Free-Space Optical (FSO) Satellite Networks Performance Analysis: Transmission Power, Latency, and Outage Probability

In the following, we introduce the laser link transmission power model and the link margin model.

In the following, we introduce the an atmosphere channel model which is under the influence of atmosphere turbulence resulting in intensity fluctuations.

In this work, we consider that the turbulence induced channel experiences an exponentiated Weibull fading channel.The probability density function (PDF) is given by [28]

and the cumulative distribution function (CDF) is given by [28]

where $\alpha$, $\beta$ are the shape parameters and $\eta$ is the scale parameter of a ground station. The three parameters can be expressed as [29]

where $\textit{g}(\textit{$\alpha$},\textit{$\beta$})$ is the $\alpha$ and $\beta$ dependent constant variable, and $\textit{$\sigma$}_{I}$ is the scintillation index of the ground station, which is expressed as [30]

where the Rytov variance $\textit{$\sigma$}_{R}$ is given by [30]

where k is the optical wave number, $\zeta$ is the zenith angle of the ground station, $\textit{h}_{S}$ is the altitude of the satellite, and $\textit{C}_{n}^{2}$ is the altitude dependent refractive index constant, which is expressed as [31]

where $\textit{v}_{r}$ is the r.m.s. ground wind speed, and $\textit{C}_{0}$ is the nominal value of the refractive index constant at ground.

In the following, we introduce the modeling of latency in FSOSNs. In optical communication networks, four types of delay can be considered as part of the end-to-end latency. These delays consist of propagation delay, transmission delay, queueing delay, and processing delay. Propagation delay of an optical link depends on its propagation distance, which we also refer to as link latency. Node delay, also referred to as node latency in this work, comprises transmission, queueing, and processing delays at a satellite.

We introduce the two satellite constellations, i.e., Starlink Phase 1 Version 3 [4] and Kuiper Shell 2 [5] constellations, as described in their FCC filings as of 2020 and 2019, respectively. For a satellite constellation, its constellation orbit parameters including number of orbital planes, number of satellites on each plane, and inclination angle can also affect network latency. Greater number of orbital planes and satellites on each plane lead to denser distribution of satellites within the constellation, thus more satellites can be connected for the same LISL range, thereby more optimal links established to achieve the shortest path over an inter-continental connection and results less propagation delay added to the network latency. Similarly, less inclination angle leads to denser satellite distribution in the space above the areas away from the pole, and less propagation delay for inter-continental connections across the low latitude areas.

To study the tradeoff between the average satellite transmission power and network latency in FSOSNs, the LISL range is a critical factor to consider. The LISL range affects the connectivity of satellites, since a satellite can only establish LISLs with other satellites that are within this range. The maximum LISL range for satellites in a constellation can be regarded as the range which is constrained only by visibility, and it can be calculated as [2]

where r is the radius of the Earth, which is generally assumed to be 6,378 km, $\textit{h}_{S}$ is the altitude of the satellite, $\textit{h}_{a}$ is the height of the Earth’s atmosphere, which is generally considered to be 80 km, and $\textit{D}_{S}$ is the maximum LISL range. For the Starlink Phase 1 Version 3 constellation, $\textit{h}_{S}$ is 550 km, and the maximum LISL range can be calculated as 5,016 km. Similarly, for the Kuiper Shell 2 constellation with an altitude of 610 km, the maximum LISL range is 5,339 km.
    To study the tradeoff between network latency and average satellite transmission power, we consider several LISL ranges for satellites in FSOSNs for the two constellations. For the Starlink Phase 1 Version 3 constellation, we assume different LISL range values, specifically 1,575 km, 1,731 km, 2,000 km, 3,000 km, 4,000 km, and 5,016 km. The 1,575 km LISL range is a reasonable minimum feasible LISL range for a satellite in this constellation, as a satellite can establish six permanent LISLs with other satellites at this range, including four with neighbors in the same orbital plane and two with its nearest left and right neighbors in adjacent orbital planes. Shorter LISL ranges are not sufficient for a satellite to establish the two permanent LISLs with neighbors in adjacent orbital planes. At the 1,731 km LISL range, a satellite can establish ten permanent LISLs with other satellites, including four with neighbors in the same orbital plane and six with its nearest neighbors in adjacent orbital planes.
    For the Kuiper Shell 2 constellation, we also consider different LISL ranges, in this case 1,515 km, 2,000 km, 3,000 km, 4,000 km, and 5,339 km. Due to its different design, the Kuiper Shell 2 constellation has a different orbital plane distribution and a different satellite distribution per orbital plane than the Starlink Phase 1 Version 3 constellation; a satellite in the former constellation needs a longer LISL range to establish permanent LISLs with neighbors in same orbital plane as compared to the LISLs with nearest neighbors in the adjacent orbital planes. At the 1,515 km LISL range, a satellite in the Kuiper Shell 2 constellation can set up eight permanent LISLs, including two with satellites in the same orbital plane and six with satellites in adjacent orbital planes. For the Kuiper Shell 2 constellation, 1,515 km is a reasonable minimum LISL range for satellites to establish a reasonable minimum connectivity, since shorter LISL ranges are not long enough for a satellite to establish permanent LISLs with neighbors in same orbital plane.

The connectivity for the uplink and downlink is affected by the range of the ground station, as the greater the ground station range, the longer the propagation distance for uplink and downlink. The distance of the link between a ground station and satellite can be calculated using (8). The uplink/downlink is longest when the elevation angle between the ground station and an ingress/egress satellite is at its minimum, and the maximum range of the ground station can be calculated for the minimum elevation angle for each constellation. For the Starlink Phase 1 Version 3 constellation, the minimum elevation angle is 25°, and the corresponding maximum range of a ground station is 1,123 km. For the Kuiper Shell 2 constellation, the minimum elevation angle is 35°, and the maximum ground station range is 1,412 km.

Network latency and average satellite transmission power should be studied based on the entire orbital period of satellites in an FSOSN, which is the time a satellite takes to complete one full orbit around the Earth. The orbital period of a satellite in an FSOSN can be calculated using [40]

where r is the radius of the Earth, which is 6,378 km, $\textit{h}_{S}$ is the altitude of the satellite in the FSOSN, G is the gravitational constant, which is $6.673\times 10^{-11}$ Nm²/kg², and $\textit{M}_{E}$ is the mass of the Earth, which is $5.98\times 10^{24}$ kg. Based on the altitude of the Starlink Phase 1 Version 3 and Kuiper Shell 2 constellations, which are 550 km and 610 km, respectively, the orbital period for a satellite in these constellations is 5,736 seconds and 5,810 seconds, respectively.

The outage probability (OP) is defined as the probability that the signal-to-noise (SNR) $\textit{$\gamma$}_{th}$ can fall below a certain threshold with acceptable communication quality. The OP in this case can be defined as

Substituting (26) into (38), the OP can be expressed as [49]

where $\bar{\textit{$\gamma$}}$ is the average SNR, and $\textit{L}_{a}$ is the atmosphere attenuation loss for uplink/downlink. In this work, we investigate the impact of cloud condition on the atmosphere attenuation and turbulence, and therefore the outage probability of optical uplink/downlink communication. We assume three cloud condition scenarios, the first one is clear weather condition with a thin cirrus cloud concentration of 0.5 cm^-3 and liquid water content as $3.128\times 10^{-4}$ g/m^-3, while the second one is clear weather condition with a cirrus cloud concentration of 0.025 cm^-3 and liquid water content as 0.06405 g/m^-3, and the third one is cloudy weather condition with a Cumulus cloud concentration of 250 cm^-3 and liquid water content as 1.0 g/m^-3, according to [10]. Based on these three different weather conditions, we calculated the OP using practical parameters and we hope these deployment scenarios can be used to provide practical insights about the system architecture, so that the reader can get a quick idea about the overall system performance.

To study the tradeoff between the average satellite transmission power and network latency in an FSOSN at different LISL ranges, we need to calculate the shortest path between the source ground station and the destination ground station over the FSOSN at an LISL range at each time slot to calculate the network latency and average satellite transmission power for that shortest path. The shortest path can be determined based on adjacency matrix containing propagation distances of laser links between satellites and between satellites and ground stations using Dijkstra’s algorithm [41]. This algorithm uses weights of the links to determine the shortest path, and these weights are the propagation distances of the optical links in our case. The following steps are used in Dijkstra’s algorithm to find the shortest path for an inter-continental connection from a source ground station to a destination ground station over an FSOSN.
1) Create a shortest path set S that keeps track of satellites that are on the shortest path, another set X for satellites that are not on the shortest path, and a vector D to track the shortest path distance for each satellite to the source ground station $\textit{g}_{s}$.
2) Assign an initial weight to all the links as their propagation distance, add $\textit{g}_{s}$ to S, and update the distance for $\textit{g}_{s}$ in D as 0, while the distances for other satellites to $\textit{g}_{s}$ are infinite in D.
3) While S does not include the destination ground station $\textit{g}_{d}$, go through the following loop.
a) Pick the satellite q with the minimum distance to $\textit{g}_{s}$, which is the closest satellite to $\textit{g}_{s}$ based on its distance to $\textit{g}_{s}$.
b) Mark q as visited and include q to S. Update the distance for q in D.
c) For every adjacent satellite $\textit{p}_{i}$ of q, check its distance to q. Since Dijkstra’s algorithm keeps tracking the currently known shortest path for each satellite from the source ground station and updates the distance if it finds a shorter one, check the new path to $\textit{g}_{s}$ through q from $\textit{p}_{i}$. For each $\textit{p}_{i}$, if the currently known distance of $\textit{p}_{i}$ to $\textit{g}_{s}$ is greater than the sum of the distance of q to $\textit{g}_{s}$ and the distance between q and $\textit{p}_{i}$, then update the distance for $\textit{p}_{i}$ to $\textit{g}_{s}$. After iterating through all the adjacent satellites $\textit{p}_{i}$ of q, go back to step a to find the next satellite to add to S.

We calculate the network latency $\textit{T}_{net}$ of a shortest path at a time slot as the end-to-end latency of the shortest path between the source ground station and the destination ground station by adding the link latency (or propagation delay) of each laser link, including the uplink, downlink, and ISLs, and the node delay (or node latency) for each satellite on the shortest path, which can be expressed as

where $\textit{T}_{up}$ and $\textit{T}_{down}$ are the link latencies of the laser uplink and downlink, respectively, and $\textit{T}_{k}$ is the link latency (or propagation delay) for kth ISL. Since there are n satellites on the shortest path for the inter-continental connection, $\textit{n}-1$ ISLs exist in total on the shortest path. As mentioned in Section III, $\textit{T}_{node}$ is the node delay for each satellite and is assumed to be 10 ms.

The propagation delay for each link on the shortest paths for first five time slots for the FSOSN with the Starlink Phase 1 Version 3 constellation at the 3,000 km LISL range for the Toronto–Sydney inter-continental connection is shown in Table 3. $\textit{T}_{up}$ and $\textit{T}_{down}$ represent the propagation delay for the uplink and downlink, and $\textit{T}_{k}$ is the propagation delay for the kth ISL. For example, the network latency of the shortest path at the first time slot is equal to the sum of $\textit{T}_{up}$, $\textit{T}_{down}$, the propagation delays of all seven LISLs, and $\textit{T}_{node}$ for each satellite on the shortest path (i.e., 20 ms × 8), and is equal to 137.22 ms.

The mean network latency $\textit{$\overline{T}$}_{net}$ is the mean of the network latency of the shortest paths at all time slots, and it can be expressed as

where j is the total number of time slots, and $\textit{T}_{net,i}$ is the network latency of the shortest path at the ith time slot. For example, $\textit{$\overline{T}$}_{net}$ for the first five time slots in Table 3 is equal to 137.22 ms.

The transmission power of each satellite is calculated by adding the transmission power required to establish the incoming link and outgoing link for that satellite. For the ingress (or first) and egress (or last) satellites on a shortest path, the incoming link and outgoing link are the uplink and downlink, respectively, and the transmission power for such satellites can be calculated as

where $\textit{L}_{PG,up}$ and $\textit{L}_{PG,down}$ is the free-space path loss for the uplink and downlink, respectively, and $\textit{L}_{PS,in}$ and $\textit{L}_{PS,out}$ is the free-space path loss for the incoming and outgoing ISLs, respectively. The optical links between satellites, i.e., LISLs, and optical links between ground stations and satellites (uplink and downlink) are bidirectional [42],[43]. That is why, we are considering the power from the satellite to the ground station for uplink and downlink for modelling transmission power of ingress and egress satellites. For other (or intermediate) satellites on the shortest path, the incoming and outgoing links are ISLs, and the transmission power for such satellites can be calculated as

The transmission power for each laser link on the shortest paths for the first five time slots for the FSOSN with the Starlink Phase 1 Version 3 constellation at an LISL range of 3,000 km for the Toronto–Sydney inter-continental connection is shown in Table 4. The satellite transmission power and average satellite transmission power for these shortest paths are given in Table 5. In Table 4, $\textit{P}_{up}$ and $\textit{P}_{down}$ represent the transmission power for the uplink and downlink, respectively, and $\textit{P}_{i}$ is the transmission power for the ith ISL. For example, $\textit{P}_{1}$ is the transmission power for the first ISL on the shortest path at a time slot. In Table 5, the transmission power for the first satellite (or ingress satellite) $\textit{P}_{TS,ingress}$ at the first time slot can be calculated by adding $\textit{P}_{up}$ and $\textit{P}_{1}$ (i.e., 89.34 mW and 198.26 mW, respectively) and is equal to 287.60 mW.

The average satellite transmission power for satellites on a shortest path at a time slot $\textit{P}_{TS,avg}$ is the average value of the transmission power of all satellites on the shortest path at that time slot, and it can be expressed as

where $\textit{P}_{TS,m}$ is the transmission power for the mth intermediate satellite on the shortest path and there are n satellites on the shortest path, and thereby $\textit{n}-2$ intermediate satellites exist in total on the shortest path. The mean of the average satellite transmission power $\textit{$\overline{P}$}_{TS,avg}$, i.e., the mean of the average satellite transmission power of the shortest paths at all time slots, can be expressed as

where j is the total number of time slots, and $\textit{P}_{TS,avg,i}$ is average satellite transmission power at the ith time slot. For instance, $\textit{P}_{TS,1}$ is the satellite transmission power for the first intermediate satellite on the shortest path at a time slot in Table 5. The average satellite transmission power $\textit{P}_{TS,avg}$ of the shortest path at the first time slot can be calculated by taking the average value of the transmission powers for the eight satellites on this shortest path, which is equal to 328.89 mW in Table 5. Finally, $\textit{$\overline{P}$}_{TS,avg}$ for the first five time slots in Table 5 is equal to 329.02 mW.

For the computation of the average satellite transmission power and network latency of the shortest path for an inter-continental connection, we first simulate the Starlink Phase 1 Version 3 and Kuiper Shell 2 constellations using the well-known satellite constellation simulator Systems Tool Kit (STK) version 12.1 [44]. We generate a satellite constellation, ground stations, and all possible ISLs and links between satellites and ground stations based on an LISL range within STK, and then we extract the data containing the links between satellites as well as the links between satellites and ground stations and the positions of satellites and ground stations from STK. We process this data in Python to compute the distance for each link at a time slot to find the shortest path between ground stations for an inter-continental connection over the FSOSN at each time slot. To this end, we first discretize the extracted data, as the data obtained from STK is continuous, and for a certain link, it contains the duration of the link for the entire orbital period. We discretize this data by dividing it into time slots and reorganize it so that it shows all links that exist at a particular time slot. The position data for each satellite is also discretized into time slots to match the discretized link data. Next, using the positions of satellites and links between them at each time slot, we compute the distance for a link at each time slot. Finally, we find the shortest path between the source and destination ground stations for an inter-continental connection over the FSOSN using the NetworkX library in Python [45].

To study the tradeoff between network latency and satellite transmission power in FSOSNs with the Starlink Phase 1 Version 3 and Kuiper Shell 2 constellations, the period of the simulation is taken as 6,000 time slots for both constellations. This encompasses the entire orbital period of a satellite in these FSOSNs, and the duration of a time slot is set to one second.

To calculate satellite transmission power in the two FSOSNs, we consider the laser link model parameters summarized in Table 6. These parameters are used in existing or practical laser satellite communication systems. We consider On-Off Keying (OOK) as the laser link’s modulation scheme. We assume the wavelength as 1,550 nm for both LISL and laser uplink/downlink, since it is widely used for optical satellite communication due to its better performance. We take both transmitter and receiver optical efficiency as 0.8, as this is the common value of these parameters for laser communication terminals (LCTs) and is used in many simulation studies as well. We assume the data rate as 10 Gbps since this is a practical link data rate for optical satellite communications and the maximum data rate that is offered by Mynaric’s LCT [46]. The transmitting divergence angle is set to 15 $\mu$rad and the receiver telescope diameter is 80 mm according to Mynaric’s LCT. We assume both transmitter and receiver pointing errors as 1 $\mu$rad, since these values are widely used in LCTs. We use a curve-fitting technique to find the receiver sensitivity of -35.5 dBm for the OOK modulation with a 10 Gbps data rate and $10^{-12}$ BER from [48]. We assume the ground station altitude as 0.1 km, since this is the general height of tall buildings and towers. We assume clear weather conditions with a thin cirrus cloud concentration of 0.5 cm^-3 and set liquid water content as $3.128\times 10^{-4}$ g/m^-3 according to [10]. We take 20 km as the height of the troposphere layer [50], since this value is practical in laser uplink/downlink. We consider no special weather condition for all the five inter-continental connection pairs. We consider the LM for inter-satellite links as 3 dB and the LM for uplink/downlink as 6 dB, since there is more turbulence and attenuation in uplink/downlink. This yields a received power $\textit{P}_{R}$ of -32.5 dBm for ISLs and -29.5 dBm for uplink/downlink according to (23). To calculate the OP for uplink/downlink, we assume the given SNR threshold as 7 dB, and elevation angle as 30 degrees. As mentioned earlier, we consider three cloud scenarios, each characterized by different parameters: thin cirrus clouds, cirrus clouds, and cumulus cloud concentrations.

To show the relationship and tradeoff between network latency and satellite transmission power, we superimpose the two curves representing $\textit{$\overline{T}$}_{net}$ and $\textit{$\overline{P}$}_{TS,avg}$ for different LISL ranges on a unified plot. For different inter-continental connections, the maximum value of $\textit{$\overline{T}$}_{net}$ (with the minimum corresponding to the $\textit{$\overline{P}$}_{TS,avg}$ case) is realized when the LISL range is at its minimum feasible value for each FSOSN. Conversely, the minimum value of $\textit{$\overline{T}$}_{net}$ (with the maximum corresponding to the $\textit{$\overline{P}$}_{TS,avg}$ case) is observed at the maximum feasible LISL range. Within the plot, the curve of $\textit{$\overline{T}$}_{net}$ gradually descends over LISL ranges, whereas the curve of $\textit{$\overline{P}$}_{TS,avg}$ rises with the increasing LISL range. The point of the intersection between these two curves signifies the juncture of relatively optimal performance between network latency and satellite transmission power can be reached at that LISL range for a given inter-continental connection. Any deviation from this identified LISL range could lead to higher energy consumption or worse latency performance, underscoring the critical importance of maintaining this delicate balance in a FSOSN.

Figs. 5 and 6 show the relationship between $\textit{$\overline{T}$}_{net}$ and $\textit{$\overline{P}$}_{TS,avg}$ for different LISL ranges for the two FSOSNs for the Toronto–Sydney inter-continental connection. For the FSOSN with the Starlink Phase 1 Version 3 constellation, the curves of $\textit{$\overline{T}$}_{net}$ and $\textit{$\overline{P}$}_{TS,avg}$ intersect when the LISL range is approximately 2,900 km, which indicates that $\textit{$\overline{T}$}_{net}$ and $\textit{$\overline{P}$}_{TS,avg}$ are balanced when the LISL range is 2,900 km for satellites in this FSOSN for this inter-continental connection, and $\textit{$\overline{T}$}_{net}$ and $\textit{$\overline{P}$}_{TS,avg}$ are around 135 ms and 380 mW, respectively, at this intersection. For the FSOSN with the Kuiper Shell 2 constellation, $\textit{$\overline{T}$}_{net}$ and $\textit{$\overline{P}$}_{TS,avg}$ intersect when the LISL range is around 3,800 km for this inter-continental connection, and these two parameters are approximately 120 ms and 700 mW, respectively.

In Figs. 7 and 8, we can see the tradeoff between $\textit{$\overline{T}$}_{net}$ and $\textit{$\overline{P}$}_{TS,avg}$ at different LISL ranges in the two FSOSNs for the Toronto–Istanbul inter-continental connection, and Figs. 9 and 10 show this tradeoff in the two FSOSNs for the Toronto–London inter-continental connection. For the Toronto–Istanbul inter-continental connection, $\textit{$\overline{T}$}_{net}$ and $\textit{$\overline{P}$}_{TS,avg}$ intersect when the LISL range is about 2,600 km for the FSOSN with the Starlink Phase 1 Version 3 constellation and 2,900 km for the FSOSN with the Kuiper Shell 2 constellation. For the Toronto to London inter-continental connection, these two parameters intersect when the LISL ranges are 3,400 km and 3,000 km for the FSOSNs with the Starlink Phase 1 Version 3 and Kuiper Shell 2 constellations, respectively. For both FSOSNs in all inter-continental connections, as the LISL range increases, higher transmission power is needed for a satellite but lower network latency is obtained. For $\textit{$\overline{T}$}_{net}$ curve in Fig. 9, it changes sightly between 3000 km to 4000 km LISL range, which is different from $\textit{$\overline{T}$}_{net}$ at the same LISL range for Kuiper Shell 2 constellation as shown in Fig. 10. This is because that the distribution of the satellites for Starlink Phase 1 Version 3 constellation is denser at high latitude comparing to Kuiper Shell 2 constellation, in this way for short inter-continental connections like Toronto to London $\textit{$\overline{T}$}_{net}$ can change unobtrusively at certain LISL ranges.

Figs. 11 and 12 show the relationship between OP and SNR for optical uplink and downlink, respectively. Based on Table 6, we consider three different weather conditions as thin cirrus, cirrus and cumulus clouds. We assume the average SNR as 30 dB, the SNR threshold as 7 dB, the elevation angle of the ground station as 30 degrees and employing Starlink Phase 1 Version 3 constellation, we plot the two figures. As depicted in Figs. 11 and 12, the OP decreases as the SNR increases under cirrus and thin cirrus cloud conditions. Notably, the thin cirrus cloud condition exhibits better performance compared to the cirrus cloud condition in terms of OP. The OP under cumulus cloud condition is 1.0, which indicates an unapproachable optical communication. Comparing the curves of the two figures under same cloud condition, we find that the OP performance for uplink is better than downlink since less atmosphere attenuation ocurrs through uplink. For upink communication, SNR should reach 45 dB to obtain $10^{-8}$ OP performance under thin cirrus cloud condition; while for downlink communication, the requirement of SNR to achieve the same OP performance is 57 dB for thin cirrus cloud condition.

Here we provide important practical insights for FSOSNs based on the analysis in this work.

$\bullet$ For LISL and laser uplink/downlink, the network latency for the shortest path for an inter-continental connection between two ground stations and the average satellite transmission power for the shortest path have an inverse relationship. As the LISL range increases, the links on the shortest path become longer, the propagation distance of the laser links increases, and the transmission power for the laser links increases along with the average satellite transmission power. At the same time, there are fewer direct links and nodes on the shortest path, and accordingly, the total propagation delay of the path decreases, the total node delay of the path decreases, and the network latency of the path decreases. Therefore, according to this reverse relationship, in order to reach less network latency to achieve better network performance, satellites needed to be designed for high energy payload to sustain higher transmission power consumption. This can be accomplished by implementing larger capacity battery, applying dynamic ISLs instead of establishing the links throughout the entire working period.

$\bullet$ Based on the results of the tradeoff analysis between $\textit{$\overline{T}$}_{net}$ and $\textit{$\overline{P}$}_{TS,avg}$ for different LISL ranges for the two FSOSNs for different inter-continental connections, we find that there is no clear relationship between the LISL range when $\textit{$\overline{T}$}_{net}$ and $\textit{$\overline{P}$}_{TS,avg}$ intersect and the length of the inter-continental connection (i.e., the end-to-end distance between the source and destination ground stations for an inter-continental connection over an FSOSN) within the same FSOSN. A shorter inter-continental connection may have a longer LISL range when $\textit{$\overline{T}$}_{net}$ and $\textit{$\overline{P}$}_{TS,avg}$ intersect. As we can observe that there are different LISL ranges where $\textit{$\overline{T}$}_{net}$ and $\textit{$\overline{P}$}_{TS,avg}$ reach a balance for different inter-continental connections over an FSOSN, all inter-continental connections over an FSOSN must be considered simultaneously while determining an appropriate LISL range for the FSOSN to balance the average satellite transmission power and network latency.

$\bullet$ In laser uplink/downlink communications, the performance of the outage probability is inevitably influenced by cloud conditions. In the case of adverse cloud conditions, such as cumulus clouds, the connectivity of the laser link becomes unavailable, irrespective of the SNR. Under these circumstances, it becomes necessary to transmit through alternative laser links between the ground station and satellites, circumventing the area affected by poor cloud conditions, even though this may cost more network latency and higher satellite transmission power. For laser downlink communication, higher SNR is required to sustain an acceptable OP comparing to uplink communication, and more satellite transmission power will be consumed.