Comparison of Homodyne and Intradyne Detection

for High-Order Modulation Schemes in Optical

Intersatellite Communication Systems

Semjon Schaefer, Werner Rosenkranz Mark Gregory

Chair for Communications
University of Kiel TESAT Spacecom
Kaiserstr. 2, D-24143 Kiel, Germany Gerberstraße 49, D-71522 Backnang, Germany
[email protected] [email protected]

Abstract—In order to increase the data rate and lower the to adjust the frequency and phase of the local oscillator (LO)
power consumption in satellite communication systems, optical to the incoming signal [3]. However, the OPLL hardware
laser terminals are an attractive alternative to the conventional complexity increases by increasing the modulation order.
RF systems. By using optical coherent detection, high receiver Therefore, the use of intradyne detection, where the local laser
sensitivity as well as full signal recovery is achieved. However, the runs un-synchronized and which is based on software-defined
local oscillator at the receiver requires a carrier recovery system. digital frequency offset compensation, seems to be a possible
In order to achieve homodyne detection, an optical phase-locked alternative for future OISL systems.
loop can be used. Each optical PLL structure is fixed to a specific
modulation format and its complexity increases with modulation In the following we numerically investigate and compare
order. Therefore, applying an intradyne detection scheme, where these two coherent detection schemes within the optical
the local laser runs un-synchronized and frequency offset and intersatellite channel. The remainder of the paper is structured
phase noise compensation is done by digital signal processing as follows: Section 2 gives an introduction to the OISL
after the optical front end, seems to be a possible alternative. We system. In section 3 the OPLL based homodyne detection
present numerical results for both coherent detection schemes scheme and in section 4 the intradyne detection scheme are
and compare them in terms of receiver sensitivity. described. Finally in section 5, we compare both schemes.
Keywords— Optical phase-locked loop; homodyne detection;
intradyne detection; carrier recovery; digital frequency offset II. OPTICAL INTERSATELLITE LINK
compensation; optical intersatellite link; Optical (inter-)satellite communication describes data
transmission between satellites using laser sources in the near-
I. INTRODUCTION infrared spectrum instead of the conventional radio frequency
The first European commercial systems providing optical (RF). Fig. 1 illustrates a typical OISL scenario. As part of an
intersatellite communication is the upcoming European Data earth observation process a satellite in a low-earth orbit (LEO)
Relay System (EDRS). This network consists of several may accumulate a high amount of data. This data shall be
satellites in different orbits and allows laser links over transmitted to a ground station (GS) on earth in real time, e.g.
distances of up to 45.000 km with data rates of up to 1.8 Gb/s in case of a fast tsunami forecast system. However, the
[1]. Due to the lower power consumption and higher data rates transmission time window of a LEO satellite is too short
optical intersatellite links (OISL) offer an attractive alternative (several minutes per day) and the RF link data rate is too low
to conventional RF-communication and enable real-time earth (~Mb/s) in order to send all data to GS during one flyover.
observations, like faster earthquake or tsunami forecasts, Hence, real-time data transmission is not possible. Instead, if
which require high-speed communication links between
satellites and between satellites and earth. Furthermore, the
narrow laser beam width ensures a better data security and the
lower weight leads to lower overall cost. However, the main
challenge is the complex alignment of both satellites to
achieve line-of-sight (LOS) connections for data transmission.
Therefore, in practical systems a pointing, acquisition and
tracking (PAT) system is needed [2].
The OISL core technology as provided for EDRS consists
of a laser communication terminal (LCT) which allows optical
coherent BPSK data transmission on satellites. The homodyne
detection structure uses an optical phase-locked loop (OPLL) Fig. 1. Conventional RF scenario (left), OISL scenario (right)
Fig. 2. Homodyne Detection setup of a state-of-the-art optical intersatellite communication system

we assume the data would be available on a satellite in the therefore realized with a fast tuneable Laser. The two signals
geostationary orbit (GEO), the time window would be large are described as follows:
(24 hours per day). However, the data need to be transmitted
fast from the LEO to the GEO satellite. Therefore, an optical s1 (t )  sˆ1 sin 0 t  1 (t )  M (t )  with 1 (t )   (t )t  1,0
link between the satellites with data rates in the range of , (1)
several Gb/s is used. s2 (t )  sˆ2 cos 0 t  2 (t )  with 2 (t )  K 0  x(t )dt  2,0

Compared to terrestrial optical communication systems,

where ŝ1,2 denote the signal amplitudes, 0 the carrier
OISL has several differences. First, the equipment has to
fulfill different quality requirements. Due to its location in frequency,  (t ) the time varying frequency offset due to the
space all parts have to be space-qualified, e.g. in terms of Doppler shift, K 0 the frequency gain of the VCO in Hz/V,
radiation and temperature hardness. Second, the narrow beam
width needs line-of-sight between the satellites, so a complex M (t ) the phase modulation, 1,0 and 2,0 a constant phase
PAT system is needed. Third, once established in space, the offset. After converting into the electrical domain by the
system must run for e.g. 15 years in a geostationary orbit photodiodes and amplification the I- and Q-signal is given as
without any maintenance. Finally, a significant difference is
the in orbit FSO channel, which is less complex than e.g. the U I (t )  G·R·sˆ1 sˆ2 cos  (t )  M (t ) 
optical fiber channel. This is different for FSO transmission , (2)
through the atmosphere [4]. U Q (t ) G·R·sˆ1 sˆ2 sin  (t )  M (t ) 

Fig. 2 illustrates the block diagram of a state-of-the-art where G denotes the gain of the transimpedance amplifier
OISL-BPSK transmission system with homodyne detection. (TIA), R the responsivity of the photodiodes and
Usually both, transmitter and receiver part, are implemented in  (t )  1 (t )  2 (t ) the phase error. To ensure a correct
one LCT for each satellite in order to guarantee bidirectional
transmission. Due to its high power efficiency, a typical laser frequency acquisition the control signal of the LO should only
source is an optical pumped neodymium-doped yttrium contain the phase error information and no data. In case of
aluminium garnet (Nd:YAG) laser at 1064 nm. The light is BPSK transmission, i.e. M (t )  0,   , we get such a data-
externally modulated by a phase-modulator (PM), driven by free error signal by multiplying UI and UQ which results in
the electrical data signal and amplified up to 2 W by an
ytterbium-doped fiber amplifier (YDFA). After transmission  (t )  U I (t )·U Q (t )
through the free-space channel the received optical signal is
  GRsˆ1 sˆ2  cos  (t )  M (t )  sin  (t )  M (t ) 
2
detected by the coherent receiver, containing a 90° optical
hybrid, which superimposes the receive light with the LO (3)
 GRsˆ1 sˆ2 
2
light. Finally, the demodulated signal is sampled and passed to  sin  2 (t )  2M (t ) 
the data recovery. Both lasers, transmitter laser and LO, will 2
show a frequency mismatch, due to the Doppler shift, natural  K D sin  2 (t )  .
frequency drift and phase noise. In order to ensure homodyne
detection this mismatch is eliminated in the optical domain by
where K D   GRsˆ1 sˆ2  2 denotes the phase discriminator
2
an OPLL structure. In case of intradyne detection the
frequency mismatch will be compensated for in the digital gain. Since KD directly influences the OPLL behavior it should
domain by digital signal processing. Both detection schemes be kept constant. However, ŝ2 continuously change during
will be described in detail in the following sections.
transmission as the distance between the satellites will vary.
Therefore, an automatic gain control (AGC) is used in
III. OPLL BASED HOMODYNE DETECTION
addition.
The presented OPLL is a Costas-loop based control system The error signal  (t) contains the residual frequency offset. It
as shown in Fig. 3. A 90°-Hybrid transforms the incoming
is passed to the loop filter F(s), which describes the control
signal s1(t) into baseband by beating with a local oscillator
element of the loop and defines the loop dynamic. A typically
s2(t). The LO is controlled by an error signal x(t ) or  (t ) and
 Fig. 5. QPSK phase discriminator
results in
Fig. 3. Optical phase-locked loop based on Costas-loop implementation  (t )  U I  U Q 
· U I  U Q  ·U I ·U Q 
(5)
 GRsˆ1sˆ2 
4
used loop filter is the lead-integrator (active) filter with the  K D sin(4 ), with K D  .
following transfer function: 4
1  sT1 Compared to BPSK the error signal for QPSK transmission
F (s)  , (4) has a two times higher phase error sensitivity. Therefore, as
sT2
the OPLL has a sinusoidal nonlinear characteristic the phase
where the time constants T1 and T2 influence the loop error variance has to be further reduced in order to achieve
linear behaviour.
dynamic. Finally, the control signal x(t) adjusts the frequency
Additionally, the system complexity increases when changing
of the fast tuneable LO based on the information of the error from BPSK to QPSK. It would be even more when applying
signal. Fig. 4 shows a typical frequency lock-in process in an 8-PSK. Therefore, in the following we present an approach for
optical intersatellite link with active loop filter observed at the intradyne detection which does not require any analog carrier
phase error signal  (t) and the frequency error (with recovery system and hence reduces hardware complexity.
K0=5.2 MHz/V, KD=0.79 V, T1=0.21 µs, T2=1.2 µs). The
frequency offset between the incoming signal and the LO at IV. DSP BASED INTRADYNE DETECTION
the beginning of the lock-in process was set to 5 MHz. In this In order to shift the hardware complexity into the digital
example the OPLL locks after ~8 µs. Both signals are domain we apply intradyne detection with digital frequency
overlapped by noise. The signals get mainly distorted by two offset compensation. The receiver structure, depicted in Fig. 6,
noise sources, phase and shot noise. As typical in optical keeps partly the same compared to Fig. 1, except the free-
coherent detection systems the laser phase noise is one of the running LO, the obsolete AGC and the now required analog-
main signal impairments. Due to their specific linewidth   to-digital converter (ADC) as well as a digital signal processor
both, the transmitter laser as well as the local oscillator, cause (DSP). The investigated compensation scheme is separated
phase rotation which influences the signal quality. Shot noise into coarse (CC) and fine compensation (FC). In the first step,
is induced in the photodiode caused by stochastic arrivals of coarse frequency offset compensation is applied, in this case
photons, which can be modelled as additive Gaussian noise. by the phase differential algorithm (PDA) [5, 6]. The sampled
In order to further increase the data rate higher-order input signal to the DSP is assumed to be
modulation formats can be used. Besides the well-known
modification needed at transmitter side the OPLL has to be X  k   I  k   jQ  k 
modified as well if changing from BPSK to QPSK. In fact the (6)
phase discriminator has to be changed, in order to get a data ˆ j   k M  k 0 n  k   n  k  ,
 Ae
free error signal. As the multiplier operation for BPSK
eliminates the data, a possible phase discriminator for QPSK where  denotes the intensity of the coherently detected signal,
is shown in Fig. 5. The error signal based on that discriminator  the phase shift due to the frequency offset f , M
Normalized Error Signal ε(t)

Frequency Error [MHz]

2 6
5
1 4
3
0 2
−1
1
0
−2 −1
0 5 10 0 5 10
Time [µs] Time [µs]

Fig. 4. Lock-in process: phase error (left) and frequency error (right) Fig. 6. Intradyne detection based on QPSK-DSP
 600 TABLE I. SIMULATION PARAMETERS
Δ f − fest Tx power PTx 32 dBm
400
Symbol rate fs 1 GS/s
Frequency [MHz]

200 Rx bandwidth Be 2 GHz

0 OPLL bandwidth BL 350 kHz
FC sample length N 32
−200
Δf
fest CC sample length L 1024
−400 Residual FO (intradyne) f 100 MHz
−600
LO power PLO 12 dBm
−500 −250 0 250 500 Linewidth (Tx/Rx) Dn 10 kHz
Frequency Offset Δ f [MHz]
blocks of length N and the mean phase shift of each block i is
Fig.7. Noise-free behaviour of the digital frequency offset compensation
computed by

the data modulation, 0 a constant phase offset, n the phase 1 ïìï N Mïü

F = arg íå (Y [ n + (i -1) N ]) ïý , (11)
noise and n[k ] the (complex) additive noise. The estimated
FC , i
M ïîï n=1 ïþï
 due to the frequency offset is computed based
phase shift F where N denotes the length of averaging.
CC

on the mean phase difference of L consecutive samples by

The resulting output of the fine compensation stage is given
1 ïì L Mïü by

F = ⋅ arg ïíå ( X [ k ]⋅ X * [ k -1]) ïý , (7)
ïîï k =1 ïþï Z  k   Y  k   exp   j 
 .
CC
M (12)
FC

where M denotes the modulation order for M-PSK, i.e. M = 4 The value of N and L influences the compensation accuracy
in case of QPSK, and L is the length of averaging. After re- [6] as well as the computational effort. Therefore, a trade-off
rotating the incoming samples the output of the coarse between accuracy and effort is required, which depends on the
compensation stage is available space-qualified DSP, signal-to-noise ratio (SNR) and
Y  k   X  k   exp   j 
 .
CC (8) initial frequency offset.
Finally, it should be mentioned that software-based frequency
The final estimated frequency offset is then calculated by compensation additionally allows real-time adaption to
different modulation formats, compensation of other

F impairments like imbalance between I and Q in the IQ-
f est = CC
. (9)
2pTs modulator as well as the usage of equalization schemes if
required.
In practical systems, the information about the estimated
frequency offset could be used to adjust the local oscillator on V. COMPARISON
the receiver satellite in order to keep the frequency mismatch Both detection schemes are numerically investigated in a
within a limited range. QPSK OISL transmission system based on Fig. 8. The system
parameters are given in Table 1. A 216 bit long two-
Fig. 7 shows the principle of the coarse compensation in dimensional pseudo random De-Bruijn sequence (PRBS) was
case of noise-free BPSK transmission (M=2) and mapped to QPSK symbols. As in the real LCT we assume that
f s  1/TS  1GHz . Obviously, the algorithm works properly in a first step the initial Doppler-shift of several GHz was
but only in a limited range, here between 250 MHz . This is adjusted in real-time by using known trajectory data of the
satellite and thermal frequency sweeping. However, a residual
due to the fact that the arg{}-operation in (4) wraps the phase
frequency offset f may remain which will be compensated in
if crossing the  border. The maximum frequency offset
which will be properly estimated is therefore the second step either by an OPLL or DSP. In case of
homodyne detection and OPLL techniques, we have to
p 1 distinguish between frequency acquisition and data
f est ,max = = , (10)
2pTs M 2MTs transmission. Since measuring the bit error ratio (BER) is only
possible after frequency acquisition was done, we assume
which conforms to the result in Fig. 7. If considering noise, perfect acquisition (except phase noise) for simulating the
the residual frequency error after CC may still reach several OPLL data demodulation performance. In case of intradyne
MHz. Hence, in a second step, fine compensation based on detection with digital frequency offset compensation the
Viterbi & Viterbi algorithm will reduce the residual frequency residual frequency offset f was set to 100 MHz. Due to the
offset and phase noise. The incoming data is separated into combination of low receive power and high LO-power the
Fig. 8. Block diagram of the QPSK simulation setup

main noise sources are shot and thermal noise at the receiver. future OISL systems. Furthermore, in order to keep the OPLL
Furthermore, phase noise due to the linewidth of Tx and LO bandwidth in an optimum range, the maximum frequency
laser is considered. Fig. 9 presents the performance evaluation offset which can be compensated for is much less in case of
of the two described detection schemes. The BER was homodyne detection (some MHz) compared to intradyne
determined by Monte-Carlo simulation for BPSK and QPSK, detection and the DSP scheme (hundreds of MHz and more).
respectively. The received input power was varied by the Hence, using an OPLL requires a longer thermal sweeping in
transmission length. Due to the cycle slip phenomena, i.e. order to reduce the initial frequency offset into the working
phase jumps of multiple  2M observed at the recovered data range and the total acquisition time is much longer than for
symbols, which occur in both receiver schemes, differential intradyne detection. Since the transmission time window in
(de)-coding was applied in order to avoid burst errors. To optical satellite systems is limited by the LOS duration, which
achieve error free transmission we aim a BER of 10-3, which will be reduced by the acquisition time, intradyne detection
allows modern forward error correction codes (FEC) to further may further extend this time window. However, one essential
decrease the BER. drawback of intradyne detection is the increased power
First of all, both schemes show the expected penalty of 3 dB consumption due to the required DSP.
when changing from BPSK to QPSK for constant symbol rate. VI. CONCLUSIONS
The low received input power results from the high free-space
We presented for the first time, to the best of our
attenuation, which requires high receiver sensitivity.
knowledge, a comparison of homodyne QPSK detection based
Furthermore, the intradyne detection shows an approx. 2 dB
on OPLL techniques and intradyne detection based on digital
better sensitivity, compared to the OPLL, if the laser linewidth frequency offset compensation in optical coherent satellite
is 10 kHz, which is a typical linewidth in state-of-the-art OISL communication systems. The simulation shows that intradyne
systems. In case of ideal homodyne detection (  = 0 kHz), detection has nearly the same behaviour as ideal homodyne
both schemes have nearly the same performance. Hence, the detection and even improves the receiver sensitivity by
linewidth requirement in case of DSP is less stringent approx. 2 dB compared to non-ideal homodyne detection. This
compared to the OPLL setup. It should be noticed, that the indicates the potential of DSP for future OISL systems.
simulation does not consider additional noise sources like
quantization noise which will further impair practical systems. REFERENCES
Nevertheless, this investigation shows the potential of DSP in [1] Heine, F. et al., “The European Data Relay System, high speed laser
based data links,“ Advanced Satellite Multimedia Systems Conference
10
−1 and the 13th Signal Processing for Space Communications Workshop
OPLL−BPSK (10 kHz) (ASMS/SPSC), 2014 7th , 8-10 Sept., Livorno, 2014
OPLL−BPSK (0 kHz)
OPLL−QPSK (10 kHz) [2] Marshalek. R. et al, in Near-earth laser communications, CRC Press,
OPLL−QPSK (0 kHz) Boca Raton, 2009
DSP−BPSK (10 kHz)
10
−2 DSP−QPSK (10 kHz) [3] Gregory, M. et al., “TESAT laser communication terminal performance
results on 5.6 Gbit coherent inter satellite and satellite to ground links,”
International Conference on Space Optics, vol. 4, 2010, p.8, Rhodes,
BER

2010
−3 2 dB 2 dB [4] Gregory, M. et al., “Three years coherent space to ground links:
10 performance results and outlook for the optical ground station equipped
with adaptive optics,” Proc. SPIE 8610, Free-Space Laser
Communication and Atmospheric Propagation XXV, 861004, March 19,
2013
−4
10 [5] Seimetz, M., High-Order Modulation for Optical Fiber Transmission,
Springer Ser. in Opt. Sciene,143, Berlin, 2009
−50 −48 −46 −44 −42 −40 −38 [6] Schaefer, S. et al., “Digital Frequency Offset Compensation in High-
P [dBm] speed Optical Intersatellite Data Transmission Systems,” European
Rx
Conference on Optical Communications 2015 (ECOC), 27th Sep. – 1st
Fig. 9. BER vs. received input power in OISL; comparison of the two detection
Oct., Valencia, Spain, 2015
schemes for BPSK and QPSK