Deep-Space Ka-band Link: Design, Continuity and Completeness. Shervin Shambayati Jet Propulsion Laboratory, California Institute of Technology 4800 Oak Grove Dr. Pasadena, CA 91109 E-mail:
[email protected] Abstract— The Mars Reconnaissance Orbiter (MRO) Ka- band Demonstration has been indefinitely postponed and its objectives must now be met through other means. One of these objectives is the evaluation of the continuity and com- pleteness performance of the deep-space Ka-band link for dif- ferent link design criteria. To meet this objective, the data from the Water Vapor Radiometers (WVR) and the Advanced Water Vapor Radiometers (AWVR) at the three Deep Space Network (DSN) communication complexes were used. Along with these data, MRO’s DSN antenna allocation schedule, Earth-MRO geometry and telecom parameters from MRO were utilized to emulate the Ka-band link performance over a ten-month period. One pass per week per complex was selected for a total of 129 passes (43 passes per complex). For each pass, at most two data rates were chosen such that the expected data return relative to the monthly atmospheric noise temperature distribution for the given complex would be maximized subject to different minimum availability re- quirements (MARs). The performance of the link was mea- sured in terms of data return, data loss, effective data rate, link availability, number of good periods, number of bad peri- ods, good and bad period duration statistics, number of passes with outages and link stability. As expected, as the MAR was increased, the link availability and link stability increased. However, even with a MAR of 99%, 16 passes suffered out- ages due to weather effects. The data return remained rela- tively the same for MARs between 10% and 80% but it de- clined rapidly as the MAR approached 99%. These results in- dicate that a simple margin policy cannot guarantee data com- pleteness and retransmissions must be used. Given that some form of retransmission has to be used with Ka-band, it is also recommended that the link be designed with approximately 80% MAR so that the data loss could be reduced substan- tially without incurring a significant penalty in data return. TABLE OF CONTENTS 1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 MRO KA-BAND CAPABILITIES . . . . . . . . . . . . . . . . . . . 2 3 METHODOLOGY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 4 PERFORMANCE METRICS . . . . . . . . . . . . . . . . . . . . . . . . 3 5 RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1 1-4244-1488-1/08/$25.00 c©2008 IEEE 2 IEEEAC Paper #1042, Final Version 1, Updated 8/11/2007. 6 CONCLUSIONS AND CAVEATS . . . . . . . . . . . . . . . . . . . . 8 ACKNOWLEDGMENTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 BIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1. INTRODUCTION Because of limited (50 MHz) spectrum allocation at 8.41- GHz X-band for deep-space missions’ downlink operations, future high-rate deep-space missions will be using the 500- MHz spectrum allocation at 32-GHz Ka-band for their down- link. Since Ka-band is much more susceptible to adverse weather events than X-band, a different operations concept was needed. This operations concept, based on maximiz- ing the expected data return during a pass, was to have been demonstrated with Mars Reconnaissance Orbiter (MRO) dur- ing its primary science phase (PSP) [3]. For this purpose, MRO carried a complete Ka-band telecommunications suite. During MRO cruise phase, the functionality of this suite was fully verified and some limited Ka-band performance data were obtained [5][6]. However, on May 24, 2006, during aer- obraking in Mars orbit, MRO’s primary Ka-band chain failed. Although a backup was available, on January 24, 2007, it was decided to suspend MRO Ka-band operational demonstration indefinitely because of possible risk to the mission. Since no Ka-band tracking was performed during the PSP, the demon- stration could not take place. Therefore, attempts were made to meet the objectives of the demonstration through other means. The main objective of the demonstration was to evaluate the performance of the deep-space Ka-band link designed ac- cording to Ka-band operations concept in terms of its data re- turn, completeness and continuity. Lacking Ka-band teleme- try from deep-space missions, it was decided to use the atmo- spheric noise temperature measurements from Water Vapor Radiometers (WVRs) and Advanced Water Vapor Radiome- ters (AWVRs) at the three Deep Space Network (DSN) com- munication complexes at Goldstone, California; near Can- berra, Australia and near Madrid, Spain. In this approach, passes were selected from MRO’s DSN schedule and Earth- MRO geometry and the parameters for MRO Ka-band com- munications capabilities were used to design the link. The WVR/AWVR data were used along with models for the gain and system noise temperature (SNT) of DSN Ka-band capa- 1 ble antennas [2] to emulate the performance of the designed Ka-band link. Using this approach, 43 passes per complex (for a total 129 passes) were selected over a ten-month period from April 1, 2006 through January 30, 2007, and the performance of the link over these passes was calculated individually, on a per complex basis and in the aggregate. The performance mea- sures include statistics on the duration and the number of good and bad periods, data return and data loss, link avail- ability, effective data rate, number of passes with outages and link “stability.” This paper is a report on the results of these calculations. Based on these results, almost regardless of the margin carried on the Ka-band link, the link will suffer out- ages due to weather and therefore, a retransmission mecha- nism must be used to insure completeness of the data. In addi- tion, different complexes seem to have different outage char- acteristics and the nature of outages at each complex should be studied further. The paper is organized as follows: In Section 2, MRO Ka- band link capabilities are discussed. Section 3 details the methodology used in this paper. Section 4 describes the met- rics used for evaluating the link performance. The results are discussed in Section 5. In Section 6, conclusions are reached and caveats are discussed. 2. MRO KA-BAND CAPABILITIES For the purpose of this study, MRO Ka-band capabilities were selected to design the link and to emulate the link perfor- mance. MRO carries a 35-W RF Ka-band Traveling-Wave Tube Amplifier (TWTA) transmitting over a 3-m parabolic antenna producing an equivalent isotropic radiated power (EIRP) of 101.3 dBm. Because the output symbol rate (the combined encoded bit rate and the parity bit rate) is limited to 6 megasymbols per second (Msps) and the turbo decod- ing speed on the ground is limited to 1.5 megabits per second (Mbps) different codes are used with different data rates. For science rates up to 1.5 Mbps (including headers but not par- ity bits) turbo codes are used. For data rates from 1.5 Mbps to 2.61 Mbps, the standard NASA (7, 1/2) code concatenated with Reed-Solomon (255,223) code with interleaving depth 5 (referred to as “concatenated coding”) is used. Concatenated coding is also used for the engineering rate of 27.8 Kbps. For data rates greater than 2.61 Mbps, Reed-Solomon-only cod- ing is used. Table 1 shows the information data rates (actual data without headers) and the associated codes that are used in this study. Note that in cases where multiple coding types could be used for the same information data rate, the most powerful code (the code with the lowest required information bit signal-to-noise ratio [Eb/N0] for the required frame error rate) is selected. 3. METHODOLOGY The methodology used in this paper is illustrated through an example. For this example, we have selected the pass on day 200, year 2006 (2006-200) at DSS-26. First the distance information is used to calculate the required gain-to-system-temperature ratio (G/T) for every data rate (Fig. 1). Table 1. MRO Data Rates and Associated Channel Codes Information Data Rate (bps) Channel Code 27846.76 Concatenated Coding 331397.95 (8920, 1/6) Turbo Code 497096.92 (8920, 1/6) Turbo Code 662795.89 (8920, 1/6) Turbo Code 745645.38 (8920, 1/2) Turbo Code 994193.84 (8920, 1/3) Turbo Code 1325591.78 (8920, 1/3) Turbo Code 1491290.75 (8920, 1/2) Turbo Code 1740422.20 Concatenated Coding 2610633.31 Concatenated Coding 2871696.64 Reed-Solomon-Only 3480844.41 Reed-Solomon-Only 5221266.61 Reed-Solomon-Only 40 45 50 55 60 65 70 10000 100000 1000000 10000000 Information Data Rate (bps) R eq ui re d G /T (d B) Figure 1. Required G/T for Different Data Rates, Day 2006- 200 After the minimum availability requirement (MAR) is se- lected, the G/T profile associated for that MAR is calculated for the pass (black curve in Fig. 2 for 90% MAR) using zenith atmospheric noise temperature statistics [1] and equations for the 34-m beam waveguide antennas system noise tempera- tures and gains [2] along with the elevation profile for the pass. Using the G/T profile at most two data rates are selected such that the expected data return over the pass is maximized sub- ject to the minimum availability requirement in the manner outlined in [3]. The required G/T for the selected data rates 2 35 40 45 50 55 60 65 70 75 17 19 21 23 25 27 29 Time (hours UTC) G/ T (dB - K^ - 1) -0.1 0.1 0.3 0.5 0.7 0.9 1.1 St at u s (G o o d= 1, Ba d= 0) Required G/T, 90% Actual G/T 90% G/T Status Figure 2. Day 2006-200, DSS-26, Required G/T, Actual G/T, 90% G/T and Link Status with 90% MAR is shown as the green curve in Fig. 2. Note that the required G/T (the green curve) is always less than or equal to the 90% G/T in Fig. 2, indicating that the selected data rates meet the 90% MAR. Once the data rates are selected, the WVR/AWVR data are then used to emulate the actual G/T during the pass (the blue curve in Fig. 2). It should be noted that the WVR/AWVR data are sampled at approximately once every five minutes. In order to achieve better temporal resolutions, the link G/T was emulated once every minute using linear interpolations between two WVR/AWVR data points. After the actual G/T is calculated the status profile of the link over the pass is determined. The status of the link is “good” when the actual G/T (the blue curve in Fig. 2) is greater than or equal to the required G/T (the green curve in Fig. 2) and is “bad” when the actual G/T is less than the required G/T. For the pass on day 2006-200, the status of the link is indicated by the red curve in Fig 2. From the link status, duration of “good” and “bad” periods as well as data return and data loss is determined for the pass. By looking at the ensemble of passes, overall outage and avail- ability statistics as well as data return and data loss for the link design approach taken are evaluated. By selecting dif- ferent MARs, the tradeoff between continuity, completeness and data return is performed. In the next section a brief de- scription of metrics used for evaluating the performance of the link is given. 4. PERFORMANCE METRICS There are several measures of performance by which the link is evaluated. The following is a brief explanation of these metrics and why they are important: 1. Data return and data loss for individual complexes and overall: the primary measures of performance of the link are data return and data loss. Data return is defined as the total amount of data received correctly by ground stations. Data loss is definied as the total amount of data transmitted by the spacecraft but not received correctly. A scheme that returns the maximum amount of data with tolerable loss of data is preferred for science missions. In addition, these two mea- sures allow us to perform a tradeoff between different MAR values. 2. Effective data rate: this quantity is obtained by dividing the data return by total track time and could be calculated per pass, per complex and overall. This measure is especially useful in comparing the performance between complexes be- cause the geometry of the link and the pass schedule could divide the total tracking time unequally among the complexes thus skewing the data return results. 3. Number of good and bad periods: this metric is impor- tant since many retransmission schemes are more concerned with the number of retransmission requests as opposed to the amount of data retransmitted. Also, in some cases, the out- ages cause the receiver to go out of lock and the receiver reac- quisition times (not taken into account in this study) lead to additional data loss. 4. Number of passes with outages: along with the number of good periods and bad periods, this metric provides informa- tion about how good periods and bad periods are distributed among passes. 5. Good period and Bad period statistics: this information in the form of average duration and standard deviation is helpful in understanding the nature of outages and good periods. 6. Availability: this quantity is the ratio of the time that the link is in the good state to the total track time. This is the standard measure to which links are designed. 7. “Stability”: this is an extension of the idea of availability. Mathematically, stability is defined as: 3 0 500 1000 1500 2000 2500 3000 3500 0 20 40 60 80 100 Minimum Availability Requirement (%) D at a Vo lu m e (G bi ts ) Actual Data Return, Goldstone Expected Data Return, Goldstone Actual Data Return, Canberra Expected Data Return, Canberra Actual Data Return, Madrid Expected Data Return, Madrid Figure 3. Expected and Actual Data Return per Complex Ψ(τ ) = Ng∑ i=1 t (g) i Iτ ( t (g) i ) ttot (1) where ttot is the total track time; Ng is the number of good periods observed during that time; t(g)i is the duration of the ith good period and Iτ (·) is an indicator function given by Iτ (t) = 1 t ≥ τ0 otherwise (2) By this definition, Ψ(τ ) is the fraction of time that the link is in a good state of duration greater than or equal to τ . This concept was first introduced in [4]. Availability is Ψ(0). 5. RESULTS The results are analyzed based on the metrics presented in the previous section. These are evaluated in terms of variations in MAR. In addition, some of the results are expanded upon by using specific passes and fixed MARs as examples. Fig. 3 shows the expected and the actual data return as func- tions of MAR. Similarly, Fig. 4 shows the expected and the actual data loss as functions of MAR. Note that because the link design method is based on maximizing the expected data return, the expected data return and data loss numbers are readily available. As seen from these figures, The data return for Goldstone and Canberra is greater than expected while for Madrid it is than expected. Conversely, the data losses for Madrid are uniformly much greater than expected and the data losses for Goldstone and Canberra are lower than ex- pected. This indicates that Madrid had worse than nominal weather and Goldstone and Canberra had better than nominal weather over the analysis period. However, when the data re- turn for all the complexes is looked at together, the expected data return and the actual data return are nearly the same. Similarly, the actual losses are only slightly higher than ex- pected (Fig. 5). Whether or not this indicates that Madrid weather is anti-correlated with the weather at Goldstone and Canberra requires further analysis. However, this clearly in- dicates that in this instance, if a retransmission scheme and large spacecraft storage were used, the link would have re- turned roughly the same amount of data as expected. For all complexes the data return is relatively the same for MARs up to 80%. Beyond 80% the data return rapidly drops off. In terms of data loss, however, there is a noticeable re- duction in loss for MARs greater than 50%. This suggests that a MAR of 80% a good design point in that it has the least amount of data loss while providing nearly maximum data return. Goldstone has the highest effective data rate for the same MAR and Madrid the lowest (Fig. 6). Goldstone and Can- berra have higher effective data rates than expected while Madrid has a lower effective data rate than expected, again in- dicating that the weather at Goldstone and Canberra was bet- ter than expected and the weather at Madrid was worse than expected. Also note that even though the data return starts falling off for MARs greater than 80%, the effective data rate does not fall off until the MAR is greater than 90%. This is due to fact that at higher MAR values, the link design algo- rithm reduces the usable track time during a pass and thus, reduces the data return. If more than two data rates were al- lowed during a pass, the results for data return and effective data rate could have been different. As expected, in general, the number of good periods and the number of bad periods decrease as the MAR increases (Fig. 7). The only exception to this is Canberra with 99% MAR. In this case there is an increase in the number of good periods 4 0 100 200 300 400 500 600 700 0 20 40 60 80 100 Minimum Availability Requirement (%) D at a Lo ss (G bi ts ) Actual Data Loss, Goldstone Expected Data Loss, Goldstone Actual Data Loss, Canberra Expected Data Loss, Canberra Actual Data Loss, Madrid Expected Data Loss, Madrid Figure 4. Expected and Actual Data Loss per Complex 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 0 20 40 60 80 100 120 Minimum Availability Requirement (%) D at a Vo lu m e (G bi ts ) Expected Data Return, Aggregate Actual Data Return, Aggregate Expected Data Loss, Aggregate Actual Data Loss, Aggregate Figure 5. Expected and Actual Data Return and Data Loss, Aggregate and in the number of bad periods. This is because with the increased MAR, one pass that was a complete loss with all other MAR values has several good periods with 99% MAR. The total number of passes with outages decreases as the MAR increases for all three complexes (Fig. 8). Madrid has by far the most number of outages and the most number of passes with outages again indicating that Madrid had expe- rienced a different weather pattern than either Canberra or Goldstone. Note that even with a MAR of 99%, there are still a number of passes with outages–five passes for Goldstone, two passes for Canberra and nine passes for Madrid– constituting more than 12% of the total number of passes. Fig. 9 provides insight into why this is the case. As seen from this figure, the ac- tual G/T is less than the required G/T for 99% MAR by more than 8 dB between 1100 and 1200 UTC. Such large drops in G/T occur more often at Ka-band than at X-band because the drops in G/T for Ka-band due to weather are roughly four times greater in dB than those for X-band [5]. Therefore, for Ka-band, a simple margin policy cannot be employed to in- sure completeness of the data and automated retransmission schemes have to be used.1 Furthermore, the spacecraft must be designed such that enough storage is available to accom- modate these retransmissions. Figs. 10 through 12 illustrate the average and the standard de- viation of good and bad period durations for Goldstone, Can- 1Long erasure correcting codes may also be useful. However, their opera- tional requirements both on the spacecraft and on the ground and their capa- bilities have not been fully investigated. 5 0 0.5 1 1.5 2 2.5 0 20 40 60 80 100 Minimum Availability Requirement Da ta Ra te (M bp s) Expected Data Rate, Goldstone Effective Data Rate , Goldstone Expected Data Rate, Canberra Effective Data Rate, Canberra Expected Data Return, Madrid Effective Data Rate, Madrid Figure 6. Actual and Expected Effective Data Rate 0 20 40 60 80 100 120 140 0 20 40 60 80 100 Minimum Availability Requirement (%) Nu m be r o fP er io ds Number of Good Periods, Goldstone Number of Bad Periods, Goldstone Number of Good Periods, Canberra Number of Bad Periods, Canberra Number of Good Periods, Madrid Number of Bad Periods, Madrid Figure 7. Number of Good and Bad Periods berra and Madrid, respectively. Madrid had poorer weather than either Goldstone or Canberra during the time period un- der consideration. Therefore, Madrid has shorter good pe- riod averages than either Goldstone or Canberra for the whole range of MARs because of more weather fluctuations. Gold- stone, having better weather than the average, has the short- est bad period average for all MAR values. Canberra, hav- ing much better than expected weather, has the largest good period average for all MAR values except 99%. Note that the good period average generally increases as the MAR in- creases except for 99% MAR for Canberra. This decrease is caused by a single pass which was a complete loss at lower MAR values but had several good periods for 99% MAR. The standard deviation of good periods is nearly as large as the average duration of the good periods. This is explained by the fact that during passes with outages, the pass consists of several good and bad periods that are relatively short, while the good periods constitute the entire pass for those tracks with no outages. This produces a wide range of values for good periods, therefore the standard deviation for the duration of good periods becomes rather large. The bad period average for all complexes is relatively small regardless of MAR, not exceeding one hour and fifteen min- utes (Canberra, 90% MAR). The standard deviation of the bad periods is slightly larger than their average duration. This indicates that while the bad periods are mostly of short dura- tion, there are several of them that are quite long. Note that good period and bad period statistics among the different complex are different from each other. Whether or 6 0 5 10 15 20 25 30 35 40 0 20 40 60 80 100 Minimum Availability Requirement (%) N u m be r o fP as se s Goldstone Canberra Madrid Figure 8. Number of Passes with Bad Periods 45 47 49 51 53 55 57 59 61 63 65 5 7 9 11 13 15 17 19 21 Time (hours UTC) G /T (d B- K^ -1) 10 15 20 25 30 35 40 45 50 55 60 El ev at io n (de g) Required G/T, 99% Required G/T, 95% Required G/T, 90% Required G/T, 10% Actual G/T 90% G/T Elevation Figure 9. Pass on Day 2006-229, DSS-55, Required G/T values for different MARs, 90% G/T, Actual G/T and Elevation. not this indicates different weather patterns for different com- plexes requires further analysis over a longer time period. Fig. 13 illustrates the expected and the actual availability for each of the complexes as well as for the three complexes ag- gregated. As seen from this figure, the aggregate expected and actual availabilities for different MAR values match each other very closely. However, this is not the case when the availability is looked on a per complex basis. At lower MAR values Madrid’s actual availability is significantly less than the expected availability (by more than 7%). This indicates that Madrid’s weather was substantially worse than expected. By contrast the weather at Canberra was much better than ex- pected since the actual availability at lower MARs is greater than expected by approximately 5%. The weather at Gold- stone was also better than expected. However since the ex- pected availability for Goldstone is already very high, the im- provement is not as substantial as that for Canberra. Note that the aggregate availability of the link over the three complexes is slightly less than expected. This matches the results in Fig. 5. Figs. 14 through 16 display the 30-minute, the one-hour and the three-hour stability of the link for different MAR values at Goldstone, Canberra and Madrid, respectively. As seen from these figures, the link is most stable at Goldstone with the three-hour stability being greater than 0.9 (i.e., the link is in a good period of three hours or longer for at least 90% of the time) even for the 10% MAR. The link at Canberra is also relatively stable with the three-hour stability exceeding 0.9 for MAR values greater than 80%. The link at Madrid is far less stable than either Goldstone or Canberra with the 7 0 1 2 3 4 5 6 7 0 20 40 60 80 100 Minimum Availability Requirement (%) D u ra tio n (ho u rs ) Avg. Good Period Sd. Good Period Avg. Bad Period Sd. Bad Period Figure 10. Average and Standard Deviation of Good Periods and Bad Periods, Goldstone 0 1 2 3 4 5 6 7 0 20 40 60 80 100 Minimum Availability Requirement (%) D u ra tio n (ho u rs ) Avg. Good Period Sd. Good Period Avg. Bad Period Sd. Bad Period Figure 11. Average and Standard Deviation of Good Periods and Bad Periods, Canberra three-hour stability exceeding 0.9 only for 99% MAR and the 30-minute and the one-hour stability exceeding 0.9 for MAR values of 90% and above. While part of this could be at- tributed to worse than expected weather at Madrid, it could also be due to differences among the weather patterns at the three complexes. However, because the period analyzed in this paper is relatively short, no such claims could be made with any degree of certainty, and further analysis is required. 6. CONCLUSIONS AND CAVEATS Conclusions In this paper a method for determining the tradeoff between data return, continuity and completeness of the deep-space Ka-band link using atmospheric noise temperature data from Water Vapor Radiometer (WVR) and Advanced Water Va- por Radiometer (AWVR) was introduced. In this method, the Ka-band link was designed using at most two data rates that maximized the average data return subject to a minimum availability requirement (MAR) and WVR/AWVR data were used to evaluate outages and data return over the link. Mars Reconnaissance Orbiter (MRO) Ka-band telecommunication capabilities along with its geometry relative to Earth and its schedule over a ten-month period were used to illustrate this method of analysis. This demonstrated that as the MAR in- creased, the total number of outage periods and the number of passes with outages tended to decrease. However, with a MAR of 99% there were still more than 12% of passes that had some outages. This indicated that it is necessary to use a retransmission scheme for deep-space missions with Ka-band as the margin cost of reliability is prohibitive. 8 0 1 2 3 4 5 6 7 0 20 40 60 80 100 Minimum Availability Requirement (%) D u ra tio n (ho u rs ) Avg. Good Period Sd. Good Period Avg. Bad Period Sd. Bad Period Figure 12. Average and Standard Deviation of Good Periods and Bad Periods, Madrid 0.8 0.82 0.84 0.86 0.88 0.9 0.92 0.94 0.96 0.98 1 0 20 40 60 80 100 Minimum Availabilty Requirement (%) O ve ra ll Av ai la bi lit y Actual Aggregate Availability Expected Aggregate Availability Actual Availability, Goldstone Expected Availability, Goldstone Actual Availability, Canberra Expected Availability, Canberra Actual Availability, Madrid Expected Availability, Madrid Figure 13. Expected and Actual Availability During the period under study, significant differences in the performance of the link among the three DSN complexes were observed. Relative to the expected performance, the Ka- band link performed better at Goldstone and Canberra and significantly worse at Madrid. However, the overall data re- turn, the overall data loss and the link availability matched expectations very closely when the passes at all three com- plexes were considered together. The data return remained relatively the same for MARs up to 80% and then decreased rapidly as the MAR increased. The data losses started to de- crease significantly for MARs above 50%. Since retransmis- sions are required even for a MAR of 99%, it is recommended that the Ka-band links be designed with a MAR value of ap- proximately 80% in order to reduce data losses significantly without incurring a significant data return penalty. It is also recommended that retransmission schemes be used to assure completeness of the data. The average duration of outages or bad periods for all com- plexes and for all MAR values was less than one hour and fifteen minutes, with Goldstone having the shortest average duration for all MAR values. The standard deviation of the bad periods, however, was about the same as the average du- ration. This indicated that some outage periods were quite long. The good periods of long duration formed a substantial frac- tion of the tracking time for Goldstone and Canberra with good periods of three hours or longer constituting more than 90% of the total track time for Goldstone for all MARs and for Canberra for MARs greater than 80%. For Madrid, how- ever, the link was not as stable with good periods of longer 9 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 0 20 40 60 80 100 Minimum Availability Requirement (%) St ab ili ty Goldstone Canberra Madrid Figure 14. 30-minute Link Stability 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 0 20 40 60 80 100 Minimum Availability Requirement (%) St ab ili ty Goldstone Canberra Madrid Figure 15. One-hour Link Stability than three hours constituting less than 90% of the total track time for all MARs less than 99%. This could indicate that the weather patterns at Madrid are inherently different from those at Canberra and Goldstone. However, this study considered a relatively short period of time and such a statement cannot be made with certainty. Caveats There are two important caveats about this study. First of all, the time resolution of the WVR and AWVR is low at roughly five minutes per sample. Therefore, these data do not capture the subsecond variations that may affect individ- ual telemetry frames at high data rates. However, since MRO Ka-band demonstration has been suspended indefinitely, the 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 0 20 40 60 80 100 Minimum Availability Requirement (%) St ab ili ty Goldstone Canberra Madrid Figure 16. Three-hour Link Stability WVR/AWVR data currently provide the best information available on the effects of the weather on the Ka-band link The second caveat concerns the duration of this study. Only ten months of data with one pass per complex per week have been considered here. These data are not enough to make general statements about the performance of the link. This study needs to be expanded to between two and five years before it could lead to any concrete conclusions. Having said this, the results of this analysis tend to validate the Ka-band link design approach outlined here and previously [3]. ACKNOWLEDGMENTS This work was performed at the Jet Propulsion Laboratory, California Institute of Technology under a contract with Na- tional Aeronautics and Space Adminstration. The author would like to thank Steve Keihm of JPL for providing him with WVR/AWVR data. The author would also like to thank JPL Interplanetary Network Directorate and Yuhsyen Shen, Wallace Tai, David Morabito, Fabrizio Pollara and Jon Hamkins of JPL for supporting this study. REFERENCES [1] Sniffin, R. W., Ed., DSMS Telecommunications Link Design Handbook (810-005, Rev. E), Mod- ule 105, Rev. B: Atmospheric and Environmental Effects, http://deepspace.jpl.nasa.gov/dsndocs/810- 005/105/105B.pdf, Jet Propulsion Laboratory, Pasadena, CA, May 26, 2006. [2] Sniffin, R. W., Ed., DSMS Telecommunications Link Design Handbook (810-005, Rev. E), Module 104, Rev. B: 34-m BWG Stations Telecommunications 10 Interfaces, http://deepspace.jpl.nasa.gov/dsndocs/810- 005/104/104B.pdf, Jet Propulsion Laboratory, Pasadena, CA, August 1, 2005. [3] Davarian, F., Shambayati, S., Slobin, S., “Deep Space Ka-Band Link Management and the MRO Demonstra- tion: Past Statistics versus Forecasting,” Proceedings of IEEE, Vol. 92 No. 12, pp. 1877-1894, December 2004. [4] Shambayati, S., “Weather Related Continuity and Com- pleteness on Deep Space Ka-band Links: Statistics and Forecasting,” IEEE Aerospace Conference 2006, pp. 1- 8, Big Sky, Montana, March 5-11, 2006. [5] Shambayati, S., Border, J. S., Morabito, D. D. and Men- doza, R, “MRO Ka-band Demonstration: Cruise Phase Lessons Learned,” IEEE Aerospace Conference 2007, pp.1-17, Big Sky, Montana, March 4-10, 2007. [6] Shambayati, S., Morabito, D. D., Border, J. S., Davar- ian, F., Lee, D. K., Mendoza, R., Britcliffe, M. and Weinreb, S., “Mars Reconnaissance Orbiter Ka-band (32 GHz) Demonstration: Cruise Phase Operations,” AIAA SpaceOps Conference 2006, pp. 1-26, Rome, Italy, June 19-23, 2006. BIOGRAPHY Shervin Shambayati obtained his Bachelors of Science degree in Ap- plied Mathematics and Engineering in 1989 from California State University, Northridge. Subsequently, he obtained his MSEE, Engineer’s Degree and Ph.D. from University of California, Los An- geles in 1991, 1993 and 2002, respec- tively. In 1993, Dr. Shambayati joined the Deep Space Communications Systems Group at Jet Propulsion Labora- tory where he took part in development and testing of Deep Space Network’s Galileo Telemetry receiver (DGT). In 1997, Dr. Shambayati joined the Information Processing Group at JPL where he has been working ever since. With that group Dr. Shambayati has been involved in various projects includ- ing Mars Global Surveyor’s Ka-Band Link Experiment II, Deep Space 1 Ka-band testing, 70m antenna Ka-band Task and the Mars Reconnaissance Orbiter Ka-band Demonstra- tion for which he was the Principal Invsetigator. His cur- rent research interests and activities include evaluating the ef- fects of weather outages on the spacecraft resources, Ka-band weather forecasting, Ka-band link design and end-to-end sys- tem architecture studies for implementation of Ka-band ser- vices in NASA’s Deep Space Network. 11