Friday, April 22, 2016

Aero Antennas: Bigger is Better!

ThinKom and QEST are both promoting Ka-band flat panel antennas that use about a 25" aperture for receive.  ARINC 791 provisions can support at least a 30" aperture, as shown by Gogo with 2Ku.  Should a Ka antenna be built as big as possible, or is there any size that is "big enough"?

As time moves forward, will service levels going up (due to evolving user behaviors) outrace the cost of services going down (due to expanding capacity and lower costs to launch capacity)?

How valuable will a larger antenna be if trends run away from expectations?

There are two fuselage mount antenna types:
  1. flat panel antennas that are fixed to the fuselage, the aperture scans without squaring the face
  2. multi-gimbal antennas square the aperture face towards a satellite, in azimuth and elevation 
The classic flat panel antenna uses a phased array.  Other methods are in the marketplace, notably flat panel antennas from Kymeta and ThinKom, that do not utilize a physical array of radiating elements per se. ThinKom VICTS has a single mechanical axis upon which platters rotate for beam steering.  All other flat panel antennas have no moving parts.

Flat panel antennas "shrink" in effective area as a satellite moves towards the horizon (scan loss).  The diameter of a flat panel aperture represents its performance with the satellite at zenith.  The effective (average) aperture is smaller.

Multi-gimbal antennas (fuselage mount) are generally limited to an aperture face 30"-36" Wide and 6"-8" High.  An 18" circular aperture is 254 sq in, which approximates the area available to a multi-gimbal antenna.   A tail-mount antenna is typically limited to 11.5".

Multi-gimbal antennas operate consistently in all pointing angles, and are especially useful when the satellite is very low towards the horizon (low elevation).  Flat panel antennas have better characteristics in the temperate zones (near the equator) than multi-gimbal antennas (skew angle effect).

In summary for this analysis:

Multi-gimbal:  Tail Mount 11.5", 18", 24", 30"
Multi-gimbal:  ARINC 791 Fuselage Mount
Flat panel:      ARINC 791 Fuselage Mount 24", 30", 42"

Four Internet service levels are provided:
  • 150 kbps
  • 250 kbps
  • 500 kbps
  • 1000 kbps
Each service level includes an additional 25% data rate on the return channel.  These average data rates are modeled to represent contemporary services (starting at 150 kbps) and ten-year hence (1,000 kbps).

The analysis is based on the assumption that each day every average channel is used by about three users in tandem (a conservative estimate), which equates to about 100 users a month per average channel. This assumes all channels are sold the average amount.  Costs are driven by peak provisioning, so every peak channel is used about three times a day.  Higher numbers of uses would lower session costs, whereas fewer uses would raise session costs.

Spectrum pricing is dropping as high throughput satellites bring more and more capacity to the marketplace.  Traditional wide beam transponders were traditionally leased at about $1.6MM per 36 MHz transponder per year ($3,800 per MHz per month).  Spot beam transponder pricing is expected closer to $1,000 per MHz per month and going down from there.

The cost relevant to a contemporary 150 kbps Internet session is expected to fall as average usage increases.  The Internet value proposition evolves hopefully with spectrum pricing as suggested by the red "trend" arrows in the following table, showing cost per each user session against varying cost of spectrum ($/MHz/month), using a 24" aperture:

Flexible satellites with large apertures and overlapping coverage across the geostationary arc will deliver lower and lower costs.  The challenge is insatiable growing demand and  level revenue per user.  I remain optimistic the market will deliver the cost efficiency the market needs, while acknowledging the challenges to deliver spectrum below $100 per MHz per month (over an abstracted coverage area).

Another perspective is to note the intersection of each Internet service level against a target user cost per session, shown below along the $1 line.  The intersecting cost of spectrum offers a guideline for the feasibility of service (e.g., once the cost of spectrum falls below the intersection for a given service).

Dreaming of a 48" Antenna

The above table represents cost per session and per airplane assuming $500 per MHz per month using a 48" (1.2m) circular aperture.

Cost per Mbps per month is based on weighted antenna performance on both the forward and return channels.  There are many factors that influence performance of an aeronautical satcom.  This analysis is focused on defining the trends, if you will the sensitivity, around increasing demand while using different antennas.  I have taken liberties to simplify the problem, to which I discuss in the Link Budget attachment at the end of this article.
Take careful note that secondary effects can be pronounced, that reality does not follow prediction especially in timeframe, and that cost predictions for relative perspective are only indicative of actual cost values.  There are other costs in providing a service that are fixed and variable.  Costs not related to the satellite service itself are not accounted for in this analysis.  This is a high-level approach to the problem that offers a sense of scale and proportion with antenna size as the variable.
For the 48" aperture, the resultant cost of $244 per Mbps is 49% of the lease cost per MHz. The Antenna Cost Efficiency ($E) of 49% is equivalent to spectral efficiency of 2.  A network built around 48" antennas would pay the headline rate per Mbps (no surcharge).   Smaller antennas will need more spectrum for the same Mbps.  A surcharge of 100% would equate to an antenna that needs twice the spectrum than a 48" antenna.  In essence the value of the antenna can be asserted by the surcharge, where the lower the surcharge the better.

For clarity:

  1. A 48" antenna is assumed spectral efficiency of 2, so cost per Mbps is one-half cost per MHz.
  2. Spectral efficiency of 1 means cost per Mbps equals cost per MHz, surcharge is 100% or twice the cost of a 48" antenna.
  3. Spectral efficiency of 0.5 means cost per Mbps is twice cost per MHz, surcharge is 300% or four times the cost of a 48" antenna.

The following table and graph show how the four service level user session cost varies as a function of the lease cost of spectrum, using the 48" aperture.

An airline has an option to offer a service for free if the session cost is low enough to be offset by modest promotions.  I generally target $1 session cost as the upper side of a free service.  For a 48" dish, in the near term, a 150 kbps service is a candidate for free service when spectrum is less than about $1,500 per MHz per month.  In the future, to offer a 1,000 kbps service the spectrum cost will need to be less than $250 per MHz per month.  The hope is over time that spectrum costs will fall faster than usage will rise, that what cost $3,800 yesterday and $1,500 today will only cost $250 ten years from now.

Ahhh, if only we could mount a 48" dish on airplanes.  It is done, for example on the Global Hawk:

Aero Antennas: 11.5" -  30"

If a 48" antenna is the standard, aero antennas are noticeably disadvantaged.  The resultant spectral efficiencies for 11.5", 14", 18", 21", 24", 30" and 42" parabolic reference antennas are calculated under a representative Ka-band spot beam satellite environment using a proprietary link budget tool with somewhat conservative modulation and coding.  Each of these apertures contribute to a trend line - for relative comparisons of the aero antennas of interest.  Each antenna is modeled with beamwidth and interference levels consistent with its diameter - see the Link Budget attachment at the end of this article for further details.  An aero antenna will be shown to operate along a trend line made up from performance of each of these reference apertures.

The satellite spot beam is presumed to have a beam center and a beam edge, with performance falling off towards beam edge.  Even in spot beam operation, a given client flies across the beam contours.  In beam hopping, an optimized beam is selected to favor beam center operation.  The EIRP and G/T for the spot beams is representative of an actual Ka-band spot beam network and is used consistently with each reference antenna.

Each table represents one antenna across five columns (or contours) from beam edge towards beam center. 

Applying a weighting function allows the spectral efficiencies to be combined for an aggregate value for each reference antenna.

Flat panel antenna performance is a function of satellite elevation angle, whereby the average recieve performance can be associated with a smaller antenna size.  As discussed in the attachment, skew effects are left as a secondary factor.

The following table shows that a 42" circular flat panel operates like a 25" aperture when the satellite is at 20 degrees above the horizon.

The average receive performance of a 42" flat panel antenna can be approximated by a 34" antenna.  Allowing for lower efficiency offers a range from 30"-34" as the average.

A 30" flat panel operates to as small as an 18" aperture and average receive performance is like a 24" antenna. Allowing for lower efficiency offers a range from 22"-24" as the average.

A 24" flat panel operates to as small as a 14" aperture and average receive performance is like an 19" antenna. Allowing for lower efficiency offers a range from 16"-19" as the average.

A zero surcharge assumes performance equal to a 48" antenna. A 100% surcharge is when the antenna costs (in spectrum) are twice those of a 48" antenna.  A 200% surcharge is when antenna costs are three times those of a 48" antenna.

As stated earlier, an ARINC 791 fuselage mount multi-gimbal antenna performs like an 18" antenna.

The following tables summarize the user session cost performance for each of the candidate aero antenna sizes as a function of cost of leasing satellite service ($/MHz/month):

11.5" Parabolic (Tail Mount)

18" Parabolic (ARINC 791 Fuselage Mount, average 24" flat panel)

24" Parabolic (average 30" flat panel) 

30" Parabolic (average 42" flat panel) 

The following tables summarize user session costs assuming $500 per MHz per month.  


There is an appreciable benefit with each successively larger antenna.  A 30" flat panel antenna offers a 30% reduction in operating costs from an ARINC 791 multi-gimbal antenna as the surcharge is reduced from 134% to 74%. A 24" flat panel antenna operating costs are similar to an ARINC 791 multi-gimbal antenna, with benefits based on coverage.

The cost savings for a larger antenna are paced by the usage for given plane.  If ten users per flight, the savings from a 30" flat panel antenna compared to multi-gimbal antenna are about $3k-$18k a year.  If 50 users per flight the savings are about $13k-$91k a year.  The range is based on the service level, where 1,000 kbps drives the highest costs. These benefits may offer a substantial return-on-investment in fitting the largest antenna possible at the appropriate time.  The benefits for retrofit a new antenna are highest when usage is highest.  Higher lease rates would create greater savings, lesser lease rates would minimize savings.


The following table (repeated for clarity) summarizes each aperture surcharge and comparable areas.  The 20 deg elevation  is provided to link to the equivalent parabolic reflector as an aid to drafting.

An 11.5" tail mount antenna has a surcharge of nearly 700%.  An antenna this small is workable but hardly efficient.

It should be emphasized that attempts to aggregate separate flat panels to face fore/aft/port/starboard and towards zenith (a house of cards) requires an effective aperture of over 250 sq inches in all directions to improve upon the ARINC 791 fuselage mount antenna.  Offering less will quickly degrade efficiency and performance.  I personally do not see a way forward for a house of cards, usually envisioned as two conformal arrays in a saddle-bag arrangement, but more likely a five sided shaped structure about the size of an ARINC 791 radome.

The ARINC 791 fuselage mount multi-gimbal array is equivalent to an 18" aperture, where the surcharge is 134% compared to a 48" antenna.  This is the baseline from which we are comparing flat panel antenna options.

A 24" flat panel antenna operates from 24" down to about 14" aperture (at 20 degrees elevation).  It should be noted that surcharge levels are not linear across this range, and low elevation performance of this antenna is particularly costly.  The average performance of a 24" aperture should be similar to the ARINC 791 fuselage mount multi gimbal antenna except for extended operations at low satellite elevation.

A 30" flat panel operates from zenith to about 20 degrees elevation, which is roughly equivalent to an 18" aperture.  The average performance is around a 24" aperture.  The 30" flat panel antenna has been shown to fit on top of ARINC 791 adapter plate while cutting the surcharge nearly half of the gimbal mount antenna.

A 42" flat panel operates from zenith to about 20 degrees elevation, which is roughly equivalent to an 24" aperture.  The average performance is around a 30" aperture.  A 42" aperture offers some margin and flexibility in operations and technologies without inordinate cost sensitivity.  It cuts the surcharge to about one-third compared to an ARINC 791 multi-gimbal antenna.  

I believe it is possible to hang a 42" aperture onto ARINC 791 fittings, especially if using an oblong/rectangular approach to favor aspect ratio and minimize frontal area.  In other words, rather than 48" circular, instead more like 30" wide by 50" long.  We already have radomes 123" long, but for ARINC 791 assemblies would ideally be in the 80" range.  It is possible to overlay both transmit and receive apertures into the same area, by the way.

A summary of the candidate flat panel antennas shows how their costs vary relative to an ARINC 791 fuselage mount multi-gimbal antenna.  The 24" flat panel antenna offers a more compact form factor for the same performance of a larger ARINC 791 multi-gimbal antenna.  However, service cost savings would favor a 30" flat panel or larger.

Combining all the antenna options into one chart offers a relative landscape for antenna sizing.  The equivalent multi-gimbal antenna is shown above each flat panel. 

Airframe and antenna suppliers are encouraged to explore options for really large apertures.  The top of the tail (today only housing an 11.5" antenna) is an ideal radiating location.  A 24", or better a 34", parabolic reflector would enable a low cost antenna with excellent performance.   A 42' parabolic reflector would consistently deliver twice the service level for the same service cost as an ARINC 791 multi-gimbal antenna.

I repeat my appeal for Ka-band antennas to be built as large as possible, within the constraints of ARINC 791 provisions.  The recurring costs and increased network capacity will reap savings and network capacity every day.  Please develop 30" or larger flat panel antennas for Ka band.  Take note of the low elevation performance when going to an aperture in the 1400 sq inch range.

Don't forget Low Elevation Performance

Flat plate antennas usually perform best when the satellite is 20 degrees above the horizon or higher.  However, it is desirable to be able to operate at lower elevation angles.  This can occur anywhere along the periphery of coverage, not necessarily only at high latitude.  The situation can also be encountered during airplane maneuvering.

Performance down to 5 and 10 degree elevation angles is shown in the table below:

While it may be conceivable to close the link with less than 100 sq. inches of aperture, but for this analysis I am assuming 100 sq. inches the limit for operation.

The 24" flat panel would be expected to operate at 15 degrees or more elevation.

The 30" flat panel would be expected operate above 10 degrees elevation.

The 42" flat panel (1385 sq in) would be expected to operate down to 5 degrees elevation.

Other factors may limit low elevation performance as a result of antenna design.

Stay tuned!

Peter Lemme
peter @

Copyright 2016
All Rights Reserved

Link Budget

The cost of a channel is calculated based on weighted performance between beam edge and beam center using a proprietary link budget tool.  The following shows the forward and return channel calculations under a typical Ka spot beam satellite with a 48" aperture.

The forward channel operates at about 480 Mbps and the return channel about 120 Mbps just for an example of using 100 MHz on the forward channel and matching the return channel with 25% of the forward data rate. The forward channel and return channel spectral efficiencies are not affected by any specific, optimized data rate.  The forward channel is bounded by downlink spectral flux density and the return channel in uplink power spectral density.  The operating point is at the threshold of the limits as modeled for an existing Ka-band spot beam service.  In this case, the forward downlink SFD is almost 10 dB below the coordinated limits, but reflects the available power from that space vehicle.  Aperture effects forward channel downlink figure of merit (G/T) and interference and uplink power spectral density.

Skew angle issues at Ka band are moderated by the exceptional discrimination.   Skew angle issues are show below to take off above 45-60 degree skew angle.  For the purpose of this analysis, skew angle is 45 degrees or less.   Higher skew angles will degrade performance as discussed in the following subsections.

Return Uplink Interference

The satellite network is modeled with only 9 dB Uplink C/I based on link budgets submitted to the FCC.  The return channel is dominated by interference holding back the benefits from a more discriminating beam.  In other words, the spectral efficiency on the return channel does not vary much with larger apertures.

The return channel performance under wide beam scenarios modeled as 16 dB C/I is shown below.  

The reduced return channel bandwidth relates to a cost reduction from 49% to 41% of the cost per MHz, or about 15%.

Return Performance as a function of Skew Angle

The following graph shows the maximum PSD from two candidate multi-gimbal antennas (30"x6") and (36"x8") at 14 Ghz and again at 29.5 GHz.  Note the 14 GHz values provided utilized the Ka-band spectral mask only for comparison purposes only.  The results are in dBW/40 kHz.
 Return channel performance does not degrade significantly until skew grows past about 40 degrees of these multi-gimbal antennas.

Forward Channel Performance as function of Skew Angle

Using a simple antenna model, C/I due to adjacent satellite interference (ASI) was calculated for both Ku band and Ka band.  The Ku band data is provided for discussion purposes.  Ku band C/I degrades steadily from a 34" aperture to a 10" aperture..  Ka band C/I does not degrade until shrinking smaller than about a 20 degree aperture, upon with C/I decreases rapidly.

Two candidate multi-gimbal antennas (30"x6" and 36"x8") C/I as a function of skew angle are shown below.

Ku band C/I performance is already degraded at zero skew angle for the 30"x6" aperture, whereas in Ka band the same antenna does not degrade until about 45 degree skew.  The larger aperture performs a bit better.

Skew angle performance of a flat panel antenna requires another variable, elevation angle.  The aperture shrinks when the satellite is low to the horizon.   The following graph shows C/I due to ASI as a function of skew angle for 90. 45, 40, 30, 20, and 10 elevation angles.

The following graph\ includes data from both the multi-gimbal and flat panel antennas.

Cost as a function of Skew Angle

Taking all of the above into account, normalized cost was calculated for an 18" aperture (approximating the multi-gimbal antenna) as a function of skew angle.  

Where 1 is the cost at zero degree skew angle, an 18" aperture cost starts to increase when C/I falls below about 10-12 degrees skew.  The costs increase 50% when C/I is 4 dB, and further reductions in C/I have a dramatic rise in cost.

Antenna Models

Here are the reference antenna models.

No comments:

Post a Comment