IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 19, NO.

12, JUNE 15, 2007 931

Tunable Terahertz Signal Generation by Chirped

Pulse Photomixing
Johan Stigwall, Member, IEEE, and Andreas Wiberg, Student Member, IEEE

Abstract—Tunable signal generation in the gigahertz to

terahertz range was demonstrated by photomixing of a chirped
optical pulse and its time-delayed replica. The chirped pulses
were created by chromatically dispersing femtosecond pulses
from a mode-locked fiber ring laser, and frequency tuning was
achieved by adjusting the time delay and/or the amount of group
velocity dispersion. It is shown that the signal phase is locked to
the femtosecond-laser pulses and that an extinction ratio close to
100% can be obtained. Analytical calculations on the signal chirp
introduced by higher order dispersion are also presented and
compared with the experimental data.
Index Terms—Fiber optics, microwave photonics, photomixing,
submillimeter-wave generation. Fig. 1. Time-frequency diagram illustrating the principle of chirped pulse
photomixing.

I. INTRODUCTION
single-mode fiber. Thus, a much larger dispersion could be ob-

T HE generation of millimeter- and submillimeter-wave

signals using photonic techniques has received a lot of
interest thanks to the extremely wide bandwidth available using
tained, allowing the creation of longer, lower frequency, and po-
tentially more narrowband millimeter (mm)-wave pulses. Since
the system operates in the erbium-band, it is also compatible
optical components. Methods based on photonic frequency- with the wide range of telecom components available and the
multiplication of lower frequency electrical signals [1] as well signal can furthermore be distributed over large distances using
as direct signal generation by, for instance, mixing of two dispersion-compensated fiber links.
independent tunable lasers [2], have previously been studied,
where the former offers no photonic tunability but low phase
noise if a high-quality electronic oscillator is used, and the II. CHIRPED-PULSE MIXING (CPM)
latter offers extremely wide tunability while having a much The tone at frequency generated through CPM arises from
larger linewidth which is limited by the frequency stability of the difference frequency (DF) between a chirped pulse and its
the lasers. In this letter, we present the results from experiments -time-shifted replica (see Fig. 1)
on a setup where a chirped pulse is mixed with its time-delayed
replica to create a widely tunable pulsed signal at frequencies
(1)
in the gigahertz (GHz) to terahertz (THz) range with a very
stable phase (locked the pulse envelope).
Mixing of chirped pulses from Ti : sapphire lasers has previ- where is the instantaneous optical frequency of a chromat-
ously been used for generation of THz radiation [3]. The chirp ically dispersed ultrashort pulse. Neglecting dispersion terms
was then obtained by using a free-space grating pair, and the above the third order, it can be written
time-shift was introduced by a Michelson interferometer. It was
shown that the pulses had a relatively small linewidth, but the
attainable dispersion and hence pulsewidth using that setup was, (2)
however, limited by the physical size of the diffraction gratings.
In this letter, we use a fundamentally mode-locked fiber-laser
where is the GVD parameter, is
with a center wavelength of 1.55 m, and instead of free space
the third-order dispersion parameter, is the propagation dis-
diffraction gratings, we use a chirped fiber Bragg grating (FBG) tance, and the copropagating time coordinate (relative to the
or group velocity dispersion (GVD) in multiple kilometers of timing of the frequency ).
After Taylor-expansion of the square root in (2) up to the
Manuscript received November 13, 2006; revised March 20, 2007. quadratic term, we obtain
The authors are with the Photonics Laboratory at Chalmers University of
Technology, SE-412 96, Göteborg, Sweden (e-mail: johan.stigwall@chalmers.
se).
(3)
Digital Object Identifier 10.1109/LPT.2007.898872

1041-1135/$25.00 © 2007 IEEE

932 IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 19, NO. 12, JUNE 15, 2007

Fig. 2. Experimental setup.

and it is clear that the DF changes approximately linearly over

time as

(4)
Fig. 3. Zoomed in oscilloscope trace from Experiment A. Horizonal time
scale: 10 ps/div.
with a center frequency of

(5)

where is the dispersion parameter. The fre-

quency chirp rate is further (assuming a negative )

(6)

In practice, the dispersion slope parameter is often used in-

stead of [4]

(7) Fig. 4. Time-frequency “spectrogram” showing the frequency chirp of the DF

pulses generated in Experiment A. Contour lines are equally spaced in electrical
signal power.
Note that the spectral width of the pulse does not influence
, but it determines the duration of the DF pulse as
A. Experiment A
(8)
In the first configuration, Figs. 3 and 4, 9.1 km of standard
single-mode fiber were used as the dispersion element and the
III. EXPERIMENTAL SETUP DGD was set to 50 ps. The DF signal was received using a pin-
photodiode connected to a sampling oscilloscope, and the center
The experimental setup that was used is illustrated in Fig. 2.
frequency was measured to be 41.5 GHz (Figs. 3 and 4) and the
A pulse train from a passively mode-locked fiber laser (Thor-
center wavelength nm. This implies a dispersion
labs FSL1010) with a full-width at half-maximum pulsewidth
parameter of ps/(nm km). The fact that a sampling
of 125 fs, center wavelength of 1550 nm, 3-dB bandwidth of
oscilloscope could be used shows that the DF signal phase is
47 nm, and a repetition rate of 80 MHz was dispersed by either
locked to the pulse envelope.
a spool of single-mode fiber or a chirped FBG. The time-shift
As can be seen in the time-frequency spectrogram (Fig. 4),
was then created by launching equal amounts of light into the
the DF signal was chirped, and the chirp rate was measured to
two axes of an adjustable differential group delay (DGD) el-
be 1.27 GHz/ns (dashed line in Fig. 4). This corresponds to
ement followed by a polarization controller and a polarizing
a dispersion slope parameter of ps/(nm km) [(7)
beam splitter (PBS) to polarize the light at 45 with respect
and (8)]. A parameter fit of (1) and (2) was also performed for
to the optical axes of the DGD element. The DGD element had
comparison, yielding the same and parameters.
a delay range of 125 to 125 ps. To the outputs of the PBS,
an optical spectrum analyzer, a sampling oscilloscope, and/or
an autocorrelator was connected. B. Experiment B
In the second experiment (Fig. 5), we used the same setup to
IV. RESULTS measure the dispersion of a chirped FBG. The DF signal had a
The above described setup was tested with different pa- center frequency of 50.0 GHz at 8.6-ps DGD, indicating a dis-
rameters and at varying difference frequencies ranging from persion of 21 ps/nm (the sign of the dispersion was obtained
sub-GHz to 0.9 THz at the most. Here, the results from three by comparing the pulse envelope with the optical power spec-
different configurations are presented. trum). The DF signal chirp rate was approximately 5 GHz/ns,
STIGWALL AND WIBERG: TUNABLE THz SIGNAL GENERATION BY CHIRPED PULSE PHOTOMIXING 933

has to be minimized, and the exact DF tuned so that the signal

phase in adjacent CPM-pulses match. The phase noise of the
signal could then be very low since the temporal jitter of the DF
signal is inherited directly from the mode-locked femtosecond
laser (free-space mode-locked lasers with temporal jitter values
below one femtosecond have been demonstrated [8]).
In some applications such as mm-wave imaging, a very broad
spectrum is, however, desired in order to enhance the resolu-
tion. In those cases, a CPM system with enhanced dispersion
could be used to create DF signals with several tens of GHz chirp
Fig. 5. Oscilloscope trace from Experiment B (50 GHz). range.
CPM in “pulsed mode” could also be useful for the characteri-
zation of, e.g., high-speed photodetectors or photonic analog-to-
digital converters. Since the signal phase is locked to the en-
velope of the pulses, it is possible to use a sampling oscillo-
scope for the analysis, and the modulation depth is always nearly
100% in an optimized setup, irrespective of the DF signal fre-
quency. Thus, the high-frequency attenuation can be calculated
directly from the reduction in detected modulation depth.

VI. CONCLUSION
Fig. 6. Autocorrelator trace from Experiment C (725 GHz). The first experimental results using a novel fiber-based
method for the generation of high-frequency signals by means
of chirped pulse photomixing were presented. The experiments
indicating a dispersion slope of ps/nm . From show excellent modulation depth and good phase stability, and
Fig. 5, it is clear that the modulation depth is very large, here the frequency of the signal can easily be tuned by changing
limited by the 50-GHz bandwidths of the photodetector and the time-delay between the chirped pulse and its replica or by
the sampling oscilloscope. changing the amount of GVD. Higher order dispersion results
in signal frequency chirp, and we have presented analytical
C. Experiment C expressions linking the GVD and the dispersion slope to the
Using the same FBG as in Experiment B but with 125-ps central DF and the chirp in DF.
DGD, a signal at 725 GHz was created and measured using auto-
correlation (Fig. 6). Note that the theoretical maximum autocor- REFERENCES
relation contrast ratio of a sinusoidal signal is 3 : 1. In this exper-
[1] A. Wiberg, P. P. Millan, M. Andres, and P. Hedekvist, “Mi-
iment, we obtained an autocorrelation contrast ratio of 2.4 : 1, crowave-photonic frequency multiplication utilizing optical four-wave
which corresponds to a modulation depth of 95%. mixing and fiber Bragg gratings,” J. Lightw. Technol., vol. 24, no. 1,
pp. 329–334, Jan. 2006.
[2] H. Ito, T. Furuta, F. Nakajima, K. Yoshino, and T. Ishibashi, “Photonic
V. DISCUSSION generation of continuous thz wave using uni-traveling-carrier photo-
diode,” J. Lightw. Technol., vol. 23, no. 12, pp. 4016–4021, Dec. 2005.
The here demonstrated method of DF generation through [3] A. S. Weling and T. F. Heinz, “Enhancement in the spectral irradiance
CPM could either be used for signal generation, for device of photoconducting terahertz emitters by chirped-pulse mixing,” J. Opt.
Soc. Amer. B, vol. 16, no. 9, p. 1455, 1999.
characterization, for generation of THz radiation, or for char- [4] G. Agrawal, Fiber-Optic Communication Systems. New York: Wiley,
acterization of dispersion. The method is particularly well 2002.
suited for quick tweaking of, e.g., the temperature gradient [5] Y. Han, O. Boyraz, and B. Jalali, “Ultrawide-band photonic
time-stretch A/D converter employing phase diversity,” IEEE Trans.
along a chirped FBG in order to flatten the dispersion curve Microw. Theory Tech., vol. 53, no. 4, pt. 1, pp. 1404–1408, Apr. 2005.
(i.e., minimize or ). is directly proportional to the [6] J. Stigwall and S. Galt, “Non-linear signal retrieval in wideband
spectral width of the DF signal. Thus, an electrical spectrum photonic time-stretch systems using the gerchberg-saxton algo-
rithm,” in Proc. Int. Topical Meeting on Microwave Photonics, 2006
analyzer can be used to monitor both and in “real time.” [Online]. Available: http://ieeexplore.ieee.org/xpls/abs_all.jsp?is-
Dispersion-flattened chirped FBGs are useful for instance in number=4127586&arnumber=4153789&count=78&index=26, Poster
photonic time-stretch systems [5], especially when considering 26
[7] J. Stigwall and S. Galt, “Signal reconstruction by phase retrieval and
advanced postprocessing methods as in [6] and [7]. optical back-propagation in phase-diverse photonic time-stretch sys-
For high-frequency signal generation, it would in many cases tems,” J. Lightw. Technol., accepted for publication.
be preferred that the output signal be a continuous wave. This [8] T. Schibli, J. Kim, O. Kuzucu, J. Gopinath, S. Tandon, G. Petrich, L.
Kolodziejski, J. Fujimoto, E. Ippen, and F. Kaertner, “Attosecond ac-
could possibly be achieved by using enough dispersion that ad- tive synchronization of passively mode-locked lasers by balanced cross
jacent pulses start to overlap. Higher order dispersion obviously correlation,” Opt. Lett., vol. 28, no. 11, pp. 947–949, 2003.