Xinhua Li,Huibin Wang,Wenbo Zhao,Qiushi Tian,Zehao Xu,Bo Yang,*

1 Space Star Technology Co.,Ltd,Zhongguancun Environmental Science and Technology Park,Haidian District,Beijing 100095,China

2 China Academy of Space Technology,Beijing 100094,China

Abstract: The next generation of High Throughput Satellite (HTS) will exploit Q/V band for the feeder link and Ka band for the user link. The exploitation of multiple gateways (GWs) is a transmit diversity measure to compensate severe rain attenuation of Q/V band feeder links. This paper introduces a new smart gateway diversity combination scheme that combinates N-active (K = 2) and N + 1 diversity method in detail from the working principle,gateway switching strategy, evaluation of system availability and implementation complexity. In addition, a feeder link outage event prediction algorithm featuring combination scheme is given detailly based on modified time-adaptive linear regression. Compared with the existing N+P scheme,the proposed scheme can provide higher system availability and more stable user’s experience under the condition of properly increasing the complexity of payload design. Compared with Nactive scheme that has been reported,it is much easier to implement in engineering and more cost-effective.As a result, the combination scheme is a realistic solution based on a trade-off between performance improvement and implementation complexity.

Keywords:Q/V frequency band;smart gateway diversity;outage prediction;gateway switching

I. INTRODUCTION

Future Very High Throughput Satellite (VHTS) will play an important role as space-based backbone nodes in space-air-ground network[1]. As the use of traditional satellite frequency band resources tends to be saturated, Q/V frequency band has the advantages of high bandwidth, narrow beam, etc., making it widely regarded as the preferred frequency band for the feeder links of VHTS[2, 3]. It is also regarded as the focus of mainstream satellite operators.The KONNECT VHTS from Eutelsat Communications and the JUPITER 3 VHTS from Hughes are expected to launch in 2022. Both of satellites utilize Q/V-band feeder links and Ka-band user links. The throughput of each system can reach 500Gbps so as to provide enough capacity to serve high speed Internet and in-flight connectivity markets at scale,offering fiber-like services both in terms of price and speed[4].

At present, the configurations of Ka and Q/V frequency band have become a consensus in the VHTS frequency scheme design. Compared with the traditional Ka band HTS, VHTS can completely assign Ka band resources to the users, the system capacity will increase greatly. Moreover, in VHTS system, a single gateway has larger capacity, hence it can manage more user beams than current HTS system. Under the same satellite capacity,fewer gateways will be needed so as to reduce the total cost of the ground segment. However, the use of Q/V band spectrum for the feeder link is more vulnerable to adverse meteorological conditions, which can reach tens of dB due to tropospheric attenuation for more than 0.5%of the connection time[5,6].

In Figure 1,an example of the Cumulative Distribution Function(CDF)of total atmospheric attenuations at 40/50GHz for an earth station located in Beijing is illustrated. In order to provide availability of 99.9%for Beijing GW,a margin of 40 dB would be required in V band by ITU model[7]. At present, the typical rain fading mitigation techniques including Uplink Power Control(ULPC)and Adaptive Coding Modulation(ACM)are fully researched and widely used in Ka band HTS networks,but they might not be sufficient to achieve the required availability(typically greater than 99.9%) on the Q/V band feeder link[8]. Smart Gateway Diversity (SGD) technique applied to Q/V band high throughput satellite systems becomes necessary to ensure the required higher feeder link availability.SGD adopts gateway handover scheme to reroute traffic from the fading GW(s) to the unfading ones exploiting the inherent presences of several spatially separated GWs, which requires gateway interconnection and ad-hoc payload architecture[9]. Currently,the research on SGD is focused on two main schemes,calledN+Pscheme andN-active scheme. In this paper,a new SGD scheme named combination scheme is proposed,which is neither standardN-active norN+P,but a mix of them.

Figure 1. CDF of total atmospheric attenuation for a GW located in Beijing,Q/V band.

II. RELATED WORKS

N+PSGD scheme featuresNoperational gateways plusPidle gateways,which can back-up up toNgateways in outage at the same time. When one of the active GWs is in outage, switching occurs and traffic of the active GW is rerouted to one of the idle GWs. In this scheme, every user beam is served by one gateway at a time. GW diversity mathematically has been analyzed in[10], where authors derive the availability inN+Pscenario. A GW switching scheme is proposed in[11–13], the main characteristic of which is the association of GWs into switching pairs based on ordered Signal-to-Noise Ratio (SNR)and the use of robust Modified Switch and Stay Combining (MSSC) strategy. The Network Control Center(NCC)obtains the SNR of forward links of all the primary and backup GWs and sorts them in descending order[11]. However, the limitation of the scheme in engineering implementation is that the SNR cannot be accurately obtained from active and idle GWs at the same time because the user terminals is connected to active GWs only and cannot receive the traffic from any of idle GWs before updating the payload switching matrix to reallocate new carriers to the beam serviced by the affected GW[14]. Another GW handover strategy is to perform the GW handover based on comparing the channel rain fading predicted value of GW feeder link and the link margin to counteract rain attenuation[15–17],but the known Linear Regression(LR)algorithms of channel rain fading prediction,such as Extreme Value algorithm,Average Value algorithm[15],Adaptive Time Window(ATW)algorithm[17], exist the problem of low sensitivity to the drastic change of rainfall intensity,or the lagging problem of the rain fading predicted value. In addition,all the users connected to the affected gateway must be reallocated to another carrier of one idle gateway during the handover,which may lead to congestion problems at the network level as a large number of users need to be reconnected at the same time and significant recovery time[18].

N-active SGD is another scheme in which a User Terminal(UT)beam is connected not just to one gateway but to a pool of gateways. If a gateway experiences very deep fading, its traffic can be routed to another gateway that has available capacity[10]. The benefit of this scheme is that allNGWs are active and there is no need for idle GWs, but its disadvantage is that the throughput of users served by GWs sharing the traffic from their affected counterpart will be reduced or each GW should have some spare capacity available in order to support other GWs in case of outage;hence GWs need to be oversized in capacity[13, 19].Also,the complexity and the on-board equipment are further increased[20]. In[8,20],transparent satellites can still be used via two options:

(1) Assign one or more separate channels to each gateway(Frequency multiplexing).

(2) Assign one or more separate time slots to each gateway(Time multiplexing).

The comparison of SGD schemes amongN+P,Nactive frequency multiplexing andN-active time multiplexing is summarized in Table 1.

From Table 1,it can be seen that:

(1)N+Pdiversity is a simple and realistic scheme,however,all the users connected to one gateway have to be re-connected to another gateway during GWs handover, which affects all the users who are connected through it. Services which are subject to high availability requirements, such as telemedicine and teleconference,are more susceptible.

(2)N-active(Frequency multiplexing)may lead to some technical difficulties as the overall number of channels can be extremely large(the number of gateways involved in the diversity multiplied by the number of user beams), although it can bring a good user experience.

(3)N-active (Time multiplexing) requires an improved communication protocol standard based on current DVB protocols and accurate time synchronization among payload,gateway and UT,which shall result in redesign of the entire system.

The above-mentioned GW diversity schemes have their own advantages and disadvantages,and it is possible to have an optimized scheme based on a trade-off between performance improvement and complexity of the payload. ForN-active (Frequency multiplexing)scheme, it can be desirable to limit the connectivity between a UT beam and the number of gateways in the pool in order to reduce the complexity of the payload. For example,K(K=2) gateways serve a UT beam whenNgateways are in the pool, namedNactive(K=2)diversity scheme.

Contributions of this paper are:

(1) We put forward a combiningN-active (K=2)withN+1 diversity scheme, abbreviated as combination scheme,which makes a trade-off between performance improvement and complexity of the payload.Building on the results fromN+Pgateway scenario,we analytically derive the average outage probability expression on the feeder link that provides critical insights into system sizing.

(2)The Modified Adaptive Time Window(MATW)algorithm is introduced for rain fading prediction that has higher sensitivity to the drastic change of rainfall intensity,which can achieve accurate prediction of link outage events to ensure GW handover smoothly.

(3)A unified analytical framework and concise formulation are presented in order to evaluate the performance of various SGD schemes in terms of average outage availability,payload complexity and UT beam throughput.

The paper is organized as follows. Section III discusses the principle of combination scheme,GW traffic status transition process and GW switching strategy. Meanwhile, its performance (in particular average outage probability) is analysed and evaluated.Section IV compares the performance ofN+P(N=10,P=2) scheme,N-active (N=10) diversity scheme and the combination scheme(N=10,K=2 andP=1). The conclusion of the paper can be found in section V.

III. COMBINATION GATEWAY TRANSMIT DIVERSITY SCHEME

3.1 Principle

The basic principle ofN-active (K=2) diversity is to serve a User Beam (UB) through the carriers of two different gateways.In other words,two gateways form one gateway cluster. All gateways belonging to the same gateway cluster are connected to an intermediate router which routes traffic to/from the Internet through the terrestrial network. On the basis ofNactive (K=2) scheme, an idle gateway is deployed,which will start to work when any one ofNactive gateways encounters heavy rain attenuation. Figure 2 shows the working status of the combination scheme in sunny days,and Figure 3 shows the working status of the combination scheme under rainfall conditions.

Figure 2. Combination scheme operation in the clear sky.

Figure 3. Combination scheme operation under rain fading.

As shown in Figure 2 and Figure 3, whenNis even, every two gateways are paired, such as GW1 and GW2, and GW (N-1) and GW (N). Every one pair of GWs,named one(1)GW cluster,serves some of the user beam clusters. WhenNis an odd number andNis not equal to 1, there is at least one (1) GW cluster which is composed of three(3)active GWs. In this case, satellite payload requires more signal splitters, filters, and signal combiners. Furthermore, the UT shall be equipped with a multi-carrier receiver so that the terminal can simultaneously receive multiple carriers from three (3) active GWs. Thus, whenNis an odd number, the complexity and cost of satellite payload and user terminal receiver will be much higher than the one asNis an adjacent even number.During overall design of VHTS,the number of active GWs prefers to be an even number as possible. In this paper,we will focus on the combination scheme(Nis an even number)and present our contributions for this scenario.

Table 1. Comparison of various diversity schemes.

3.2 Traffic Status Transition Process

In general, the traffic status transition process during GWs handover can be decomposed in three steps:

(1) Any one ofNactive gateways experiencing heavy fading will switch to the idle gateway.

(2)When another active gateway is in outage at the same time as the first, the users need to switch to the paired gateway.

(3)Upon the resume of the first affected GW or the second outage GW,the traffic is migrated back to the original carrier.

The traffic status transition process during GWs handover is illustrated in Figure 4. GW A and GW B are working in V frequency band for the forward link,and the center frequencies of the transmitting carriers are C1 and C2 respectively as depicted in status a). T1 and T2 represent the flows to the user terminals managed by the respective GW A and GW B. Under the condition of b1), GW A is in outage state. When the idle gateway is available,the traffic managed by GW A is migrated to the carrier C3 of the idle gateway.In the status of b2), when the GW B and GW A are experiencing outage simultaneously,traffic managed by GW B is migrated to idle GW on carrier C3. Since now,idle GW has to serve temporarily traffic T1 and T2,both of which may have to adapt the Quality of Service (QoS) policy to the current circumstances. The data rate of individual UT that connects idle GW drops by half accordingly. If the GW B recovers firstly from status b2),the traffic T2 is migrated back to C2,which is shown with the dotted line in Figure 4.When GW A is available in status c),the traffic T1 is migrated back to the original carrier C1 from status b1).

Figure 4. Traffic status transition process GWs handover.

When the idle gateway has already occupied or been in outage due to the rainfall attenuation and GW A is in outage because of heavy rain fading, traffic managed by GW A is migrated to carrier C2 managed by GW B,which is shown in b3). As soon as GW A is available,traffic T1 is migrated back to the original carrier C1,as shown in c).

The GW traffic state transition process for the return link is similar with the forward link.

3.3 GW Switching Strategy

3.3.1N+1 Diversity GW Handover Procedure

In this paper, the GW handover strategy based on the sorting of rain fading predicted value is adopted.All GWs predict their channels in advance and sort the GWs in accordance with the predicted rain fading value to make the decision by threshold, so that the GW complete the handover before the outage of feeder link and ensure the stability of traffic transmission. A MATW algorithm is used to predict the channel rain fading. By introducing a judgement strategy of predicted value inflection point, the algorithm becomes sensitive to the drastic changes of rainfall intensity,and the predicted value of rain fading is closer to the actual value.

The Gateway Switching Controller (GSC) is deployed in NCC which plays the important role during gateways handover. The Figure 5 shows the procedure of gateways handover. GSC is to predict the rain fading and make a switchover judgement based on the obtained date from antenna tracking receiver of each active GW (from step 1 to step 3). Then, GSC will send the switchover instruction to the affected GW to mute its RF and send the switchover instruction to Satellite Control Center(SCC)to switch feeder beam from satellite (step 4 and step 5). Finally, GSC will send the instruction to the idle GW to enable its RF(step 6). Switching operation latency is the time required by the system to complete all necessary handover procedure of traffic between two GWs during the switchover (step 4 to step 6 in Figure 5), which can be set to 30s to evaluate the performance of the switching algorithm[15]. In order to ensure that the switching process of two GWs is completed before the actual GW outage,the channel prediction duration should be larger than switching operation latency,and 60s is recommended[15].

Figure 5. Gateways handover procedure.

The GSC is deployed in NCC, including two modules of channel prediction and switching decision.The flow chart of switchover decision strategy is shown in Figure 6.

Figure 6. Flow chart of switchover decision strategy for N+1 diversity.

Channel prediction module obtains the current power level of tracking receiver of each GW in real time, the predicted value of rain fading of each GW at a certain time in the future is calculated through proposed algorithm, and the value is sent to the switchover decision module as reference information for handover. Its working process and principle are as follows:

(1)Obtain the power level of tracking receiver: The channel prediction module collects the tracking receiver level vectorH1(t),H2(t),...,HN(t) of active GWs and the level vectorH′(t)of the idle GW in the past time period [t-W,t] by Equipment Monitoring System (EMS), unit is dBm,Wis the sampling time window,unit is s.

(2) Calculate value of rain fading: Take the maximum level value recorded by tracking receivers of each GW in sunny days as the reference valuehjbase,subtract the receiving level vectorHj(t)of tracking receivers of each GW from this value, then get the rain fading value vectorIj(t)=hjbase−Hj(t)of each GW in the past time period[t-W,t],the unit is dB.

(3) Predict value of rain fading: The channel prediction module adopts the MATW algorithm to predict the rain fading value ofj(t+T) at the timet+Tbased on the rain fading vectorIj(t)=[Aj(t −W)...Aj(t)],whereTis the channel prediction time window.

Switchover judgment module performs GW switchover or switchover back in accordance with the predicted value of rain fading provided by the channel prediction module. Its working process and principle are as follows:

(1) Switchover or switchover back judgment: The GSC determines to start a sub-flow of active-idle GW switchover or a sub-flow of idle-active GW switchover back based on the availability of the idle GW.

(2)Active-idle GW switchover(whenB′=0)

a. Sort predicted value of rain fading: sort the predicted rain fading values of all active GWs in ascending order. For each active GW,its corresponding predicted rain fading value is denoted asSN, and the GW is denoted asGSN. Thus, the definition can be obtained:

b. Match predicted value of rain fading: Match the idle GW′GS′to the active GW′GSN′under the largest rain attenuation,the matched result is shown in Table 2.

Table 2. Matched result of active and idle GW.

c. GW switchover judgment: The predicted rain fading value is compared with the switching threshold, among which the switching thresholds of different GWs are different. The switching threshold can be set according to typical link budget of VHTS system,and the typical value is 25dB[17]. Once the predicted rain fading value of the active GW exceed the switching threshold value (SN>AthN) for the first time,and the predicted rain fading value of the idle GW is lower than the switching threshold value,switchover will be performed fromGSNtoand the switchover information will be recorded.

(3)Idle-active GW switchover back(whenB′=1)

a. Obtain traffic status of the GW(to be switchover back): obtain traffic status of the GW that is to be switchover back and its predicted rain fading value.

b. GW switchover back judgment: The predicted rain fading value is compared with the switching threshold. If the observed condition of the original active GW satisfies the condition that its predicted rain fading values of the active GW are lower than the threshold value(SN

The MATW algorithm is based on the linear regression model,the model formula is as follows:

The optimal sampling time windowW∗can be defined as the following form:

whereσis variance of White Gaussian Noiseξ(t),σ1andσ0are standard deviations of White Gaussian Noise Random Walk[17].,ais the sampling interval(sampling interval of tracking receiver level of each GW,typical value is 5s),nis the number of sampling points in sampling time windowW,n=[W/a].

In (2) and (3) equations, [W] is the largest integer that does not exceedW;Y(i) is the log of the rain fading value of the sampling point in the time sampling window (log base e),iis the timing within the time window,i ∈[t −[W]+1,t −[W]+2,...,t].

The operation process of the MATW algorithm is shown in Figure 7, with detailed description as follows:

Figure 7. The operation process of the MATW algorithm.

(1) Set the initial sampling time window (W0),channel prediction time sampling window (T),switching threshold(Ath).

(2)According to the input requirements of the linear regression model, logarithmic transformation is performed on the rain fading values of sampling points in the time sample windowY(i) = ln(A(i)),i ∈[t −[W] + 1,t −[W] + 2,...,t], then get (Y(t −[W]+1),Y(t −[W]+2),...,Y(t)).

(3)In a sampling time window, calculate the linear regression parameters based on the linear regression parameter formula(m(t),q(t),σ,σ1,σ0),wherem(t)∼N(0),q(t)∼N(0),ξ(t)∼N(0,t·σ2).

(4) Definingt0as present timing,m(t0) andq(t0)is obtained by the result of the previous step(3). According to(t0+T)=m(t0)∗(t0+T)+q(t0),the result of(t0+T)is calculated to get the logarithm of rain fading prediction value.

(5) According to the sampling window parameter formula, the optimal sampling window length is determined. The optimal sampling time windowW∗can be calculated in formula(4).

(6) Judge whether the result of predicted value(t+T)has changed monotonously atXconsecutive points within(T-∆t)time(Xis the largest integer not exceeding (T−∆t)/a, ∆tis required time for GW switchover,typical values is 30s;typical values for′T′is 60s;typical values for′a′is 5s;so typical values for′X′is 6). If the monotony gets bigger or smaller (no inflection point), the sampling windowWis updated to the calculation result of the previous stepW∗, and the counting number is cleared. If there is an inflection point withinXpoints(refers to no longer changing monotonously), then the sampling window is not updated and the counting number is cleared.When the counting number is less thanX,the sampling window is not updated and the counting number is increased by one(1).

(7)The logarithmic inverse transformation of(t+T)is performed to obtain the rain fading value(t+T)=.

(8) By calling this algorithmjtimes, the predicted rain fading values(t+T)at timet+TofjGWs can be calculated,jgoes from 1 toN+1. For sake of simplicity, set=, then the predicted values ofN+1 GWs at the same time can be expressed as.

The ATW algorithm is to update the sampling window in real time[17]. It leads to the waste of computing resources and the lag of prediction trend. The proposed solution of the MATW algorithm is to introduce the judgment of the change trend of the predicted value within 30 seconds. When the continuous predicted value within 30 seconds is higher than the previous predicted value, or when the continuous predicted value within 30 seconds is lower than the previous predicted value,the sampling window is updated. Otherwise, the length of the sampling window remains unchanged,and the last sampling value in the sampling window is the real value at the current moment. This sampling method is more suitable for the actual requirement of prediction,especially for the inflection point of rainfall,and it can accurately predict the variation trend of rain fading.

Using the ATW algorithm which have been published before and the MATW algorithm in this paper,to predict the rain fading value at Chengdu GW of Chinasat-16 within a certain period of time, the predicted and actual value curves are shown in Figure 8. The x-coordinate of the 2000s in this figure corresponds to the 17h31m09s,August 30th,2019.The initial sampling window of rain fading prediction is 40s,and the sampling interval is 5s. It can be seen from the figure that the ATW algorithm has lag phenomenon in predicting the overall increase and decrease trend of rain fading,while the MATW algorithm introduced in this paper has little lag phenomenon and has higher sensitivity to the drastic change of rainfall intensity,which is closer to the actual value of rain fading.

3.3.2N-active (K= 2) Diversity Handover Procedure

Figure 8. Comparison among predicted values of rain attenuation using ATW algorithm and MATW algorithm,and actual values of rain attenuation.

Assuming that the idle gateway has already been occupied or in outage due to the heavy rain fading,NCC makes the decision to re-route the data traffic from the outage GW to the other GW in the same GW cluster when the predicted rain fading value of either GWs exceeds the switching threshold. The implementation of the change of carriers is to be achieved by signalling it in the overhead of the protocol,and the routing tables in the terrestrial network shall be updated simultaneously.

Figure 9 and Figure 10 illustrate the change of traffic flows during GW handover process. As shown in Figure 9,under normal operating conditions in sunny days, each gateway transmits two sub-streams flowing to user beam 1 and user beam 2 respectively. The carrier switching payload on the satellite directly allocates these sub-streams to an appropriate carrier of the corresponding beam. Assuming that GW1 is in outage due to heavy rainfall as illustrated in Figure 10,its data traffic will be rerouted to GW2 via the intermediate router which receives the rerouting command sent by NCC. With that, all traffics are forwarded to user beams through GW2 using the carriers that have already allocated to transmit the normal traffic towards the corresponding user beam, and have to adapt the QoS to the current circumstances. The switching process needs to be supported by a cross-layer mechanism based on the Multi-path Transmission Control Protocol(MP-TCP)[21,22].

Figure 9. The two carriers of a GW are switched by the satellite towards different user beams in the clear sky.

Figure 10. All traffic flows go through GW2 under GW1 rain fading.

3.4 Average Outage Probability

For theN+Pscheme,the probability of a user inside a UT beam to be in outage can be obtained as(5)and(6)[10].

If up toPgateways are simultaneously in outage,all user beams can still be served in theN+Pscheme. IfP+igateways are in outage simultaneously,a fractioni/Nof the user beams will not be served by any gateway. From the view of UT, it results in a probabilityi/Nof outage ifP+igateway are unavailable, as it can be assumed that an idle gateway can be allocated randomly to any of theigateways in outage.

The equation (6) cannot be used to calculate the probability of user outage in the combination scheme,because the number of outage gateways in the system is no longer necessarily related to whether the user is outage or not. In other words,the number of gateways in outage is no longer the key factor to judge whether the user is in outage situation. The decisive factor is whether the paired gateways in the system are in outage status at the same time due to fading events. The user will be in outage status only when the paired gateways are in outage and the idle gateway is in outage or occupied as well. Therefore, the availability calculation of the combination scheme can be divided into two steps:

(1) Calculate the outage probability of the paired gateways.

(2) Consider the paired gateways and an idle gateway asN+Psystem,and then the availability model ofN+Pscheme is used for calculation.

The equivalent model of outage probability is shown in Figure 11. After pairing, the number of the active gateway pairs isN′assuming that the number of the active gateways isN= 2N′. In order to simplify the calculation and find out the optimal situation that the system can achieve under the equivalent model, the following assumptions are made:

Figure 11. The equivalent model of outage probability calculation for combination scheme.

(1) The same probability distribution of rainfall events and outage probability are adopted for active and idle gateways,which is set asp0.

(2)The distance between gateways(including active gateways and an idle gateway) is more than 100km,and rainfall events are independent of each other[11].

According to the above assumptions,it can be concluded that:

Therefore,the probability of outageipairs of gateways andjidle gateways in the system is as follows:

where,i ∈[0,N′]∧i ∈N∗,j ∈[0,P]∧j∈N∗.

Therefore, the outage probability of combination scheme(P=1)is:

wherep0is the outage probability of the active or idle gateway(s),N′is the number of the active gateway pairs in the system.

Assuming that the number of active gatewaysN=10,20 and 50, we draw the curve of availability of a user varying with the outage probability of a single gateway,which is shown in Figure 12.The availability of a user (ideal UT link) is greater than 99.9% even when the outage probability of each gateway is equal to 7%and the number of active gateways reaches 50.

Figure 12. A user’s availability vs. gateway’s outage probability for combination scheme.

With an idle gateway, the maximum number of active gatewayNthat the system can support has been analyzed, while ensuring that the user’s availability higher than 99.9%. Using equation (9), we calculate the user availability under the condition that the unavailability of each gateway is 7%, 5%, and 3% respectively. Some calculation results are shown in the Table 3.

Table 3. The availability of a user varying outage probability of each gateway for combination scheme.

The results show that when the outage probability of each gateway is less than 7%, the system can support at least 132 active gateways theoretically,meeting the requirement that the availability of a user (ideal UT link)is greater than or equal to 99.9%. In fact, it barely needs so many active gateways to construct a VHTS system, so an idle gateway is sufficient if the combination scheme is adopted.

IV. PERFORMANCE COMPARISION OF VARIOUS DIVERSITY SCHEMES

In order to compare the performance among the proposed combination scheme,publishedN+Pscheme andN-active scheme,the following assumptions for a VHTS system are made:

(1) ForN-active scheme, the number of the active gatewaysN=10.

(2)ForN+Pscheme,the number of the active gatewaysN=10 and the member of idle gatewaysP=2.

(3)For combination scheme, the number of the active gatewaysN=10 and the number of idle gatewaysP=1.

(4)N-active gateways andPidle gateways are more than 100 km apart from each other,their rainfall events can be considered irrelevant.

(5) The system availability is greater than 99.9%.The feeder link availability of each gateway is equal,ranging from 97%to 98%,and the UT link is ideal.

(6) The number of user beams in the systemM=160.

(7)Uplink of feeder link: 8GHz bandwidth per GW(47.2～50.2GHz and 50.4～51.4GHz,total 4GHz per polarization).

(8) Downlink of feeder link: 4GHz bandwidth per GW(37.5～39.5GHz,total 2GHz per polarization).

(9) Uplink of user link: 250MHz per beam (8 frequency/polarization reuse).

(10) Downlink of user link: 500MHz per beam (8frequency/polarization reuse).

4.1 Average Outage Probability Comparison

ForN+Pscheme,according to equation(6),the outage probability of a user inside a UT beam can be calculated, and the calculation results are shown in the Table 4.

Table 4. The availability of a user varying outage probability of each gateway for N+P scheme.

From Table 4, we can see that at least two (2) idle gateways are needed to ensure that the system availability of is greater than 99.9% (ideal UT link), assuming the availability of each gateway is from 97%to 98%.

ForN-active scheme, the outage probability of a user inside a UT beam equals to the probability of all active gateways that serve it to experience an outage.In this case, each gateway has an outage probabilitypi. Then:

Thus, we draw the user’s availability curve changing with the outage probability of a single gateway for the three (3) schemes. The results are shown in Figure 13. The simulation results show that the combination scheme is close to theN-active scheme,and it is much better than theN+P(P=2)scheme in terms of system availability.

Figure 13. A user’s availability vs. gateway’s outage probability for combination scheme.

4.2 Payload Complexity Comparison

In this paper, the payload design of different diversity schemes with the forward link payload design are compared, because it is the most challenging part of the diversity schemes. The concepts discussed are intended to be also applicable to the return link payload design.

4.2.1N+P(P=2)Scheme Forward Link Payload

For the forward link,the special feature of theN+Pscheme payload is the microwave switching matrix.The function of the switching matrix is to switch between the active gateways and idle gateways. After the signal transmitted by gateways is selected by the switching matrix, the signal is amplified by the Low Noise Amplifier (LNA), and then the frequency will be transformed from the V-band signal to the Ka band signal. The Ka band signal passes through the channel filter,the interference will be filtered out. Finally,it is sent to the free space through the antenna after being amplified by TWTA high power amplifier. The forward payload block diagram is shown in Figure 14.

Figure 14. Forward link payload block diagram of N+P(P=2)scheme.

4.2.2N-active Scheme Forward Link Payload

Under theN-active scheme,if the microwave network is chosen to implement the payload,the bandwidth of a single user beam with 500MHz needs to be divided into 10 sub-bands. Each sub-band has a bandwidth of 50MHz and a protection band of 5MHz. Therefore,the utilization rate of available frequency resources decreases. In addition,if theN-active is adopted,a user beam corresponds toNgateways. With the increase ofN,the payload system complexity continues to increase. As a result, the number of forward payload channels will increaseNtimes,and the number of return payload channels will increaseNtimes as well.Assuming that VHTS system has 160 user beams,and the number of channels is normally 160 channels,both the number of forward payload channels and the number of return payload channels areN*160 after the usage ofN-active scheme. Frequency resources and weight resources will be seriously wasted. Furthermore,it is almost impossible to implement practically.

N-active scheme forward link payload diagram is shown in Figure 15. The V-band signal is converted to low-frequency signal by receiving frequency conversion module after the LNA so as to realize the payload through Digital Transparent Process(DTP)technology. DTP namely Spaceflex is an advanced repeater sub-systems having analogue to-digital (ADC)and digital-to-analogue (DAC) channelizers on their input and output accesses and making extensive use of digital processing to support channel routing with fine bandwidth granularity[23]. There are 10 active gateways in the VHTS system architecture, and each gateway has 12GHz frequency resources including 8GHz uplink bandwidth and 4GHz downlink bandwidth. Due to the current generation of DTP (Space-Flex 24)presents up to 48 implemented I/O ports with up to 500 MHz bandwidth interface and 24 GHz routing capacity[23],five(5)digital transparent processors can meet the system requirements under this architecture. It can guarantee the smaller channel granularity and the exchange flexibility through the digital transparent processor. However, the disadvantage of this digital device is its exorbitant price.

Figure 15. Forward link payload block diagram of N-active scheme.

4.2.3 Combination Scheme Forward Link Payload

The key payload of combination scheme includes microwave switching matrix, power divider and fusion microwave network. 10 out of 11 gateways are selected by microwave switching matrix as active gateways to receive 20 channels of dual-polarization Vband signals. The received V-band signals are amplified by LNA,and each signal is divided into two channels by power divider. After the power division, the two channel signals enter into the fusion microwave network and are synthesized with the two channel signals from paired active gateway to form a new frequency combination. After that, the two channel signals with recombination frequency are transformed into Ka frequency band through V/Ka frequency converter. Finally, it is sent to the free space through the antenna after being amplified by TWTA high power amplifier. The forward payload diagram is shown in Figure 16.

Figure 16. Forward link payload block diagram of combination scheme.

Based on the above payload design and relevant engineering experience,payload budget results of different diversity schemes are calculated, which is shown in Table 5.

Table 5. Payload budget comparison table(N=10).

4.3 Other Performance Comparison

The performance comparison among various diversity schemes can also be conducted from the user side,including the UT beam throughput, complexity of UT and user experience.

In term of UT beam throughput, when all the gateways are operational, the throughput of a UT beam is 100% of its nominal throughput. The UT beam throughput will be reduced to a fraction of its nominal throughput due to the outage of gateway(s). For the combination scheme, the CDF of the fraction of the normal throughput for one beam can be calculated by:

wherepis the outage probability of each gateway,bis the instantaneous fraction of a UT beam throughput provided by the paired gateways.

According to equation (12), the UT beam throughput can achieve 100% of its nominal throughput for 99.88%of the time and no less than 50%of the normal throughput for more than 99.99%of the time assuming that each gateway is operational for 98% of the time.ForN-active scheme, the CDF of the fraction of the normal throughput for one beam and the simulation results are given in[18]. Considering that each gateway is operational for 98% of the time, no less than 70%of the normal throughput can still be provided for more than 99.99%of the time,and 100%of its nominal throughput can be achieved only for 81.71%of the time. Thus, in terms of UT beam throughput performance,the combination scheme andN-active scheme are roughly equivalent. ForN+Pscheme, the fraction of UT beam normal throughput is either 100%or 0%. It is the most volatile and unstable compared with the other schemes.

In term of the complexity of UT and user experience, the combination scheme is a medium scheme compared with other SGD schemes,which can be seen in Table 6.

Table 6. Other performance comparison table.

4.4 Summary of the Above Comparison

ForN+Pscheme,the satellite payload,gateways and the terminals are simplest to be realized. However,the gateways handover will cause the interruption of the user terminal business. Moreover,when the number of active GWs is given, the more idle GWs are needed to achieve the same system availability than other schemes.

ForN-active scheme, the user terminal experience is the best. However, the system design complexity,satellite payload and user terminal complexity will be increased greatly.

The combination scheme can be considered as acompromise design between theN+Pscheme andN-active scheme. It guarantees the user terminal experience under the premise of slightly increased complexity on both satellite payload and the user terminal.

V. CONCLUSION

The next generation VHTS system will be built on the SGD architecture utilizing Q/V-band feeder links and Ka-band user links so as to achieve very high system availability that is higher than 99.9%.In this paper,the combination scheme ofN-active(K=2)withN+1 diversity is proposed including its working principle and traffic status transition process during GWs handover.The GW handover strategy based on the sorting of rain fading predicted value is adopted. The MATW algorithm introduced in this paper for rain fading prediction will not lead to lag phenomenon and has higher sensitivity to the drastic change of rainfall intensity,which is closer to the actual value of rain fading. In addition, the system availability formula is deduced and simulated, and simulation results show that an idle gateway is sufficient for the combination scheme.Compared withN+PandN-active scheme,the combination scheme is a feasible and realistic scheme,which can well balance the system design complexity, payload complexity, system availability and user experience.

ACKNOWLEDGEMENT

This work was supported by the Key Pre-Research Project for Civil Space Technology(No. B0106).