A time-domain photonic arbitrary waveform

generator
Jinxin Liao, He Wen, Xiaoping Zheng,* Hanyi Zhang, and Bingkun Zhou
State Key Laboratory on Integrated Optoelectronics / Tsinghua National Laboratory for Information Science and
Technology, Department of Electronic Engineering, Tsinghua University, Beijing, 100084, China
*[email protected]

Abstract: A time domain photonic arbitrary waveform generator (PAWG)

scheme based on multi-wavelength optical differential quadrature phase
shift keying modulation in combination with differential detection is
proposed and experimentally demonstrated. The time domain method shows
advantages of large time-bandwidth product, good flexibility, fast
waveform refreshing rate, and high waveform quality over the frequency
domain method. In contrast with other proposed time domain PAWGs or
photonic digital-to-analog converters, our PAWG proposal shows a greater
dynamic range and a larger noise margin due to its bipolar output, and
possesses good scalabilities both in resolution and sampling rate. Assisted
with the integration technology, this PAWG presents a good prospect for
broad range practical applications in future.
©2012 Optical Society of America
OCIS codes: (070.2025) Discrete optical signal processing; (070.6020) Continuous optical
signal processing; (060.5625) Radio frequency photonics; (060.4510) Optical communications.

References and links

1. J. Chou, Y. Han, and B. Jalali, “Adaptive RF-photonic arbitrary waveform generator,” IEEE Photon. Technol.
Lett. 15(4), 581–583 (2003).
2. I. S. Lin, J. D. McKinney, and A. M. Weiner, “Photonic synthesis of broadband microwave arbitrary waveforms
applicable to ultra-wideband communication,” IEEE Micro. Wirel. Compon. Lett. 15(4), 226–228 (2005).
3. V. Torres-Company, J. Lancis, P. Andrés, and L. R. Chen, “Reconfigurable RF-waveform generation based on
incoherent-filter design,” J. Lightwave Technol. 26(15), 2476–2483 (2008).
4. C. B. Huang, D. E. Leaird, and A. M. Weiner, “Time-multiplexed photonically enabled radio-frequency arbitrary
waveform generation with 100 ps transitions,” Opt. Lett. 32(22), 3242–3244 (2007).
5. R. P. Scott, N. K. Fontaine, C. Yang, D. J. Geisler, K. Okamoto, J. P. Heritage, and S. J. B. Yoo, “Rapid
updating of optical arbitrary waveforms via time-domain multiplexing,” Opt. Lett. 33(10), 1068–1070 (2008).
6. V. Torres-Company, A. J. Metcalf, D. Leaird, and A. M. Weiner, “Multichannel radio-frequency arbitrary
waveform generation based on multiwavelength comb switching and 2-D line-by-line pulse shaping,” IEEE
Photon. Technol. Lett. 24(11), 891–893 (2012).
7. V. Torres-Company and L. R. Chen, “Radio-frequency waveform generator with time-multiplexing capabilities
based on multi-wavelength pulse compression,” Opt. Express 17(25), 22553–22565 (2009).
8. R. P. Scott, N. K. Fontaine, J. P. Heritage, and S. J. B. Yoo, “Dynamic optical arbitrary waveform generation and
measurement,” Opt. Express 18(18), 18655–18670 (2010).
9. C. H. Lee, Microwave Photonics (Boca Raton, FL: CRC Press, 2007), Chap. 6.
10. P. K. Kondratko, A. Leven, Y. K. Chen, J. Lin, U. V. Koc, K. Y. Tu, and J. Lee, “12.5-GHz optically sampled
interference-based photonic arbitrary waveform generator,” IEEE Photon. Technol. Lett. 17(12), 2727–2729
(2005).
11. Y. K. Chen, A. Leven, T. Hu, N. Weimann, K. Tu, V. Houtsma, R. Kopf, and A. Tate, “Photonic generation of
microwave and millimeter-wave arbitrary waveforms,” in IEEE Lasers and Electro-Optics Society, Annual
Meeting, 419–420 (2008).
12. A. Yacoubian and P. K. Das, “Digital-to-analog conversion using electrooptic modulators,” IEEE Photon.
Technol. Lett. 15(1), 117–119 (2003).
13. X. Yu, K. Wang, X. Zheng, and H. Zhang, “Incoherent photonic digital-to-analogue converter based on
broadband optical source,” Electron. Lett. 43(19), 1044–1045 (2007).
14. Y. Peng, H. Zhang, Y. Zhang, and M. Yao, “Photonic digital-to-analog converter based on summing of serial
weighted multiwavelength Pulses,” IEEE Photon. Technol. Lett. 20(24), 2135–2137 (2008).
15. T. Nishitani, T. Konishi, H. Furukawa, and K. Itoh, “All-optical digital-to-analog conversion using pulse pattern
recognition based on optical correlation processing,” Opt. Express 13(25), 10310–10315 (2005).
16. S. Oda and A. Maruta, “All-optical digital-to-analog conversion using nonlinear optical loop mirrors,” IEEE
Photon. Technol. Lett. 18(5), 703–705 (2006).

#165140 - $15.00 USD Received 21 Mar 2012; revised 7 May 2012; accepted 12 May 2012; published 18 May 2012
(C) 2012 OSA 21 May 2012 / Vol. 20, No. 11 / OPTICS EXPRESS 12631
17. A. H. Gnauck and P. J. Winzer, “Optical phase-shift-keyed transmission,” J. Lightwave Technol. 23(1), 115–130
(2005).
18. J. Liao, H. Wen, X. Zheng, H. Zhang, Y. Guo, and B. Zhou, “A 2N-bit bipolar photonic digital-to-analog
converter based on multi-wavelength optical DQPSK modulation in combination with balanced detection,” in
Optical Fiber Communication Conference, Technical Digest (CD) (Optical Society of America, 2012), paper
JW2A.78.
19. IEEE standard for terminology and test methods for analog-to-digital converters, IEEE Std 1241–2000 (2001).
20. Y. M. Greshishchev, D. Pollex, S.-C. Wang, M. Besson, P. Flemeke, S. Szilagyi, J. Aguirre, C. Falt, N. Ben-
Hamida, R. Gibbins, and P. Schvan, “A 56GS/S 6b DAC in 65nm CMOS with 256×6b memory,” in proceedings
of IEEE Solid-State Circuits Conference, pp.194–196, 2011.
21. S. Savory and A. Hadjifotiou, “Laser linewidth requirements for optical DQPSK systems,” IEEE Photon.
Technol. Lett. 16(3), 930–932 (2004).
22. S. W. Corzine, P. Evans, M. Fisher, J. Gheorma, M. Kato, V. Dominic, P. Samra, A. Nilsson, J. Rahn, I.
Lyubomirsky, A. Dentai, P. Studenkov, M. Missey, D. Lambert, A. Spannagel, R. Muthiah, R. Salvatore, S.
Murthy, E. Strzelecka, J. L. Pleumeekers, A. Chen, R. Schneider, R. Nagarajan, M. Ziari, J. Stewart, C. H.
Joyner, F. Kish, and D. F. Welch, “Large-scale InP transmitter PICs for PM-DQPSK fiber transmission
systems,” IEEE Photon. Technol. Lett. 22(14), 1015–1017 (2010).
23. M. Kroh, G. Unterbörsch, G. Tsianos, R. Ziegler, A. G. Steffan, H. G. Bach, J. Kreissl, R. Kunkel, G. G.
Mekonnen, W. Rehbein, D. Schmidt, R. Ludwig, K. Petermann, J. Bruns, T. Mitze, K. Voigt, and L.
Zimmermann, “Hybrid integrated 40 Gb/s DPSK receiver on SOI,” in Optical Fiber Communication Conference,
Technical Digest (CD) (Optical Society of America, 2009), paper OMK3.

1. Introduction
High speed arbitrary waveform generation has grown as an important area which finds
extensive commercial and military applications, such as testing, large capacity
communications, high speed signal processing, and military defense electronics etc. As the
optical domain processing technology possesses inherent advantages of large bandwidth, high
frequency and immunity to electromagnetic interferences which are difficult for the all-
electronic approach to achieve, the photonic arbitrary waveform generator (PAWG) has
received significant attention. At present, most of the existing PAWG schemes employ the
frequency-domain method known as a fourier-transform based technique. This method allows
the amplitude and phase of discrete optical spectral lines to be independently controlled and
enables arbitrary waveform to be synthesized in a time aperture [1–8]. In these frequency-
domain PAWG apparatus [1–3], a grating followed by a spatial light modulator (SLM) with a
large number of spectral controlled elements is usually implemented as an optical pulse
shaper. It shows good performance in generating complex radio frequency waveform but with
two significant drawbacks. One is that the waveform refreshing time is limited by the
response time of the SLM (~10ms). This problem can be released by utilizing the time
multiplexing technique [4–7] or replacing the SLM with a much more short response time
shaper, such as the electro-optical modulator array [8]. The other one significant drawback is
that the maximum time aperture of the output temporal waveform (~2ns) directly associated
with the lowest attainable frequency, is determined by the minimum resolution (~1GHz) of
the optical pulse shaper employed in the system [9]. The time multiplexing technique can
increase the time aperture but only to some extent. The reason is that large numbers of the
multiple branches in time-multiplexing based schemes are required to achieve a PAWG
capable of fast and flexible waveform refreshing along with infinite waveform time aperture.
The number corresponding to the ratio of the waveform refreshing time and the waveform
time aperture is tremendous (~105). It is very difficult to expand so many multiple branches to
meet the requirement.
However, the PAWG employing the time-domain method which is similar to the
traditional electrical AWG does not have these problems [10, 11]. The time-domain PAWG
generates samples to approximate the target waveform at each temporal point. The waveform
time aperture is determined by the product of the memory depth and the sampling interval,
which can be long enough to several milliseconds. It can be much longer with the real-time
sequencing technique. Thus, the time-domain method can provide a larger time-bandwidth
product compared with the frequency-domain method. Furthermore, the flexibility, the quality
and the refreshing rate of the time-domain arbitrary waveform generation are excellent. The
time-domain PAWG can be regarded as a photonic digital-to-analog converter (PDAC). But

#165140 - $15.00 USD Received 21 Mar 2012; revised 7 May 2012; accepted 12 May 2012; published 18 May 2012
(C) 2012 OSA 21 May 2012 / Vol. 20, No. 11 / OPTICS EXPRESS 12632
so far, all the proposed PDAC are unipolar, with only positive value [10–16]. From the
perspective of dynamic range and noise margin, a bipolar PDAC (BPDAC) may have an
advantage over the unipolar PDAC and can achieve higher resolution more easily. This is
illustrated by comparing the constellation diagram of the bipolar output against that of the
unipolar output [17].
In [18], we have recently put forward a 2N-bit BPDAC based on multi-wavelength optical
differential quadrature phase shift keying (ODQPSK) modulation in combination with
differential detection. In this paper, this 2N-bit BPDAC is applied to achieve a time domain
PAWG with good flexibility, fast waveform refreshing rate and high waveform quality. This
PAWG shows three attractive features. The first is that the bipolar output brings about a
greater dynamic range and a larger noise margin compared with the unipolar output of other
proposed time domain PAWGs or PDACs. The second is that it has good scalabilities both in
the sample resolution and the sampling rate. The sample resolution can be increased by using
multi-wavelength laser array, and the sampling rate can be increased by using time-
multiplexing technique. The third is that it consisting of a multi-wavelength DQPSK
transmitter and a DPSK receiver is fully compatible with the mature popular DQPSK system.
It can be easily integrated on a chip or constructed with the available photonic integrated
circuit (PIC). We set up a proof-of-principle PAWG experiment with 4-bit resolution at
2.5GS/s to demonstrate this scheme. The generated arbitrary waveforms are very close to the
ideal waveforms. The average spurious-free dynamic range (SFDR) results of single tone and
two tone signals are 31.4dB and 27.9dB, respectively. We also experimentally verify the time-
multiplexing PAWG with multiple sampling rates. Furthermore, we stress some specifications
of the essential components in this PAWG, and give a summary on this PAWG proposal.
2. Operation principle
Figure 1(a) shows the architecture of the PAWG with 2N-bit resolution. It is composed of two
main parts, the multi-wavelength ODQPSK modulation part and the differential demodulation
part. In the multi-wavelength ODQPSK modulation part, N channels of independent light with
different wavelengths pass through N dual parallel MZMs (DP-MZM) respectively. Each of
the DP-MZM is driven by two channels of non-return-zero (NRZ) digital signal, generating
one ODQPSK signal. Combined the N branches together, the multi-wavelength ODQPSK
signal is delivered to the demodulation part where a 1-bit delay-time interferometer (DI) acts
as a differential demodulator and its two outputs are balanced detected. We first illustrate how
the samples with 2-bit resolution are generated by using single wavelength laser, then we
detail the implementation of the PAWG with 2N-bit resolution and present the time-
multiplexing technique assisted PAWG.
(a) (b)
I1
D2n D2 n 1 D4 D3 D2 D1 18.4 Re{E}
Vout
DP-MZM1


P
1 11
90 precoder
Q1
I2 DP-MZM2 Qn I n Q2 I 2 Q1 I1 10
4P
2 Im{E}
……
t
90
Q2
T Vout 01
In -
n -1
4 P
DP-MZMn

n LPF
00
90 DI BD
Qn

Fig. 1. (a)Architecture of the PAWG with 2N-bit resolution. (b) The changing form of the
analog output Vout varies with the phase shift ∆φ in the 2-bit PAWG. When ∆φ18.4°, Vout
presents bipolar four-equally-spaced discrete values.

#165140 - $15.00 USD Received 21 Mar 2012; revised 7 May 2012; accepted 12 May 2012; published 18 May 2012
(C) 2012 OSA 21 May 2012 / Vol. 20, No. 11 / OPTICS EXPRESS 12633
2.1 Principle of PAWG with 2-bit resolution based on ODQPSK modulation in combination
with differential detection
In the 2-bit PAWG using a single wavelength, the ODQPSK signal can be expressed as
E  t  ei , where    4,3 4,5 4,7 4 . After differential detection, the amplitude of
the output sample Vout is given by:

Vout  E1  t  T  ei1  E2  t  ei2 ei  E1  t  T  ei1  E2  t  ei2 ei

2 2

(1)
 4 E1  t  T  E2*  t  cos      ,    2  1   0,  2,  ,3 2

where E1  t  T  ei1 is the ODQPSK signal delayed by 1 bit period in the longer arm of DI,
E2  t  ei2 ei is the ODQPSK signal in the short arm of DI, ∆φ denotes the phase shift
between the two arms, and T is the digital bit duration. From Eq. (1), it can be inferred that
Vout varies with the phase shift ∆φ in a form of cosine function, where ∆θ determines the
initial phase of the function. Figure 1(b) illustrates this phenomenon. As long as the phase
shift ∆φ meets the condition cos  sin   1 3 or3 , Vout presents bipolar four-equally-
spaced discrete values, In Fig. 1(b),   18.4 is taken for example. As Vout is directly
associated with ∆θ which denotes the changing form between the adjacent input digital codes,
a precoder is required to map the input digital codes to the bipolar four-equally-spaced
discrete values. The coding algorithm of the precoder is given in Eq. (2), and the one-to-one
mapping between the input and the output is shown in Table 1. Thus, a 2-bit PAWG is
achieved. Profiting from the optical phase modulation combined with balanced detection, this
PAWG generates bipolar samples, and shows advantages of a greater dynamic range and a
larger noise margin over the unipolar PAWG or PDAC based on the intensity modulation
together with direct detection [12–14]. The reason is that the minimal distance between
adjacent two symbols of the phase modulated signal is larger than that of the intensity
modulated signal. This advantage is analogous to the 3dB optical signal-to-noise ratio
(OSNR) advantage that the phase modulation format owns over the intensity modulation
format in optical communication system.

 I n  D2 n 1  D2 n   D2 n 1  I n -1    D2 n 1  D2 n    D2 n  Qn -1 

 (2)
Qn  D2 n 1  D2 n   D2 n  Qn -1    D2 n 1  D2 n    D2 n 1  I n -1 

Table 1. The One-To-One Mapping Between the Input and the Output
Input digital codes ∆θ Output analog value
00 0 0
01 3π/2 1
10 π/2 2
11 π 3

2.2 Implementation of PAWG with 2N-bit resolution

The resolution of this PAWG can be scaled from 2-bit to 2N-bit by using N channels of
incoherent light with different wavelengths, where the optical power ratio is set as
1: 4 : : 4N 1 . Similar weighted and summing methods have been reported in [12–14]. For
multi-wavelength laser array, there are some requirements on the frequency difference ∆f
between any two light waves that should be specified. First, ∆f should be an integral multiple
of the free spectral range (FSR) of DI which is equal to 1/T, to ensure that each wavelength
light wave obtains the same phase shift of ∆φ in DI, since the transmission function of the DI
is periodic and the phase shift ∆φ is directly associated with the lightwave frequency [17].

#165140 - $15.00 USD Received 21 Mar 2012; revised 7 May 2012; accepted 12 May 2012; published 18 May 2012
(C) 2012 OSA 21 May 2012 / Vol. 20, No. 11 / OPTICS EXPRESS 12634
Second, ∆f should be greater than the detector’s bandwidth to ensure that the beating noise of
any two light waves is removed. Figure 2 illustrates these requirements. The optical power
ratio of different wavelengths can also be realized by using a multi-wavelength laser array
with equal optical power of each wavelength followed by a special weighted coupler with
split ratio of 1: 4 : : 4N 1 .

4n -1 P

Transm. [dB]
Δφ
4P
P
1T 18.4 Δf
Frequency
Fig. 2. Requirements on the multi-wavelength laser array.

2.3 Time multiplexing PAWG with 2N-bit resolution

Time Multiplexing Unit (× m)
3 2
1
90
 1 0
13
2
90
2 12
8 6
3
8
m TMU1 0 4
m
90
WDM DEMUX

4n -1 P
4P TMU2
P
time- and wavelength-
interleaved short pulse T Output
generator -
 LPF
TMUn
DI BD

Fig. 3. Architecture of the time multiplexing PAWG with 2N-bit resolution.

With the optical domain time passive multiplexing method which has advantages of low
loss, stable delay time, and immunity to electromagnetic interferences etc, the multi-level
optical short pulse can be easily temporal interleaved to increase the sampling rate of this
PAWG. The architecture of the 2N-bit PAWG with m multiple expanded sampling rate is
presented in Fig. 3. In the apparatus, the multi-wavelength time-interleaved short pulse
generator with very small time jitter (<300fs) is based on phase modulating the multi-
wavelength pulse source and subsequent compressing the pulse in the dispersive medium [7].
The generated optical short pulse trains with different wavelengths are delivered to a
wavelength demultiplexer. The separated single wavelength short pulse train with repetition
of Fs-GS/s enters the time multiplexing unit (TMU) array, respectively. In the TMU, the
single wavelength short pulse train is spilt into m parallel DPMZMs and is DQPSK
modulated, respectively. The bit duration T of the NRZ digital driving signal is equal to 1/Fs.
By setting the corresponding time delay of the m channels as  , 2 , , m (m  T ) , the
repetition of the output DQPSK modulated pulse train is increased to m∙Fs-GS/s, which
means that the sampling rate is m multiple expanded. Then, by synchronizing and summing
all the N channels of DQPSK modulated short pulse with different wavelengths together and
setting the corresponding optical power ratio as 1: 4 : : 4N 1 , the sample resolution is scaled
to 2N-bit. Thus, a 2N-bit PAWG with m multiple expanded sampling rate is built. It can be
seen that in this m∙Fs-GS/s 2N-bit PAWG, 2∙m∙N channels of NRZ digital driving signal with
a lower bit rate of Fs-GS/s, m∙N DPMZMs are required.

#165140 - $15.00 USD Received 21 Mar 2012; revised 7 May 2012; accepted 12 May 2012; published 18 May 2012
(C) 2012 OSA 21 May 2012 / Vol. 20, No. 11 / OPTICS EXPRESS 12635
3. Results and discussions
I1
DP-MZM1
MZM

PC1 MZM 90
10GHz VOA
CW1 Q1
I2
DP-MZM2
MZM
AM PM TDCM
PC2 MZM 90 TTD
CW2 short pulse generator Q2

50:50
T
EDFA
-
Oscilloscope 
BD DI

Fig. 4. Experiment setup of a PAWG with 4-bit resolution at 2.5GS/s.

Limited by only one independent channel of 10Gbit/s signal, we just set up a 4-bit 2.5-
GS/s PAWG experiment as shown in Fig. 4 to verify the proposed 2N-bit PAWG. In the
experiment, two continuous wave (CW) lights with different wavelengths enter into two DP-
MZMs and are quadrature modulated respectively. The 2.5Gbit/s non-return-zero driven
digital signals are obtained by demultiplexing a 10Gbit/s digital signal from a pulse pattern
generator (PPG). The generated dual-wavelength ODQPSK signal is directed into a 1-bit DI
followed by a BD after being amplified. The DI is welded by two 2 × 2 couplers where the
delay time difference between the two arms is 400 ps. Its two output ports are balanced
detected, and the detected output electrical signal is sent to an oscilloscope for measuring.
Then, the entire system is calibrated as follows. First, switch one laser on only and regulate
∆φ around 18.4° to build a 2-bit PAWG. Second, switch the other laser on only and tune its
wavelength to accommodate the built 2-bit subsystem. Third, synchronize the two ODQPSK
signals with a tunable optical time delay (TTD) module and adjust the optical power ratio
between the two ODQPSK signals as 1:4 with a variable optical attenuator (VOA). After this,
the 4-bit PAWG at 2.5GS/s is built.
Programming the four channels of input digital bit sequences, we can synthesize arbitrary
waveform in the time domain. Figure 5(a) shows the generated raw waveforms of the
sawtooth wave (with 16 ramp), the triangle wave (with 16 ramp), the sinusoidal wave (f =
99.82MHz), and the two-tone wave (f1 = 399.22MHz / f2 = 439.14MHz). The burr of
waveform is a fundamental issue which is the result of the phase modulation together with
differential detection. More specifically, from Eq. (1), we know that only when ∆θ = 0 which
means that both the two channels of digital bits keep continuous in transition time, would the
E1  t  T  E2*  t  and the output analog amplitude remains unchanged in transition time. In
other three changing forms where at least one digital bit changes, the E1  t  T  E2*  t  would
depress in transition time, and the burr of waveform emerges. The lowpass filtering operation
can remove the burr but pay a cost of lowering the highest attainable frequency of the output
analog signal, as the 3dB cut-off frequency should be smaller than half of the sampling rate.
Fig. 5(b) shows the comparison between the corresponding waveforms after being filtered by
a software filter (3 order Bessel lowpass filter, f3dB cut-off1GHz) and the ideal waveforms. The
corresponding cross correlation coefficients are 0.9987, 0.9962, 0.9905, and 0.9982. It can be
seen that the generated waveforms are very close to the ideal waveforms.
To further evaluate the performance of this 4-bit PAWG, we sweep synthesize single-tone
signals and use the spectrum analyzer to measure the corresponding SFDR. Normal detector
without average operation is used. We also employed signal-to-noise and distortion ratio
(SINAD) test method shown in [19] along with (SINAD-6.02)/1.76 to compute the effective
number of bits (ENOB). The SFDR and ENOB results are shown in Fig. 6(a) and the average
SFDR value is 31.4dB. We also synthesize two-tone signals at three groups of frequency (f1 =

#165140 - $15.00 USD Received 21 Mar 2012; revised 7 May 2012; accepted 12 May 2012; published 18 May 2012
(C) 2012 OSA 21 May 2012 / Vol. 20, No. 11 / OPTICS EXPRESS 12636
598.82MHz / f2 = 658.72MHz; f1 = 608.72MHz / f2 = 648.80MHz; f1 = 399.22MHz / f2 =
439.14MHz). The spectrums are shown in Fig. 6(b) and the corresponding SFDR are 28.1dB,
27.3dB and 28.3dB with an average of 27.9dB. The spurious spectrum and the loss of the
ENOB mainly come from two aspects, the burr between samples and the fluctuation of
samples. The burr mainly brings the high frequency interference that can be relieved by
lowpass filtering. The fluctuation of samples originates from the nonideal characteristic (e.g.
drifting or distortion) of the amplified driven signal, which is concerned with the programmed
data pattern because of the AC coupled characteristic of the driver. By relieving the nonideal
characteristic of the amplified driven signal, the SFDR and ENOB results can be improved.

Fig. 5. (a)Generated raw waveforms of the sawtooth wave (with 16 ramp), the triangle wave
(with 16 ramp), the sinusoidal wave (f = 99.82MHz), and the two-tone wave (f1 = 399.22MHz /
f2 = 439.14MHz). (b)Comparison between the corresponding filtered waveforms (blue line)
and the ideal waveforms (red line).

#165140 - $15.00 USD Received 21 Mar 2012; revised 7 May 2012; accepted 12 May 2012; published 18 May 2012
(C) 2012 OSA 21 May 2012 / Vol. 20, No. 11 / OPTICS EXPRESS 12637
 Fig. 6. (a) SFDR and ENOB results. (b) Spectrum of two-tone signals at three groups of
frequency.

Adding an optical short pulse generator, we can transform the built 4-bit PAWG to a time-
multiplexing 20-GS/s 2-bit PAWG. The two parallel DP-MZMs can be regarded as a TMU( ×
2). In the optical short pulse generator, the CW light is modulated by an amplitude modulator
(AM), generating a pulse train with a full width of half maximum (FWHM) of 30ps and a
repetition rate of 10GHz. The pulse train is put into a phase modulator (PM) followed by a
tunable dispersion compensation module (TDCM). With high power phase modulation and
appropriate dispersion, the pulse train is compressed with FWHM of 12ps. Then the short
pulse train enters into the TMU( × 2). The entire experiment needs recalibration, including
regulating the TTD module to multiplex the pulse trains and adjusting the VOA to make the
two channels have the same optical power. When only one channel is used, the sampling rate
of the generated pulse train is 10GHz, as shown in Fig. 0.7(a). When both channels are used,
the sampling rate of the generated pulse train turns into 20GHz, as shown in Fig. 7(b). It can
be seen that the peaks of the pulse train appear bipolar four-equally-spaced values. Since the
bandwidth of commercial DP-MZM has achieved more than 40GHz, this PAWG can easily
realize 40GS/s by using higher speed driving digital signals, and surpass 80GS/s assisted by
the optical domain time multiplexing technique. Such a high speed is very difficult for the
current electronic AWG or DAC to achieve [20].

Fig. 7. (a) Generated pulse train at 10GHz when only one channel is used. (b) Generated pulse
train at 20GHz when two channels are used.

For this PAWG proposal, the performance depends on the stabilization of (∆θ + ∆φ).
Therefore, to meet this requirement, some specifications of the essential components should
be specified. Seen from Fig. 1(b), we know that the condition of cos  sin   1 3 or3
guarantees the perfect linearity of the digital-to-analog conversion, and a small phase detuning
of ∆φ in DI significantly affects the linearity since the inner sample value and the outer
sample value move in opposite direction with different rates of change. From Eq. (1), we
obtain the amplitude deviation ratio of discrete sample values as

dVout Vout   tan(   )  d   =18.4

(3)

#165140 - $15.00 USD Received 21 Mar 2012; revised 7 May 2012; accepted 12 May 2012; published 18 May 2012
(C) 2012 OSA 21 May 2012 / Vol. 20, No. 11 / OPTICS EXPRESS 12638
 It can be inferred that the inner sample value (∆φ = π/2 or 3π/2) is more vulnerable to the
phase detuning. We calculate that if the required amplitude deviation ratio is less than 5%, the
tolerable phase detuning is about ± 1°. Such a stringent requirement on phase shift
stabilization requires the application of DI have precise temperature control and athermal
package. Besides, the disturbance of ∆θ mainly resulting from the phase noise of the laser
source and the chirp occurred in the modulation process brings an undesirable affect on the
signal-to-noise ratio (SNR) of the analog output. However, the chirp can be reduced distinctly
by using common push-pull type MZMs. The impact of the phase noise related to the laser
linewidth can also be neglected by using narrow linewidth lasers, especially at much higher
sampling rate [21].
As discussed before, it can be seen that this PAWG is fully compatible with the mature
popular DQPSK system. This feature brings many benefits. A main one is that the greatly
developed photonic integrated circuit (PIC) technologies on the mature ODQPSK system
enable this PAWG to be integrated on a chip or constructed with available PICs. As reported
in [22] and [23], a differential detection receiver and a large-scale transmitter PIC consisting
of 10 tunable distributed-feed-back (DFB) lasers followed by 10 DP-MZMs corresponding to
the two main parts of this PAWG were implemented. The integration of this PAWG would
make it more stable and more practical.
4. Conclusion
We have put forward a time domain PAWG scheme based on multi-wavelength ODQPSK
modulation in combination with differential detection. Proof-of-principle experiments of 4-bit
PAWG at 2.5GS/s and 2-bit time-multiplexing PAWG at 20GS/s are set up to demonstrate
this PAWG scheme. The generated arbitrary waveform is very close to the ideal waveform.
The measured average SFDR of single-tone signals and two-tone signals are 31.4db and
27.9db, respectively. Compared with the frequency domain PAWG, this time domain PAWG
shows advantages of large time-domain product, good flexibility, fast waveform refreshing
rate and high waveform quality. In contrast with other proposed time domain PAWGs or
PDACs, this PAWG shows a greater dynamic range and a larger noise margin due to its
bipolar output, and has good scalabilities both in resolution and sampling rate. Furthermore,
the PIC technology has allowed this PAWG consisting of a multi-wavelength DQPSK
transmitter and a DPSK receiver to be integrated on a chip, and presents a good prospect of
this PAWG’s wide practical applications in future.
Acknowledgments
This work was supported in part by National Nature Science Foundation of China (NSFC)
under grant No. 60736003, 61025004, and 61032005; National Basic Research Program of
China under the grant No 2012CB315603-04; Foundation of the Key State Lab of Integrated
Optoelectronics under grant No. 2010KFB007; the Ph.D. Programs Foundation of Ministry of
Education of China under grant No. 20100002110039, and the Postdoctoral Science
Foundation of China under grant No. 2012M510442.

#165140 - $15.00 USD Received 21 Mar 2012; revised 7 May 2012; accepted 12 May 2012; published 18 May 2012
(C) 2012 OSA 21 May 2012 / Vol. 20, No. 11 / OPTICS EXPRESS 12639