Friday, February 28, 2020

The Third-Generation of Broadband Aeronautical Satellite Communications

As a new decade arrives, broadband aeronautical satellite communications are moving into a third generation. Multiple modems and multi-beam antennas are creating a multi-dimensional interface challenge that only fiber optic cables can support, which will lower the weight along with expanding the horizons.
Additional Modem Unit

The First Generation – ARINC 791

Focusing on Ku/Ka band (as L band is inherently narrowband), the first generation, really gained a foothold with the advent of Connexion-by-Boeing (CbB) and Skylink by ARINC Direct starting by 2003. 

CbB Phased Array Receive Aperture
CbB introduced its air-transport service using a Boeing phased array antenna, only to transition to the MELCO reflector when low-elevation performance dogged service on North Pacific routes. The initial use of CDMA allowed arguably the most benign interference methodology, best suited by leasing whole transponders.

MELCO Antenna
The first-generation technology was a logical adaptation of a generic satellite terminal. Since then, the most significant change has been in modem/hub technology, allowing for slicing up a whole transponder.

Standardization of first-generation equipment began with the first meeting of what is now the SAE-ITC AEEC Ku/Ka satcom subcommittee in January 2008. ARINC 791 part 1 and ARINC 791 part 2 were issued to reflect the industry consensus of “box” or LRU positions, functions, and interwiring. The characteristics were very flexible, allowing variation of implementation using standardized provisions. 

ARINC 791 was crafted around the concept of a single antenna assembly, with combined Rx and Tx aperture, that was physically steerable in elevation and azimuth.

There have been some interesting antenna variations along this model: 
  • back-to-back multi-band antenna with a Ku-band aperture on one side and a Ka-band aperture on the other side

  • a multi-panel antenna with two or three apertures installed in tandem. The aft panels are shadowed at low elevation. At high elevation, the panels can be combined.

  • the tri-axis antenna with skew offset and intelligent beam management

In all cases, a single antenna was accommodated by using six or seven standardized lug locations.


The Second Generation – ARINC 792

The advent of flat-panel antennas, notably Gogo/ThinKom 2Ku, marks the advent of the second generation of Broadband Aero Satcom. The hallmark feature of a flat-panel antenna is in the name, it is a flat panel that does not involve a “swept volume”, or vacant space set aside for the antenna to scan. A flat-panel antenna represents an assembly – what you see is what you get, when it comes to radome design. A flat-panel antenna requires much less volume than an ARINC 791 style antenna.
On the issue of mechanical steering. There are three gimbals in the tri-axis antenna. There are two gimbals in a mechanically-steerable azimuth/elevation antenna.  These are what would be considered an ARINC 791 style antenna (one fixture with an antenna aperture that is “squared” to the servicing satellite).

A flat-panel antenna is oriented with the aperture facing straight up.  The aperture can now fill the swept volume, as much as tripling the antenna gain. The issue is that the satellite signal suffers scan loss: optically the gain falls off with the cosine of the scan angle (or sine of the elevation angle). Other factors can increase the order of the cosine roll off based on element design and impedance matching.

Only NGSO constellations, notably LEO, have the option for greatly reducing the scanning range if they are capable of service without any need to combine coverage with GEO. A reduced scan range favors a phased-array antenna.

ARINC 792 flat panel antennas are typically crafted into two apertures: one for transmit and a second aperture for receive. The receive aperture is large to increase receive gain – there is no other option. Transmit aperture is a combination of gain and beamwidth, where amplifiers can offset lower aperture gain and interference is managed adequately by power spectral density.

Size matters.

The larger the aperture, the more the gain, the higher the spectral efficiency. The higher the spectral efficiency, the lower the cost, the higher the performance.

Scan loss considerations favor a larger aperture. ARINC 792 was designed to accommodate up to two 42” apertures. A 42” aperture scan loss can match an ARINC 791 aperture (18” equivalent) down to five degrees of elevation. A 42” aperture could perform better than any ARINC 791 antenna, but there are some gotchas.

One option to improve scan rolloff is to tilt the entire aperture slightly towards the satellite. This is a mechanical addition to electronic beam steering. 

The benefit from a 5" tilt is compared to the flat orientation. The angle of tilt is dependent on the size of the aperture, smaller apertures can tilt more in the same height as large apertures. Fixed tilt degrades zenith performance slightly.  

The charts below (Ku band and Ka band) show the performance without tilt on the left and with 5" of tilt on the right side. The figure of merit is calculated at zenith, 45, 30, 20, and 10 degree elevation angle. The color coding potrays red as inadequate, yellow as marginal and green as acceptable. As apparent, the tilt allows good performance to 10 degree elevation for 21"-24" apertures while the flat equivalent would be 34" - 37". 

Ku-band Benefit from Tilt

Ka-band Benefit from Tilt

ARINC 792 was the first “clean-sheet” design where the objective was to develop a forward-looking standard (where ARINC 791 was a backwards-looking standard). The assumption was that the outside antenna equipment (OAE) would incorporate the high-power amplifier (HPA), thus vanquishing the KRFU and all the difficulties it presented.

With two apertures to support, a Tx and Rx connector arrangement was limited to a coaxial signal interface, power and control data. 

Beam-steering was presumed in the OAE, so the KANDU was stripped down into the KPSU, the power supply unit. Based on industry input, up to 1 kW per aperture is available.


The Third Generation – ARINC 792A

Phased-array antenna design comes in many flavors. A fundamental characteristic of a phased-array antenna is the method of beam steering. A mechanically steerable array physically orients a fixed beam. An electronically steerable array steers the beam without squaring up the aperture face. In this context, 2Ku VICTS is an ESA, but it relies on a single gimbal to allow overlapping platters to rotate relative to each other. 

A factor with any antenna is agility, or how quickly the beam can be moved to acquire or to track a target satellite. An ESA does not guarantee infinite agility. Each antenna type may have different limitations.

A phased-array antenna formulates a signal by varying in phase (or preferably, time-of-arrival) across the aperture. Carrier bandwidth complicates the proposition. The ideal method uses true-time delay. Fixed phase shifting creates beam-walk at the ends of the usable bandwidth. 

Optical methods are inherently true-time delay as evidenced by VICTS in 2Ku.  There are other methods of incorporating a true-time delay feed-network. 

If the feed network is passive, then a single amplifier input is used, and the antenna does not consume much extra power.

ARINC 791 antennas use a single input (for Tx) and a single output (Rx) with passive distribution across a horn array. The fixed feed network ensures the horn array radiation pattern is optimized in gain and beamwidth, relying on mechanical steering.  The situation is complicated by polarization. Linear polarization antennas have a Vertical and Horizontal interface and skew angle is achieved by mixing them proportionally. Circular polarization antennas may be fixed in Left-Hand or Right-Hand, or switched.

Solid State Phased Array Antenna

Integrated circuitry has opened up a new category of phased array antenna. Each antenna element becomes a full-blown radio which manages phase-shifting by command. The feed network is greatly simplified. 

Implementation involves beam-forming ICs that each power some number of attached elements. For example, one beam-forming IC may manage just one element, or two elements or four elements and so on. The number of elements is driven by the area of the aperture, the frequency range of the carrier, and the scan range. A beam-forming network is a set of beam-forming ICs that power the elements. It takes a complete beam-forming network to formulate one beam. The more elements one beam forming IC can power, the fewer beam-forming ICs.

Multi-beam refers to the ability to create more than one usable beam. Fundamentally, multi-beam can be done in time-slots, where the aperture switches between beams, or in a full-time manner. If a full-time manner, the set of elements can be split into two smaller sets of elements that each can independently form a beam (but at half the gain), or a second beam-forming network can be overlaid, allowing each beam to operate at full gain. There is no limit to the number of overlapping beam-forming networks. A single set of elements could generate 32 beams, for example.

Multi-band refers to the ability to create beams in Ku band and in Ka band. Multi-band creates a restriction in that the element layout has to support the highest frequency band (the densest array). It is possible to formulate a Ku-band signal on an element array that is laid out for Ka band by only using some of the elements.

Phased-Array Antenna Performance

Antenna transmit performance is scaled largely by beamwidth, which relates to aperture size. HPA can offset gain, and spectrum can offset beamwidth.

Antenna receive performance is fundamentally limited by gain and noise figure (or noise temperature): Figure of Merit (G/T).  Receive gain is simply a factor of the number of elements and element gain offset by scan loss.

Passive antenna arrays may operate at significantly lower noise temperatures than solid state arrays, which suffer from a high noise temperature. The best a solid state antenna is approach the performance of a well-designed passive antenna, but typically it will operate 1-2 dB down. This factor acerbates the low elevation performance further.

The following chart is a tool to quickly get a sense of how many elements and figure of merit for any sized aperture between 11" and 42". The technique translates G/T to number of elements (by its x-intercept) with either the passive or active array lines. For discussion sake, the passive array is about the highest you can expect and the active array is about two dB down. The same chart translates from the number of elements to the diameter of a circular aperture.  Ku band and Ka band charts are provided.
Calculating the erformance of passive and active arrays of a given size aperture

Ku-band Phased Array Performance Chart

Ka-band Phased Array Performance Chart

Phased-Array Antenna - Modem Interface

There are examples of moving the modem into the solid-state phased array antennas emerging in the marketplace.  I will discuss this further below, along the lines of needing to support more than one modem.

Every beam-forming integrated circuit (IC) has a transmit side that has some form of RF up-conversion and amplification and a receive side that has some form of RF down-conversion and amplification.

Depending on the design, the beam-forming signal could be digital (data), baseband (I/Q), or intermediate frequency (IF).

Digital Data Mode

In the digital data mode, the beam-forming IC would attempt to modulate or demodulate the signal based on its reception. Of course, a single element would not have the gain necessary for the carrier signal, but possibly it could pick up a beacon. The real opportunity is the digital date interface to the aperture. Effectively, the modem is moved from the Modman to the Outside Antenna Equipment. This is the simplest antenna, a fully integrated terminal – just power and data.

Baseband Mode

Satellite communications rely on an imperfect transmitting medium that suffers from rapid fade (or varying signal strength). Typically, some form of phase-shift keying is used with a margin to demodulation, to manage fade. The choice of modulation and coding adapts to the signal strength available, starting with binary phase-shift keying (BPSK) and then QPSK, 8-PSK and so forth. In this context the bits are conveyed by symbols (one bit per symbol, two bits per symbol, four bits per symbol, etc.). This is baseband i/q symbol modulation.

The modem normally converts from digital data to I/Q to IF. Baseband interface means that the modem takes the digital data to i/q and communicates I/Q to the antenna for conversion straight to RF (no IF at all). All antenna designers seem to favor moving to baseband interface over IF. 

Baseband is a digital data interface. The I/Q signals are each communicated at the symbol rate with enough bits to reflect their analog value with enough discrimination. The symbol rates are moving to as high as 600 Msps and it takes something like 4 bits for each I or Q, or about 5 Gbps. That could easily double to 10 Gbps when multi-beam is involved and doubled again for both Receive and Transmit. Baseband quickly leaps beyond the capability of copper-based interfaces and into the realm of fiber optic interfaces.

Fiber Optic Interfaces

Enter the third generation of broadband aeronautical satellite communications. ARINC 792A is being proposed for project initiation at the upcoming AEEC general session to build to the equipment described in ARINC 792 but abandon the Ethernet and coaxial interfaces in favor of fiber optic.

Fiber optic cables and connectors have evolved significantly over the last 50 years. There is no reluctance to using fiber optic cabling based on manufacturing or maintenance concerns. Fiber optic cable can be used for transmitting analog signals (IF/RF over fiber) or very high digital data rates.

The preferred implementation dedicates a fiber strand to a single (simplex) direction of information flow, but it is possible to combine information flow in both directions (duplex) over one strand. 

The beauty of fiber optic bundles is that they can include almost an infinite number of strands. Each strand opens another interface.  Instead of complex methods to combine signals on a single coax, each signal can have its own strand. 

There are many interesting growth opportunities that can be provisioned by including enough strands without changing the penetration through the pressure bulkhead, again an option unthinkable with coax. 

How many strands, which types, and how to terminate them are all for ARINC 792A deliberations to come. One option is to forgo a bulkhead connector, and instead push the entire fiber optic bundle through and seal the hole around it – similar to the ARINC 741 HGA top-mounted interface. 

Multiple Modems

The modem is a beachhead. It represents one-half of the network interface; the hub is the other end. The hub not only manages the network but also is the gateway to the terrestrial internet. The hub is part of a teleport. A given satellite operator will operate hubs in a homogenous manner, to ensure that all remote client modems are compatible. However, modem/hub suppliers are very competitive and do not interoperate. The modem onboard an airplane is compatible with only one hub supplier. If a service provider wants to extend service between two satellite operators, they face the challenge of needing two different modems. That issue is paramount with the highly integrated spot beam networks and NGSO networks. Where before you leased spectrum and hosted your own hub in the carriers assigned, now you are accessing a network that relies on you using their modem.

Multi-carrier applications run into multiple modems as well.  The simple handoff modem (two demodulators and one modulator) is not a multiple modem application – the network is homogenous, there is only one transmit carrier. Multi-modems are needed with full time multi-beam – where the carriers are not homogenous and where there may be two transmit carriers. Note, we have never fully analyzed the issue of intermodulation between carriers other than spurious noise (Ku/Ka aero) and there may be some surprises awaiting us.

With the need for two or more modems, an ad hoc working group formed to consider the options.

There are three obvious multiple modem architectures:
1.     Software-Defined Radio
2.     Card-Level Standard
3.     Additional Modem Unit

Software-Defined Radio (SDR)

The first and foremost is moving to a software-defined radio (SDR) modem. In fact, every modem is or will be an SDR, the simplicity is programmability.  The problem, there is no agreement on a performance standard or an API. It is apparent that modems are migrating towards the antenna – whether i/q or digital data interface.  What we call a modem today is really two parts – the network manager and the modulation and coding (modcod). The network manager is highly proprietary and completely customized. The modcods are of a type, with some standards and logical extensions. Satellite selection, beam pointing, tuning, and power control may remain in the Modman; the data to signal functions may move to the antenna.

An SDR modem standard will emerge when the suppliers are agreeable; that moment is not now.

Card-Level Modem Standard

The second option was to push the aero modems into a specific form factor. With this in hand, the Modman can be built in a standardized manner. Today the Modman is customized to host a specific modem card. The SDR advantage is prevalent here, if SDR suppliers set on a standard form factor and edge connector than the modems would follow. There was no proposed standard offered. The card-level modem option is deferred and may be overtaken when SDR programmability is in hand.

Additional Modem Unit

The option for hosting multiple modems that is most obvious and practical is to create a box-level (LRU) characteristic. The Modman is a 4 MCU LRU. It has to host a processor card (the manager) and a power supply, plus possibly a frequency reference. The Modman gets a budget of 100 Watts of power and cooling.  While large scale integration is forthcoming, today there is simply not enough room to host another modem card, plus the power and heat demands it would entail.

As the SDR approach seems the future, the working group agreed to propose an Additional Modem Unit (AMU) as an adjunct to an existing Modman (A792 required, A791 desired) – that would not require the Modman to be replaced (it would have to be reprogrammed, of course).

Setting aside the options relating to fiber optic interfaces to 792A work, various options were considered as they impact three interfaces: keyline, network (Ethernet) and signal (Coax). A daisy-chain approach provides a simple way to add the AMU to an existing Modman. Up to two AMU’s can be added in this manner.

While aircraft rack space is at a premium, the working group believes that a 2 MCU AMU can be accommodated. A 2 MCU enclosure affords 50 Watts of cooling and ship’s power. A smaller 1 MCU enclosure gets half the budget and is considered inadequate for near term equipment. However, a future application may be suitable for 1 MCU.

The AMU connectors inserts are either size 1 or size 2. Size 1 Quadrax inserts are not currently listed in ARINC 600 and may not be currently approved and stocked by aircraft manufacturers. Size 1 inserts can support either a 1 MCU or a 2 MCU encloser.  Size 2 inserts for the AMU are listed and approved but prevent the AMU from shrinking from 2 MCU to 1 MCU.  No decision has been made on which insert to embrace, but the 2 MCU solution has no barriers to agreements, has considerable growth including fiber optic, and only uses two inserts Versus three inserts for size 1 (and an alternative setup if fiber optic). The 2 MCU solution allows for multiple modems within the enclosure. 1 MCU LRUs are a bit flimsy and require some form of bracing to survive environmental qualification.

Network Security and Quality of Service (QoS)

ARINC 791 Part 2 describes a multi-domain level VLAN. The supported domains include Passenger Owned Devices (POD), Passenger Information and Entertainment Services, and Aircraft Information Services (AIS). Each onboard Ethernet interface is assigned to a given domain and is isolated from the other domains. VLAN subnetworks manage the isolation of traffic types and networks within a single serial data stream over the satellite data link. The proprietary radio layer adds additional features that VLAN rides on top of. By necessity, the satellite link combines all traffic into one physical connection.

Logically, the traffic priority is arbitrated, from within the Modman and by the Hub between all the other users of the network (other airplanes). Key to managing traffic congestion is priority and precedence, which relates the import of each data gram against another. 5G network slicing is a concept of running virtual networks across a single physical network. In this manner, the traffic is assigned to logical networks and the arbitration occurs between the logical networks, effectively assigning network resources intelligently rather than arbitrarily. 

AEEC Project Paper 848, which work was begun in 2014 and is nearing publication, professes the Media Independent Secure Offboard Network (MISON), which establishes a virtual private network (VPN) extending a secure subnetwork interface between and airplane onboard local area network and a single Enterprise terrestrial LAN. The initial VPN technology is IPSec.

IPSec will obscure any markings of the payload used to arbitrate QoS. PP848 proposes to provide sufficient marking between the MISON onboard instance and the radio interface to permit network slicing across the radio layer. A MISON VPN instance is already constrained to a single Enterprise and a single QoS. A MISON VPN is first established by the onboard MISON instance, but then permits asynchronous bidirectional communication.

PP848 does not profess the means to assign traffic to a given MISON instance. Once the MISON has segregated the traffic types into MISON VPN instances, the MISON must signal to the radio the QoS to apply. PP848 will defer to the radio to define the method for marking QoS.

The typical methods for marking are connection-oriented (effectively based on subnetwork) and connection-less (based on marking inherent in the datagram). 

Layer 3 Connection-less marking includes using the Type of Service (ToS) field in IPv4 headers to designated Differentiated Services Code Point (DSCP).  DSCP is a mutable field that relates per-hop behavior.

Layer 2 Connection-less marking includes the Class of Service (CoS) that is prevalent in VLAN. The proposal for ARINC 791 Part 2 is to use CoS for QoS marking, and to use GSMA standard for defining the designated traffic types and assignments.  

Application DHCP (PHB) categorizations are listed below.

The proposal for ARINC 791 Part 2 is to continue use of VLAN subnetwork segregation to isolate domain traffic. The specific subnetwork assignment will be configurable for each installation. 

ARINC 791 Part 2 supplement 2 is being prepared for summer 2021 publication.

More info about the subcommittee can be found here.

Stay tuned!

Peter Lemme

peter @
Follow me on twitter: @Satcom_Guru
Copyright 2020 All Rights Reserved

Peter Lemme has been a leader in avionics engineering for 38 years. He offers independent consulting services largely focused on avionics and L, Ku, and Ka band satellite communications to aircraft. Peter chaired the SAE-ITC AEEC Ku/Ka-band satcom subcommittee for more than ten years, developing ARINC 791 and 792 characteristics, and continues as a member. He contributes to the Network Infrastructure and Interfaces (NIS) subcommittee developing Project Paper 848, standard for Media Independent Secure Offboard Network.

Peter was Boeing avionics supervisor for 767 and 747-400 data link recording, data link reporting, and satellite communications. He was an FAA designated engineering representative (DER) for ACARS, satellite communications, DFDAU, DFDR, ACMS and printers. Peter was lead engineer for Thrust Management System (757, 767, 747-400), also supervisor for satellite communications for 777, and was manager of terminal-area projects (GLS, MLS, enhanced vision).

An instrument-rated private pilot, single engine land and sea, Peter has enjoyed perspectives from both operating and designing airplanes. Hundreds of hours of flight test analysis and thousands of hours in simulators have given him an appreciation for the many aspects that drive aviation; whether tandem complexity, policy, human, or technical; and the difficulties and challenges to achieving success. 

1 comment:

  1. Hello Peter. Is there any discussion on the team about a next gen satellite antenna (ESA) standards for GA/BA? I don't see A791 or A792 as practically applicable to GA/BA outside a very small number of BBJ/ACJ-type aircraft or military/gov applications.