JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 21, NOVEMBER 1, 2013 3395
Sub-Nyquist Sampled Analog-to-Digital Conversion
Based on Photonic Time Stretch and Compressive
Sensing With Optical Random Mixing
Ying Chen, Hao Chi, Member, IEEE, Tao Jin, Shilie Zheng, Xiaofeng Jin, and Xianmin Zhang, Member, IEEE
Abstract—An approach to realizing wideband analog-to-digital
conversion based on the techniques of photonic time stretch (PTS)
and compressive sensing (CS) is proposed. In the system, a mul-
titone signal within a wide bandwidth (spectrally sparse) signal
is slowed down in the time domain by a photonic time stretcher.
The stretched signal is then down-sampled and reconstructed by a
random-demodulator-based CS scheme, in which random mixing
is realized in an optical domain. Thanks to the techniques of PTS
and CS, wideband spectrally sparse signals can be acquired with a
sampling rate far below the Nyquist rate of the original signal. The
optical random mixing applied in the system has the advantages of
lower distortions and larger bandwidth compared to its electrical
counterpart. In order to construct a Gaussian measurement ma-
trix with zero mean, balanced detection is applied after the optical
mixer. In addition, in order to eliminate the dc component and
the even-order harmonics of the stretched signal, we propose to
use balanced PTS technique in the system. We demonstrate that
a system with a time stretch factor 20 and a compression factor 4
can effectively acquire a spectrally sparse wideband signal, which
means a sampling rate as low as 1/80 of the Nyquist rate.
Index Terms—Analog-to-digital converter (ADC), balanced de-
tection, compressive sensing (CS), optical mixing, photonic time
stretch (PTS), sub-Nyquist sampling.
I. INTRODUCTION
A
NALOG-TO-DIGITAL CONVERTERS (ADCs) play a
major role in modern communications and electronic sys-
tems as the signal processing and transmission in a digital
domain offers flexibility, reconfigurability, and higher perfor-
mance. Fast ADCs are highly demanded in many applications
such as widebandwireless communications, advanced radar sys-
tem, radio astronomy, and ultrawideband signal measurement.
However, the improvement in ADCs largely lags behind that of
digital signal processing (DSP) due to the limits in timing is-
Manuscript received May 24, 2013; revised August 4, 2013 and September 9,
2013; accepted September 12, 2013. Date of publication September 17, 2013;
date of current version October 11, 2013. This work was supported in part by the
National BasicResearchProgram of China under 973 Program 2012CB315703),
in part by the National Natural Science Foundation of China under Grants NSFC
61177003, 61171036, and 61275027, in part by the Program for New Century
Excellent Talents in University under Grant NCET-10-0723, and in part by
the State Key Laboratory of Advanced Optical Communication Systems and
Networks.
The authors are with the Department of Information Science and Elec-
tronic Engineering, Zhejiang University, Hangzhou 310027, China (e-mail:
chihao@zju.edu.cn).
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/JLT.2013.2282088
sues and other inherent device technologies. The use of photonic
technologies to improve electronic ADCs has attracted much in-
terest since 1970s due to the advantages of high time-bandwidth
product and immunity to electromagnetic interference offered
by photonics [1]. By now, some photonics-assisted ADC tech-
nologies have been proposed and demonstrated [1], [2], which
include the ADCs based on Mach–Zehnder modulators (MZMs)
or interferometers [3]–[9], the photonic time-stretched ADCs
[10]–[13], and the photonic quantizers based on optical nonlin-
earities [14]–[16].
An alternative approach is to use photonic technologies in
sub-Nyquist sampled ADCs to alleviate the demand on the
sampling rate, which includes the techniques based on optical
down-conversion [17], [18], optical undersampling [19]–[21],
nonuniform optical sampling [22], [23], and multirate optical
sampling [24]. The optical down-conversion is realized by mix-
ing a high speed signal with a microwave local oscillator in
the optical domain using an MZM. The down-converted signal
can be digitized and processed by an electronic ADC with a
relatively lower sampling rate [17], [18]. The technique of opti-
cal undersampling employs a pulsed laser source to uniformly
sample the input bandpass signal at a sub-Nyquist rate, and
then a low frequency alias can be filtered out and digitized with
much lower sampling rate [19]–[21]. The techniques of optical
down-conversion and optical undersampling are effective when
processing signals with high carrier frequency and narrow band-
width. The scheme of nonuniform optical sampling [22], [23]
is to acquire a multitone and band-limited signal through pseu-
dorandom sampling, whose average rate can be much less than
the Nyquist rate. The major limitation of the nonuniform optical
sampler lies in its high sensitivity to timing jitter and relatively
low signal-to-noise ratio (SNR) of the acquired signal. In the
scheme of a multirate optical sampled ADC, multiband sig-
nals are sampled by an array of asynchronous optical pulsed
sources with different sub-Nyquist sampling rates with the help
of the wavelength division multiplexing [24]. The multirate op-
tical sampler involves multiple pulsed laser sources and optical
links, which make the system complicated.
Another technique to realize optical sub-Nyquist sampling
is the photonics-assisted compressive sensing (CS). According
to the theory of CS, a wideband spectrally sparse signal can
be recovered from a number of sampling points that is far less
than the Nyquist sampling points. The main idea of CS is to
employ nonadaptive linear projections to preserve the informa-
tion of the signal with small number of samples. As a result,
the input signal can be captured at a sampling rate far below
0733-8724 © 2013 IEEE