The demand for emerging space network systems is rapidly increasing with an increase in the accessibility, affordability, and availability of small satellite constellation technologies [1–3]. Optical communication is a promising candidate for an optical network bridge between low-earth orbit (LEO), medium-earth orbit (MEO), and geostationary orbit (GEO) satellites as well as deep space probes [4,5]. Free-space optical (FSO) communication systems not only overcome the resource regulatory constraints of radio frequency communication systems [6] but also satisfy high-speed, high-throughput requirements [7] and further improve security through the use of quantum technology [8–10]. With the development of numerous worldwide inter-satellite links (ISLs) [11], ground-satellite links (GSLs) [12–14], and their imitation of ground air [15–17] and terrestrial links [18,19], the enormous potential of optical communication as an optical network bridge is gradually becoming a reality. As part of these efforts, researchers have leveraged FSO communication technology to develop spaceborne [20–22], airborne [15,16], and ground terminals [23–26] and have improved their performance in the past decades.

Near-earth and deep-space FSO communication systems, which commonly use intensity modulation techniques such as on–off keying (OOK) or Manchester, rely on signal-dependent encoding, making it difficult to establish reliable communication links in noisy environments [27,28]. Furthermore, finding ways to improve the signal-to-noise ratio (SNR) is even more challenging for GSLs, which are hampered by atmospheric effects such as beam spreading, beam wander, scintillation, angle of arrival fluctuations, and signal power constraints imposed by satellite payload limitations [29–31]. Therefore, efforts have been made to theoretically quantify the SNR using predicted background noise in order to identify the technical requirements for ground terminals [32–36]. Furthermore, researchers have sought to develop hardware and software solutions as realistic means of enhancing the SNR of terminals. One approach involves improving the signal by utilizing adaptive optics [37,38] or spatial diversity receivers [39,40] to mitigate atmospheric impacts or by using straightforward free-space-optics-based or multimode-fiber-based systems [41] with a larger numerical aperture than that of single-mode fiber coupling systems to increase the light-gathering efficiency. Another approach involves reducing the noise by using various filtering methods [10,35,42] or noise reduction algorithms [43,44]. Following the implementation of some of these ideas, further attempts have been made to demonstrate daytime optical communication, even under conditions where sunlight causes a substantial amount of background noise [10,45,46].

To establish communication with LEO satellites for our extended space mission, in this study, we designed and developed an all-free-space ground terminal that accounts for solar radiation noise. This facility is located in Daejeon, Republic of Korea, and is henceforth referred to as the Agency for Defense Development’s optical ground station (ADD-OGS). We evaluated the background noise power of the ADD-OGS, including temporal and spatial variations during the day, as well as potential changes in background noise that may arise during LEO satellite-to-ground communication. Based on our daytime observations, we investigated its influence on the beacon detection and communication performance of the ADD-OGS. Finally, we demonstrated daytime optical communication of a few gigabits per second between the developed ADD-OGS—designed to account for solar noise—and a mobile FSO terminal located approximately 7 km away from the ADD-OGS, which is approximately equivalent to the effective thickness of the atmosphere between the satellites and the ground [8]. The collection and analysis of background noise, together with the practical demonstration of the results, may contribute to the expansion of daytime scenarios for different laser communication links. We anticipate that these experimental findings will help advance scientific and industrial optical data transmission research for global optical communication space networks that connect future satellite constellations to the earth.

Figure 1 depicts the assembly of the ADD-OGS for daytime optical communication between the ground and an LEO satellite. It uses a all-free-space arrangement to eliminate the optics-to-fiber coupling loss caused by poor beam quality owing to high turbulence over long link distances. The ADD-OGS is primarily intended to satisfy the optical communications terminal standard defined by the Space Development Agency (SDA) [27]. Therefore, the ADD-OGS can support data rates of up to 2.5 Gbps using a nonreturn-to-zero (NRZ) OOK modulation technique. Notably, we attempted to reduce solar noise levels using several approaches. First, the ADD-OGS was designed to be operated at 1550 nm, which can lower the background noise of sunlight by up to 3% compared with that achieved at an operational wavelength of 800 nm [10]. Second, the transmitting and receiving wavelengths of the ADD-OGS were set at 1536 and 1553 nm, respectively, to isolate the receiving wavelength from the influence of background noise and the transmitting beam. Third, various filtering techniques were employed to suppress solar noise across different domains. This was achieved by combining: (i) spatial filtering using a pinhole and barrel structure to reject off-axis light [47,48]; (ii) spectral filtering using narrowband optical filters and solar-band rejection filters to isolate the receiving wavelength [10,45,46]; and (iii) spectral filtering using electronic filters to suppress high-frequency noises and interference components [49,50]. However, performance deterioration still occurs even if precautions are taken to mitigate the impact of solar noise, as discussed in Section 3.

As presented in Table 1, the ADD-OGS was designed to have a margin of 23.69 dB for the downlink and 13.84 dB for the uplink. In this configuration, the satellite terminal employed a fiber-coupled positive-intrinsic-negative (PIN) photodiode (PD) with an optical preamplifier, which provided a receiver sensitivity of approximately −43.9dBm, whereas the ground terminal used a free-space avalanche photodiode (APD), with a receiver sensitivity of approximately −36.6dBm under identical conditions of a 2.5-Gbps data rate and a bit error rate (BER) of 10−6. To estimate the link margin, the amount of received light PR is calculated as follows [51]: PR=PT+GT−LT−LFS−Latm−LP+GR−LR−Limpl,(1)where PT is the transmitter power, GT is the transmitter gain, LT is the transmitter optical loss, LFS is the free-space loss, Latm is the atmospheric loss, LP is the pointing loss, GR is the receiver gain, LR is the receiver optical loss, and Limpl is the implementation loss. The total channel loss Ltotal is the sum of the following losses: implementation loss, pointing loss, receiver optical loss, atmospheric loss, free-space loss, and transmitter optical loss. In this calculation, atmospheric loss includes losses due to absorption, scattering, and scintillation. Atmospheric signal attenuation from absorption and scattering can be computed using the Beers–Lambert law [52,53] and databases observed at the German Aerospace Center (DLR) with a visibility of 23 km [54]. In addition, the scintillation fading loss can be quantified based on the scintillation index and aperture diameter of the receiver [55,56]. Consequently, both the uplink and downlink for communication and beacon use were designed to satisfy the SDA standard, with a link margin of 3 dB or more.Table 1.

Link Budget Design

Figure 1(c) shows the optical configuration of the ADD-OGS, which was constructed using a Ritchey–Chrétien telescope (RC700, Planewave Instruments, USA), with a focal length of 8.4 m and consisting of a hyperbolic primary mirror (M1) and a hyperbolic secondary mirror (M2). The tertiary mirror of the telescope (M3) can be rotated to utilize the two Nasmyth foci. One Nasmyth focus port is used to acquire the initial location and attitude information of the ADD-OGS through star tracking using a visible-light camera. The other Nasmyth focus port is used for communication, with the communication beam transferred to the optical system of a separate Coudé table via the Coudé path (M4–M7). In this manner, separating the majority of the optical system from the telescope circumvents the payload constraints of the telescope. Only lightweight optical systems used to transmit and receive the beacon beam are piggybacked on top of the telescope.

Figure 1(d) shows the optical configuration of beacon transmission and reception by the optical system, which rides on top of the telescope. The seed source for the beacon is a distributed feedback laser diode (DFB-LD) with an output power of 40 mW and a central wavelength of 1553 nm. The seed source is intensity-modulated at a frequency of tens of kilohertz to distinguish it from the background noise. The modulated light is amplified to 40 W using an erbium-doped fiber amplifier (EDFA) and is finally emitted with a divergence of 500 μrad at an output collimator. The beam is spirally scanned by controlling a fast steering mirror (FSM) in an open-loop manner to obtain an uncertainty cone of a few milliradians. Simultaneously, it is designed such that the satellite beacon beam could pass through the catadioptric telescope and be detected by a 640×512-pixels short-wavelength infrared (SWIR) camera.

Figure 1(e) depicts the optical layout of communication transmission and reception by the optical system on the Coudé table. A DFB-LD with a central wavelength of 1553 nm serves as the laser source of the NRZ-OOK transmitter, which is subsequently modulated with a 2.5-Gbps NRZ-OOK via a Mach–Zehnder modulator (MX-LN-05, iXblue, France). Most of the light from the transmitter is amplified up to 30 W using a two-stage EDFA, whereas a small part is employed to control the bias voltage to compensate for the path difference drift caused by temperature changes. The amplified beam is reflected by a steering mirror to produce a point-ahead angle, passed through the Coudé path, and released from the Ritchey–Chrétien telescope to reach the satellite terminal on the opposite side.

Conversely, the satellite communication beam passes through the telescope and is transmitted to the Coudé table in the opposite order, where it is reflected by numerous mirrors, including the FSM, and reaches the PD. During this process, off-axis light is blocked by barrel and pinhole structures along the optical reception path [47,48]. The received optical beam then passes through a filter array comprising an optical long-pass filter, which primarily blocks strong solar radiation below 1000 nm, and an optical band-pass filter, which selectively transmits wavelengths centered at 1536 nm with a bandwidth of several hundred gigahertz, before impinging on the photodetector [10,45,46]. Finally, 90% of the received beam reaches the APD, and 10% of the beam is split into a PIN-based quadrant PD (QPD). To ensure that the beam is focused on the APD, the FSM is adjusted using position data from the QPD.

Figure 2 shows the results of satellite tracking and background noise measurement tests. Here, we present the findings of a test conducted on August 28, 2024, which was one of several successful evening tracking tests of the bright LEO satellites [Figs. 2(a) and 2(c)–2(f)]. Initially, we performed automated calibration of the ADD-OGS using more than 50 photographs collected by visible cameras throughout the sky. After calibration, the ground station was pointed at the LEO satellite SL-16 R/B, which was orbiting at approximately 850 km, using the trajectory calculated from two-line element (TLE) data. The ADD-OGS was closed-loop-controlled to retain the satellite target at the center of the camera’s field of view (FOV) [Fig. 2(c)]. This was achieved by developing a proportional, integral, and derivative feedback system that adjusts the offsets of the telescope’s two rotation axes based on visual data while maintaining the anticipated satellite trajectory [Figs. 2(e) and 2(f)]. Consequently, the ADD-OGS maintained the target satellite centered in the camera’s FOV with an instantaneous FOV of 24 μrad, which falls within the FOV of the APD.

Following an intensive validation of the pointing, acquisition, and tracking performance of the ADD-OGS for LEO satellites, we examined variations in solar noise under the assumption that the satellite-to-ground link is operational [Figs. 2(b), 2(g)–2(j)]. Hence, we investigated the background noise that may occur when tracking the LEO satellite STARLINK-31539, which passed by the Republic of Korea’s sky on August 13, 2024, at around noon. In this experiment, we open-loop tracked the satellite rather than image-based closed-loop tracking. The ADD-OGS followed the predicted satellite orbit derived from the TLE information, as illustrated in Figs. 2(h)–2(j), showing high consistency between the observed and calculated paths. It should be noted that, in addition to our current setup, implementing closed-loop control with a satellite equipped with an onboard laser beacon will be necessary in order to fully characterize solar background noise during satellite-to-ground optical links.

Figure 2(b) shows an enlarged image of the QPD-based signal detection device that monitored the solar background noise for this daytime experiment. Four QPD cells were used to detect background noise current signals, which were then converted to voltage signals using transimpedance amplifiers (TIAs) and monitored via a data acquisition (DAQ) system. The QPD was selected for its ability to detect faint solar noise signals on the order of hundreds of picowatts, whereas the AGC(automatic gain control)-based TIAs were designed with variable gain to enable a wide dynamic range. Furthermore, electronic filters with cutoff frequencies of several kilohertz were employed to suppress high-frequency noise [49,50]. The DAQ system continuously captured the solar noise signals along with the ADD-OGS’s altitude, right ascension, and declination, with all data time synchronization using the GPS receiver to ensure acquisition within a unified time domain.

Figure 2(g) shows a representative case of rapid changes in solar noise during satellite tracking. This experiment was conducted around midday under cloud-negligible conditions and low air mass, when the clearness index—defined as the ratio of observed ground-level solar irradiance to the estimated extraterrestrial irradiance—was close to unity. These conditions were intentionally chosen to avoid atmospheric effects responsible for short-term irradiance fluctuations [57–59]. Furthermore, it was believed that transitory solar activities such as flares had little impact on the measurement results since they primarily affect X-ray, radio, and ultraviolet wavelengths, with only a minor contribution to near-infrared [50,60,61]. In the subsequent Fig. 3, the satellite trajectory of Fig. 2(g) is analyzed using spatial datasets to determine the reason for this abrupt rise. The rapid increase in solar noise observed in Fig. 2(g) was attributed to the combined spatial and temporal dependencies as shown in Fig. 3(i). As this result shows, the ADD-OGS may scan sky regions with elevated solar background noise while tracking specific satellite trajectories, and this has the potential to temporarily deteriorate optical communication link performance.

Figure 3 shows the results of temporal and spatial solar background noise variation measurement tests. In these trials, the previously reported QPD-based detection system of the ADD-OGS was used to record solar noise. First, we evaluated the sunlight incident on the ADD-OGS for 12 h to examine the temporal dependence of solar noise [Figs. 3(a) and 3(c)–3(f)]. Solar noise data were collected while the ADD-OGS tracked the trajectory of the sun with a constant right ascension offset ϕ of 1.67° (100 arcmin), as shown in Fig. 3(a). Figures 3(d)–3(f) compare the ADD-OGS’s altitude, right ascension, and declination angles to those of the sun, with Fig. 3(e) indicating an intended offset in right ascension relative to the sun. Consequently, we obtained tens to hundreds of nanowatts of solar noise in Fig. 3(c), and the areas of the graph where notable short-term power outages can be observed were caused by passing clouds that blocked the sun. In particular, the simulation [62] showed that the noise power produced by direct sun irradiance peaked at midday, whereas the actual collected noise power caused by total solar irradiance peaked at approximately 3 p.m. This is because the total (global) solar irradiance received by the ADD-OGS was calculated as the sum of direct irradiance Ib, diffuse irradiance Id, and reflected irradiance Ir [63]: It=Ib+Id+Ir,(2)It=Ib,ncosφ+Id,h1+cosβ2+ρ(Ib,ncosθz+Id,h)1−cosβ2,(3)where Ib,n is the direct normal irradiance, Id,h is the diffuse horizontal irradiance, β is the tilted angle of the aperture surface of the ADD-OGS relative to the horizontal plane, φ is the angle of incidence, ρ is the albedo, and θz is the solar zenith angle. The difference between the two graphs shown in Fig. 3(c) became larger at 10 a.m. and 3 p.m. owing to the higher tilted angle β with respect to the horizontal and higher albedo value ρ brought on by reflection from surrounding terrain features. This effect arises from the distinct topography around the ADD-OGS installation site, with eastern mountainous terrain and western man-made building structures enhancing morning (10 a.m.) and afternoon (3 p.m.) albedo-induced reflection, both of which contribute to greater solar noise levels.

Second, we assessed the spatial dependency of solar noise by monitoring the incidence of sunlight on the ADD-OGS while adjusting the angle of the ADD-OGS around noon [Figs. 3(b) and 3(g)–3(i)]. The measurements were taken at increasing angles φ to φ′ between the ADD-OGS and the sun, as shown in Fig. 3(b), and the trends of solar noise at different angles were observed [Fig. 3(g)]. Furthermore, to investigate a wider range of solar noise distributions, we obtained a three-dimensional (3D) map that displayed solar noise at various right ascension and declination angles of the ADD-OGS relative to the sun [Fig. 3(h)]. The noise was saturated inside the FOV of the QPD, as shown in Figs. 3(g) and 3(h). At other angles, it appeared to be influenced by the surrounding terrain rather than by the angle relative to the sun. Figure 3(i) presents a two-dimensional (2D) projection of the 3D dataset from Fig. 3(h), overlaid with the satellite trajectory corresponding to Fig. 2(g). Figure 3(i) shows that the background noise increased when the ADD-OGS passed through terrain-induced high-noise regions during satellite tracking. This result indicates that there should be no highly reflective reflectors or terrain features near the ground terminals. Otherwise, it is preferable to evaluate background noise reflections from the surrounding topographical features beforehand to determine the optimal operation orientation of ground terminals in a daytime communication scenario. These findings highlight the necessity of incorporating local background noise distribution analysis and target satellite trajectory mapping into the planning phase.

Shot noise is caused by irregular current fluctuations during the optical signal detection process of a photodetector. Shot noise variances σshot2 in the PIN-PD and APD can be calculated as follows [64]: σshot,PIN2=2qIPB=2qRPRB,(4)σshot,APD2=2qIPBM2F=2qRPRBM2F,(5)where q is the electric charge, IP is the output current of the detector, B is the electrical bandwidth, M is the avalanche gain, F is the excess noise factor, R is the responsivity, and PR is the received optical power.

In addition, even in the absence of photons reaching the detector, random electron-hole production induces dark current noise. Dark current noise variances σdark2 in the PIN-PD and APD are represented as follows [64]: σdark,PIN2=2qIdB,(6)σdark,APD2=2qM2FIdB,(7)where Id is the dark current of the detector. The surface dark current noise is usually ignored in APDs because avalanche multiplication is a bulk effect.

Moreover, thermal noise is caused by irregular fluctuations in free electrons, and the thermal noise variance σthermal2 can be expressed as follows [64]: σthermal2=4kBTBRL,(8)where kB is the Blotzmann constant, T is the temperature, and RL is the load resistance.

Furthermore, signal–background and background–background beat noises are caused by incoherent solar radiation and can be expressed as follows [65]: σsig-back2=4R2PRNbackB,(9)σback-back2=R2nNback2(2Δλ−B)B,(10)where Nback is the peak power spectral density per mode of the background radiation, n is the number of spatial and polarization modes of the background light, and Δλ is the bandwidth of the optical bandpass filter: Nback=Nfλ2,(11)n=2ΩFOV/ΩDL (if ΩFOV<ΩS),(12)where Nf is the spectral radiance function that provides the radiated power per emitting area to one steradian with a bandwidth of 1 Hz, ΩS is the solid angle under which the receiver sees the extended background source, ΩFOV is the receiver’s FOV, and ΩDL is the diffraction-limited value of the antenna’s FOV.

First, we used Eqs. (4)–(12) to calculate the noise variances and SNRs of the QPD and APD as the received optical power of the telescope increased. The power that the telescope received was split in this analysis: 10% of the incident power went to the QPD and 90% went to the APD (Fig. 1). When the received signal power of the telescope was −23dBm or less, the influence of the background–background beat noise dominated in the QPD, as shown in Fig. 4(a). However, as the received optical power increased, shot noise and signal–background noise gradually became dominant in the QPD. By contrast, Fig. 4(d) illustrates that, in the APD, background–background beat noise dominates when the received power of the telescope is less than −32dBm, whereas the effects of shot and signal background noises gradually become dominant as the received optical power increases. Figures 4(b) and 4(e) illustrate how the SNRs of the QPD and APD vary in response to changes in the received signal power.

Next, the error Δx, error variance σΔx, and accuracy σx0 of beacon detection on the x-axis are calculated as a function of the QPD’s SNR as follows [43,66]: Δx=erf(2x0ω),(13)σΔx=1+erf2(2x0ω)SNR,(14)σx0≈πω221SNR1+erf2(2x0ω),(15)where x0 is the position of the beam center, and ω is the beam radius. The analysis results indicate that −42dBm of optical power is required to achieve an error variance of 40 μm, which is the threshold for focusing light onto the 80-μm APD aperture [Fig. 4(c)].

The bit error probability pB on the APD can be estimated using Eq. (16) as a function of the quality factor Q in the NRZ-OOK scheme as follows [65] [Fig. 4(f)]: pB=12erfc(Q2),(16)Q=I1−I0σ1+σ0,(17)where I1 and I0 are the electric signal currents for a “1”-bit and a “0”-bit, respectively, and σ1 and σ0 are the standard deviations associated with these signal levels, respectively.

We then experimentally examined the bit error rate (BER) performance of the ADD-OGS during the day, as shown in Figs. 4(g) and 4(h). Because the satellite terminal had not yet been launched, communication tests were conducted using a mobile FSO terminal [67] instead of the satellite terminal. The experiment was conducted over a 7-km ground-to-ground link, which approximately corresponds to the effective atmospheric thickness between an LEO satellite and the ground [8]. During the experiment, the atmospheric visibility was approximately 18 km, and the average refractive index structure constant was 3×10−14m−2/3, as measured by a scintillometer. The total atmospheric loss under these conditions is shown in Table 3, which compares the mean link margins between the terrestrial test configuration and a GSL downlink. The total channel loss of our 7-km test link was measured at 230.38 dB, approximately 43.61 dB lower than the projected 273.99 dB total channel loss for LEO satellite-to-ground links. Following the strategy proposed by Moll et al. [68], we adjusted the laser output power and beam divergence to deliberately reduce both the transmitter power and transmitter gain, thereby creating a more demanding link margin condition than that of the GSL design and enabling a rigorous evaluation of communication performance under minimal link margin scenarios with a 0.54-dB margin. For performance verification in this test, a 2.5-Gbps NRZ-OOK-coded pseudo-random binary sequence data signal with a length of 27−1 was provided by the mobile FSO terminal. The test results revealed an average BER of 5.85×10−8 at a data rate of 2.5 Gbps for approximately 300 s when the ADD-OGS received −36dBm of power. Therefore, the BER results from the practical demonstration are in good agreement with the simulation results.Table 3.

Mean Link Margin Comparison between the Terrestrial Test Link and the GSL (Downlink)^a

In this study, we constructed an all-free-space optical ground station called ADD-OGS, which achieved link margins of 23.69 dB for the downlink and 13.84 dB for the uplink. We first conducted evening satellite tracking experiments with an LEO satellite, SL-16 R/B, in an 850-km orbit to confirm the ADD-OGS’s pointing, acquisition, and tracking capabilities. We then used open-loop tracking based on TLE data to assess the daytime background noise received by the ADD-OGS under simulated LEO satellite tracking conditions. We observed a sharp increase in solar noise while tracking one of the LEO satellites, STARLINK-31539.

We also comprehensively examined the temporal and spatial dependency of solar noise to identify the source of noise in the sample satellite trajectory scenarios involving rapid solar noise changes. Temporal variations were observed over a 12-h period while maintaining a constant angle between the sun and the ADD-OGS, whereas spatial variations were assessed while varying the angle between the sun and the ADD-OGS at midday. Both the higher noise levels observed between 10 a.m. and 3 p.m. in the temporal data and the noise variations related with 3D topographical features suggest that surrounding terrain contributes to elevated solar noise. Moreover, an examination of the sample satellite’s trajectory revealed that it passed through a part of high-noise regions caused by surrounding topographic features. Therefore, we plan to conduct in-depth quantitative research on the influence of this albedo-induced reflection in the future.

In addition, we investigated the impact of background noise on the beacon detection and communication performance of the ADD-OGS using noise analysis. According to the beacon detection analysis, −42dBm of optical power is required to achieve error variance when focusing light into the APD aperture. Additionally, the communication simulation results closely agree with experimental data, showing a BER of 5.85×10−8 at 2.5 Gbps and a received power of −36dBm. In the noise analysis, the majority of noise in both the APD and QPD was caused by beat noise components, which are the result of background radiation beating with the signal or background radiation beating with itself.

As a result, we analytically and experimentally verified that 2.5-Gbps data transmission up to 7 km is feasible, even in expected daytime satellite tracking scenarios with high background noise, by the proposed ADD-OGS with spectral and spatial filtering to achieve an acceptable SNR. However, the current system does not include channel error correction codes, limiting its communication performance under dynamic noise conditions. Future work will integrate forward error correction (FEC) into ADD-OGS to improve link robustness and better reflect realistic models of optical communication environments.

In the future, we plan to conduct communication tests between the developed ADD-OGS and the LEO satellite terminal that will be launched. We also plan to construct a next-generation optical ground station that is transportable, with the goal of enhancing site diversity to ensure communication reliability in a variety of weather conditions [69–71]. The experimental results provide guidance for the practical daytime implementation of FSO terminals and are applicable to a wide range of channels, including ground-to-ground, ground-to-air, unmanned aerial vehicle (UAV)-to-satellite, and underwater laser communication links. Furthermore, we believe that our study may be beneficial to furthering a variety of applications such as satellite communication networks, space exploration, atmospheric sciences, climate studies, and the global renewable energy industry.