, the simulation 

Fig. 4. Standard deviation of the signal levels at the maximum eye opening in 

dependence of an equivalent PRBS order after the bandpass filter without additional 

CW signal (corresponding to Fig. 3; solid lines: true PRBS and dashed 

lines: critical sequence). 

shows that after a PRBS order of 4 the distortions due to bit pattern 

effects stagnate. 

When injecting the data signal with an additional CW signal 

that is orthogonally polarised into the UL-SOA, a clearly 

opened eye diagram can be observed at the output of the 

UL-SOA (Fig. 5). Also, the data signal has been amplified 

by 12 dB. The amplification is about half as much as the 

amplification for the case with the exclusive transmission of 

the data signal through the UL-SOA because the gain equally 

distributes to the data signal and the CW signal. Furthermore, 

the ER of the data signal has been reduced from 6 dB to 1.5 dB. 

The investigation of the signal levels’ standard deviation shows 

that for the case with the orthogonally polarised additional CW, 

nearly no bit pattern effects can be observed (Fig. 6). Only a 

very slight increase between a PRBS order of 1 and 2 can be 

observed so that the influence of bit pattern effects is negligible. 

For this reason, the additional CW signal can be regarded as a 

holding beam successfully suppressing bit pattern effects. 



Fig. 5. Eye diagram for the converted signal at ; the input signal has been 

a PRBS with the length of 2 0 1 at ; compared to the input signal, the 

output signal has been amplified but the ER has been decreased to 1.9 dB. 

Fig. 7. Carrier density distribution along the propagation direction for infinite 

long constant signal levels of the data signal. 

8 mm-long device in [21]. As a result, there is less back-propagating 

ASE, shifting the carrier density peak towards the front 

of the UL-SOA contributing to a stable operation point. 

In the eye diagram of Fig. 5 still a broadening of the data 

signal’s traces at the UL-SOA’s output can be seen although the 

bit pattern effects are negligible. Since the trace broadening can 

also be observed for sine modulated signals where no bit pattern 

effects are expected at all because half the sine’s time period is 

much shorter than the carrier life time. When investigating the 

field intensity’s 

relative standard deviation of the 

“0”-level and the “1”-level and their distribution, for both signal 

levels a Gaussian distribution with an equal relative standard 

deviation can be observed. As a result, the trace broadening can 

be ascribed to ASE. 

Fig. 6. Standard deviation of the optical signal levels at the maximum eye 

opening in dependence of an equivalent PRBS order after the bandpass filter 

with co-polarised additional CW signal (corresponding to Fig. 5; solid lines: 

true PRBS and dashed lines: critical sequence). 

To prove the efficiency of the CW signal as a holding beam, 

the carrier density distribution inside the UL-SOA along the 

propagation direction has to be regarded. Fig. 7 shows the 

steady-state carrier density for infinite long constant signal 

levels. For the case with an additional CW signal, the carrier 

density peak would shifts by 1.5% of the device length when 

a worst-case transition after a very long constant signal state 

appears. For the case without additional CW signal, the carrier 

density peak shifts by 9.5% of the device length resulting in 

a mentionable change of the operation point. Moreover, the 

smallish shifts of the carrier density peak also have to be related 

to the UL-SOA’s driving conditions and the optimised device. 

First of all, the low coupling losses and the high optical input 

power deeply saturate the UL-SOA even for a data signal’s 

“0”-level. Moreover, the low input extinction ratio only results 

in a small shift of the carrier density peak for a worst case 

transition. Another aspect is the optimised geometry of the 

device, increasing the efficiency of the nonlinear intraband 

effects. Hence, the UL-SOA can be shortened compared to the 

V. AOWC WITH ER IMPROVEMENT 

In the previous section it has been shown that due to an additional 

CW signal also transmission rates above the carrier lifetime 

are possible. Unfortunately, due to saturation effects the 

ER of the data signal has been decreased dramatically (Fig. 9) 

making the data signal useless for further transmission. 

In [14], [21] has been presented when injecting the addition 

CW signal parallel to the data signal into the UL-SOA, 

an ER improvement can be achieved due to a Bogatov-like 

effect. Since the scheme is sensitive to the input power ratio 

, Fig. 9 shows the signal’s ER after the UL-SOA 

and the bandpass filter at different wavelengths for two different 

wavelength alignments of the input signals. The results 

illustrate that wavelength conversion with ER improvement 

due to XGM or FWM as wavelength conversion mechanism is 

possible. In order to select the AOWC mechanism that should 

be used, the wavelength alignment and the CW signal’s power 

have to be set correctly. 

A. AOWC With ER Improvement Due to XGM 

In [21], it has been demonstrated that the data signal has to 

be located on the shorter wavelength side of the CW signal in 

order to obtain an ER improvement for the data signal (circle- 



Fig. 8. Extinction ratio measured at different wavelength in dependence of the input signals’ power ratio for a 100 Gb/s RZ50% OOK signal (left: < 

and right: < ), is the first FWM product on the shorter wavelength side of the input signals and is the first FWM product on the longer 

wavelength side of the input signals. 

symbol lines of Fig. 8(b)). For the opposite wavelength alignment, 

the data signal’s ER decreases as the circle-symbol lines 

in Fig. 8(a)) show. For AOWC with ER improvement due to 

XGM the wavelength alignment has to be swapped (up-trianglesymbol 

lines of Fig. 8(a)). The reason for the opposite wavelength 

alignment can be explained by Fig. 1. Similar to the solo 

ER improvement case (without AOWC) described in [14], the 

gain-suppressing part of the probe signal’s amplification asymmetry 

(positive wavelength detuning) is used to achieve the ER 

improvement. When investigating how the signals’ power develop 

inside the UL-SOA, the data signal becomes the stronger 

pump signal and its modulation influences the weaker probe 

(CW signal). Since the CW signal has been cross-gain modulated 

in the UL-SOA’s saturated section, its modulation is inverse 

to the data signal’s modulation. Furthermore, the intensity 

of the Bogatov-like effect is dependent on the pump signal’s 

power (data signal’s power). For this reason the CW’s “1”- 

level is less suppressed than the “0”-level resulting in an increasing 

ER. 

Moreover, Fig. 8 shows that the ER improvement efficiency 

for the wavelength conversion case is approximately 3 dB better 

than for the solo ER improvement case. In both cases, the ER 

improved signal is the probe and for this reason the ER improvement 

efficiency is dependent on the pump’s modulation 

depth (ER). In the second part of the UL-SOA’s saturated section 

where the ER improvement takes place the pump’s modulation 

is strong for the wavelength conversion case. For the 

solo ER improvement case, the pump signal (CW) is dependent 

on the XGM efficiency of the first part of the UL-SOA’s 

saturated section wether for the wavelength conversion case the 

pump (data signal) is modulated from the beginning. 

Fig. 9 shows an eye diagram of a wavelength converted and 

ER improved 100 Gb/s RZ-data signal for an input power ratio 

of 1 dB. After the bandpass filter where the signal has been 

filtered from ASE and the original data signal, the converted 

signal’s “1”-level has been amplified by 7.5 dB and the ER has 

been increased from 6 dB to 13 dB. 

Fig. 9. Eye diagram for a PRBS with the length of 2 0 1 after the bandpass 

filter; the mechanism for the wavelength conversion has been XGM so that the 

data signal was inversely converted to . 

B. AOWC With ER Improvement Due to FWM 

The AOWC with ER improvement due to XGM from the 

previous subsection has only the ability to convert to a shorter 

wavelength. When using FWM as wavelength conversion mechanism 

instead of XGM, up and down conversion with ER improvement 

can be realised. Moreover, the converted signal is 

not inversely modulated to the original input data signal. 

1) Converting to Longer Wavelengths: When increasing the 

input power of the CW signal to 9 dBm and setting the wavelength 

alignment to 

, AOWC to longer wavelengths 

with ER improvement can be achieved. The power of 

the signal after the UL-SOA and the bandpass is about equal 

compared to the data signal’s input power and the ER slightly 

increased from 6 dB to 7 dB so that the ER improvement mainly 

compensates the ER degeneration due to the saturation effects. 

The reason for the worse ER improvement compared to the 

AOWC with ER improvement due to XGM and the solo ER 

improvement can be figured out when the mechanism of the 



Fig. 10. Output spectra of the UL-SOA for a FWM wavelength converted 100 Gb/s RZ50% OOK signal for an input power ratio (P =P ) of 8 dB (left: 

conversion to higher wavelengths and right: conversion to lower wavelengths), the red graph illustrates the relevant Bogatov-like effect causing the ER improvement 

at . 

ER improvement is further investigated. In the UL-SOA’s saturated 

section the signal is FWM converted to . Since 

the data signal’s ER decreases while propagating through the 

UL-SOA also the ER of the wavelength converted signal is low. 

In the second part of the UL-SOA’s saturated section, the ER 

is improved by the Bogatov-like effect. Opposite to the previously 

discussed ER improvement mechanisms, the amplifying 

part of the probe’s amplification asymmetry of Fig. 1 (negative 

wavelength detuning) is used to achieve the ER improvement 

(Fig. 10(a)). During the whole conversion, the converted signal 

has less power than the original data signal. For this reason. 

the converted signal will be refereed to as the probe and to the 

original signal as the pump. Since the probe and the pump have 

the same modulation and the efficiency of the Bogatov-like effect 

is dependent on the pump’s power, the probe’s “1”-level 

is more amplified than the “0”-level resulting in an increasing 

ER. Because of the Bogatov-like effect is small for a negative 

wavelength detuning compared to the positive wavelength detuning 

also the efficiency of the ER improvement decreases. Although, 

the CW signal in Fig. 10(a) has more power than the 

data signal, its Bogatov-like effect does not cause the ER improvement. 

First of all, the Bogatov-like effect is negligible over 

such a large negative wavelength detuning and secondly the CW 

signal is inversely modulated to the converted signal resulting in 

a decreasing ER in combination with a negative wavelength detuning. 

2) Converting to Shorter Wavelengths: Keeping the input 

power of the CW signal at 9 dBm but changing the wavelength 

alignment to 

, AOWC to shorter wavelengths 

with ER improvement is achieved. The power of the signal after 

the UL-SOA and the bandpass has been decreased by about 1 

dB compared to the data signal’s input power and the ER also 

slightly increased from 6 dB to 7 dB so that the ER improvement 

mainly compensates the ER degeneration due to the saturation 

effects. Again, the small ER improvement efficiency can be revealed 

when a closer look is taken on how the Bogatov-like effect 

influences the converted signal. Similar to the other FWM 

case, the signal is FWM converted to 

with a decreased 

ER in the UL-SOA’s saturated section while the ER 

improvement only takes place in the second part of the saturated 

section. In general, one would expect that the converted 

signal with low power (probe) is influenced by the stronger data 

signal but due to the higher input power ratio, the data signal 

is strongly suppressed by the Bogatov-like effect of the CW 

signal (Fig. 10(b)). Therefore, the influence of the data signal on 

the converted signal due to the Bogatov-like effect is negligible 

because of the data signal’s low power. Instead the cross-gain 

modulated and strong CW signal can be regarded as the pump 

causing the ER improvement. Similar to the solo ER improvement 

case, the attenuating part of the Bogatov-like effect in 

combination with the inversely modulated pump increases the 

ER of the converted signal. So far, the Bogatov-like effect was 

only discussed for a signals’ detuning of but in this case 

the Bogatov-like effect interacts over decreasing the efficiency 

of the ER improvement. 

The bit pattern effects are also negligible for the presented 

AOWC with ER improvement schemes in this section. As 

stressed in Section IV, the inexistence of the bit pattern effects 

is caused by the high optical power of the input signals and 

the low ER of the data signal. For the XGM case none of 

these parameters has changed (only the CW signal is parallelly 

polarised to the data signal) and for the FWM case the optical 

power of the holding beam (CW signal) had to be increased by 

8 dB reducing bit pattern effects even better. 

VI. DISCUSSION CONSIDERING 2R-REGENERATION 

Since the signal in Fig. 9 has been amplified (re-amplification) 

and its ER has been improved (first step in direction 

of re-shaping), the presented scheme has the potential for 2R 

regeneration. But unfortunately, the traces of the output signal 

have been broadened so that the re-shaping condition has not 

been met completely because of missing noise compression. 

The broadening of the traces can be ascribed to the operating 

point of the UL-SOA. Under these driving conditions, the 

steep slope of the transfer function is used for a maximum ER 

improvement. As a result, a small fluctuation of the optical 

power due to ASE will increase the fluctuation because of the 

transfer characteristic’s steep slope. A solution for this problem 



would be clipping of the signal levels, being typically obtained 

for the “1”-level due to the saturation effects in SOAs. The 

inexistence of the gain saturation for the presented scheme is 

also indicated by the signals’ output power levels in Fig. 5 and 

Fig. 9 being about 5 dB less for the parallel polarised case where 

the Bogatov-like effect is enabled. Since the attenuating part 

of the Bogatov-like effect is used for the ER improvement, the 

output signal’s “1”-level for the AOWC cases with conversion 

to shorter wavelengths and for the solo ER improvement case 

do not saturate. 

For the conversion to longer wavelengths, the amplifying 

part of the Bogatov-like effect is used to achieve the ER improvement 

so that in general a saturation should be possible but 

unfortunately this mechanism is very inefficient (see discussion 

in Section V-B). Hence, the UL-SOA’s length needs to be 

increased and it might not even be observable in MQW-devices 

because of their minor -factor. When decreasing the 

-factor, the probe’s gain asymmetry becomes more and more 

symmetric and will also lead to attenuation for a negative 

wavelength detuning [14] so that the ER improvement mechanism 

for conversion to longer wavelengths might no more 

workout. In such a case, the wavelength alignment must be 

in order that the inverse modulated CW also 

improves the ER of the wavelength converted signal. Furthermore, 

in MQW-devices the mode confinement is much smaller 

also decreasing the efficiency of the effects due to the nonlinear 

gain [20] so that an ER improvement might not be observable 

in MQW-devices at all. 

VII. CONCLUSION 

We presented three different schemes for AOWC with ER improvement 

in bulk UL-SOAs and proved their high-speed potential 

with 100 Gb/s RZ PRBS signals. Numerically, an ER improvement 

of 7 dB could be achieved while the signal has been 

converted over 4 nm to a shorter wavelength due to XGM with 

a 4 mm highly nonlinear bulk device. Unfortunately, this wavelength 

conversion is limited to down conversion, but when using 

FWM, also up conversion is possible. In this case, however, the 

ER improvement efficiency decreases and mainly compensates 

degrading effects. 

It has been shown that for an optimised device under proper 

driving conditions no bit pattern effects will arise up to a word 

length of 25 bits. The suppression of the bit pattern effects can 

be ascribed to the additional injected CW signal needed for the 

wavelength conversion and the Bogatov-like effect causing the 

ER improvement. 

In the proposed schemes, small fluctuations increase, since 

the ER improvement is realised with a steep slope of the nonlinear 

transfer characteristic. For later AOWC applications with 

2R-regeneration, a clipping of the signal mark levels have to be 

performed. 

APPENDIX 

CRITICAL BIT SEQUENCES 

To investigate the impact of bit pattern effects on the presented 

scheme, critical bit sequences are used because of 

lower computational effort compared to true PRBSs. Fig. 11 

Fig. 11. Flowchart for generating a critical bit sequence of the Nth PRBS order; 

the generation rule can be applied for even PRBS orders; for odd PRBS orders 

the alternating bit sequences have to be extended or shortened by one bit in 

order to avoid bit subsequences of the same consecutive bits longer than the 

PRBS order. 

shows the flowchart how these critical sequences are created. 

As an example the 4th order critical bit sequence is: 

000010001010000111101110101111. 

For the simulation in this article the flowchart has passed 

through until sequences of different PRBS orders had a length 

of 263 bits. 

REFERENCES 

[1] S. B. Yoo, “Optical packet and burst switching technologies for the 

future photonic Internet,” J. Lightw. Technol., vol. 24, no. 12, pp. 

4468–4492, 2006. 

[2] S. Diez, C. Schmidt, R. Ludwig, H. G. Weber, K. Obermann, S. Kindt, 

I. Koltchanov, and K. Petermann, “Four-wave mixing in semiconductor 

optical amplifiers for frequency conversion and fast optical switching,” 

IEEE Sel. Topics Quantum Electron., vol. 3, no. 5, pp. 1131–1145, 

1997. 

[3] A. Kelly, A. Ellis, D. Nesset, R. Kashyap, and D. Moodie, “100 Gb/s 

wavelength conversion using FWM in an MQW semiconductor optical 

amplifier,” Electron. Lett., vol. 34, no. 20, pp. 1955–1956, 1998. 



[4] T. Durhuus, B. Mikkelsen, C. Joergensen, S. L. Danielsen, and K. 

Stubkjaer, “All-optical wavelength conversion by semiconductor optical 

amplifiers,” J. Lightw. Technol., vol. 14, no. 6, pp. 942–954, 1996. 

[5] C. Joergensen, S. Danielsen, K. Stubkjaer, M. Schilling, K. Daub, P. 

Doussiere, F. Pommerau, P. Hansen, H. Poulsen, A. Kloch, M. Vaa, B. 

Mikkelsen, E. Lach, G. Laube, W. Idler, and K. Wunstel, “All-optical 

wavelength conversion at bit rates above 10 Gb/s using semiconductor 

optical amplifiers,” IEEE Sel. Topics Quantum Electron., vol. 3, no. 5, 

pp. 1168–1180, 1997. 

[6] D. Wolfson, A. Kloch, T. Fjelde, C. Janz, B. Dagens, and M. Renaud, 

“40 Gb/s all-optical wavelength conversion, regeneration and demultiplexing 

in an SOA-based all-active Mach-Zehnder interferometer,” 

IEEE Photon. Technol. Lett., vol. 12, no. 3, pp. 332–334, 2000. 

[7] G. Contestabile, R. Proietti, N. Calabretta, and E. Ciaramella, 

“Cross-gain compression in semiconductor optical amplifiers,” J. 

Lightw. Technol., vol. 25, no. 3, pp. 915–921, 2007. 

[8] H. Chayet, S. B. Ezra, N. Shachar, S. Tzadok, S. Tsadka, and J. 

Leuthold, “Regenerative all-optical wavelength converter based on 

semiconductor optical amplifier and sharp frequency response filter,” 

in Proc. OFC, 2004, ThS2. 

[9] R. J. Manning, X. Yang, R. P. Webb, R. Giller, F. C. Garcia Gunning, 

and A. D. Ellis, “The ‘Turbo-Switch’—A novel technique to increase 

the high-speed response of SOAs for wavelength conversion,” in Proc. 

OFC, 2006, paper OWS8. 

[10] J. Leuthold, L. Möller, J. Jaques, S. Cabot, L. Zhang, P. Bernasconi, 

M. Cappuzzo, L. Gomez, E. Laskowski, E. Chen, A. Wong-Foy, and 

A. Griffin, “160 Gb/s SOA all-optical wavelength converter and assessment 

of its regenerative properties,” Electron. Lett., vol. 40, no. 9, pp. 

554–555, 2004. 

[11] J. Leuthold, W. Freude, G. Boettger, J. Wang, P. Vorreau, and A. Marculescu, 

“Trends in the field of all-optical wavelength conversion and 

regeneration for communication up to 160 Gb/s,” in Proc. ECOC, 2005, 

Tu.3.3.6. 

[12] U. Busolt, G. Bramann, H.-J. Wünsche, C. Schmidt, B. Sartorius, 

H.-P. Nolting, and T. Rosin, “Two-wave competition for all-optical 

signal processing: First experimental verification in ultra-long SOAs,” 

in Proc. ECOC, Rimini, 2003, Th3.5.1. 

[13] G. Bramann, H.-J. Wünsche, U. Busolt, C. Schmidt, M. Schlak, B. Sartorius, 

and H.-P. Nolting, “Two-wave competition in ultralong semiconductor 

optical amplifiers,” IEEE J. Quantum Electron., vol. 41, no. 

10, pp. 1260–1267, 2005. 

[14] P. Runge, R. Elschner, C.-A. Bunge, and K. Petermann, “Extinction 

ratio improvement in ultralong semiconductor optical amplifiers,” 

IEEE J. Quantum Electron., vol. 45, no. 6, pp. 578–585, 2009. 

[15] A. Bogatov, P. Eliseev, and B. Sverdlov, “Anomalous interaction of 

spectral modes in a semiconductor laser,” IEEE J. Quantum Electron., 

vol. 11, no. 7, pp. 510–515, 1975. 

[16] A. Uskov, J. Mørk, and J. Mark, “Wave mixing in semiconductor laser 

amplifiers due to carrier heating and spectral-hole burning,” IEEE J. 

Quantum Electron., vol. 30, no. 8, pp. 1769–1781, 1994. 

[17] A. Mecozzi and J. Mørk, “Saturation effects in nondegenerate fourwave 

mixing between short optical pulses in semiconductor laser amplifiers,” 

IEEE J. Quantum Electron., vol. 3, no. 5, pp. 1190–1207, 

1997. 

[18] G. Toptchiyski, S. Kindt, K. Petermann, E. Hilliger, S. Diez, and H. 

G. Weber, “Time-domain modeling of semiconductor optical amplifiers 

for OTDM applications,” J. Lightw. Technol., vol. 17, no. 12, pp. 

2577–2583, 1999. 

[19] J. Leuthold, M. Mayer, J. Eckner, G. Guekos, H. Melchior, and C. 

Zellweger, “Material gain of bulk 1.55 m InGaAsP/InP semiconductor 

optical amplifiers approximated by a polynomial model,” J. 

Appl. Phys., vol. 87, no. 1, pp. 618–620, 2000. 

[20] P. Runge, R. Elschner, and K. Petermann, “Optimising four-wave 

mixing in ultralong SOAs,” in Proc. NUSOD ’09, Gwangju, 2009. 

[21] P. Runge, R. Elschner, C.-A. Bunge, K. Petermann, M. Schlak, W. 

Brinker, and B. Sartorius, “Operational conditions for the extinction 

ratio improvement in ultralong SOAs,” IEEE Photon. Technol. Lett., 

vol. 21, no. 2, pp. 106–108, 2009. 

[22] P. Runge, K. Petermann, W. Brinker, M. Schlak, and B. Sartorius, “Supercontinuum 

generating in ultralong SOAs—Theory and experiment,” 

in Proc. ECOC, Vienna, Austria, 2009. 

Patrick Runge was born in Berlin, Germany, in 1979. He received the Dipl.-Ing. 

degree in computer science from the Technische Universität Berlin, Germany, 

in 2005. 

From 2005 to 2007 he was with Hymite GmbH, where he was involved in 

the design and measurement of optoelectronic packages for optical communication. 

In 2007, he returned to the Technische Universität Berlin to pursue the 

Ph.D. degree, where he is currently researching the physics and applications of 

ultralong semiconductor optical amplifiers. 

Christian-Alexander Bunge (S’01–A’02–M’03) was born in Berlin, Germany, 

in 1973. He received the Dipl.-Ing. degree in electrical engineering from university 

of technology Berlin (TU Berlin) in 1999. Until 2002, he was with TU 

Berlin’s photonics group of Prof. Petermann, where he received his Ph.D. 

From 2002 to 2004, he was with the Polymer Optical Fiber Application 

Center (POF-AC) in Nuremberg, where he was responsible for fibre modelling 

and short-haul systems design. In 2004, he joined the TU Berlin as senior scientist, 

working on high-speed optical transmission systems, studying nonlinear 

optics, and modelling of optical components. In 2009, he became professor 

at Deutsche Telekom’s university for telecommunication in Leipzig, where 

his main research interests are short-haul and access transmission techniques, 

multimode glass and polymer-fibre links, and signal processing. 

Klaus Petermann (M’76–SM’85–F’09) was born in Mannheim, Germany in 

1951. He received the Dipl.-Ing. degree in 1974 and the Dr.-Ing. degree in 1976, 

both in electrical engineering from the Technische Universität Braunschweig, 

Germany. 

From 1974 to 1976, he was a Research Associate at the Institut für Hochfrequenztechnik, 

Technische Universität Braunschweig, where he worked on 

optical waveguide theory. From 1977 to 1983, he was with AEG-Telefunken, 

Forschungsinstitut, Ulm, Germany, where he was engaged in research work 

on semiconductor lasers, optical fibers, and optical fiber sensors. In 1983, he 

became a Full Professor at the Technische Universität Berlin, Germany. His 

current research interests include optical fiber communications and integrated 

optics. 

Dr. Petermann is a member of the Berlin-Brandenburg Academy of Science. 

He is the recipient of the Leibniz Award from the Deutsche Forschungsgemeinschaft 

in 1993, and the Distinguished Lecturer Award from the Laser and 

Electro-Optics Society within the IEEE in 1999/2000. From 1999 to 2004, he 

was an Associate Editor of the IEEE PHOTONICS TECHNOLOGY LETTERS. From 

1996 to 2004, he was a member of the board of the Verband Der Elektrotechnik 

Elektronik Informationstechnik (VDE). From 2004 to 2006, he was the Vice 

President for research at the Technische Universität Berlin.

All-Optical Wavelength Conversion With Extinction Ratio ...

Create successful ePaper yourself

Delete template?

Save as template?