Estimates to produce the current SpaceX Starlink dish antenna, an advanced marvel of RF electronics, range between $2000 and $2,400 in quantity, with Business Insider stating it cost $2,400 per dish in quantity one (1) million. SpaceX is currently licensed to operate up to 1 million dishes in the U.S. and is seeking approval to operate up to 5 million dishes in its latest Federal Communications Commission (FCC) filing.

Without the dish, Starlink’s current 10,000-plus consumer user beta wouldn’t exist. But what exactly is the dish capable of in terms of delivering speeds and what other products and improvements could emerge from the initial hardware?

Dish 1.0 – What we know

The current (3/1/2021) SpaceX Starlink beta single user “kit” is comprised of a dish, its mount, a low-cost router, and cabling to connect everything together, along with a graphic diagram showing how everything fits together. No user manual, either in the box or online. It is designed to be self-installed, with different antenna mounting options available through the company’s website.

Broadband speed delivery is dependent upon multiple factors, according to the company, including the total number of satellites in orbit, dish software, and network software and configuration. In initial statements, SpaceX founder and CEO Musk said the system would be capable of gigabit speeds, then he and other SpaceX officials started resetting expectations more conservatively to being “sufficient for online gaming.”

Beta customers were told to expect download speeds between 50 Mbps to 150 Mbps, but some Reddit posters have seen upwards of 175 Mbps.  Musk tweeted in February that users could see speeds of up to 300 Mbps by the second half of 2021, the time the company should have around 1440 satellites in orbit for full global coverage.  One U.S. Air Force test reportedly provided up to 600 Mbps to an aircraft in flight, with a Ball Aerospace antenna, while a second test planned in the spring of 2021 will also use Ball antenna hardware.

The basic Starlink dish is powered using a custom version of Power over Ethernet (PoE), drawing between 90 to 100 watts in operation. Broadcast power output is such that the heat generated from operations acts as a deicer.  How well the dish holds up in hotter environments as well as all environments over time has yet to be seen.

SpaceX is testing the Starlink system in various civilian mobility applications, including at sea aboard its fleet of recovery ships and onboard private jets. The dish has been seen mounted onboard SpaceX ships. The (presumably) unmodified dish is being used in the business aviation tests.

Beta consumer dishes are “geo-locked” to a specific service area or “cell.” Move much outside the cell area and service stopped.  Geo-lock is tied to a software/firmware switch in the dish, with an onboard GPS processor providing dish location during operations. The March 5, 2021 FCC filing implies that location data is or could be checked on a frequent basis for aviation applications to ensure the dish only operates in airspace where it is licensed.  Consumers will likely pay more for mobility, even if it is simply moving a dish between one service cell and another without any permanent vehicle mounting.

What we don’t know

We don’t know if the current dish can delivery up to gigabit downlink speeds since SpaceX has not made any statements or provided any documentation.  The use of gigabit PoE for dish data output suggests it might, but there’s no information available to verify.  Gigabit speeds would presumably require more elaborate RF encoding than currently being used in the beta or in the 2H 2021 upgrades.

The top end of uplink speeds are a big unknown. Beta customers seem to be getting 10 Mbps and faster uplink speeds, but it is unknown if they will be formally capable of getting 20 Mbps or faster uplink speeds in the second half of 2021. There are certainly upper-end uplink speed limits for consumer-designed equipment due to FCC license limits on broadcast power and RF safety practices.  The author freely admits his lack of depth in RF regulation and antenna magic.

Software upgrades/controls in the works

Dish power consumption is a known issue/feature, with the device pulling over 90 watts during operation according to beta users on Reddit, giving it a de facto self-deicing capability.  An off-the-shelf 1000VA lead-acid battery UPS would be able to support full power dish operations for a few minutes, exclusive of other devices being connected.

SpaceX is in the process of making Starlink a regulated phone company to access various FCC subsidy programs.  As a regulated phone company, Starlink will have to provide some sort of “lifeline” capability for voice dial tone if regular power is interrupted and the company said it will provide a backup battery designed to power the dish for up to 12 hours.

Since normal power operations would quickly drain any affordable UPS option dry, the Starlink dish will need at least one low-power mode to provide low-bandwidth services for voice in a lifeline capability.  This low-power/low-bandwidth mode is likely to be in the 64 Kbps range, enabling the use of an off-the-shelf IP phone using a G.722 HD voice codec for outbound calls and keeping the line alive for inbound calls.   (Yes, there are better codecs that provide higher quality voice at lower bandwidth usage, but G.722 is supported in practically all IP desktop handsets making it an off-the-shelf solution with plenty of options along with the ability to transcode to other formats if needed. My friends in the IP voice community are welcome to fight me on this.)

More speed through bonding

Government and business customers could “bond” multiple dishes together with a suitable router for increased bandwidth.  Bonding multiple data channels is a well-established practice in the telecom world and would enable a customer site to gain access to an aggregate 600 Mbps downlink with two dishes or over 1 Gbps with four dishes, assuming 300 Mbps top speeds per dish by the second half of 2021. No customized hardware and minimal to no customized software would be required, making this a feasible solution for larger businesses or as a scalable solution in cellular backhaul.   Modified cabling may/will be required for direct installation on taller cell towers.

The only limiting factor with bonding would be enough power to support the operation of multiple dishes and associated equipment and sufficient physical space with a clear view of the sky for multiple dishes.

COTS dishes: Maybe?

For various reasons, the U.S. military would prefer multiple sources of ground station hardware. The U.S. Air Force wants an open satellite ground station architecture, so it and the other service branches are likely to experiment with different hardware outside of the basic Starlink kit. Instead of packing multiple different systems for GEO and LEO, the services want a single piece of hardware that can be reconfigured on the fly with software to access what is needed.

Dish 1.x

Some changes to existing Starlink dish design and manufacture are no doubt already in the works, focusing on improving performance and lowering cost for large scale production.  SpaceX is licensed for up to 1 million dishes in the United States alone, but it has likely shipped substantially less than 5,000 dishes per month during the initial beta period, given the company’s early 2021 statements of “over 10,000” users with the “private” beta starting in August 2020 and opening to a public beta in October 2020.

The Tesla-Starlink paradox…

Initial media reports of the Starlink March 5, 2021 FCC filing for mobile operations indicated that Tesla vehicles would soon get the capability to be linked via satellite.  SpaceX and Tesla CEO Elon Musk was quick to dispel this notion, tweeting the existing dish was “much too big” and more appropriate for installation on aircraft, ships, large trucks, and RVs.

But “too big” is a quite simple answer, implying physical form factor is the only issue involved. Certainly, mounting a satellite dish on top of the roof of an aerodynamically smooth Tesla would be awkward and decrease overall mileage/range per charge due to more weight and increased wind resistance.  Powering up and operating the dish is a much more significant factor, given the 90 watts it consumes while in full-power operation.

No doubt Tesla engineers are evaluating incorporation of Starlink-esque satellite hardware into future vehicles by integrating a “deconstructed” dish into the roof, trunk, and hood, but the option may not be cheap.  An unintegrated dish 1.0 currently lists between $2000 to $2400, based on various industry and media estimates and that’s before the long lead time necessary to integrate antenna elements into a vehicle.

… and possible solutions (for many things beyond Tesla)

If you take Dish 1.0 with high-speed performance and high-power requirements as a starting point, there are three potential areas where it could evolve to be seamless integrated into a vehicle with minimal impact.  Overall size, form factor, and use of higher-performance electronics.  Two of the three areas would lower cost while the third would add it.

Overall size

The current Starlink dish is 59 centimeters (23 inches) across.  If you don’t need to deliver 300 Mbps or faster speeds, the dish “size” could be smaller with fewer elements, which also brings lower cost. 

Form factor

Integrating a single antenna on the roof would be the simplest approach but would not provide redundancy.  There might also be connectivity advantages to integrating antennas on the hood and trunk areas, providing a total of three to track satellites and deliver bandwidth, but at the cost of additional hardware, wiring, and integration with the vehicle electronics. Three wired antennas would also mean more weight.

Higher-performance electronics

Military radars use gallium arsenide (GaAs) and Gallium Nitride (GaN) semiconductor materials to pack higher output power and performance into a smaller space over silicon-based circuits. 

Until recently, GaN was a relatively exotic material to incorporate into consumer electronics, but the material is now being used in chargers and laptop power “bricks,” reducing part count, overall size of the device, and more efficient/faster charging, resulting in less heat.

Migrating Starlink RF electronics (and other electronic bits, for that matter) to GaAs or GaN would provide significant dividends in power usage and physical size. It wouldn’t be cheap, but it is not out of the range of possibility, especially in military and space applications where power and size are critical.

On the other hand, SpaceX’s (and Elon Musk’s) design philosophy is to do one thing in house and very very well.  A leap to widespread use of gallium-based materials may not be in the cards unless it could provide benefits in Starlink dishes and Tesla cars and solar panels.

From 1.0 to a family of dish options

If we place the current Starlink dish as a “middle” option for customers, there are potential opportunities for low-end and higher-end antennas by working with the combination of overall size, form factor, and potentially higher-performance electronics.

Smaller

There are numerous applications that don’t require 100 Mbps or higher speeds, but the business case question for SpaceX is if these smaller/cheaper and lower-cost opportunities would generate sufficient average revenue per user to make it worth the engineering time to build a smaller, more affordable antenna.

For example, a small flat panel patch antenna could enable various IoT and monitoring applications, as well as usage in smaller UAV/drones. Such services would put Starlink in direct competition with Iridium and ORBCOMM, but the spectrum may hold more value by simply being used for high-bandwidth applications.

If there are plans to make Starlink services a part of Tesla vehicles, smaller-sized antennas consuming less power with more flexible/conformal electronics would be necessary.

Larger

Enterprise and government customers with the need for faster downlink speeds, faster uplink speeds, and symmetrical bandwidth all need a bigger dish, especially if they don’t want the overhead and real estate of bonding multiple 1.0 dishes together.  Significantly faster uplink speeds will likely require a larger dish and different FCC licensing.

A larger dish may be required for 10 Gbps downlink services, but there is not enough information to assert if this speed will be achieved through a larger dish or a combination of speed improvements on a per-dish basis in combination with bonding multiple dishes together.

Faster

Incorporating GaAs or GaN into RF and other elements of the Dish 1.0 design would presumably increase the overall performance of a single antenna to support higher data rates up to 10 Gbps without any substantial increases of form factor and electrical power for operation. Overall broadband delivery for 10 Gbps per user would also be dependent upon network and satellite capability.

Narrowing the possibilities

Out of many potential options, a smaller-sized antenna would seem to be the easiest to create from an engineering perspective by simply reducing the transmit/receive elements, but the sophisticated cost and layering to produce Dish 1.0 and the lack of a (currently) business model would seem to preclude any urgency.  However, SpaceX’s discussion with enterprise customers such as the oil and gas industry may change what takes place.

A more power-efficient antenna incorporating gallium-based electronics must be on the roadmap for mobility, national defense, and off-the-grid applications. It maybe nicknamed “The Gold Dish” due to its potentially cost and potentially higher performance, depending on what decisions are made between system power, size, and broadcast power.  The only thing holding back use of gallium would be cost and a large scale commitment to silicon-based tech across Musk’s other projects.

Producing a larger dish at some point would also seem to be a logical step, but there would seem to be less urgency given the ability to bond multiple dishes together for larger amounts of bandwidth in the near term and potentially further Starlink network and RF encoding improvements in the 2022-2023.