Wednesday, June 24, 2020

Questions on the impact of trees on SpaceX Starlink

The Starlink Web site says "Starlink is targeting service in the Northern U.S. and Canada in 2020, rapidly expanding to near-global coverage of the populated world by 2021," but the rollout will not be uniform. The initial coverage will be centered around 53 degrees latitude and locations with unobstructed views of the sky will have an advantage. Like many folks, I am wondering about coverage at my particular location which is in a wooded mountain area at 34.8462° N latitude.

Elon Musk says Starlink setup will be simple -- plug it in and point it to the sky and motors will find the optimal orientation for the antenna. That will work well if you are in the desert with a clear view of the sky, but as you see my house is surrounded by trees. I will have a clear view of satellites when they are directly overhead, but they will disappear over the irregular "tree horizon" as they orbit the Earth.

When might I have uninterrupted connectivity at my house and how fast will it be?

In the most recent publication of which I am aware (plans change), Spacex says that in the very early phases of constellation deployment, when they wish to expand coverage at the expense of capacity, satellites will be able to communicate with terminals at an elevation angle as low as 25 degrees and, as more satellites are launched, they will increase that to 40 degrees, as shown below.

Steerable service range at full deployment (left) and initial launch (right). Source.

A casual glance around my house and consideration of the coverage shown in Serge Eagleson's simulation late last year convinces me that I will not have continuous coverage at 25-degree elevation and quite possibly will not until well into the 40-degree phase. I can't say exactly when I will have coverage because there are too many variables and questions I don't have answers to, like:
  • How quickly will satellite density increase at my latitude?
  • How rapidly will the 10.7-14.5 GHz signals attenuate near the tree-top horizon?
  • Will the satellites dynamically adjust transmit power or is it fixed?
  • When the signal begins to weaken, will the satellite simply drop the link or will it maintain connectivity at reduced speed by retransmission of error packets or changing the modulation algorithm?
  • Will the terminal motors dynamically re-orient the dish while tracking a satellite? (That sounds too expensive).
  • Will the terminal motors automatically re-orient the dish when the constellation configuration changes? (That sounds reasonable).
The impact of trees is not just my personal concern. The economic viability of Starlink will depend upon revenue during the startup years and Elon Musk has stated that Starlink's greatest challenge is the "fully considered" cost of the user terminal -- the cost of hardware, setup and maintenance, and SpaceX President Gwynne Shotwell agrees, saying "hopefully we don't have to hire a million people running around in little white vans fixing or installing user terminals on folkses house."

I don't know what proportion of potential Starlink customers are in forested areas but if SpaceX wants to get us online soon, they might consider offering a mast-mount or rooftop alternative to the plug-and-play terminal -- something along the lines of this OneWeb prototype:

Update 6/26/2020

A reader asked about the effect of rain on Ku-band propagation from a satellite and I found this project, that concluded that "as air temperature, relative humidity, air pressure, and wind speed increased or reduced there is no signal attenuation for weather parameters in both the wet and dry seasons under consideration. There were no statistical correlations between Ku downlink signal strength and the weather parameters under investigation". I'd take that as an optimistic quick answer, but I bet SpaceX engineers could give a more definitive answer.

I wonder how snow would affect the "flying saucer antennas".

Saturday, June 13, 2020

Rural broadband subsidy -- what's the rush?

We are poised to give out $16 billion less than a week before election day ... This approach is not thoughtful policy, it’s rush-it-out-the-door electioneering.
FCC Commissioner Jessica Rosenworcel

The Federal Communications Commission (FCC) has adopted procedures for Phase I of the Rural Digital Opportunity Fund (RDOF) auction, which will award up to $16 billion in support over 10 years for the deployment of fixed broadband networks to homes and businesses in census tracks that are unserved by voice and broadband with download speeds of at least 25 Mbps. Successful bidders in the October reverse auction will have to provide a minimum of 25/3 Mbps up/download speed and the FCC will prioritize low-latency (sub-100 ms) networks when awarding funding. They are prioritizing high speeds and low latency so users will "be able to use tomorrow’s Internet applications as well as today’s."

The FCC considers geostationary satellite providers as high-latency and conventional terrestrial providers as low-latency and last month Ars Technica reported that the FCC planned to classify SpaceX and other low-Earth orbit (LEO) satellite operators as high-latency, saying "the providers haven't proven they can deliver low-latency broadband." However, the FCC order issued on June 11 equivocated a bit, saying the FCC has"serious doubts that any LEO networks will be able to meet the short-form application requirements for bidding in the low-latency tier." (The short-form applications are due by July 15, 2020).

If the goal is to support applications that are in use ten years from now, it seems they should consider technologies that will be available ten years from now and the FCC is well aware of the LEO broadband projects of SpaceX, Amazon, Telesat, and several Chinese companies and the possibility that OneWeb will be resurrected. These efforts may turn out to be market failures, but we have both theoretical and experimental evidence to expect that, if successful, they will be able to deliver sub-100 ms latency.

For example, this simulation predicts latencies well under 100 ms even without the inter-satellite laser links that SpaceX plans to introduce in the future and Elon Musk has also reported low latencies with their initial test satellites.

More recently, Telefónica, a major Internet service provider in Latin America and Europe has completed tests with Telesat. (Telefonica owns the Movistar, O2 and Vivo brands). The test scenarios included high-definition video streaming, video conferencing with teams, remote desktop connection to seamlessly manage a remote computer, a VPN connection without any delay or outages, FTP-encrypted file transfers of 2 GB in both directions, and IPSec tunnel encryption with no reduction in the performance of the link. This was done without TCP acceleration or data compression and hey achieved round-trip latencies of 30-60ms with no packet loss.

What's the rush?

Why not allocate $1.6 billion of the ten-year fund this year and the remainder a year or two later when we will have operational data from one or more LEO satellite ISPs? Doing so might require some bureaucratic adjustment and would complicate the planning and bidding process for potential ISPs, but if we are thinking ten-years in the future, why not consider technology that will quite possibly be important in ten years?

Sadly, the rush is political.

For a start, the FCC Chairman came from Verizon and I bet the terrestrial Internet service providers have contributed to Trump and their senators and lobbied the FCC to classify LEO satellites as high-latency.

Furthermore, three of the five FCC commissioners are Republicans and you can read their statements on the RDOF bidding process here. It turns out that the two Democrats were able to convince one of the Republicans to change the draft ruling to allow for the unlikely possibility that LEO satellites might qualify as low-latency providers as noted above.

However, the Democrats were unable to convince any Republican colleagues to postpone the funding until accurate coverage data could be collected and wrote "dissents in part." Commissioner Geoffrey Starks expressed concern over the decision "to spend such a large portion of the budget -- over such a long term of support -- based on broadband maps that are not accurate" and Commissioner Jessica Rosenworcel wrote a stronger dissent including suggestions for addressing the urban and rural digital divides and pointing out that "We are poised to give out $16 billion less than a week before election day ... This approach is not thoughtful policy, it’s rush-it-out-the-door electioneering."

Republican candidates will rush to take credit for bringing the Internet to millions of their rural constituents. Cunning.

Monday, June 08, 2020

Can SpaceX launch 30,000 second-generation Starlink satellites? Maybe.

The bottom line is that success is not guaranteed, but neither is failure -- there is a non-zero probability of success.

On May 26th, SpaceX applied for permission to launch 30,000 "second-generation" Starlink broadband Internet satellites. (Note that the software on Starlink satellites is updated about once a week). The application narrative states that the second-generation satellites will be configured as follows:                                                                                                                                                                                                                                                                                                                                     

(The offsets of the single-satellite planes are set to form a uniform pattern so when a satellite crosses the equator another satellite in the adjacent plane will cross the equator a short, constant time later). The following snapshots from a simulation created by Richard Cole illustrate the coverage.

The application narrative describes the ground and space segments and addresses the problems of debris mitigation and interference with astronomical observation. Here are some points that caught my eye
  • The second-generation satellites "will have three times the data capacity of SpaceX’s current satellites." 
  • Polar regions are covered.
  • The second-generation satellites will have optical inter-satellite links.
  • Low altitudes "will enable smaller spot beams and greater satellite diversity, achieving the intensive frequency reuse needed to heighten capacity available anywhere in the world."
  • Low altitudes will allow SpaceX to use high E-band frequencies for communication with ground stations.
  • Low altitudes will reduce latency.
  • Terminal setup will be plug and play -- "point it at the sky and plug it in."
  • SpaceX promises to make sure that "Starlink has no material effect on discoveries in astronomy."
  • SpaceX will extend its debris data sharing and collision-avoidance activities and "encourages all operators to follow these same practices."
Can they pull it off?
The FCC demands that half of the satellites for an approved constellation be launched within 6 years and all to be launched within 9 years. Can SpaceX manufacture, launch, and fund 30,000 second-generation satellites that quickly while continuing to launch first-generation satellites and replacements for those that are de-orbited after approximately five years of useful life?
When asked about Starlink during an interview at the Satellite 2020 Conference in March, Elon Musk said his goal for Starlink was to remain in the "not bankrupt category." If Elon is not sure, I can't be either, but they have a few positive things going for them.
Today, SpaceX is manufacturing about 120 satellites per month, which is far too few to satisfy the FCC. That being said, it is safe to say that Elon Musk knows more about modern, automated manufacturing than anyone alive today, having learned from his experience making cars, solar tiles, batteries, satellites, and rockets. He has learned the importance of building "the machine that builds the machines," which includes the factory, equipment, staffing, processes, and supply chain. SpaceX may already be able to make satellites faster than 120 per month and they will surely improve the manufacturing "machine" and the design of second-generation satellites for manufacturing ease. If that is not sufficient, they can open another factory -- maybe acquire OneWeb's.
How about launching the satellites? SpaceX President and COO Gwynne Shotwell said in a recent interview that if their Starship was not ready to take astronauts to the International Space Station within three years, it would be "a major company fail" and in an earlier interview, she said that Starships will be able to carry 400 satellites at a time. In his Satelite 2020 interview, Elon Musk set a goal of three launches per day with Starships landing back at the launch sites within a few days of launch. If they can achieve that cadence, launching will not be a problem.
Can they fund the project? It will take around three years for SpaceX to complete the first phase of the first-generation satellites and Shotwell and Musk have both recently said that low-cost end-user terminals are critical for global success and those are perhaps three years away. During those three years, SpaceX will have income from their launch business, government contracts, and Starlink customers as well as private investment.
They have been collaborating on Starship with NASA for six years and recently flew two astronauts to the International Space Station. Colonel Eric Felt, head of the Air Force Research Laboratory’s Space Vehicles Directorate, characterized that as the "culmination of perhaps the most successful private-public partnership of all times” so it seems safe to say they will be getting funding from the Defence Department as well as NASA.
How about private investment? SpaceX has succeeded in attracting Starlink investors recently, but Starlink's financial performance during the first three years of operation will impact their ability to continue raising capital. During the next three years, the majority of the satellites they launch will be at latitudes that deliver the most capacity in the relatively affluent regions of North America and Europe, and Internet service in the US and Canada is relatively expensive. Hopefully, they will be able to serve those markets without losing too much on expensive terminals while improving terminal technology. (They may also be interested in acquiring OneWeb antenna technology).
Debris is my biggest worry. The SpaceX application says their propulsion system will allow them to autonomously avoid collisions with tracked objects. Furthermore, over 85% of their satellites will be lower than the International Space Station and will be relatively quick to burn up in the atmosphere. They also promise to share all ephemeris data and encourage others to do the same.
But, what about objects that are too small to track? What if another operator is actively trying to avoid collisions -- don't they need to coordinate with SpaceX? What about other companies that are planning to launch low-Earth orbit satellites? With the numbers of satellites being launched, it seems that international regulation and coordination are required and even then a debris solution is not obvious. Orbiting debris is clearly a global (not US) tragedy of the commons.
The bottom line is that success is not guaranteed, but neither is failure -- there is a non-zero probability of success.