Heriot-Watt University Research Gateway

NASA's alphasat propagation terminals: Milan, Italy, andEdinburgh, Scotland

Citation for published version:Zemba, M, Nessel, J, Riva, CG, Luini, L & Goussetis, G 2019, 'NASA's alphasat propagation terminals:Milan, Italy, and Edinburgh, Scotland', International Journal of Satellite Communications and Networking,vol. 37, no. 5, pp. 502-512. https://doi.org/10.1002/sat.1296

Digital Object Identifier (DOI):10.1002/sat.1296

Link:Link to publication record in Heriot-Watt Research Portal

Document Version:Peer reviewed version

Published In:International Journal of Satellite Communications and Networking

Publisher Rights Statement:This article has been accepted for publication and undergone full peer review but has not been through thecopyediting, typesetting, pagination and proofreading process which may lead to differences between thisversion and the Version of Record. Please cite this article as doi: 10.1002/sat.1296

General rightsCopyright for the publications made accessible via Heriot-Watt Research Portal is retained by the author(s) and /or other copyright owners and it is a condition of accessing these publications that users recognise and abide bythe legal requirements associated with these rights.

Take down policyHeriot-Watt University has made every reasonable effort to ensure that the content in Heriot-Watt ResearchPortal complies with UK legislation. If you believe that the public display of this file breaches copyright pleasecontact open.access@hw.ac.uk providing details, and we will remove access to the work immediately andinvestigate your claim.

Download date: 10. Jul. 2021

https://doi.org/10.1002/sat.1296https://doi.org/10.1002/sat.1296https://researchportal.hw.ac.uk/en/publications/2da372e1-8175-4ea6-ad6f-219782914c31

S P E C I A L I S S U E PA P E R

NASA’s Alphasat Propagation Terminals:Milan, Italy and Edinburgh, Scotland

M. Zemba1 | J. Nessel2 | C. Riva2 | L. Luini2 | G.Goussetis3

1AdvancedHigh Frequency Branch, NASAGlenn Research Center, Cleveland, Ohio,44135, USA2Dipartimento di Elettronica, Informazionee Bioingegneria, Politecnico diMilano,Milan, Italy3Institute of Sensors, Signals & Systems,Heriot-Watt University, Edinburgh,Scotland, UK

CorrespondenceMichael J. Zemba, AdvancedHighFrequency Branch, NASAGlenn ResearchCenter, Cleveland, Ohio, 44135USAEmail: michael.j.zemba@nasa.gov

Funding informationN/A

SinceMay of 2014, NASA’s Glenn Research Center has oper-atedmeasurement campaigns for theAlphasatAldoParaboniPropagation Experiment alongside the European commu-nity of propagation experimenters. Presently, three NASAstations have been deployed to distinct climatological re-gions across Europe. NASA’s participation in the campaignbegan in 2014 through a collaborative effort with the Po-litecnico diMilano (POLIMI) to jointly operate a 20/40 GHzground terminal at the POLIMI campus inMilan, Italy. Subse-quently, a single-channel 40 GHz terminal was deployed toEdinburgh, Scotland in March 2016 in collaboration withHeriot-Watt University (HWU). A third terminal was de-ployed toNASA’sMadridDeep SpaceCommunicationsCom-plex (MDSCC) inMarch 2017 with NASA’S Jet PropulsionLaboratory (JPL), also observing the 40GHz beacon. In addi-tion, a fourth station is planned for deployment to Andøya,Norway by early 2019 in collaboration with the NorwegianDefence Research Establishment (FFI). This paper will detailthe design and results of the twomost established terminals,Milan and Edinburgh, which together comprise 11 stationyears of propagationmeasurements.K E YWORD S

Alphasat experiment, atmospheric effects, measurements,microwave propagation, propagation losses

1

2 M. ZEMBA ET AL.

1 | INTRODUCTIONAs the finite spectrum available for satellite communications continues to grow congested, there is commensuratedemand for higher frequency capabilities. This demand is simultaneously driven by the appeal of link capacities thataremuch higher than presently achievable within lower frequency allocations. However, one significant impedimentto the implementation of higher frequency systems is the increased sensitivity of the link to atmospheric effects. Aconsiderable increase in attenuation due to rain, clouds, and gases necessitates the use of intelligent system designthrough mitigation techniques such as adaptive power control and site diversity, while increased phase distortionpresents a substantial challenge in the implementation of microwave uplink arrays [1]. Thus, a thorough understandingof atmospheric propagation at these frequencies of interest and within the pertinent climatological zones is required toeffectively implement satellite communications links at Ka-band and above.

To this end, the Inmarsat communications satellite Alphasat was launched in July 2013 with the hosted AldoParaboni Technology Demonstration Payload (TDP) #5[2]. The Aldo Paraboni experiment was developed by the ItalianSpace Agency (ASI) and European Space Agency (ESA) and, in addition to a Q/V-band communications experiment,features coherent continuous-wave (CW) Ka- andQ-band beacons for the purpose of assessing atmospheric effects(rain attenuation, scintillation, depolarization, etc.) on links operating in these frequencies. In addition, the experimentwill assist in refining physical models for predictions of atmospheric attenuation within theQ-band.[3, 4].

NASA’s vision for the agency’s future space communication architecture turns primarily to Ka-band and opticalcommunications in pursuit of higher bandwidth and link security. Such an architecturemust utilize extensive cognitivenetworking and fademitigation techniques to circumvent weather-related impairments and ensure efficient manage-ment of network resources. In addition, the next generation successor to the Tracking andData Relay Satellite System(TDRSS) will require significantly higher bandwidths than available in the current Ku-band allocation, and the agency isthusly investigating the use of available allocations in the Q-band (37-42 GHz) and V/W-bands (74-84 GHz) as downlinkoptions tomeet these requirements[5].

F IGURE 1 NASA’s current and planned Alphasat propagation terminals: Milan, Italy, installed in 2014 at PolitecnicodiMilano; Edinburgh, Scotland, installed in 2016 at Heriot-Watt University; and Robledo de Chavela, Spain, installed in2017 at NASA’sMadrid Deep Space Communications Complex. A fourth terminal is planned for Andøya, Norwaywiththe Norwegian Defence Research Establishment.

M. ZEMBA ET AL. 3

As depicted in Figure 1, NASA presently maintains three Alphasat propagation terminals. The first of these wasinstalled in April 2014 at the Politecnico diMilano campus inMilan, Italy and observes both the Ka- andQ-band beacons[6, 7, 8]. This station was also recently updated in September 2017, which is further detailed in Section 2.1. After twoyears of operation inMilan, inMarch 2016, a secondQ-band receiver was installed at Heriot-Watt University (HWU)in Edinburgh, Scotland. Given the high latitude (55.9123◦N), this installation presently yields the lowest elevationangle (21◦) measurements within the framework of the Alphasat experiment. In 2018, HWUupdated the terminal withKa-bandmeasurement capability. Most recently, another Q-band receiver was deployed to the NASAMadrid DeepSpace Communications Complex (MDSCC) in Robledo de Chavela, Spain to provide atmospheric characterization forlinks near 40 GHz through the Deep Space Network (DSN). Lastly, plans are underway for a fourth terminal in AndoyaNorway in collaboration with the Norwegian Defence Research Establishment (FFI). This station is intended to yield co-and cross-polarizationmeasurements at both Ka- andQ-band andwould provide unique low-elevation angle data in acoastal environment above the arctic circle. While the aforementioned receivers sharemuch in terms of their designand data processing techniques, Section 2will elaborate on the unique details of the design and operation of the twosites covered in this paper, Milan (2.1) and Edinburgh (2.2).

2 | EXPERIMENT DESIGN2.1 | Milan, ItalyTheMilan receivers were installed at the POLIMI campus in April 2014 atop the Dipartimento di Elettronica, Infor-mazione e Bioingegneria (DEIB) building (Figure 2). In addition to the Ka-band andQ-band beacon receivers, a suite ofcollocated weather instrumentation includes an R.M. Youngweather station providingmeasurements of temperature,pressure, humidity, wind speed, wind direction, rain accumulation bymeans of a tipping bucket, as well as a Thies Climalaser disdrometer which yields droplet size and velocity distributions. Previous work has taken advantage these addi-tional measurements to investigate frequency scaling from 20 to 40GHz using the drop size distribution (DSD)[9, 10],as well as the impact of the scattering model on the specific attenuation as derived from the DSD[11]. Additionally,a Radiometer Physics Gmbh (RPG) water vapor radiometer was installed in November 2016, providing radiometricmeasurements that will be used to validate the digital radiometric measurement recently added to the receivers.

F IGURE 2 The Ka/Q-band Alphasat receivers at Politecnico diMilano inMilan, Italy. On the left, a photo of thereceivers and the associated weathermonitoring equipment including weather station, laser disdrometer, and tippingbucket. On the right, an overhead view of the receiver location atop the roof of the Dipartimento di Elettronica,Informazione e Bioingegneria (DEIB) building.

4 M. ZEMBA ET AL.

LO20.129 GHz

LO19.203 GHz

Reference10 MHz

Ka-Band RF

Q-Band RF

IF & ADC

ADCNI PCI-5124

fs = 3.077 MHz

N = 217

CH0

CH1

LO65 MHz

BPF70 MHz

LNA+35 dB

19.701 GHz

39.402 GHz

BPF20.2 GHz

LO19.203 GHz

BPF70 MHz

BPF5 MHz

BPF5 MHz

BPF5 MHz

LNA+35 dB

IF Amp.+25 dB

IF Amp.+25 dB

IF Amp.+25 dB

BPF5 MHz

IF Amp.+25 dB

IF Amp.+25 dB

F IGURE 3 A block diagram of theMilan Ka- andQ-band beacon receivers, which share a common intermediatefrequency (IF) downconversion stage. The station was updated in September 2017 to adapt the final IF from 455 kHz to5MHz and to widen the IF bandwidth to accommodate a digital noise powermeasurement.

TABLE 1 Installation summary for NASA’sMilan andEdinburgh Alphasat beacon receivers.Item Milan EdinburghInstallation Date April 2014 March 2016Latitude 45.4787◦N 55.9123◦NLongitude 9.2327◦E 3.3223◦EAltitude 138m 130mNominal Elevation 35◦ 21◦Nominal Azimuth 158◦ 147◦Frequencies 19.701GHz39.402GHz 39.402GHz

Dynamic Range 35 dB (Ka)45 dB (Q) 35 dB (Q)

TheMilan beacon receivers, like all of NASA’s Al-phasat terminals, were designed and developed atNASAGRC. As originally deployed, theMilan receiversdownconvertedeachbeacon to afinal frequencyof 455kHz; this designwas selected to utilize legacy hardwarethat was readily available for this configuration. As sub-sequent receivers were designed for Edinburgh andMadrid, a novel digital radiometric measurement wasimplementedwhich led to the downconversion stagesfor the receivers being designed for a final IF of 5MHz.To accomodate the newmeasurement, a higher IF wasdesirable to widen the bandwidth of the system andallow for integration of the noise power over a widerfinal output band. To add this functionality inMilan, theIF was thusly reconfigured in September 2017 – Figure3 shows the block diagram of the receivers in their present and upgraded configuration. The system consists of 1.2m(Ka) and 0.6m (Q) Cassegrain reflector antennas, each with beamwidths of 0.9◦andwith gains of 45.6 dBi. This narrowbeamwidth, in conjunction the inclined orbit of Alphasat (0◦- 3◦), necessitates active tracking of the beacons which isaccomplished through the use of electronic pan/tilt positioners that update the pointing of the antennas once every 60seconds with a pointing accuracy of 0.01◦. The first downconversion of the beacon occurs directly at the feed of eachantenna. The Ka-band channel is downconverted from 19.701GHz to 70MHz in one stage, while theQ-band is down-converted in two stages from 39.402 GHz to 20.199 GHz and then to 70MHz. All local oscillators (LOs) are referencedto a common ultra-stable 10MHz citrine oscillator. The RF electronics aremounted to a temperature-controlled coldplate that is stabilized within ± 0.01◦C through thermoelectric cooling (TEC) tiles. The low-noise amplifier (LNA), whichis mounted to the waveguide output of the antenna rather than the cold plate, is maintained to ± 0.1 ◦C through itsown TEC tile. The air temperature within the enclosures is not controlled directly but remains stable within ± 2 ◦C as a

M. ZEMBA ET AL. 5

byproduct of the plate and LNA control. After downconversion to 70MHz, both channels continue to a second, commonIF stage < 2 meters from the antennas which is also temperature controlled to within ± 1 ◦C. In this stage, the finaldownconversion to 5MHz occurs, followed by additional filtering and amplification prior to transmission of the signal tothe digitizer over coaxial cabling. Discussion of the digital signal processing beyond this point is discussed in Section 3.

2.2 | Edinburgh, ScotlandThe Edinburgh beacon receiver was installed at the HWU campus inMarch 2016 on the roof of the Earl Mountbattenbuilding (Figure 4) and observes Alphasat at an elevation of 21◦, presently the lowest elevation anglemeasurementswithin the Alphasat campaign. A nearby Campbell Scientific weather station is located approximately 70 m to thesouth of the receiver and 42◦west of the path to Alphasat, again providing the standardmeteorological measurementsof temperature, pressure, humidity, wind speed, wind direction, and rain accumulation via a tipping bucket. Thedesign of the HWU receiver is largely derived from theQ-band channel of theMilan terminal with a few incrementalimprovements such as the elimination of the secondary IF enclosure and the addition of the noise powermeasurement.

F IGURE 4 TheQ-band Alphasat receiver at Heriot-Watt University in Edinburgh, Scotland, UK. On the left, a photoof theQ-band beacon receiver with tracking system, and on the right, an overhead view of the receiver location atopthe roof of the Earl Mountbatten building.

LO19.203 GHz

+17 dBmMiteq DLCRO

SN 1936107

Reference10 MHz

+19 dBmWenzel 501-24055

SN 20641A003

Q-Band RF

39.402 GHz

BPF20.2 GHz

K&L 5WG-201900

SN SG351-1

LO20.129 GHz

+19 dBmMiteq DLCRO

SN 1932144

LO65 MHz

+10 dBmNexyn NXOS-PLXO

SN 094105560

BPF70 MHzK&L 3C20-70

SN 1

Marki M9-0942LN Miteq DB0426LW1 Mini-Circuits ZFM-3-S+

S N595501527 BPF5 MHz

TTE KC5-5M-1M

M4715

BPF70 MHzLark XMC70

28090-03

LPF10 GHzK&L 6L250

SN 10

LNA+35 dB

5 dB NFCernex CNL

IF Amp.+25 dB

Mini-Circuits ZFL-500LN+

S F233701211

Atten.-6 dB

Atten.-6 dB

Atten.-10 dB

IF A mp.+25 dB

Mini-Circuits ZFL-500LN+

S F233701211

Atten.-6 dB

Split.-3 dB

Split.-3 dB

Atten.-6 dB

ADCNI PCI-5124

fs = 1.111 MHz

N = 217

CH0

CH1

F IGURE 5 A block diagram of the EdinburghQ-band beacon receiver, which downconverts from 39.402GHz to 5MHz in three stages within the outdoor, temperature-controlled enclosure. A 1MHz bandwidth is used toaccommodate noise powermeasurement.

6 M. ZEMBA ET AL.

As inMilan, the Scotland beacon receiver utilizes a 0.6 mCassegrain reflector with a 0.9◦beamwidth, 45.6 dBi gain,and mechanical tracking with a resolution of 0.01◦. At all NASA Alphasat stations, the positioning system employsopen-loop tracking utilizing Orbital EphemerisMessage (OEM) data which is updated weekly. Temperature stability forthe RF electronics is achieved in the samemanner with control of the cold plate to the same tolerances, although theLNA does not have independent temperature control in this system. Another change as compared toMilan is that thedownconversion is implemented entirely within the enclosure at the antenna feed (from 39.402 GHz to 20.199 GHz to70MHz to 5MHz). The LNA used in the Edinburgh system also has a slightly higher noise figure as compared to theMilan receiver (3.4 dB vs 2.7 dB at POLIMI), resulting in a slight reduction of dynamic range (35 dB vs. 40 dB at POLIMI).

3 | DIGITAL SIGNAL PROCESSING3.1 | Frequency Tracking &WindowingIn both Milan and Edinburgh, the downconverted 5 MHz beacon signals are digitized using a National InstrumentsPCI-5124 oscilloscope card. Measurement of the signal power is accomplished through a novel frequency estimationtechnique[12, 13] using a variant of theQuinn-Fernandes (QNF) frequency estimator[14, 15], which interpolates theFast Fourier Transform (FFT). Measurements are recorded at a rate of 10 Hz and are also averaged in real time to 1 Hz.This allows for characterization of atmospheric scintillation with the 10Hz data while the 1Hz data remains availablefor characterization of attenuation and other long timescale phenomena without affecting the computational load.The sampling frequency and number of points acquired by the digitizer for eachmeasurement are adjustable and areprimarily set to maintain a measurement bandwidth near to 10 Hzwithin the processing capabilities of the digitizer andworkstation (e.g. a sampling frequency of fs = 3.07MHz andN = 218 for ameasurement bandwidth of fs/N = 11.7 Hz).

Several digital processing techniques are also applied tomaximize the dynamic range of the receivers by improvingtheir performance in low signal-to-noise ratio environments: the simplest of these is the use of an a priori frequencywindow to limit themeasurement to a small frequency range known to contain the beacon. Because theQNF algorithmbegins with a simple peak search, it is susceptible to favoring spurious noise peaks over the true signal when the SNRis low. This can be reduced by restricting the bandwidth of the peak search as signal power decreases. While the

14 14.5 15 15.5 16 16.5 17 17.5 18 18. 9

Time (Hours, UTC)

-50

-40

-30

-20

-10

0

To

tal A

tte

nu

atio

n (

dB

)

Ka

Q

(a) Milan, total attenuation2 3 4 5 6 7 8 9 10

Time (Hours, UTC)

-35

-30

-25

-20

-15

-10

-5

0

To

tal A

tte

nu

atio

n (

dB

)

Q

(b) Edinburgh, total attenuation

14 14.5 15 15.5 16 16.5 17 17.5 18 18.5 19

Time (Hours, UTC)

461

461.5

462

Ka R

eceiv

ed IF

(kH

z)

Ka

Q

467

468

469

Q R

eceiv

ed IF

(kH

z)

(c) Milan, observed IF frequency2 3 4 5 6 7 8 9 10

Time (Hours, UTC)

149.5

150

150.5

151

151.5

152

Ka R

eceiv

ed IF

(kH

z)

Q

(d) Edinburgh, observed IF frequency

F IGURE 6 Timeseries examples of heavy attenuation events in inMilan (left, 2017-09-01), and in Edinburgh (right,2017-11-16), demonstrating the full dynamic range of the receivers and the frequency tracking algorithms.

M. ZEMBA ET AL. 7

beacon frequency does drift gradually due to Doppler shift, it does not change significantly over the duration of extremefades. This windowing procedure is implemented bymaintaining an 2-minute average of themost recent frequencyobservations as long as the beacon is above a predefined high-power threshold. As the power approaches the noisefloor and falls below this threshold, the peak search of the QNF algorithm is restricted to a small bandwidth around themean of the previously observed frequency. The size of the windowmust be larger than themaximum expected dopplerdrift over the duration of any event that fully exceeds the system dynamic range, such that the beacon is immediatelyreacquired as it reappears above the noise.

A second technique is also employed inMilan, where both the Ka- andQ-band channels are being observed. Withmeasurements of both channels available, the Ka-band signal can be used to track theQ-band beacon duringmoderatefades due to the coherency of the two beacons. In moderate attenuation events where theQ-band channel is near thenoise floor but the Ka-band channel is still clearly visible, the frequency of the Q-band beacon can be predicted from anaccuratemeasurement of the Ka-band beacon frequency. In this case, a similar a prioriwindow is applied as describedabove, but here the center of the band is defined by the prediction calculated from the observed Ka-band frequency.This results in amodest increase in dynamic range by eliminating erroneous estimates caused by spurious peaks in theinitial peak search of theQNF algorithm.

Figure 6 presents timeseries examples of heavy attenuation events for each receiver, Milan (left) and Edinburgh(right). In both examples, the total attenuation (top) exceeds the full dynamic range of the receivers, and the effect of theaforementioned tracking algorithms can be observed in the recorded IF frequency of the beacons (bottom). InMilan, theQ-band receiver fades below the noise floor for over 10minutes, but the receiver is able to utilize the Ka-band signal tocontinue tracking the anticipated frequency of the Q-band beacon. While this does not impact performance for periodswhere the signal level is completely below the noise floor, it is able to slightly improve dynamic range at the fringes ofthe event by reducing the effect of noise on the frequency estimation algorithm. It is only as the Ka-band signal alsoloses lock for approximately 2minutes that the algorithm is unable to accurately predict the location of either beacon.In the Edinburgh example, the Q-band signal is tracked through use of past frequencymeasurements. As the powerfades below the noise floor, the past frequency observations are used to restrict the frequency estimate to a narrowband around the expected beacon location. This approach is also implemented inMilan for conditions when the Ka bandbeacon is unavailable, although it as not as readily visible in this example due to use of a wider bandwidth window.

3.2 | Filtering, Decimation &Digital RadiometricMeasurementWhile the 5 MHz output of each beacon receiver would require a sampling rate of ≥10 MHz to meet the Nyquistsampling requirement, the extremely narrow-band nature of the CW signals allows for several alternative approaches.

0 4 8 12 16 20 24

Time (Hours, UTC)

-45

-40

-35

-30

-25

-20

-15

-10

-5

0

5

To

tal A

tte

nu

atio

n (

dB

)

Q

Noise Power

-67

-66

-65

Inte

gra

ted

No

ise

Po

we

r (d

Bm

)

F IGURE 7 Example noise powermeasurement of amoderate attenuation event in in Edinburgh, 2018-01-23.

8 M. ZEMBA ET AL.

In Milan, the signals are bandpass sampled; the channels are bandpass filtered, and the sampling rate is set belowthe Nyquist rate such that the image of the high-frequency components is observed at baseband. This reduces thecomputational load as opposed to sampling at a much higher rate, which requires the processing of orders of magnitudemore data that is not required to characterize very narrowband beacon signals. Beginning in Edinburgh, a process offiltering and decimationwas designed to similarly expedite processing timewhile also enabling a digital radiometricmeasurement. In the Edinburgh implementation, the spectrum is fully Nyquist sampled (fs = 11.11MHz) and a digitalbandpass filter (50 kHz, 10th order Type 2 Chebyshev) is applied to isolate the beacon signal. The frequency tracking ofthe signal, detailed in Section 3.1, is also used here to center the passband of the Chebyshev filter at the current trackedfrequency. The resulting filtered spectrum is then digitally decimated by a factor of 32 to reduce processing time (giventhe processing requirement of 0.1 seconds mandated by the 10 Hz measurement rate). In parallel with the beaconpower measurement, the original, fully sampled spectrum is also filtered with a notch filter centered at the beaconfrequency to eliminate the beacon. Here, a 10th order Type II Chebyshev bandstop filter with a 250Hz bandwidth isapplied. This is done to null the beacon signal and allow for an integrated noise power measurement over the full IFbandwidth of 1MHz. With the beacon signal removed, the noise power within the final 1MHz bandwidth is integratedand recorded at 10Hz and averaged over a 1 sec integration time, resulting in an estimated radiometric resolution of:

∆T =Ts y s

B ∗ τ =908K

√106Hz ∗ 1sec

= 0.908K

A1MHz bandwidth is still very narrow for a radiometermeasurement, but previous analysis [16] has demonstratedit to have utility alongside the beaconmeasurements, particularly given the very limited cost and overhead required toimplement it. Calibration of the digital radiometermeasurement is also detailed in previouswork [16]. Figure 7 presentsan example timeseries measurement of the integrated noise power along with the received Q-band beacon powerduring amoderate attenuation event in Edinburgh, demonstrating the expected inverse correlation between two.

3.3 | Data CalibrationMeasured signal attenuation is calibrated using vertical profiles of pressure, humidity and temperature from theEuropean Centre for Medium-RangeWeather Forecast (ECMWF) and the MPM93Mass Absorption Model[17] tocalculate reference clear-sky attenuation levels. This reference level is compared to clear-sky conditions at the receivers(identified using the on-site rain gauges) and a calibration offset is calculated on amonthly basis. In the event that asingle calibration is not valid for a full month (e.g. due to changes in operation such as the 2017 receiver modifications),the offset is calculated for shorter periods as necessary. In addition, erroneous data (power outages, loss of satellitetracking) are removed from the data bymanual inspection.

4 | RESULTS4.1 | Milan, ItalyAfter calibrating the datasets as described in Section 3.3, the total attenuation from each site was statistically analyzedfor characterization of the attenuation statistics over the presently available data. Milan was characterized fromMay2014 through June 2018 for a total of four years and onemonth. The annual attenuation statistics forMilan are shownin Figure 8a (Ka) and 8b (Q) in the form of complementary cumulative distribution functions (CCDFs). Each curve

M. ZEMBA ET AL. 9

0 5 10 15 20 25 30 35 40 450.01

0.1

1

10

100

2014

2015

2016

2017

2018

Total

ITU-R Model

19.701 GHz Attenuation (dB)

Perc

en

t o

f T

ime V

alu

e is E

xce

eded

(%

)

(a) CCDF of total attenuation at Ka-band inMilan.

0 5 10 15 20 25 30 35 40 450.01

0.1

1

10

100

2014

2015

2016

2017

2018

Total

ITU-R Model

39.402 GHz Attenuation (dB)

Perc

en

t o

f T

ime V

alu

e is E

xce

eded

(%

)

(b) CCDF of total attenuation at Q-band inMilan.

0 5 10 15 20 25 30 35 40 450.01

0.1

1

10

100

January

February

March

April

May

June

July

August

September

October

November

December

Total

ITU-R Model

19.701 GHz Attenuation (dB)

Perc

en

t o

f T

ime V

alu

e is E

xce

eded

(%

)

(c) CCDF of monthly averages for the Ka-band channel.

0 5 10 15 20 25 30 35 40 450.01

0.1

1

10

100

January

February

March

April

May

June

July

August

September

October

November

December

Total

ITU-R Model

39.402 GHz Attenuation (dB)

Perc

en

t o

f T

ime V

alu

e is E

xce

eded

(%

)

(d) CCDF of monthly averages for theQ-band channel.

F IGURE 8 Complementary Cumulative Distribution Functions (CCDFs) of total attenuation inMilan at Ka- (19.701GHz) andQ-band (39.402GHz), by year (top) and averaged bymonth (bottom).

represents the noted calendar year, with the exception of 2014 and 2018, which are a partial years at the beginning andend of the dataset (fromMay 2014 forward, and from June 2018 backward). The total CCDF over the entire datasetis also shown, as well as a comparisonwith the ITU-Rmodel[18, 19, 20], which shows very good agreement. For 99%availability, the associated linkmargin was 2.56 dB (Ka) and 6.73 dB (Q).

In addition, in Figure 8c and 8d, a seasonal analysis is presented in the form of monthly average CCDFs. Thesecurves represent the average of each calendarmonth for all data in the data set – e.g., the curve for ’May’ (dark green)is the average of all data collected during the month of May from 2014 to 2018. Again, the ITU model shows goodagreement with the total curve, with the wetter summermonths (June, July August) experiencing a statistically higherattenuation than the averagemodel, and the drier winter months experiencing less (December, January, February). Thisis true of both the Ka- andQ-band channels, although there is amoremarked variability between thewet/dry seasons atQ-band due to the significantly higher impact of rain attenuation in the 40 GHz band. The highest average attenuation,observed in June, corresponded to a 99% availability margin of 3.87 dB (Ka) and 9.54 dB (Q). The lowest attenuation,observed in December, corresponded to 1.22 dB (Ka) and 1.95 dB (Q) at 99% availability.

Figure 9 also presents plots of the power spectral density (PSD) during strong, characteristic events of rain attenu-

10 M. ZEMBA ET AL.

10-3

10-2

10-1

100

101

10-4

10-3

10-2

10-1

100

101

102

Frequency (Hz)

Pow

er

spectr

al density (

dB

2/H

z)

Ka

Q

Attenuation (-20 dB/dec)

Scintillation (-80/3 dB/dec)

(a) Attenuation event

10-3

10-2

10-1

100

101

10-4

10-3

10-2

10-1

100

Frequency (Hz)

Pow

er

spectr

al density (

dB

2/H

z)

Ka

Q

Attenuation Slope (-20 dB/dec)

Scintillation Slope (-80/3 dB/dec)

(b) Scintillation event

F IGURE 9 The power spectral density (PSD) of a characteristic attenuation (left) and scintillation (right) event inMilan demonstrating the characteristic PSD slopes for attenuation and scintillation.

ation and scintillation on both the Ka- andQ-band channels inMilan, demonstrating concurrencewith the expectedtheoretical PSD slopes of -20 dB/dec for attenuation and -80/3 dB/dec for scintillation. Further analysis on the statisticsof scintillation fades is presented in Nessel et al. [21].

4.2 | Edinburgh, ScotlandThe Edinburgh dataset was characterized fromApril 2016 throughMay 2018 for a total of two years and onemonth.The annual attenuation statistics at Q-band are shown in Figure 10a. As in Figure 8b, each curve represents the notedcalendar year, with the exception of those at the beginning and end of the dataset (2016 and 2018) which are partialyears. The total CCDF over the entire dataset is also shown, as well as a comparison with the ITU-R model. Here,

0 5 10 15 20 25 30 35 40 450.01

0.1

1

10

100

2016

2017

2018

Total

ITU-R Model

39.402 GHz Attenuation (dB)

Perc

en

t o

f T

ime V

alu

e is E

xce

ed

ed

(%

)

(a) CCDF of total attenuation at Q-band in Edinburgh.

0 5 10 15 20 25 30 35 40 450.01

0.1

1

10

100

January

February

March

April

May

June

July

August

September

October

November

December

Total

ITU-R Model

39.402 GHz Attenuation (dB)

Perc

en

t o

f T

ime V

alu

e is E

xce

eded

(%

)

(b) CCDF of monthly averages for Q-band in Edinburgh.

F IGURE 10 Complementary Cumulative Distribution Functions (CCDFs) of total attenuation in Edinburgh atQ-band (39.402GHz), by year (left) and averaged bymonth (right).

M. ZEMBA ET AL. 11

agreement with the ITU-Rmodel is not as good as inMilan, and it is postulated that this is due to an overprediction ofrain rate within themodel. While rain in Edinburgh is extremely common, it tends to rain frequently at low rain rates(< 20mm/hr) andmore rarely at higher rain rates, whichmay be the cause of the discrepancy between themodel andthe collected data. Themodel curve plotted in 8bwas generated by interpolation of the ITU-R P.837-7 rainmaps[19],yielding an 0.01% rain rate of R0.01 = 49.0mm/hr. Data analyzed from the HWU tipping bucket suggests amuch lowerR0.01 = 19.5 mm/hr. When this lower rate is incorporated, agreement is better, although there is still a noticeableoverprediction. The observedmargins over the dataset from 2016 to 2018were 3.19 dB, 5.96 dB, and 12.12 dB for95%, 99% and 99.9% availability, respectively.

In addition, in Figure 8c and 8d, themonthly averages are presented for Edinburgh. As in Figure 8d, these curvesagain represent the average of each calendarmonth for all data in the data set. Here, it can be observed that the ITUmodel shows better agreement with the rainier months (June, July, August, September), but still tends to overpredict ascompared to the average and particularly as to the drier months (November, December, January, February). The highestaverage attenuation, observed in August, corresponded to a 95%, 99%, and 99.9% availability margin of 3.61 dB, 8.39dB, and 25.90 dB, respectively. The lowest attenuation, observed in December, corresponded to 2.86 dB, 4.42 dB, and9.11 dB at 95%, 99%, and 99.9% availabilities.

5 | CONCLUSIONSNASA’s participation in the Alphasat propagation experiment has thus far yielded 11 station years of propagation databetween the Ka- andQ-bandmeasurements inMilan, Italy, and Edinburgh, Scotland. The data collection campaigns areexpected to continue for a minimum of five years at each location, and may be extended contingent upon hardwarelifespan and beacon payload availability. Herein we have presented the design, operation, and first several years ofstatistics of the propagation data collected at POLIMI and HWU. In addition, a third Q-band receiver was recentlydeployed to NASA’s Madrid Deep Space Communications Complex, and a fourth Alphasat terminal is planned forinstallation in Andøya Norway by 2019, which is expected to supersede Edinburgh as the highest latitude Alphasatinstallationwith an elevation of approximately 11◦ . These lowelevation angle polarmeasurementswill also complementthe ongoing ESA/NASA Ka-band measurement campaign in Svalbard [22]. Future publications will explore second-order propagation statistics such as rain fade duration and fade slope, as well as further statistical characterization ofscintillation.

ACKNOWLEDGEMENTS

The authors would like to acknowledge the European Space Agency (ESA) and Agenzia Spaziale Italian (ASI) for theirdevelopment of the Alphasat Aldo Paraboni payload and propagation experiment, as well as the faculty at the Air ForceResearch Laboratory, Politecnico di Milano, Heriot-Watt University, and the Madrid Deep Space CommunicationsComplex who have provided a great deal of assistance in the deployment and operation of the NASA terminals. We alsoexpress appreciation for the strong community of propagation experimenters who continue to tirelessly advance thefield of microwave propagation through Alphasat and other ongoing propagation campaigns.

REFERENCES

[1] ZembaM, Morse J, Nessel J. Ka-band Atmospheric Phase Stability Measurements in Goldstone, CA; White Sands, NM;and Guam. In: 8th European Conference on Antennas and Propagation TheHague, Netherlands; April 6 - 11, 2014. .

12 M. ZEMBA ET AL.

[2] Rossi T, Sanctis MD, Ruggieri M. Satellite communication and propagation experiments through the Alphasat Q/Vband Aldo Paraboni Technology Demonstration Payload. IEEE Aerospace and Electronic Systems Magazine June2016;31(3):18–27.

[3] Paraboni A, Vernucci A, Zuliani L, Colzi E, Martellucci A. A New Satellite Experiment in the Q/V band for the Verifica-tion of Fade Countermeasures Based on the Spatial Non-Uniformity of Attenuation. In: 2nd European Conference onAntennas and Propagation Edinburgh, United Kingdom; November 11 - 16, 2007. .

[4] Cola FD, Pandolfi A, Paolo GD, Rivera J, Benzi E, Martellucci A, et al. Alphasat Aldo Paraboni payload IOT campaign andstatus after the first year of operation. In: 2016 IEEE Aerospace Conference Big Sky, Montana;March 5 - 12, 2016. .

[5] NASA Space Communications and Navigation Office, NASA’s Space-Based Relay Study: Overview and Direction. Wash-ington, DC; 2013. Accessed: 2018-08-02. https://www.nasa.gov/sites/default/files/files/SBRSWhitePaper3-7-2013.pdf.

[6] Nessel J, ZembaM, Houts J. Design of a K/Q-band Beacon Receiver for the Alphasat TDP#5 Experiment. In: 2014 IEEEInternational Symposium on Antennas and PropagationMemphis, TN; July 6 - 11, 2014. .

[7] Nessel J, Morse J, Zemba M, Riva C, Luini L. Preliminary Results of the NASA Beacon Receiver for the Alphasat AldoParaboni TDP5 Propagation Experiment. In: 20th Ka and Broadband Communications Conference Salerno, Italy; Octo-ber 1 - 3, 2014. .

[8] Nessel J, Morse J, Zemba M, Riva C, Luini L. Three Years of Atmospheric Characterization at Ka/Q-band with theNASA/POLIMI Alphasat Receiver in Milan, Italy. In: 12th European Conference on Antennas and Propagation London,United Kingdom; April 9 - 13, 2018. .

[9] Nessel J, Zemba M, Luini L, Riva C. Comparison of Instantaneous Frequency Scaling from Rain Attenuation and Opti-cal Disdrometer Measurements at K/Q bands. In: 21st Ka and Broadband Communications Conference Bologna, Italy;October 12 - 14, 2015. .

[10] Zemba M, Nessel J, Houts J, Luini L, Riva C. Statistical Analysis of Instantaneous Frequency Scaling Factor as Derivedfrom Optical Disdrometer Measurements at K/Q Bands. In: 10th European Conference on Antennas and PropagationDavos, Switzerland; April 10 - 15, 2016. .

[11] ZembaM, Luini L, Nessel J, Riva C, Houts J. Impact of ScatteringModel onDisdrometer Derived Attenuation Scaling. In:22nd Ka and Broadband Communications Conference Cleveland, Ohio; October 17 - 20, 2016. .

[12] Zemba M, Morse J, Nessel J. Frequency Estimator Performance for a Software-based Beacon Receiver. In: 2014 IEEEInternational Symposium on Antennas and PropagationMemphis, TN; July 6 - 11, 2014. .

[13] Nessel J, Zemba M, Houts J. Software-Defined Beacon Receiver Using Frequency Estimation Algorithms. NASA TechBriefs June 2015;39(6):58–59. LEW-19222-1.

[14] QuinnBG. Estimating Frequency by Interpolation. IEEETransactions on Signal ProcessingMay1994;42(5):1264 –1268.[15] Quinn BG, Fernandes JM. A Fast Technique for the Estimation of Frequency. Biometrika September 1991;78:489 – 497.[16] Nessel J, Goussetis G, Zemba M, Houts J. Design and Preliminary Results from Edinburgh, UK Alphasat Q-Band Prop-

agation Terminal. In: 22nd Ka and Broadband Communications Conference Cleveland, Ohio; October 17 - 20, 2016..

[17] LiebeH, HuffordGA, CottonMG. PropagationModeling ofMoist Air and SuspendedWater/Ice Particles at FrequenciesBelow 1000GHz. In: Proc. NATO/AGARDWave Propagation Panel, 52ndMeeting, No. 3/1-10Mallorca, Spain;May 17 -20, 1993. .

[18] ITU-R Recommendation P.836-6: Water Vapour: Surface Density and Total Columnar Content. Geneva, Switzerland:International Telecommunications Union; 2017.

M. ZEMBA ET AL. 13

[19] ITU-R Recommendation P.837-7: Characteristics of Precipitation for Propagation Modelling. Geneva, Switzerland: In-ternational Telecommunications Union; 2017.

[20] ITU-RRecommendation P.840: AttenuationDue toClouds and Fog. Geneva, Switzerland: International Telecommunica-tions Union; 2017.

[21] Nessel J, ZembaM,Morse J, Luini L, Riva C. Preliminary Results from theNASAAlphasat Beacon Receiver inMilan, Italy.In: 9th European Conference on Antennas and Propagation Lisbon, Portugal; April 12 - 17, 2015. .

[22] ZembaM,Morse J,Nessel J. Design of aKa-BandPropagationTerminal forAtmosphericMeasurements in Polar Regions.In: 10th European Conference on Antennas and Propagation Davos, Switzerland; April 10 - 15, 2016. .

MICHAEL ZEMBA received his Bachelor’s of Science degree in Electrical Engineering in 2011 and hisMaster’s of Science in Electrical Engineering in 2013, both from The University of Akron in Akron,Ohio. Michael joined the Advanced High Frequency branch of NASA’s Glenn Research Center in2011 and has since worked in the fields of atmospheric propagation andmicrowave remote sensing,satellite communications, antenna design andmetrology, and 3Dprinting. AlongwithDr. JamesNessel,

he is the co-investigator of NASA’s Atmospheric Propagation Studies project, which has operated over a dozentropospheric propagationmeasurement campaigns for the agency at Ka/Q-band (20 & 40GHz), V/W-band (70 & 80GHz), and optical (1550 nm).

JAMESNESSEL received the B.S. andM.S. degrees in electrical engineering fromArizona State Univer-sity (ASU) in 2002 and 2004, respectively, and received his Ph.D. in 2014 from the University of Akronin the area of active phase compensation of widely distributed antenna arrays. At ASU, he specializedin semiconductor device theory where his research involved the development of models for predictingthe effects of gamma radiation on semiconductormicroelectromechanical systems (MEMS) devices

with Los Alamos National Laboratories. Since 2004, he has been an Electronics Engineer with the AdvancedHighFrequency Branch of the National Aeronautics and Space Administration Glenn Research Center in Cleveland, OH,USA. He is presently co-investigator of NASA’s Atmospheric Propagation Studies project, and his research interestsinclude Ka-band propagation, microwave remote sensing, and active phase correction for transmit arraying ofmicrowave signals.

CARLO G. RIVAwas born in 1965. He received the Laurea Degree in Electronic Engineering and thePhD degree in Electronic and Communication Engineering, from Politecnico diMilano,Milano, Italy,in 1990 and 1995, respectively. In 1999, he joined the Dipartimento di Elettronica, Informazione eBioingegneria, Politecnico diMilano, where, since 2006, he has been an Associate Professor of elec-tromagnetic fields. He participated in theOlympus, Italsat and (the running) Alphasat Aldo Paraboni

(for whose propagation experiment he has been appointed Principal Investigator by ASI in 2012) propagationmeasurement campaigns, in the COST255, COST280, and COSTIC0802 international projects on propagation andtelecommunications and in the Satellite Communications Network of Excellence (SatNEx). He supports the ITU-RStudyGroups activities and he is Chairman ofWP3J of SG3 (“Propagation fundamentals”). He is the author of about

14 M. ZEMBA ET AL.

200 papers published in international journals or international conference proceedings. His main research activitiesare in the fields of atmospheric propagation of millimeter-waves, propagation impairmentmitigation techniques,and satellite communication adaptive systems.

LORENZO LUINI was born in Italy, in 1979. He received the Laurea Degree in TelecommunicationEngineering in 2004 and the PhD in Information Technology in 2009 both from Politecnico di Mi-lano, Italy. He is currently an assistant professor at DEIB (Dipartimento di Elettronica, Informazionee Bioingegneria). Since 2004, his research activities have been relative to E.M. wave propagationthrough the atmosphere, both at radio and optical frequencies: physical modeling and synthesis of the

meteorological environment (atmospheric gases, clouds and precipitation); development and implementation ofmodels for the remote sensing of atmospheric constituents using radiometric data; physical and statistical modelingfor E.M. propagation applications (rain intensity, gaseous absorption, attenuation due to rain/ice particles, wavedepolarization, scintillations, attenuation dynamics, attenuation due to clouds, expected performance of SatComand terrestrial wireless links, radio interference, spatial correlation of phenomena); analysis and dimensioningof wireless terrestrial and SatCom (GEO, MEO, LEO) systems operating in the 1 to 100 GHz range; design andsimulation of systems implementing FadeMitigation Techniques (site/time diversity, reconfigurable systems, powercontrol); assessment of the impact of the atmosphere on Free SpaceOptics Earth-space (to satellite or deep spaceprobes) systems; assessment of the impact of atmospheric constituents on Ka-band Synthetic Aperture Radars(SAR); analysis of the performance of space-borne GNSS receivers. He has been involved in several European COSTprojects, in the European Satellite Network of Excellence (SatNEx), and in several projects commissioned to theresearch group by the European Space Agency (ESA) and the USAAir Force Laboratory. Lorenzo Luini also workedas a System Engineer in the Industrial Unit—Global Navigation Satellite System (GNSS) Department—at ThalesAlenia Space Italia S.p.A.

GEORGE GOUSSETIS received the Diploma degree in Electrical and Computer Engineering from theNational Technical University of Athens, Greece, in 1998, and the Ph.D. degree from the Universityof Westminster, London, UK, in 2002. In 2002 he also graduated B.Sc. in physics (first class) fromUniversity College London (UCL), UK. In 1998, he joined the Space Engineering, Rome, Italy, as RFEngineer and in 1999 theWireless Communications Research Group, University ofWestminster, UK,

as a Research Assistant. Between 2002 and 2006 he was a Senior Research Fellow at Loughborough University,UK. Hewas a Lecturer (Assistant Professor) with Heriot-Watt University, Edinburgh, UK between 2006 and 2009and a Reader (Associate Professor) with Queen’s University Belfast, UK, between 2009 and 2012. In 2013 hejoined Heriot-Watt as a Reader. He has authored or co-authored over 150 peer-reviewed papers three bookchapters and two patents. Prof. Goussetis has held a research fellowship from theOnassis foundation in 2001, aresearch fellowship from the UKRoyal Academy of Engineering between 2006-2011 and a EuropeanMarie-Curieexperienced researcher fellowship in 2011-12. In 2010 he was visiting Professor in UPCT, Spain. He is the co-recipient of the 2011 European Space Agency young engineer of the year prize, the 2011 EuCAP best student paperprize and the 2012 EuCAP best antenna theory paper prize.